JP3639145B2 - Abnormal point detection device - Google Patents

Description

【００２０】
で表わされる周波数特性Ｈ₁（ｆ）を持ち、漏洩音の受信信号から求めたクロススペクトルをフィルタリングするためのフィルタＡと、
上記コヒーレンシィγ（ｆ）の絶対値に関する予め決められた閾値をγ_l（ｆ）として、
【００２１】
【数２６】

【００２２】
で表わされる周波数特性Ｈ₂（ｆ）を持つフィルタＢと、
ｊを０以上のある実数、上記３つの超音波センサの内、２つのセンサで受信した受信信号から求めたクロススペクトルをＣ（ｆ）として、
【００２３】
【数２７】

【００２４】
で表わされる周波数特性Ｈ₃（ｆ）を持つフィルタＣと、
ｋを１以上のある実数として、
【００２５】
【数２８】

【００２６】
で表わされる周波数特性Ｈ₄（ｆ）を持つフィルタＤと、
上記４つの式の内、少なくともｌｏｇ_j（|Ｃ（ｆ）|）と|Ｃ（ｆ）|^1/kのいずれか１つの周波数特性と、ＳＣＯＴ（Smoothed Coherence Transform）フィルタの周波数特性１／√（Ｃ₁₁（ｆ）・Ｃ₂₂（ｆ））とをそれぞれ重み付けし、重み付けしたこれらの周波数特性を乗じた周波数特性を持つフィルタＥと
の内、一つまたは二つ以上のフィルタを含む前処理部と、
上記前処理部により前処理された信号から相互相関関数を演算する相関処理部とを備えたことを特徴とするものである。
【００２７】
また、上記フィルタＥは、｜γ（ｆ）｜ⁱの周波数特性、ｌｏｇ_j（|Ｃ（ｆ）|）、および|Ｃ（ｆ）|^1/kの内、いずれか１つの周波数特性と、ＳＣＯＴフィルタの周波数特性とをそれぞれ重み付けし、重み付けしたこれらの周波数特性を乗じた周波数特性を持つフィルタ、またはｌｏｇ_j（|Ｃ（ｆ）|）、および|Ｃ（ｆ）|^1/kの内、いずれか１つの周波数特性において、コヒーレンシィγ（ｆ）の絶対値がある閾値以下になる周波数成分が０とするような周波数特性と、ＳＣＯＴフィルタの周波数特性とをそれぞれ重み付けし、重み付けしたこれらの周波数特性を乗じた周波数特性を持つフィルタのいずれかであることを特徴とするものである。
【００２８】
また、上記前処理部は、下限周波数をｆｓ、上限周波数をｆｅとして、
【００２９】
【数２９】

【００３０】
で表わされる周波数特性Ｈ₅（ｆ）を持つバンドパスフィルタをさらに含むことを特徴とするものである。
【００３１】
また、上記前処理部は、異常箇所が存在する場合としない場合の予備実験から得られる統計データに基づいて上記実数ｉ、ｊ、ｋおよび上記コヒーレンシィγ（ｆ）の絶対値に関する予め決められた閾値γ_l（ｆ）を決定することを特徴とするものである。
【００３２】
また、上記信号処理部は、上記相関処理部により演算された相互相関関数の包絡線を求め、さらに予め決められた繰り返し回数だけ求められた上記相互相関関数の包絡線を平均化する後処理部をさらに備えたことを特徴とするものである。
【００３３】
また、上記後処理部は、上記平均化された相互相関関数の包絡線のピーク値と予め決められたある閾値との大小関係から異常箇所の有無を判定することを特徴とするものである。
【００３４】
また、上記後処理部は、３つの超音波センサの内、ある２つの超音波センサで受信した受信信号についての上記平均化された相互相関関数の包絡線がピークになる時間と、３つの超音波センサの内、上記２つの超音波センサの片方のセンサと上記２つの超音波センサ以外の超音波センサとで受信した受信信号についての上記平均化された相互相関関数の包絡線がピークになる時間と、上記３つの超音波センサの各離間距離とから異常箇所の位置を特定することを特徴とするものである。
【００３５】
また、上記後処理部は、異常箇所が存在する場合としない場合の予備実験から得られる統計データに基づいて上記平均化された相互相関関数の包絡線のピーク値に関する閾値を決定することを特徴とするものである。
【００３６】
また、他の発明に係る異常箇所検出装置は、被検査管に異常箇所が存在することにより発生する漏洩音を受信するための２つの超音波センサを備えると共に、これら超音波センサによる漏洩音の受信信号をそれぞれ受信部を介して入力し信号処理する信号処理部に、ｆを受信信号の周波数、ｉを０以上のある実数、上記２つの超音波センサで受信した受信信号から求めたコヒーレンシィをγ（ｆ）、上記２つのセンサの内の片方の超音波センサで受信した受信信号から求めたパワースペクトルをＣ₁₁（ｆ）、上記２つの超音波センサの他方で受信した受信信号から求めたパワースペクトルをＣ₂₂（ｆ）として、
【００３７】
【数３０】

【００３８】
で表わされる周波数特性Ｈ₁（ｆ）を持ち、漏洩音の受信信号から求めたクロススペクトルをフィルタリングするためのフィルタＡと、上記コヒーレンシィγ（ｆ）の絶対値に関する予め決められた閾値をγ_l（ｆ）として、
【００３９】
【数３１】

【００４０】
で表わされる周波数特性Ｈ₂（ｆ）を持つフィルタＢと、ｊを０以上のある実数、上記２つのセンサで受信した受信信号から求めたクロススペクトルをＣ（ｆ）として、
【００４１】
【数３２】

【００４２】
で表わされる周波数特性Ｈ₃（ｆ）を持つフィルタＣと、ｋを１以上のある実数として、
【００４３】
【数３３】

【００４４】
で表わされる周波数特性Ｈ₄（ｆ）を持つフィルタＤと、上記４つの式の内、少なくともｌｏｇ_j（|Ｃ（ｆ）|）と|Ｃ（ｆ）|^1/kのいずれか１つの周波数特性と、ＳＣＯＴフィルタの周波数特性１／√（Ｃ₁₁（ｆ）・Ｃ₂₂（ｆ））とをそれぞれ重み付けし、重み付けしたこれらの周波数特性を乗じた周波数特性を持つフィルタＥとの内、一つまたは二つ以上のフィルタを含む前処理部と、上記前処理部により前処理された信号から相互相関関数を演算する相関処理部とを備えたことを特徴とするものである。
【００４５】
また、上記フィルタＥは、｜γ（ｆ）｜ⁱの周波数特性、ｌｏｇ_j（|Ｃ（ｆ）|）、および|Ｃ（ｆ）|^1/kの内、いずれか１つの周波数特性と、ＳＣＯＴフィルタの周波数特性とをそれぞれ重み付けし、重み付けしたこれらの周波数特性を乗じた周波数特性を持つフィルタ、またはｌｏｇ_j（|Ｃ（ｆ）|）、および|Ｃ（ｆ）|^1/kの内、いずれか１つの周波数特性において、コヒーレンシィγ（ｆ）の絶対値がある閾値以下になる周波数成分が０とするような周波数特性と、ＳＣＯＴ（Smoothed Coherence Transform）フィルタの周波数特性とをそれぞれ重み付けし、重み付けしたこれらの周波数特性を乗じた周波数特性を持つフィルタのいずれかであることを特徴とするものである。
【００４６】
また、上記前処理部は、下限周波数をｆｓ、上限周波数をｆｅとして、
【００４７】
【数３４】

【００４８】
で表わされる周波数特性Ｈ₅（ｆ）を持つバンドパスフィルタをさらに含むことを特徴とするものである。
【００４９】
また、上記前処理部は、異常箇所が存在する場合としない場合の予備実験から得られる統計データに基づいて上記実数ｉ、ｊ、ｋおよび上記コヒーレンシィγ（ｆ）の絶対値に関する予め決められた閾値γ_l（ｆ）を決定することを特徴とするものである。
【００５０】
また、上記信号処理部は、上記相関処理部により演算された相互相関関数の包絡線を求め、さらに予め決められた繰り返し回数だけ求められた上記相互相関関数の包絡線を平均化する後処理部をさらに備えたことを特徴とするものである。
【００５１】
また、上記後処理部は、上記平均化された相互相関関数の包絡線のピーク値と予め決められたある閾値との大小関係から異常箇所の有無を判定することを特徴とするものである。
【００５２】
また、上記後処理部は、２つの超音波センサで受信した受信信号についての上記平均化された相互相関関数の包絡線がピークになる時間と、被検査管を漏洩音を伝搬するときの伝搬速度と、上記２つの超音波センサの各離間距離とから異常箇所の位置を特定することを特徴とするものである。
【００５３】
さらに、上記後処理部は、異常箇所が存在する場合としない場合の予備実験から得られる統計データに基づいて上記平均化された相互相関関数の包絡線のピーク値に関する閾値を決定することを特徴とするものである。
【００５４】
また、他の発明に係る異常箇所検出装置は、被検査管に異常箇所が存在することにより発生する漏洩音を受信するための３つの超音波センサを備えると共に、これらの超音波センサによる漏洩音の受信信号をそれぞれ受信部を介して入力し信号処理する信号処理部に、ｆを受信信号の周波数、ｉを０以上のある実数、上記３つの超音波センサの内、ある２つのセンサで受信した受信信号から求めたコヒーレンシィをγ（ｆ）、上記２つのセンサの内の片方の超音波センサで受信した受信信号から求めたパワースペクトルをＣ₁₁（ｆ）、上記２つの超音波センサの他方で受信した受信信号から求めたパワースペクトルをＣ₂₂（ｆ）として、
【００５５】
【数３５】

【００５６】
で表わされる周波数特性Ｈ₁（ｆ）を持ち、漏洩音の受信信号から求めたクロススペクトルをフィルタリングするためのフィルタＡと、上記コヒーレンシィγ（ｆ）の絶対値に関する予め決められた閾値をγ₁（ｆ）として、
【００５７】
【数３６】

【００５８】
で表わされる周波数特性Ｈ₂（ｆ）を持つフィルタＢと、ｊを０以上のある実数、上記３つの超音波センサの内、２つのセンサで受信した受信信号から求めたクロススペクトルをＣ（ｆ）として、
【００５９】
【数３７】

【００６０】
で表わされる周波数特性Ｈ₃（ｆ）を持つフィルタＣと、ｋを１以上のある実数として、
【００６１】
【数３８】

【００６２】
で表わされる周波数特性Ｈ₄（ｆ）を持つフィルタＤと、σ_abs（ｆ）を上記３つの超音波センサの内、ある２つのセンサで受信した２つの受信信号の周波数スペクトルの絶対値の比の分割された複数個の受信信号間の分散とし、σ_φ（ｆ）を上記２つのセンサで受信した２つの受信信号の周波数スペクトルの位相差の分割された複数個の受信信号間の分散として、
【００６３】
【数３９】

【００６４】
で表わされる周波数特性Ｈ₇（ｆ）を持つフィルタＥと、上記３つの超音波センサの内、ある２つのセンサで受信した２つの受信信号の周波数スペクトルの絶対値の比の分割された複数個の受信信号間の分散σ_abs（ｆ）と、上記２つのセンサで受信した２つの受信信号の周波数スペクトルの位相差の分割された複数個の受信信号間の分散σ_φ（ｆ）との積に関する閾値をσ_l（ｆ）とし、
【００６５】
【数４０】

【００６６】
で表わされる周波数特性Ｈ₈（ｆ）を持つフィルタＦと、上記６つの式の内、少なくともｌｏｇ_j（|Ｃ（ｆ）|）と|Ｃ（ｆ）|^1/kのいずれか１つの周波数特性と、ＳＣＯＴ（Smoothed Coherence Transform）フィルタの周波数特性とをそれぞれ重み付けし、重み付けしたこれらの周波数特性を乗じた周波数特性を持つフィルタＧとの内、一つまたは二つ以上のフィルタを含む前処理部と、上記前処理部により前処理された信号から相互相関関数を演算する相関処理部とを備えたことを特徴とするものである。
【００６７】
また、上記フィルタＧは、｜γ（ｆ）｜ⁱおよび｛１／（σ_abs（ｆ）・σ_φ（ｆ））｝ⁱの内、いずれか一つの周波数特性と、ｌｏｇ_j（|Ｃ（ｆ）|）、および|Ｃ（ｆ）|^1/kの内、いずれか一つの周波数特性と、ＳＣＯＴフィルタの周波数特性とをそれぞれ重み付けし、重み付けしたこれらの周波数特性を乗じた周波数特性を持つフィルタ、または、ｌｏｇ_j（|Ｃ（ｆ）|）、および|Ｃ（ｆ）|^1/kの内、いずれか一つの周波数特性において、コヒーレンシィγ（ｆ）の絶対値が予め決められたある閾値以下になる周波数成分が０とするような周波数特性、もしくは、上記３つの超音波センサの内、ある２つのセンサで受信した２つの受信信号の周波数スペクトルの絶対値の比の分割された複数個の受信信号間の分散σ_abs（ｆ）と、上記２つのセンサで受信した２つの受信信号の周波数スペクトルの位相差の分割された複数個の受信信号間の分散σ_φ（ｆ）との積関する閾値σ_l（ｆ）がある閾値以上になる周波数成分が０とするような周波数特性と、ＳＣＯＴフィルタの周波数特性とをそれぞれ重み付けし、重み付けしたこれらの周波数特性を乗じた周波数特性を持つフィルタのいずれかであることを特徴とするものである。
【００６８】
また、上記前処理部は、下限周波数をｆｓ、上限周波数をｆｅとして、
【００６９】
【数４１】

【００７０】
で表わされる周波数特性Ｈ₅（ｆ）を持つバンドパスフィルタをさらに含むことを特徴とするものである。
【００７１】
また、上記前処理部は、異常箇所が存在する場合としない場合の予備実験から得られる統計データに基づいて上記実数ｉ、ｊ、ｋおよび上記コヒーレンシィγ（ｆ）の絶対値に関する予め決められた閾値γ_l（ｆ）、上記３つの超音波センサの内、ある２つのセンサで受信した２つの受信信号の周波数スペクトルの絶対値の比の分割された複数個の受信信号間の分散σ_abs（ｆ）と、上記２つのセンサで受信した２つの受信信号の周波数スペクトルの位相差の分割された複数個の受信信号間の分散σ_φ（ｆ）との積に関する閾値σ_l（ｆ）を決定することを特徴とするものである。
【００７２】
また、上記信号処理部は、上記相関処理部により演算された相互相関関数の包絡線を求め、さらに予め決められた繰り返し回数だけ求められた上記相互相関関数の包絡線を平均化する後処理部をさらに備えたことを特徴とするものである。
【００７３】
また、上記後処理部は、上記平均化された相互相関関数の包絡線のピーク値と、予め決められたある閾値との大小関係から異常箇所の有無を判定することを特徴とするものである。
【００７４】
また、上記後処理部は、３つの超音波センサの内、ある２つの超音波センサで受信した受信信号についての上記平均化された相互相関関数の包絡線がピークになる時間と、３つの超音波センサの内、上記２つの超音波センサの片方のセンサと上記２つの超音波センサ以外の超音波センサとで受信した受信信号についての上記平均化された相互相関関数の包絡線がピークになる時間と、上記３つの超音波センサの各離間距離とから異常箇所の位置を特定することを特徴とするものである。
【００７５】
また、上記後処理部は、異常箇所が存在する場合としない場合の予備実験から得られる統計データに基づいて上記平均化された相互相関関数の包絡線のピーク値に関する閾値を決定することを特徴とするものである。
【００７６】
また、さらに他の発明に係る異常箇所検出装置は、被検査管に異常箇所が存在することにより発生する漏洩音を受信するための２つの超音波センサを備えると共に、これら超音波センサによる漏洩音の受信信号をそれぞれ受信部を介して入力し信号処理する信号処理部に、ｆを受信信号の周波数、ｉを０以上のある実数、上記２つの超音波センサで受信した受信信号から求めたコヒーレンシィをγ（ｆ）、上記２つのセンサの内の片方の超音波センサで受信した受信信号から求めたパワースペクトルをＣ₁₁（ｆ）とし、上記２つの超音波センサの他方で受信した受信信号から求めたパワースペクトルをＣ₂₂（ｆ）として、
【００７７】
【数４２】

で表わされる周波数特性Ｈ₁（ｆ）を持ち、漏洩音の受信信号から求めたクロススペクトルをフィルタリングするためのフィルタＡと、上記コヒーレンシィγ（ｆ）の絶対値に関する予め決められた閾値をγ_l（ｆ）として、
【００７８】
【数４３】

【００７９】
で表わされる周波数特性Ｈ₂（ｆ）を持つフィルタＢと、ｊを０以上のある実数、上記２つのセンサで受信した受信信号から求めたクロススペクトルをＣ（ｆ）として、
【００８０】
【数４４】

【００８１】
で表わされる周波数特性Ｈ₃（ｆ）を持つフィルタＣと、ｋを１以上のある実数として、
【００８２】
【数４５】

【００８３】
で表わされる周波数特性Ｈ₄（ｆ）を持つフィルタＤと、σ_abs（ｆ）を上記２つのセンサで受信した２つの受信信号の周波数スペクトルの絶対値の比の分割された複数個の受信信号間の分散とし、σ_φ（ｆ）を上記２つのセンサで受信した２つの受信信号の周波数スペクトルの位相差の分割された複数個の受信信号間の分散として、
【００８４】
【数４６】

【００８５】
で表わされる周波数特性Ｈ₇（ｆ）を持つフィルタＥと、上記２つのセンサで受信した２つの受信信号の周波数スペクトルの絶対値の比の分割された複数個の受信信号間の分散σ_abs（ｆ）と、上記２つのセンサで受信した２つの受信信号の周波数スペクトルの位相差の分割された複数個の受信信号間の分散σ_φ（ｆ）との積に関する閾値をσ_l（ｆ）とし、
【００８６】
【数４７】

【００８７】
で表わされる周波数特性Ｈ₈（ｆ）を持つフィルタＦと、上記６つの式の内、少なくともｌｏｇ_j（|Ｃ（ｆ）|）と|Ｃ（ｆ）|^1/kのいずれか１つの周波数特性と、ＳＣＯＴ（Smoothed Coherence Transform）フィルタの周波数特性とをそれぞれ重み付けし、重み付けしたこれらの周波数特性を乗じた周波数特性を持つフィルタＧとの内、一つまたは二つ以上のフィルタを含む前処理部と、上記前処理部により前処理された信号から相互相関関数を演算する相関処理部とを備えたことを特徴とするものである。
【００８８】
また、上記フィルタＧは、｜γ（ｆ）｜ⁱおよび｛１／（σ_abs（ｆ）・σ_φ（ｆ））｝ⁱの内、いずれか一つの周波数特性と、ｌｏｇ_j（|Ｃ（ｆ）|）、および|Ｃ（ｆ）|^1/kの内、いずれか一つの周波数特性と、ＳＣＯＴフィルタの周波数特性とをそれぞれ重み付けし、重み付けしたこれらの周波数特性を乗じた周波数特性を持つフィルタ、またはｌｏｇ_j（|Ｃ（ｆ）|）、および|Ｃ（ｆ）|^1/kの内、いずれか一つの周波数特性において、コヒーレンシィγ（ｆ）の絶対値が予め決められたある閾値以下になる周波数成分が０となるような周波数特性、もしくは上記２つのセンサで受信した２つの受信信号の周波数スペクトルの絶対値の比の分割された複数個の受信信号間の分散σ_abs（ｆ）と、上記２つのセンサで受信した２つの受信信号の周波数スペクトルの位相差の分割された複数個の受信信号間の分散σ_φ（ｆ）との積に関する閾値σ_l（ｆ）が、予め決められたある閾値以上になる周波数成分が０とするような周波数特性と、ＳＣＯＴフィルタの周波数特性とをそれぞれ重み付けし、重み付けしたこれらの周波数特性を乗じた周波数特性を持つフィルタのいずれかであることを特徴とするものである。
【００８９】
また、上記前処理部は、下限周波数をｆｓ、上限周波数をｆｅとして、
【００９０】
【数４８】

【００９１】
で表わされる周波数特性Ｈ₅（ｆ）を持つバンドパスフィルタをさらに含むことを特徴とするものである。
【００９２】
また、上記前処理部は、異常箇所が存在する場合としない場合の予備実験から得られる統計データに基づいて上記実数ｉ、ｊ、ｋ、上記コヒーレンシィγ（ｆ）の絶対値に関する予め決められた閾値γ_l（ｆ）、および上記２つのセンサで受信した２つの受信信号の周波数スペクトルの絶対値の比の分割された複数個の受信信号間の分散σ_abs（ｆ）と、上記２つのセンサで受信した２つの受信信号の周波数スペクトルの位相差の分割された複数個の受信信号間の分散σ_φ（ｆ）との積に関する閾値σ_l（ｆ）を決定することを特徴とするものである。
【００９３】
また、上記信号処理部は、上記相関処理部により演算された相互相関関数の包絡線を求め、さらに予め決められた繰り返し回数だけ求められた上記相互相関関数の包絡線を平均化する後処理部を備えたことを特徴とするものである。
【００９４】
また、上記後処理部は、上記平均化された相互相関関数の包絡線のピーク値と予め決められたある閾値との大小関係から異常箇所の有無を判定することを特徴とするものである。
【００９５】
また、上記後処理部は、２つの超音波センサで受信した受信信号についての上記平均化された相互相関関数の包絡線がピークになる時間と、被検査管を漏洩音が伝搬するときの伝搬速度と、上記２つの超音波センサの各離間距離とから異常箇所の位置を特定することを特徴とするものである。
【００９６】
さらに、上記後処理部は、異常箇所が存在する場合としない場合の予備実験から得られる統計データに基づいて上記平均化された相互相関関数の包絡線のピーク値に関する閾値を決定することを特徴とするものである。
【００９７】
【発明の実施の形態】
実施の形態１．
この発明の実施の形態１に係る異常箇所検出装置について図１から図３４を参照しながら説明する。
図１はこの発明の実施の形態１に係る異常箇所検出装置を示す構成図である。図１において、１は水、油、あるいはその他の液体またはガス等気体を通す被検査管としての導管、２は上記導管１の異常箇所、６は地中である。
【００９８】
また、図１に示す異常箇所検出装置は、異常箇所２が存在することにより発生する漏洩音を受信するための超音波センサ３ａ、３ｂおよび３ｃと、受信装置７とを備えている。
この図１においては、上記導管１が上記地中６に埋もれている場合について示しているが、上記導管１は、そのすべての部分もしくはその一部分が上記地中６より上に存在していても構わない。また、上記異常箇所２が１箇所である場合について述べているが、上記異常箇所２は１箇所でなくても、複数箇所でも構わない。
【００９９】
また、図１において、Ｌ１は超音波センサ３ａと超音波センサ３ｂの間の導管１に沿った距離、Ｌ２は超音波センサ３ｂと超音波センサ３ｃの間の導管１に沿った距離であり、ｘは超音波センサ３ａから異常箇所２までの導管１に沿った距離である。
【０１００】
なお、図１において、導管１の中の水、油、あるいはその他の液体またはガス等の気体は、流れていても、流れていなくても構わない。また、流れている場合には、流れの方向はどちら向きでも構わない。
【０１０１】
なお、超音波は、人間の耳に聞こえない程度に高い周波数の音波や弾性波を指す言葉として使われるが、この発明では、周波数は特に規定しないものとする。すなわち、この発明における「超音波」という文言には、人間の耳で聞こえる周波数の上限の限界よりも高い周波数の音波や弾性波に限らず、この上限よりも低い周波数の音波や弾性波も含めた波という意味を含んでおり、無論、人間の耳で聞こえる周波数の下限の限界よりも低い周波数の音波や弾性波という意味も含む。
【０１０２】
また、図１においては、超音波センサ３ａ、３ｂおよび３ｃが導管１に当てて置かれている場合を示しているが、上記超音波センサ３ａ、３ｂおよび３ｃは、上記導管１の３箇所において漏洩音を受信することが目的であり、この目的が達成できるならば、上記超音波センサ３ａ、３ｂおよび３ｃは、上記導管１に直接接触していなくても構わない。また、この目的が達成できるならば、上記超音波センサ３ａ、３ｂおよび３ｃは、上記導管１の内部に配置されても構わない。
【０１０３】
ここで、異常箇所２と、超音波センサ３ａ、３ｂ、および３ｃとの位置関係について説明する。
３つの超音波センサの内の２つの超音波センサ、例えば図１における超音波センサ３ｂと３ｃとの間は、地表上に露出していて、その間に異常箇所がないことが目視により判断できる領域等の、異常箇所２がないことが既知である領域である。異常箇所２は、３つの超音波センサの内の上記２つの超音波センサ以外の超音波センサ、例えば図１における超音波センサ３ａと、上記２つの超音波センサの内、上記２つの超音波センサ以外の超音波センサに近い方の超音波センサ、例えば、図１における超音波センサ３ｂとの間に位置している。
【０１０４】
図１において、受信装置７は、３つの受信部７１と、信号処理部７２と、報知手段としての表示部７３と、制御部７４とを含む。
超音波センサ３ａ、３ｂおよび３ｃは、受信部７１に接続されている。受信部７１は信号処理部７２に接続されている。信号処理部７２は表示部７３に接続されている。
【０１０５】
制御部７４は、受信部７１、信号処理部７２、および表示部７３に接続されており、検査を行うための情報やコマンドが入力され、また、受信部７１、信号処理部７２、および表示部７３に対し、これらの動作を制御するための制御信号や、検査の進行状況の情報に関する信号を、逐次送受信してこれらの制御を司る。
【０１０６】
また、受信部７１は、図示はしないが、受信信号を増幅するためのアンプと、Ａ／Ｄ変換部とを含む。
信号処理部７２は、受信信号をフィルタリングするための前処理部７２ａと、超音波センサ３ａ、３ｂおよび３ｃで受信した３つの受信信号の内の２つの信号の相互相関関数を計算するための相関処理部７２ｂと、相互相関関数に包絡線検波等の処理を行うための後処理部７２ｃとを含む。前処理部７２ａは、相関処理部７２ｂに接続されており、相関処理部７２ｂは、後処理部７２ｃに接続されている。また、上記信号処理部７２は、図示はしていないが、内部にメモリを有する。このメモリに演算処理された種々の結果が適宜記憶される。
【０１０７】
また、信号処理部７２における前処理部７２ａは、周波数をｆとし、ｉを０以上のある実数とし、上記超音波センサ３ａ、３ｂ、および３ｃの内、ある２つのセンサで受信した受信信号から求めたコヒーレンシィをγ（ｆ）とし、上記２つのセンサの内の一方で受信した受信信号から求めたパワースペクトルをＣ₁₁（ｆ）とし、上記２つのセンサの内の他方で受信した受信信号から求めたパワースペクトルをＣ₂₂（ｆ）として、
【０１０８】
【数４９】

【０１０９】
で表わされる周波数特性Ｈ₁（ｆ）を持つフィルタと、
コヒーレンシィγ（ｆ）の絶対値に関する閾値をγ_l（ｆ）として、
【０１１０】
【数５０】

【０１１１】
で表わされる周波数特性Ｈ₂（ｆ）を持つフィルタと、
ｊを０以上のある実数とし、上記超音波センサ３ａ、３ｂ、および３ｃの内、上記２つのセンサで受信した受信信号から求めたクロススペクトルをＣ（ｆ）として、
【０１１２】
【数５１】

【０１１３】
で表わされる周波数特性Ｈ₃（ｆ）を持つフィルタと、
ｋを１以上のある実数として、
【０１１４】
【数５２】

【０１１５】
で表わされる周波数特性Ｈ₄（ｆ）を持つフィルタと、
式（１）における｜γ（ｆ）｜ⁱの周波数特性と、式（３）におけるｌｏｇ_j（|Ｃ（ｆ）|）、および式（４）における|Ｃ（ｆ）|^1/kの内、いずれか一方の周波数特性と、ＳＣＯＴフィルタの周波数特性とを重み付けし、これらを乗じた周波数特性を持つフィルタや、式（３）におけるｌｏｇ_j（|Ｃ（ｆ）|）、および式（４）における|Ｃ（ｆ）|^1/kの内、いずれか一方の周波数特性において、式（２）のように、コヒーレンシィγ（ｆ）の絶対値がある閾値以下になる周波数成分が０とするような周波数特性と、ＳＣＯＴフィルタの周波数特性とを重み付けし、これらを乗じた周波数特性を持つフィルタと
の内、いずれか一つあるいは２つ以上を含んでいる。
さらに、下限周波数をｆｓとし、上限周波数をｆｅとして、
【０１１６】
【数５３】

【０１１７】
で表わされる周波数特性Ｈ₅（ｆ）を持つバンドパスフィルタを含んでいる。
【０１１８】
上記前処理部７２ａが含んでいるフィルタ特性において、式（５）における下限周波数ｆｓおよび上限周波数ｆｅは、予備実験の結果に応じて決定され、制御部７４にこれら値が入力される。また、式（１）における実数ｉ、式（２）におけるコヒーレンシィの絶対値に関する閾値γ_l（ｆ）、式（３）における実数ｊ、および式（４）における実数ｋも、同様の予備実験の結果に応じて決定される。
【０１１９】
この予備実験は、異常箇所２が存在する場合と、実質上、存在しない場合について、この実施の形態１に係る異常箇所検出装置と同じか、または、同様の異常箇所検出装置を用いて行われる。このような予備実験から得られた統計データから、検査の状況に応じて異常箇所２の有無や異常箇所２の位置の検出を最も精度よく行えるような、式（１）における実数ｉ、式（２）におけるコヒーレンシィの絶対値に関する閾値γ_l（ｆ）、式（３）における実数ｊ、式（４）における実数ｋ、および式（５）における下限周波数ｆｓと上限周波数ｆｅが決められる。
【０１２０】
なお、後述するが、漏洩音が導管１を伝搬するときの伝搬速度が既知であれば、超音波センサ３ａ、３ｂ、３ｃの３つの内、２つの超音波センサの間が、地表上に露出していてその間に異常箇所がないことが目視により判断できる領域等の、異常箇所２がないことが既知である領域である２つの超音波センサの内の一つ、例えば図１における超音波センサ３ｃを取り除いても、異常箇所２の有無、異常箇所２の位置の特定を行うことができる。このような場合であれば、超音波センサ３ａ、３ｂ、および３ｃの内の一つを取り除くことにより、検査を容易にし、検査システムを廉価にできるという効果がある。
【０１２１】
ここで、漏洩音の特徴および雑音の特徴について説明する。
漏洩音には、一定位置の発信源からの時間的に連続な信号であるという特徴がある。
一方、例えば水道管の漏水検査においてしばしば受信される自動車の走行音、一時的な水道使用音等の雑音は、移動する発信源からの信号であるか、もしくは、時間的に不連続的な信号であるかの少なくともどちらかの特徴を持つ。
【０１２２】
また、例えば水道管の漏水検査においてしばしば受信される工場の機械音や停止している車のエンジン音のように、一定位置の発信源からの時間的に連続であるという特徴を持つ雑音も存在する。漏洩音と、このような雑音とは、信号の連続性や、発信源が時間的に移動するかどうかといった特徴からは、識別することはできない。
【０１２３】
しかし、上記工場の機械音や、上記停止している車のエンジン音のように、一定位置の発信源からの時間的に連続な信号の発信源に関しては、その発信源が近くに存在していれば、検査に先立って、検査現場附近の地図や、または、検査現場附近を観察することから、その発信源の位置を知ることは容易であり、工場主に工場の機械音を一時的に発生させないように依頼したり、車の運転手にエンジン音を一時的に発生させないように依頼することにより、そのような発信源からの雑音が発生しないようにすることが可能である。
【０１２４】
また、そのような発信源が遠方にあり、検査に先立って、その存在を知ることが容易でないような場合もあるが、そのような場合には、その信号のレベルは微弱であるという特徴を持つ。
上述したように、漏洩音と、雑音とは、信号の連続性や、発信源位置の時間的な安定性や、信号レベルに、それぞれ特徴を持つ。
【０１２５】
次に、図１に示した異常箇所検出装置の動作について、図２のフローチャートを用いて説明する。
まず、制御部７４に、漏洩音の周波数帯域、超音波センサ３ａと超音波センサ３ｂの離間距離、超音波センサ３ｂと超音波センサ３ｃの離間距離、１回の受信で取り込むデータの時間長さ、１回の受信で取り込んだデータを時間軸上で分割する分割数Ｎ、後述する異常箇所２の位置の特定精度を安定させるための、データ取込みの繰り返し回数Ｍ、および漏洩の有無を判断するための予め決められた閾値が入力される（ステップＳ１）。なお、上記漏洩音の周波数帯域が、未知である場合には、上記周波数帯域が未知であるという情報が入力される。
【０１２６】
また、上記１回の受信で取り込むデータの時間長さ、および上記データ取込みの繰り返し回数Ｍは、許容される検査の安定性、および検査にかかる許容時間により設定され、上記１回の受信で取り込むデータの長さを大きくし、上記データ取込みの繰り返し回数Ｍを多くすることにより、検査の安定性を増加させることができる。また、許容される検査時間が短ければ、上記取り込むデータの長さを小さくし、上記データ取込みの繰り返し回数Ｍを少なくすればよい。
【０１２７】
また、制御部７４においては、検査を行った日付け、検査を行った時間、超音波センサ３ａ、３ｂ、および３ｃの種類やシリアルナンバー、検査を行った導管の地図上の位置等の情報の内、すべての情報、あるいは、上記の情報の内のいずれか１つあるいは２つ以上を入力できるようにし、さらに入力した情報を表示部７３に表示できるようにし、さらに上記入力した情報を、記録、保管しておけば、検査の安定性が増すばかりでなく、異常箇所２の発生傾向に関するデータベースの構築に役立つ。さらに、ある一定期間後の定期検査の際の参照データとして役立つ作用効果を奏する。また、ある一定期間を過ぎて、再度の検査を行う際に、検査データの再現性の確認や、経時変化の調査に役立てることもできる。
【０１２８】
次に、検査者により、検査開始のタイミングが入力される。上記タイミングの入力は、スイッチのオン・オフにより行われても良いし、例えば検査者の音声や検査者が手を叩いた音をマイクで拾うことにより行われても良い。また、制御部７４に、タイマーを備えることにより、上記タイミングが入力されてからある時間が経過した時点で、検査が開始されるようにしてもよい。
【０１２９】
図１において、導管１に異常箇所２が存在すると、上記異常箇所２から漏洩が生じ、それに伴って漏洩音が発生する。発生した漏洩音は、導管１を伝搬して超音波センサ３ａ、３ｂおよび３ｃで雑音とともに受信される。
【０１３０】
次に、制御部７４から、超音波センサ３ａ、３ｂおよび３ｃでの受信を開始するための信号を受信部７１に送信し、超音波センサ３ａ、３ｂおよび３ｃで受信した受信信号を受信部７１に取り込む（ステップＳ２）。受信部７１において、受信信号は、増幅された後にＡ／Ｄ変換され、信号処理部７２に含まれる前処理部７２ａに送られる。
【０１３１】
前処理部７２ａにおいて、３つの受信信号は、時間軸上で予め決められた個数であるＮ個に各々分割される。次に、分割されたＮ個のデータそれぞれがフーリエ変換され、Ｎ個のデータそれぞれについての周波数スペクトルが計算される。
【０１３２】
超音波センサ３ａで受信した信号に関して、１回の受信で取り込むデータの時間長さをＴとし、上記Ｎ個に分割されたデータの第ｎ番目のデータにおいて、ある周波数ｆの成分における上記周波数スペクトルの絶対値をＡn とし、その位相をφanとして、共役複素数を表わす記号を＊とし、
【０１３３】
【数５４】

【０１３４】
で表わされるパワースペクトルＣ_aa（ｆ）が計算される。また、超音波センサ３ｂで受信した信号に関して、上記Ｎ個に分割されたデータの第ｎ番目のデータにおける、ある周波数ｆの成分における上記周波数スペクトルの絶対値をＢn とし、その位相をφbnとして、
【０１３５】
【数５５】

【０１３６】
で表わされるパワースペクトルＣ_bb（ｆ）が計算される。また、超音波センサ３ｃで受信した信号に関して、上記Ｎ個に分割されたデータの第ｎ番目のデータにおける、ある周波数ｆの成分における上記周波数スペクトルの絶対値をＣn とし、その位相をφcnとして、
【０１３７】
【数５６】

【０１３８】
で表わされるパワースペクトルＣ_CC（ｆ）が計算される（ステップＳ３）。
【０１３９】
次に、２つの超音波センサ３ｂ、３ｃの間の領域が地表上に露出していて、その間に異常箇所がないことが目視により判断できる領域等の、異常箇所２がないことが既知である領域である２つの超音波センサの内の一つと、前記２つの超音波センサとは異なる他の一つの超音波センサ３ａとの組み合わせ、例えば図１における超音波センサ３ａと超音波センサ３ｂの組み合わせが選択される。
【０１４０】
さらに、２つの超音波センサ３ｂ、３ｃの間の領域が地表上に露出していて、その間に異常箇所がないことが目視により判断できる領域等の、異常箇所２がないことが既知である領域である２つの超音波センサの内、上記の超音波センサの組み合わせにおいて選択されなかった超音波センサ３ｃと、２つの超音波センサの間の領域が地表上に露出していてその間に異常箇所がないことが目視により判断できる領域等の、異常箇所２がないことが既知である領域である２つの超音波センサの他の一つの組み合わせ、例えば図１における超音波センサ３ａと超音波センサ３ｃの組み合わせが選択される（ステップＳ４）。
【０１４１】
ここで、上記２つの超音波センサの組み合わせ、超音波センサ３ａと３ｂの組み合わせと、超音波センサ３ａと３ｃの組み合わせの内、超音波センサ３ａと３ｂの組み合わせについての信号処理方法について説明する。
２つの超音波センサ３ａおよび３ｂで受信した受信信号のクロススペクトルＣ_ab（ｆ）が、
【０１４２】
【数５７】

【０１４３】
で計算される（ステップＳ５）。クロススペクトルＣ（ｆ）は、式（１）においてＣ₁₁（ｆ）＝Ｃ_aa（ｆ）、Ｃ₂₂（ｆ）＝Ｃ_bb（ｆ）とし、式（１）の周波数特性を持つフィルタによって、さらに、フィルタリングされる。
【０１４４】
ここで、式（１）の周波数特性を持つフィルタの効果について詳しく説明する。
式（１）は、ＳＣＯＴフィルタの周波数特性に、さらに、コヒーレンシィγ（ｆ）の絶対値を実数ｉ乗したものを乗じたものである。
【０１４５】
まず、コヒーレンシィγ（ｆ）の特性について説明する。
コヒーレンシィγ（ｆ）は、
【０１４６】
【数５８】

【０１４７】
で計算される（ステップＳ６）。式（１０）で表わされるコヒーレンシィγ（ｆ）は、式（１０）に式（６）、式（７）および式（９）を代入することから分かるように、
【０１４８】
【数５９】

【０１４９】
の条件が成立すると、その絶対値が１となる値であり、式（１１）の条件が成立しなければ、その絶対値が１以下となる関数である。なお、Ａとφ_aは超音波センサ３ａで受信される受信信号の周波数スペクトルの振幅と位相を示し、同様に、Ｂとφ_bは超音波センサ３ｂで受信される受信信号の周波数スペクトルの振幅と位相（添字は時間データ）を示す。
【０１５０】
式（１１）の条件が成立するということは、上記周波数ｆにおいて、２つの超音波センサ３ａおよび３ｂで受信される受信信号の周波数スペクトルの振幅比と位相差とが時間軸上で分割されたＮ個のデータ間で一定であるということを意味している。
これは、上記周波数ｆの成分は、一定位置の発信源からの時間的に連続的な信号であるということを意味する。
つまりは、この周波数ｆの信号には、漏洩音や、例えば工場の機械音や停止している車のエンジン音のような雑音が含まれているということを意味する。
【０１５１】
また、式（１１）の条件が成立しないということは、上記周波数ｆにおいて、２つの超音波センサ３ａおよび３ｂで受信される受信信号の周波数スペクトルの振幅比と位相差とが時間軸上で分割されたＮ個のデータ間で一定でないということを意味している。つまりは、この周波数ｆの信号は、移動する発信源からの信号であるか、時間的に不連続的な信号であるかの、少なくともどちらかの性質を持つ信号であるということを意味する。
さらには、上記周波数ｆの成分には、漏洩音や、例えば工場の機械音や停止している車のエンジン音のような雑音の他に、自動車の走行音、水道使用音や管内流水音等の雑音も含まれているということを意味する。
【０１５２】
また、ある周波数ｆにおいて、漏洩音や、例えば工場の機械音や停止している車のエンジン音のような、一定位置の発信源からの時間的に連続的な信号が支配的であれば、式（１１）は近似的には成立し、コヒーレンシィγ（ｆ）の絶対値は近似的に１となるが、自動車の走行音、水道使用音や管内流水音といった雑音等の移動する発信源からの信号であるか、時間的に不連続的な信号であるかの、少なくともどちらかの性質を持つ信号が支配的になるにつれ、式（１１）は近似的にも成立しなくなり、コヒーレンシィγ（ｆ）の絶対値はそれに応じて小さい値をとる。
【０１５３】
以上述べたように、式（１０）のコヒーレンシィγ（ｆ）は、上記周波数ｆにおいて、２つの超音波センサ、ここでは、超音波センサ３ａおよび３ｂで受信される受信信号の周波数スペクトルの各周波数成分において、漏洩音や、工場の機械音や停止している車のエンジン音のような、一定位置の発信源からの時間的に連続的な信号と、自動車の走行音、水道使用音や管内流水音といった雑音等の移動する発信源からの信号であるか、時間的に不連続的な信号であるかの、少なくともどちらかの性質を持つ信号とのレベル比を反映した関数である。
【０１５４】
次に、ＳＣＯＴフィルタの特性決定について説明する（ステップＳ７）。
ＳＣＯＴフィルタは、フィルタリングされたクロススペクトルＣ'（ｆ）の絶対値を、フィルタリングされる前のクロススペクトルＣ（ｆ）の絶対値とは無関係に、一定位置の発信源からの時間的に連続的な信号の、移動する発信源からの時間的に不連続な信号に対するレベル比を反映した関数であるコヒーレンシィγ（ｆ）の絶対値と同じにするフィルタである。
【０１５５】
したがって、一定位置の発信源からの時間的に連続的な信号が漏洩音で、移動する発信源からの時間的に不連続な信号が雑音であれば、フィルタリングされたクロススペクトルＣ'（ｆ）の絶対値を、フィルタリングされる前のクロススペクトルＣ（ｆ）の絶対値とは無関係に、漏洩音の対雑音比を反映した関数にするので、雑音が支配的な周波数成分を抑圧する効果を持つ。
【０１５６】
また、漏洩音の周波数スペクトルが雑音のそれと比較して広帯域である場合であれば、フィルタリングされたクロススペクトルＣ'（ｆ）を、それに応じて広帯域化し、相互相関関数のサイドローブを低減し、異常箇所２の位置の特定に関する分解能を向上させる効果を持つ。
したがって、式（１）の周波数特性を持つフィルタは、上述のＳＣＯＴフィルタの特徴効果を持つ。
【０１５７】
しかし、ＳＣＯＴフィルタには、以下のような問題点がある。
まず、一定位置の発信源からの時間的に連続的な信号が漏洩音で、移動する発信源からの信号であるか、もしくは、時間的に不連続な信号であるかの、少なくともどちらかの特徴を持つ信号が雑音であり、ある周波数成分において、この雑音が支配的であったとしても、例えば移動する発信源の移動速度が小さければ、フィルタリングされたクロススペクトルの絶対値Ｃ'（ｆ）は、漏洩音が支配的な周波数成分の絶対値とほぼ同じレベルになってしまうので、漏洩音の対雑音比を十分に反映した関数にはならない。
【０１５８】
つまり、上記雑音が支配的な周波数成分を十分に抑圧することはできない。また、ＳＣＯＴフィルタは、フィルタリングされたクロススペクトルＣ'（ｆ）の絶対値を、一定位置の発信源からの時間的に連続的な信号の、移動する発信源からの時間的に不連続な信号に対するレベル比を反映した関数であるコヒーレンシィγ（ｆ）の絶対値と同じにするので、工場からの機械音や、停止している車のエンジン音のような、一定位置からの時間的に連続な雑音に関しては、上記工場や上記停止している車がはるか遠方にあり、このような一定位置からの時間的に連続な雑音が支配的な周波数成分が、フィルタリングされる前のクロススペクトルＣ（ｆ）において、本来ならば無視できる程小さいレベルであっても、この周波数成分を漏洩音が支配的な周波数成分とほぼ同じレベルまで増幅する逆効果を生じてしまう。
【０１５９】
以上のＳＣＯＴフィルタの問題点を解決するために、式（１）の周波数特性を持つフィルタは、ＳＣＯＴフィルタの周波数特性に、さらにコヒーレンシィγ（ｆ）の絶対値を実数ｉ乗したものを乗じた特性を持っている。ＳＣＯＴフィルタの周波数特性に、さらにコヒーレンシィγ（ｆ）の絶対値を実数ｉ乗したものを乗じることは、コヒーレンシィγ（ｆ）の絶対値が大きい値をとる周波数成分を、さらに強調して抽出し、コヒーレンシィγ（ｆ）の絶対値が小さい値をとる周波数成分を、さら抑圧する効果がある。
【０１６０】
したがって、式（１）の周波数特性を持つフィルタは、一定位置の発信源からの時間的に連続的な信号が漏洩音で、移動する発信源からの時間的に不連続な信号が雑音であれば、移動する発信源の移動速度が小さく、ＳＣＯＴフィルタのみの使用では、雑音が支配的な周波数成分を十分に抑圧できないような場合であっても、ＳＣＯＴフィルタの効果を損なうことなく、雑音が支配的な周波数成分を十分に抑圧することができる。
【０１６１】
なお、コヒーレンシィγ（ｆ）の絶対値に関してある閾値γ_l（ｆ）を予め決めておき、式（２）においてＣ₁₁（ｆ）＝Ｃ_aa（ｆ）、Ｃ₂₂（ｆ）＝Ｃ_bb（ｆ）として、式（２）の周波数特性を持つフィルタを使用してもよい。式（２）の周波数特性を持つフィルタは、コヒーレンシィγ（ｆ）の絶対値がコヒーレンシィγ（ｆ）に関する上記閾値γ_l（ｆ）以下になる周波数成分を完全に除去するので、一定位置の発信源からの時間的に連続的な信号が漏洩音で、移動する発信源からの時間的に不連続な信号が雑音であれば、移動する発信源の移動速度が小さく、ＳＣＯＴフィルタのみの使用では、雑音が支配的な周波数成分を十分に抑圧できないような場合であっても、ＳＣＯＴフィルタの効果を損なうことなく、雑音が支配的な周波数成分を十分に抑圧することができる。したがって、式（１）の周波数特性を持つフィルタを使用した場合と同等の効果が得られる。
【０１６２】
また、クロススペクトルＣ（ｆ）は、式（３）においてＣ₁₁（ｆ）＝Ｃ_aa（ｆ）、Ｃ₂₂（ｆ）＝Ｃ_bb（ｆ）として、式（３）の周波数特性を持つフィルタによってフィルタリングされてもよい（ステップＳ８）。
【０１６３】
ここで、式（３）の周波数特性を持つフィルタの効果について詳しく説明する。
式（３）は、ＳＣＯＴフィルタの周波数特性に、さらに、クロススペクトルＣ（ｆ）の絶対値の対数を乗じたものである。
したがって、式（３）の周波数特性を持つフィルタは、上述したＳＣＯＴフィルタの効果を持つ。
【０１６４】
さらに、クロススペクトルＣ（ｆ）の絶対値の対数を乗じることにより、クロススペクトルＣ（ｆ）の絶対値の大小、つまり、受信信号のレベルの大小に応じた関数を乗じていることになるので、例えば、コヒーレンシィγ（ｆ）の絶対値が大きい値をとる周波数成分において、漏洩音でなく、工場からの機械音や停止している車のエンジン音のような一定位置からの時間的に連続な雑音が支配的になっていても、そのレベルが低ければ、その周波数成分のレベルを、漏洩音が支配的な周波数成分のレベルと同程度まで増幅して通過させてしまうことはない。さらに、絶対値そのものでなく、絶対値の対数を乗じているので、クロススペクトルＣ（ｆ）と比較して広帯域なスペクトルが得られるので、ＳＣＯＴフィルタの効果を損なうことはなく、サイドローブの小さい相関波形を得ることができる。
【０１６５】
なお、式（４）においてＣ₁₁（ｆ）＝Ｃ_aa（ｆ）、Ｃ₂₂（ｆ）＝Ｃ_bb（ｆ）として、式（４）の周波数特性を持つフィルタを用いても構わない。これにより、クロススペクトルの絶対値Ｃ（ｆ）の大小、つまり、受信信号のレベルの大小に応じた関数を乗じていることになるので、例えば、コヒーレンシィγ（ｆ）の絶対値が大きい値をとる周波数成分において、漏洩音でなく、工場からの機械音や停止している車のエンジン音のような一定位置からの時間的に連続な雑音が支配的になっていても、そのレベルが低ければ、その周波数成分のレベルを、漏洩音が支配的な周波数成分のレベルと同程度まで増幅して通過させてしまうことはない。
【０１６６】
さらに、クロススペクトルＣ（ｆ）の絶対値そのものでなく、絶対値を１／ｋ乗したものを乗じているので、クロススペクトルＣ（ｆ）と比較して広帯域なスペクトルが得られるので、ＳＣＯＴフィルタの効果を損なうことはなく、サイドローブの小さい相関波形を得ることができる。したがって、式（３）の周波数特性を持つフィルタを使用した場合と同等の効果が得られる。
以上、式（１）、式（２）、式（３）、および式（４）の周波数特性を持つフィルタの効果について述べた。
【０１６７】
さらに、この発明の実施の形態１における異常箇所検出装置の信号処理部の前処理部７２ａにおいては、式（１）における｜γ（ｆ）｜ⁱの周波数特性と、式（３）におけるｌｏｇ_j（|Ｃ（ｆ）|）、および式（４）における|Ｃ（ｆ）|^1/kの内、いずれか一方の周波数特性と、ＳＣＯＴフィルタの周波数特性とを重み付けし、これらを乗じた周波数特性を持つフィルタや、式（３）におけるｌｏｇ_j（|Ｃ（ｆ）|）、および式（４）における|Ｃ（ｆ）|^1/kの内、いずれか一方の周波数特性において、式（２）のように、コヒーレンシィγ（ｆ）の絶対値がある閾値以下になる周波数成分が０とするような周波数特性と、ＳＣＯＴフィルタの周波数特性とを重み付けし、これらを乗じた周波数特性を持つフィルタを用いても良い。
【０１６８】
このようなフィルタを用いることにより、式（１）や式（２）の周波数特性のフィルタが持つ効果と、式（３）や式（４）の周波数特性のフィルタが持つ効果とを兼ね備えた効果が得られる。なお、上記周波数特性に対する重みの係数については、上述したものと同様の予備実験からきめればよい。
【０１６９】
この発明の実施の形態１においては、以上述べた前処理部７２ａにおけるフィルタリングにより、漏洩音の周波数帯域が未知な場合であっても、漏洩音を効率良く抽出できるので、式（５）の周波数特性を持つバンドパスフィルタは、特に用いなくても良い。したがって、前処理部７２ａにおいて、式（５）の周波数特性を持つバンドパスフィルタは取り除いても良い。これにより、前処理部７２ａの構成を簡易にし、装置全体を廉価にできるという効果がある。
なお、漏洩音の周波数帯域が既知であれば、フィルタリングされたクロススペクトルを、さらに式（５）のバンドパスフィルタによってフィルタリングする。これにより、漏洩音の周波数帯域の範囲外の周波数成分を持つ雑音を完全に除去することができる。
【０１７０】
フィルタリングされたクロススペクトルＣ'（ｆ）は、相関処理部７２ｂに送られ、逆フーリエ変換を示す記号Ｆ^-1として逆フーリエ変換されて、超音波センサ３ａでの受信時間に対する遅延時間をτとし、式（１２）の相互相関関数φ（τ）が計算される（ステップＳ９）。
相関処理部７２ｂで計算された相互相関関数φ（τ）の計算結果は、後処理部７２ｃに出力される。
【０１７１】
【数６０】

【０１７２】
次に、後処理部７２ｃの動作について説明する。
後処理部７２ｃにおいて、上記相互相関関数φ（τ）は包絡線検波される（ステップＳ１０）。
上記包絡線は、その都度メモリに保存される。それと同時に、包絡線をメモリに保存した回数をカウントする。この回数は、制御部７４から受信部７１に対してデータを取り込むための制御信号を送信した回数と同じである。もし、包絡線をメモリに保存した回数が、予め決められたデータ取込みの繰り返し回数Ｍより小さい値であれば、後処理部７２ｃから制御部７４に、再度データの取り込みを行うことを要求する信号を送信する。それに従い、制御部７４からは、受信部７１にデータを取り込むための制御信号を送信する。
【０１７３】
以上述べた繰り返しを、上記包絡線をメモリに保存した回数が、予め決められたデータ取込みの繰り返し回数Ｍと等しくなるまで行う。上記包絡線をメモリに保存した回数が、予め決められたデータ取込みの繰り返し回数Ｍと等しくなったら、この繰り返しを終了する（ステップＳ１１）。
【０１７４】
次に、メモリに保存された上記包絡線は平均化された後、平均化された包絡線のピーク値Ａabと、上記平均化された包絡線がピークになるときの、超音波センサ３ａでの受信時間に対する遅延時間τabが求められる（ステップＳ１２）。
【０１７５】
以上述べた信号処理は、超音波センサ３ａおよび超音波センサ３ｂの組み合わせに関して行われたが、超音波センサ３ａおよび３ｂの組み合わせに関して行われるのと平行して、超音波センサ３ａおよび超音波センサ３ｃの組み合わせに関しても行われる。
これにより、超音波センサ３ａおよび超音波センサ３ｃの組み合わせに関しても、上記平均化された相互相関関数φ（τ）の包絡線のピーク値Ａaｃと、上記平均化された包絡線がピークになるときの、超音波センサ３ａでの受信時間に対する遅延時間τaｃが得られる。
【０１７６】
次に、超音波センサ３ａおよび超音波センサ３ｂの間の平均化された相互相関関数の包絡線のピーク値Ａabと、超音波センサ３ａおよび超音波センサ３ｃの間の平均化された相互相関関数の包絡線のピーク値Ａaｃの内、いずれか一方、もしくは両方の値が、予め決められた異常箇所２の有無を判定するための閾値に比べて、大か小かを判定する。これにより、異常箇所２の有無が判定される（ステップＳ１３，Ｓ１４）。
【０１７７】
異常箇所２の有無を判定するための上記閾値は、超音波センサ３ａおよび超音波センサ３ｂの間の平均化された相互相関関数の包絡線のピーク値Ａabに関する閾値であっても、また、超音波センサ３ａおよび超音波センサ３ｃの間の平均化された相互相関関数の包絡線のピーク値Ａaｃに関する閾値であってもよい。また、上記閾値は、超音波センサ３ａおよび超音波センサ３ｂの間の平均化された相互相関関数の包絡線のピーク値Ａabと、超音波センサ３ａおよび超音波センサ３ｃの間の平均化された相互相関関数の包絡線のピーク値Ａaｃとの積や、和に関する閾値であっても構わない。
【０１７８】
また、異常箇所２の有無を判定するための上記閾値は、平均化された相互相関関数の包絡線のピーク値に関する閾値でなくても、受信信号のレベルに関する閾値であっても良いし、その周波数スペクトルの絶対値に関する閾値、フィルタリングでれたクロススペクトルＣ（ｆ）の絶対値に関する閾値であってもよい。その場合は、前処理部７２の段階で異常箇所２の有無が判定される。
【０１７９】
また、異常箇所２の有無に関する判定は、これら全ての閾値の内、いずれか一つによって判定してもよいし、２つあるいは３つ以上の情報を組み合わせて判定してもよい。
【０１８０】
上記閾値は、予備実験により決められている。この予備実験は、異常箇所２が存在する場合と、実質上存在しない場合について、この実施の形態１に係る異常箇所検出装置と同じか、または、同様の異常箇所検出装置を用いて行われる。このような予備実験から得られた統計データから、異常箇所２の有無を判定するための上記閾値データが予め決められている。
【０１８１】
また、上記閾値は、異常箇所２の有無の判定だけでなく、異常箇所２の形状やサイズをクラス分けするためのものであってもよい。この場合、上記予備実験は、異常箇所２の形状やサイズに関してクラス分けされる各場合について、この実施の形態１に係る異常箇所検出装置と同じか、または、同様の異常箇所検出装置を用いて行われる。
【０１８２】
なお、上記の大小関係に関する情報の内、より多くの情報を組み合わせて異常箇所の有無の判定を行えば、より確度の高い判定を行うことができる効果が得られる。特に、各判定結果について異なる重み付けを行って重み付け多数決の論理を使って判定すれば、上記の３つの判定結果がバラバラの判定結果になったとき、より確度の高い判定結果を得ることができる。重み付け多数決の判定に使う重みの係数については、上述したものと同様の予備実験からきめれば、判定の確度はさらに高くできる。
【０１８３】
異常箇所２があると判定された場合、超音波センサ３ａと３ｂとの間の離間距離と、超音波センサ３ｂと超音波センサ３ｃの間の離間距離と、超音波センサ３ａと３ｂとの間の平均化された相互相関関数の包絡線がピークになるときの、超音波センサ３ａでの受信時間に対する遅延時間τabと、超音波センサ３ａと３ｃとの間の平均化された相互相関関数の包絡線がピークになるときの、超音波センサ３ａでの受信時間に対する遅延時間τaｃとから、超音波センサ３ａから異常箇所２までのの導管１に沿った距離ｘが式（１３）により特定される。
【０１８４】
【数６１】

【０１８５】
なお、漏洩音が導管１を伝搬するときの伝搬速度が既知であれば、超音波センサ３ａ、３ｂ、および３ｃの内、２つの超音波センサの間が、地表上に露出していてその間に異常箇所がないことが目視により判断できる領域等の、異常箇所２がないことが既知である領域である２つの超音波センサの内の一つ、例えば、図１における超音波センサ３ｃを取り除いても、異常箇所２の有無、異常箇所２の位置の特定を行うことができる。例えば、超音波センサ３ｃを取り除いた場合、超音波センサ３ａから異常箇所２までの距離ｘは、漏洩音が導管１を伝搬するときの伝搬速度をｖとし、式（１４）で与えられる。これにより、検査を容易にし、検査システムを廉価にできるという効果がある。
【０１８６】
【数６２】

【０１８７】
このように求められた異常箇所２の有無、および特定された異常箇所２の位置は、表示部７３に出力される（ステップＳ１５）。
表示部７３では、異常箇所２の有無、および特定された異常箇所２の位置を表示する。これらの情報は、単に表示するだけでなく、検査結果の記録として、記録、保管しておけば、異常箇所の発生傾向に関するデータベースの構築に役立つだけでなく、ある一定期間後の定期検査の際の参照データとして役立つ作用効果を奏する。
【０１８８】
また、表示部７３には、上述したように、異常箇所２の有無に関する情報が入力された。この情報は２値の情報である。したがって、これを光のオン・オフや、表示のほかに警報音のオン・オフなど、検査者の五感に反応する形式で検査者に報知できるように表示部以外に他の報知手段を設けて伝えるようにしても良い。また、検査にかかる許容時間が多い場合には、異常箇所２の有無や、異常箇所２の位置の特定に関する判断を、検査者が平均化された包絡線を目で見て判断しても良い。このような場合には、言うまでもないが、表示部２において異常箇所２の有無や特定した漏水箇所の表示を行う機能を取り除いても構わない。これにより、装置が低廉化できる作用効果が得られることは言うまでもない。
【０１８９】
また、表示部７３において、平均化された包絡線、平均化された包絡線のピーク値、平均化された包絡線がピークになる時間、超音波センサ３ａ、３ｂ、および３ｃで得られた受信信号波形、上記受信信号の周波数スペクトル、上記受信信号から求めたコヒーレンシィなどの内、すべての情報、あるいは、上記の情報の内のいずれか１つあるいは２つ以上を表示し、これらの情報を、さらに記録、保管しておけば、検査の安定性が増すばかりでなく、異常箇所の発生傾向に関するデータベースの構築に役立つ。さらに、ある一定期間後の定期検査の際の参照データとして役立つ作用効果を奏する。また、ある一定期間を過ぎて、再度の検査を行う際に、検査データの再現性の確認や、経時変化の調査に役立てることもできる。
【０１９０】
次に、以上述べた異常箇所２の位置の特定方法に関して、その効果を確認した実験結果を図３から図３４を用いて示す。
図３は、上記実験の実験系を示すものであり、８は消火栓、９はレコーダー、１０は計算機である。また、導管１はここでは水道管であり、超音波センサ３ｂと超音波センサ３ｃの間は、導管１が地上に出ていて異常箇所２が存在しないことが目視により判断できる領域である。また、超音波センサ３ａと超音波センサ３ｂの間の導管１に沿った離間距離Ｌ１は１９．８ｍであり、超音波センサ３ｂと超音波センサ３ｃの間の導管１に沿った離間距離Ｌ２は１０ｍである。また、超音波センサ３ａから異常箇所２までの導管１に沿った離間距離ｘを実測したところ、この値は９ｍであった。
【０１９１】
図３において、３つの超音波センサ３ａ、３ｂ、および３ｃで、異常箇所２からの漏洩音を受信し、レコーダー９に受信信号を記録した。レコーダー９で受信した受信信号は、計算機１０に取り込み、計算機１０上で、図１に示した前処理部７２ａ、相関処理部７２ｂ、後処理部７２ｃにおける処理と同じ処理を行い、異常箇所２の位置を特定した。なお、漏洩音の周波数帯域は未知であるとし、１回の受信で取り込むデータの時間長さを１秒とし、１回の受信で取り込んだデータを時間軸上で分割する分割数Ｎを１２とし、データ取り込みの繰り返し回数Ｍを１０とした。また、この実験は、異常箇所２が存在するという前提のもとで行ったので、異常箇所の有無を判定するための閾値は設定しなかった。
【０１９２】
図４〜図３４は、図３の実験系において、超音波センサ３ａ、３ｂ、および３ｃで得られた受信信号波形、受信信号の周波数スペクトルの絶対値、受信信号のパワースペクトルの絶対値、上記３つの超音波センサの内の２つの超音波センサで受信した受信信号から求めたクロススペクトルの絶対値、フィルタリングされたクロススペクトルの絶対値、相互相関関数の包絡線、および平均化された相互相関関数の包絡線を示す図である。
【０１９３】
上記実験においては、異常箇所２が存在するという前提のもとで行い、異常箇所の有無を判定するための閾値は設定しなかったので、図４〜図３４における受信信号波形、受信信号の周波数スペクトルの絶対値、受信信号のパワースペクトルの絶対値、上記３つの超音波センサの内の２つの超音波センサで受信した受信信号から求めたクロススペクトルの絶対値、フィルタリングされたクロススペクトルの絶対値、相互相関関数の包絡線、および平均化された相互相関関数の包絡線の値は、異常箇所の有無を判定するための閾値との大小関係を比較されることはない。
【０１９４】
したがって、図４〜図３４における受信信号波形、受信信号の周波数スペクトルの絶対値、受信信号のパワースペクトルの絶対値、上記３つの超音波センサの内の２つの超音波センサで受信した受信信号から求めたクロススペクトルの絶対値、フィルタリングされたクロススペクトルの絶対値、相互相関関数の包絡線、および平均化された相互相関関数の包絡線の値は、図を見易くするために、適当な値を１として規格化し、相対振幅として示されている。
【０１９５】
図４は、超音波センサ３ａ、３ｂおよび３ｃで１回の受信で受信した受信信号波形である。図４（ａ）は超音波センサ３ａで受信した受信信号波形であり、図４（ｂ）は超音波センサ３ｂで受信した受信信号波形であり、図４（ｃ）は超音波センサ３ｃで受信した受信信号波形である。
【０１９６】
図５は、超音波センサ３ａ、３ｂおよび３ｃで１回の受信で受信した受信信号を時間軸上でＮ個に分割し、Ｎ個に分割したその１個目のデータに関して求めた周波数スペクトルの絶対値であり、図５（ａ）は超音波センサ３ａで１回の受信で受信した受信信号に関するものであり、図５（ｂ）は超音波センサ３ｂで１回の受信で受信した受信信号に関するものであり、図５（ｃ）は超音波センサ３ｃで１回の受信で受信した受信信号に関するものである。
【０１９７】
図６は、超音波センサ３ａ、３ｂおよび３ｃで受信した受信信号の周波数スペクトルと、式（６）〜式（８）とを用いて求めたパワースペクトルの絶対値である。図６（ａ）は、超音波センサ３ａで受信した受信信号から求めたパワースペクトルの絶対値であり、図６（ｂ）は、超音波センサ３ｂで受信した受信信号から求めたパワースペクトルの絶対値であり、図６（ｃ）は、超音波センサ３ｃで受信した受信信号から求めたパワースペクトルの絶対値である。
【０１９８】
図７は、超音波センサ３ａと超音波３ｂの組み合わせ、および超音波センサ３ａおよび３ｃの組み合わせに関するクロススペクトルの絶対値であり、図７（ａ）は、超音波センサ３ａと超音波センサ３ｂの組み合わせから求めたクロススペクトルの絶対値、図７（ｂ）は、超音波センサ３ａと超音波センサ３ｃの組み合わせから求めたクロススペクトルの絶対値である。
【０１９９】
前処理部７２ａにおけるフィルタに、式（１）の周波数特性を持つフィルタを用いてクロススペクトルをフィルタリングして求めたフィルタリングされたクロススペクトルの絶対値を図８に示す。図８（ａ）は超音波センサ３ａと超音波センサ３ｂの組み合わせに関するもの、図８（ｂ）は超音波センサ３ａと超音波センサ３ｃの組み合わせに関するものである。式（１）における実数ｉは１とした。
【０２００】
図８のフィルタリングされたクロススペクトルを逆フーリエ変換して求めた相互相関関数を図９に示す。図９（ａ）は超音波センサ３ａと超音波センサ３ｂの組み合わせに関するもの、図９（ｂ）は超音波センサ３ａと超音波センサ３ｃの組み合わせに関するものである。
【０２０１】
図９の相互相関関数を包絡線検波して求めた包絡線を図１０に示す。図１０（ａ）は超音波センサ３ａと超音波センサ３ｂの組み合わせに関するもの、図１０（ｂ）は超音波センサ３ａと超音波センサ３ｃの組み合わせに関するものである。
【０２０２】
図１０（ａ）において、包絡線がピークになるときの超音波センサ３ａでの受信時間に対する遅延時間τabは０．２５ｍｓ、図１０（ｂ）において、包絡線がピークになるときの、超音波センサ３ａでの受信時間に対する遅延時間τaｃは６．５４ｍｓである。これらの遅延時間から、図３における異常箇所２の位置を式（１３）により特定すると、異常箇所２の超音波センサ３ａからの導管１に沿った距離ｘは９．７０ｍとなった。
【０２０３】
図１０の包絡線を求める過程を予め決められたデータ取込みの繰り返し回数Ｍだけ繰り返し、それらを平均化して求めた、平均化された包絡線を図１１に示す。図１１（ａ）は超音波センサ３ａと超音波センサ３ｂの組み合わせに関するもの、図１１（ｂ）は超音波センサ３ａと超音波センサ３ｃの組み合わせに関するものである。
【０２０４】
図１１（ａ）において、平均化された包絡線がピークになるときの超音波センサ３ａでの受信時間に対する遅延時間τabは０．５４ｍｓ、図１１（ｂ）において、平均化された包絡線がピークになるときの超音波センサ３ａでの受信時間に対する遅延時間τaｃは６．２９ｍｓである。これらの遅延時間から、図３における異常箇所２の位置を式（１３）により特定すると、異常箇所２の超音波センサ３ａからの導管１に沿った距離ｘは９．４３ｍとなった。
【０２０５】
このように、式（１）の周波数特性を持つフィルタを用いてクロススペクトルをフィルタリングすることにより、異常箇所２の位置を精度良く特定することができた。また、予め決められた繰り返し回数Ｍだけ受信を繰り返し、包絡線を平均化することにより、特定精度がさらに向上する効果が見られた。
【０２０６】
前処理部７２ａにおけるフィルタに、式（２）の周波数特性を持つフィルタを用いてクロススペクトルをフィルタリングして求めたフィルタリングされたクロススペクトルの絶対値を図１２に示す。図１２（ａ）は超音波センサ３ａと超音波センサ３ｂの組み合わせに関するもの、図１２（ｂ）は超音波センサ３ａと超音波センサ３ｃの組み合わせに関するものである。式（２）におけるコヒーレンシィγ（ｆ）の絶対値に関する閾値γ_l（ｆ）は０．５とした。
【０２０７】
図１２のフィルタリングされたクロススペクトルを逆フーリエ変換して求めた相互相関関数を図１３に示す。図１３（ａ）は超音波センサ３ａと超音波センサ３ｂの組み合わせに関するもの、図１３（ｂ）は超音波センサ３ａと超音波センサ３ｃの組み合わせに関するものである。
【０２０８】
図１３の相互相関関数を包絡線検波して求めた包絡線を図１４に示す。図１４（ａ）は超音波センサ３ａと超音波センサ３ｂの組み合わせに関するもの、図１４（ｂ）は超音波センサ３ａと超音波センサ３ｃの組み合わせに関するものである。
【０２０９】
図１４（ａ）において、包絡線がピークになるときの超音波センサ３ａでの受信時間に対する遅延時間τabは０．５４ｍｓ、図１４（ｂ）において、包絡線がピークになるときの超音波センサ３ａでの受信時間に対する遅延時間τaｃは６．１７ｍｓである。これらの遅延時間から、図３における異常箇所２の位置を式（１３）により特定すると、異常箇所２の超音波センサ３ａからの導管１に沿った距離ｘは９．４２ｍとなった。
【０２１０】
図１４の包絡線を求める過程を予め決められた繰り返し回数Ｍだけ繰り返し、それらを平均化して求めた平均化された包絡線を図１５に示す。図１５（ａ）は超音波センサ３ａと超音波センサ３ｂの組み合わせに関するもの、図１５（ｂ）は超音波センサ３ａと超音波センサ３ｃの組み合わせに関するものである。
【０２１１】
図１５（ａ）において、平均化された包絡線がピークになるときの超音波センサ３ａでの受信時間に対する遅延時間τabは０．５４ｍｓ、図１５（ｂ）において、平均化された包絡線がピークになるときの超音波センサ３ａでの受信時間に対する遅延時間τaｃは６．１７ｍｓである。これらの遅延時間から、図３における異常箇所２の位置を式（１３）により特定すると、異常箇所２の超音波センサ３ａからの導管１に沿った距離ｘは９．４２ｍとなった。
【０２１２】
このように、式（２）の周波数特性を持つフィルタを用いてクロススペクトルをフィルタリングすることにより、異常箇所２の位置を精度良く特定することができた。
【０２１３】
前処理部７２ａにおけるフィルタに、式（３）の周波数特性を持つフィルタを用いてクロススペクトルをフィルタリングして求めたフィルタリングされたクロススペクトルの絶対値を図１６に示す。図１６（ａ）は超音波センサ３ａと超音波センサ３ｂの組み合わせに関するもの、図１６（ｂ）は超音波センサ３ａと超音波センサ３ｃの組み合わせに関するものである。式（３）における実数ｊは自然対数の底であるｅ＝２．７１８２８１８・・・とした。
【０２１４】
図１６のフィルタリングされたクロススペクトルを逆フーリエ変換して求めた相互相関関数を図１７に示す。図１７（ａ）は超音波センサ３ａと超音波センサ３ｂの組み合わせに関するもの、図１７（ｂ）は超音波センサ３ａと超音波センサ３ｃの組み合わせに関するものである。
【０２１５】
図１７の相互相関関数を包絡線検波して求めた包絡線を図１８に示す。図１８（ａ）は超音波センサ３ａと超音波センサ３ｂの組み合わせに関するもの、図１８（ｂ）は超音波センサ３ａと超音波センサ３ｃの組み合わせに関するものである。
【０２１６】
図１８（ａ）において、包絡線がピークになるときの超音波センサ３ａでの受信時間に対する遅延時間τabは０．５０ｍｓ、図１８（ｂ）において、包絡線がピークになるときの超音波センサ３ａでの受信時間に対する遅延時間τaｃは６．６３ｍｓである。これらの遅延時間から、図３における異常箇所２の位置を式（１３）により特定すると、異常箇所２の超音波センサ３ａからの導管１に沿った距離ｘは９．４９ｍとなった。
【０２１７】
図１８の包絡線を求める過程を予め決められた繰り返し回数Ｍだけ繰り返し、それらを平均化して求めた平均化された包絡線を図１９に示す。図１９（ａ）は超音波センサ３ａと超音波センサ３ｂの組み合わせに関するもの、図１９（ｂ）は超音波センサ３ａと超音波センサ３ｃの組み合わせに関するものである。
【０２１８】
図１９（ａ）において、平均化された包絡線がピークになるときの超音波センサ３ａでの受信時間に対する遅延時間τabは０．５４ｍｓ、図１９（ｂ）において、平均化された包絡線がピークになるときの超音波センサ３ａでの受信時間に対する遅延時間τaｃは６．４６ｍｓである。これらの遅延時間から、図３における異常箇所２の位置を式（１３）により特定すると、異常箇所２の超音波センサ３ａからの導管１に沿った距離ｘは９．４４ｍとなった。
【０２１９】
このように、式（３）の周波数特性を持つフィルタを用いてクロススペクトルをフィルタリングすることにより、異常箇所２の位置を精度良く特定することができた。また、予め決められた繰り返し回数Ｍだけ受信を繰り返し、包絡線を平均化することにより、特定精度がさらに向上する効果が見られた。
【０２２０】
前処理部７２ａにおけるフィルタに、式（４）の周波数特性を持つフィルタを用いてクロススペクトルをフィルタリングして求めたフィルタリングされたクロススペクトルの絶対値を図２０に示す。図２０（ａ）は超音波センサ３ａと超音波センサ３ｂの組み合わせに関するもの、図２０（ｂ）は超音波センサ３ａと超音波センサ３ｃの組み合わせに関するものである。式（４）における実数ｋは２とした。
【０２２１】
図２０のフィルタリングされたクロススペクトルを逆フーリエ変換して求めた相互相関関数を図２１に示す。図２１（ａ）は超音波センサ３ａと超音波センサ３ｂの組み合わせに関するもの、図２１（ｂ）は超音波センサ３ａと超音波センサ３ｃの組み合わせに関するものである。
【０２２２】
図２１の相互相関関数を包絡線検波して求めた包絡線を図２２に示す。図２２（ａ）は超音波センサ３ａと超音波センサ３ｂの組み合わせに関するもの、図２２（ｂ）は超音波センサ３ａと超音波センサ３ｃの組み合わせに関するものである。
【０２２３】
図２２（ａ）において、包絡線がピークになるときの超音波センサ３ａでの受信時間に対する遅延時間τabは０．６３ｍｓ、図２２（ｂ）において、包絡線がピークになるときの超音波センサ３ａでの受信時間に対する遅延時間τaｃは６．２９ｍｓである。これらの遅延時間から、図３における異常箇所２の位置を式（１３）により特定すると、異常箇所２の超音波センサ３ａからの導管１に沿った距離ｘは９．３４ｍとなった。
【０２２４】
図２２の包絡線を求める過程を予め決められた繰り返し回数Ｍだけ繰り返し、それらを平均化して求めた平均化された包絡線を図２３に示す。図２３（ａ）は超音波センサ３ａと超音波センサ３ｂの組み合わせに関するもの、図２３（ｂ）は超音波センサ３ａと超音波センサ３ｃの組み合わせに関するものである。
【０２２５】
図２３（ａ）において、平均化された包絡線がピークになるときの超音波センサ３ａでの受信時間に対する遅延時間τabは０．７１ｍｓ、図２３（ｂ）において、平均化された包絡線がピークになるときの超音波センサ３ａでの受信時間に対する遅延時間τaｃは６．２９ｍｓである。これらの遅延時間から、図３における異常箇所２の位置を式（１３）により特定すると、異常箇所２の超音波センサ３ａからの導管１に沿った距離ｘは９．２６ｍとなった。
【０２２６】
このように、式（４）の周波数特性を持つフィルタを用いてクロススペクトルをフィルタリングすることにより、異常箇所２の位置を精度良く特定することができた。また、予め決められた繰り返し回数Ｍだけ受信を繰り返し、包絡線を平均化することにより、特定精度がさらに向上する効果が見られた。
【０２２７】
次に、前処理部７２ａにおけるフィルタに、式（１）における｜γ（ｆ）｜ⁱの周波数特性と、式（３）におけるｌｏｇ_j（|Ｃ（ｆ）|）、および式（４）における|Ｃ（ｆ）|^1/kの内、いずれか一方の周波数特性と、ＳＣＯＴフィルタの周波数特性とを重み付けし、これらを乗じた周波数特性を持つフィルタや、式（３）におけるｌｏｇ_j（|Ｃ（ｆ）|）、および式（４）における|Ｃ（ｆ）|^1/kの内、いずれか一方の周波数特性において、式（２）のように、コヒーレンシィγ（ｆ）の絶対値がある閾値以下になる周波数成分が０とするような周波数特性と、ＳＣＯＴフィルタの周波数特性とを重み付けし、これらを乗じた周波数特性を持つフィルタの例として、式（１）における｜γ（ｆ）｜ⁱの周波数特性と、式（３）におけるｌｏｇ_j（|Ｃ（ｆ）|）の周波数特性と、ＳＣＯＴフィルタの周波数特性とを１：１：１で重み付けし、これらを乗じた式（１５）の周波数特性Ｈ₆（ｆ）を持つフィルタを用いてクロススペクトルをフィルタリングして求めたフィルタリングされたクロススペクトルの絶対値を図２４に示す。
【０２２８】
図２４（ａ）は超音波センサ３ａと超音波センサ３ｂの組み合わせに関するもの、図２４（ｂ）は超音波センサ３ａと超音波センサ３ｃの組み合わせに関するものである。式（１５）における実数ｉは１、実数ｊは自然対数の底であるｅ＝２．７１８２８１８・・・とした。
【０２２９】
【数６３】

【０２３０】
図２４のフィルタリングされたクロススペクトルを逆フーリエ変換して求めた相互相関関数を図２５に示す。図２５（ａ）は超音波センサ３ａと超音波センサ３ｂの組み合わせに関するもの、図２５（ｂ）は超音波センサ３ａと超音波センサ３ｃの組み合わせに関するものである。
【０２３１】
図２５の相互相関関数を包絡線検波して求めた包絡線を図２６に示す。図２６（ａ）は超音波センサ３ａと超音波センサ３ｂの組み合わせに関するもの、図２６（ｂ）は超音波センサ３ａと超音波センサ３ｃの組み合わせに関するものである。
【０２３２】
図２６（ａ）において、包絡線がピークになるときの超音波センサ３ａでの受信時間に対する遅延時間τabは０．５４ｍｓ、図２６（ｂ）において、包絡線がピークになるときの超音波センサ３ａでの受信時間に対する遅延時間τaｃは６．４２ｍｓである。これらの遅延時間から、図３における異常箇所２の位置を式（１３）により特定すると、異常箇所２の超音波センサ３ａからの導管１に沿った距離ｘは９．４４ｍとなった。
【０２３３】
図２６の包絡線を求める過程を予め決められた繰り返し回数Ｍだけ繰り返し、それらを平均化して求めた平均化された包絡線を図２７に示す。図２７（ａ）は超音波センサ３ａと超音波センサ３ｂの組み合わせに関するもの、図２７（ｂ）は超音波センサ３ａと超音波センサ３ｃの組み合わせに関するものである。
【０２３４】
図２７（ａ）において、平均化された包絡線がピークになるときの超音波センサ３ａでの受信時間に対する遅延時間τabは０．５８ｍｓ、図２７（ｂ）において、平均化された包絡線がピークになるときの超音波センサ３ａでの受信時間に対する遅延時間τaｃは６．２５ｍｓである。これらの遅延時間から、図３における異常箇所２の位置を式（１３）により特定すると、異常箇所２の超音波センサ３ａからの導管１に沿った距離ｘは９．３９ｍとなった。
【０２３５】
このように、式（１）における｜γ（ｆ）｜ⁱの周波数特性と、式（３）におけるｌｏｇ_j（|Ｃ（ｆ）|）、および式（４）における|Ｃ（ｆ）|^1/kの内、いずれか一方の周波数特性と、ＳＣＯＴフィルタの周波数特性とを重み付けし、これらを乗じた周波数特性を持つフィルタや、式（３）におけるｌｏｇ_j（|Ｃ（ｆ）|）、および式（４）における|Ｃ（ｆ）|^1/kの内、いずれか一方の周波数特性において、式（２）のように、コヒーレンシィγ（ｆ）の絶対値がある閾値以下になる周波数成分が０とするような周波数特性と、ＳＣＯＴフィルタの周波数特性とを重み付けし、これらを乗じた周波数特性を持つフィルタの例として、式（１）における｜γ（ｆ）｜ⁱの周波数特性と、式（３）におけるｌｏｇ_j（|Ｃ（ｆ）|）の周波数特性と、ＳＣＯＴフィルタの周波数特性とを１：１：１で重み付けし、これらを乗じた式（１５）の周波数特性Ｈ₆（ｆ）を持つフィルタを用いてクロススペクトルをフィルタリングすることにより、異常箇所２の位置を精度良く特定することができた。また、予め決められた繰り返し回数Ｍだけ受信を繰り返し、包絡線を平均化することにより、特定精度がさらに向上する効果が見られた。
【０２３６】
なお、比較のため、前処理部７２ａにおいて、クロススペクトルをフィルタリングするためのフィルタを用いずに処理を行った時の結果を以下に示す。
前処理部７２ａにおいて、クロススペクトルをフィルタリングするためのフィルタを用いずに、クロススペクトルを逆フーリエ変換して求めた相互相関関数を図２８に示す。図２８（ａ）は超音波センサ３ａと超音波センサ３ｂの組み合わせに関するもの、図２８（ｂ）は超音波センサ３ａと超音波センサ３ｃの組み合わせに関するものである。
【０２３７】
図２８の相互相関関数を包絡線検波して求めた包絡線を図２９に示す。図２９（ａ）は超音波センサ３ａと超音波センサ３ｂの組み合わせに関するもの、図２９（ｂ）は超音波センサ３ａと超音波センサ３ｃの組み合わせに関するものである。
【０２３８】
図２９（ａ）において、包絡線がピークになるときの、超音波センサ３ａでの受信時間に対する遅延時間τabは４．４２ｍｓ、図２９（ｂ）において、包絡線がピークになるときの超音波センサ３ａでの受信時間に対する遅延時間τaｃは６．２９ｍｓである。これらの遅延時間から、図３における異常箇所２の位置を式（１３）により特定すると、異常箇所２の超音波センサ３ａからの導管１に沿った距離ｘは−１．９１ｍとなった。
【０２３９】
図２９の包絡線を求める過程を予め決められた繰り返し回数Ｍだけ繰り返し、それらを平均化して求めた平均化された包絡線を図３０に示す。図３０（ａ）は超音波センサ３ａと超音波センサ３ｂの組み合わせに関するもの、図３０（ｂ）は超音波センサ３ａと超音波センサ３ｃの組み合わせに関するものである。
【０２４０】
図３０（ａ）において、平均化された包絡線がピークになるときの、超音波センサ３ａでの受信時間に対する遅延時間τabは４．４２ｍｓ、図３０（ｂ）において、平均化された包絡線がピークになるときの超音波センサ３ａでの受信時間に対する遅延時間τaｃは６．２１ｍｓである。これらの遅延時間から、図３における異常箇所２の位置を式（１３）により特定すると、異常箇所２の超音波センサ３ａからの導管１に沿った距離ｘは−２．４６ｍとなった。
【０２４１】
このように、前処理部７２ａにおいて、クロススペクトルをフィルタリングするためのフィルタを用いずに処理を行った場合、包絡線の平均化を行わない場合、行った場合とも、異常箇所２の位置を精度良く特定することはできなかった。
【０２４２】
さらに、比較のため、前処理部７２ａにおけるフィルタにＳＣＯＴフィルタのみを用いて処理を行った時の結果を以下に示す。
前処理部７２ａにおけるフィルタに、ＳＣＯＴフィルタのみを用いてクロススペクトルをフィルタリングして求めたフィルタリングされたクロススペクトルの絶対値を図３１に示す。図３１（ａ）は超音波センサ３ａと超音波センサ３ｂの組み合わせに関するもの、図３１（ｂ）は超音波センサ３ａと超音波センサ３ｃの組み合わせに関するものである。
【０２４３】
図３１のクロススペクトルを逆フーリエ変換して求めた相互相関関数を図３２に示す。図３２（ａ）は超音波センサ３ａと超音波センサ３ｂの組み合わせに関するもの、図３２（ｂ）は超音波センサ３ａと超音波センサ３ｃの組み合わせに関するものである。
【０２４４】
図３２の相互相関関数を包絡線検波して求めた包絡線を図３３に示す。図３３（ａ）は超音波センサ３ａと超音波センサ３ｂの組み合わせに関するもの、図３３（ｂ）は超音波センサ３ａと超音波センサ３ｃの組み合わせに関するものである。
【０２４５】
図３３（ａ）において、包絡線がピークになるときの超音波センサ３ａでの受信時間に対する遅延時間τabは０．６３ｍｓ、図３３（ｂ）において、包絡線がピークになるときの超音波センサ３ａでの受信時間に対する遅延時間τaｃは−０．０４ｍｓである。これらの遅延時間から、図３における異常箇所２の位置を式（１３）により特定すると、異常箇所２の超音波センサ３ａからの導管１に沿った距離ｘは１４．６ｍとなった。
【０２４６】
図３３の包絡線を求める過程を予め決められた繰り返し回数Ｍだけ繰り返し、それらを平均化して求めた平均化された包絡線を図３４に示す。図３４（ａ）は超音波センサ３ａと超音波センサ３ｂの組み合わせに関するもの、図３４（ｂ）は超音波センサ３ａと超音波センサ３ｃの組み合わせに関するものである。
【０２４７】
図３４（ａ）において、平均化された包絡線がピークになるときの超音波センサ３ａでの受信時間に対する遅延時間τabは０．５０ｍｓ、図３４（ｂ）において、平均化された包絡線がピークになるときの超音波センサ３ａでの受信時間に対する遅延時間τaｃは−０．０８ｍｓである。これらの遅延時間から、図３における異常箇所２の位置を式（１３）により特定すると、異常箇所２の超音波センサ３ａからの導管１に沿った距離ｘは１４．２１ｍとなった。
【０２４８】
このように、前処理部７２ａにおけるフィルタにＳＣＯＴフィルタのみを用いて処理を行った場合、包絡線の平均化を行わない場合、行った場合とも、異常箇所２の位置を精度良く特定することはできなかった。
【０２４９】
以上の実験結果から、この発明の実施の形態１における信号処理部７２の前処理部７２ａにおける信号処理により、導管の異常箇所を、従来よりも精度良く特定できることが実験的にも確認された。さらに、後処理部７２ｃにおける信号処理により、導管の異常箇所の特定を従来よりも安定して特定できることが、実験的にも確認された。
【０２５０】
また、以上述べた実験結果から、この発明の実施の形態１における信号処理法により、漏洩音の周波数帯域が未知な場合であっても、異常箇所２の位置を精度良く特定できることが実験的にも確認された。
【０２５１】
この発明の実施の形態１においては、従来とは異なり、相関処理の前処理に、式（１）の周波数特性を持つフィルタと、式（２）の周波数特性を持つフィルタと、式（３）の周波数特性を持つフィルタと、式（４）の周波数特性を持つフィルタと、式（１）における｜γ（ｆ）｜ⁱの周波数特性と、式（３）におけるｌｏｇ_j（|Ｃ（ｆ）|）、および式（４）における|Ｃ（ｆ）|^1/kの内、いずれか一方の周波数特性と、ＳＣＯＴフィルタの周波数特性とを重み付けし、これらを乗じた周波数特性を持つフィルタや、式（３）におけるｌｏｇ_j（|Ｃ（ｆ）|）、および式（４）における|Ｃ（ｆ）|^1/kの内、いずれか一方の周波数特性において、式（２）のように、コヒーレンシィγ（ｆ）がある閾値以下になる周波数成分が０とするような周波数特性と、ＳＣＯＴフィルタの周波数特性とを重み付けし、これらを乗じた周波数特性を持つフィルタとの内のいずれか一つ、または、二つ以上の周波数特性を持つフィルタを持ち合わせているので、漏洩音の周波数帯域が未知な場合であっても異常箇所２の位置を精度良く特定することができる。
【０２５２】
また、一定位置の発信源からの時間的に連続的な信号が漏洩音で、移動する発信源からの時間的に不連続な信号が雑音であれば、移動する発信源の移動速度が小さく、ＳＣＯＴフィルタのみでは、雑音が支配的な周波数成分を十分に抑圧できないような場合であっても、漏洩箇所の位置を精度良く特定することができる。
【０２５３】
また、一定位置の発信源からの時間的に連続的な信号が、漏洩音だけでなく、工場からの機械音や停止している車のエンジン音のような雑音であり、この雑音が支配的になっている周波数成分に関しても、精度良く漏洩箇所の特定を行うことができる。
【０２５４】
さらに、相互相関関数の包絡線を、複数回の繰り返し測定から平均化するという平均化処理をさらに行うことによって、漏洩音や雑音が、ランダムで周期性のない信号であるという問題を克服することができ、異常箇所２の位置の特定を安定して行うことができる。
【０２５５】
実施の形態２．
次に、この発明の実施の形態２に係る異常箇所検出装置について説明する。
この実施の形態２に係る異常箇所検出装置は、図１に示す実施の形態２と同様な構成を備えるが、信号処理部７２内の前処理部７２ａは、式（１）で表される周波数特性Ｈ₁（ｆ）を持つフィルタと、式（２）で表される周波数特性Ｈ₂（ｆ）を持つフィルタと、式（３）で表される周波数特性Ｈ₃（ｆ）を持つフィルタと、式（４）で表される周波数特性Ｈ₄（ｆ）を持つフィルタの他に、後述する周波数特性を持つフィルタの内、いずれか一つまたは二つ以上のフィルタを含んでいる。
【０２５６】
すなわち、信号処理部７２内の前処理部７２ａは、周波数特性Ｈ₁（ｆ）を持つフィルタと、周波数特性Ｈ₂（ｆ）を持つフィルタと、周波数特性Ｈ₃（ｆ）を持つフィルタと、周波数特性Ｈ₄（ｆ）を持つフィルタと、σ_abs（ｆ）を上記３つの超音波センサ３ａ、３ｂ及び３ｃの内、ある２つのセンサで受信した２つの受信信号の周波数スペクトルの絶対値の比の分割された複数個の受信信号間の分散とし、σ_φ（ｆ）を上記２つのセンサで受信した２つの受信信号の周波数スペクトルの位相差の分割された複数個の受信信号間の分散として、
【０２５７】
【数６４】

【０２５８】
で表わされる周波数特性Ｈ₇（ｆ）を持つフィルタと、
上記３つの超音波センサ３ａ、３ｂ及び３ｃの内、ある２つのセンサで受信した２つの受信信号の周波数スペクトルの振幅の絶対値の比の分割された複数個の受信信号間の分散σ_abs（ｆ）と、上記２つのセンサで受信した受信信号の周波数スペクトルの位相差の分割された複数個の受信信号間の分散σ_φ（ｆ）との積に関する閾値をσ_l（ｆ）とし、
【０２５９】
【数６５】

【０２６０】
で表わされる周波数特性Ｈ₈（ｆ）を持つフィルタと、
式（１）における｜γ（ｆ）｜ⁱ、および式（１６）における｛１／（σ_abs（ｆ）・σ_φ（ｆ））｝ⁱの内、いずれか一方の周波数特性と、式（３）におけるｌｏｇ_j（|Ｃ（ｆ）|）及び式（４）における|Ｃ（ｆ）|^1/kの内、いずれか一方の周波数特性と、ＳＣＯＴフィルタの周波数特性とを重み付けし、これらを乗じた周波数特性を持つフィルタや、式（３）におけるｌｏｇ_j（|Ｃ（ｆ）|）、および式（４）における|Ｃ（ｆ）|^1/kの内、いずれか一方の周波数特性において、式（２）のように、コヒーレンシィγ（ｆ）の絶対値が予め決められた閾値以下になる周波数成分が０とするような周波数特性、もしくは、式（１７）のように、上記３つの超音波センサの内、ある２つのセンサで受信した２つの受信信号の周波数スペクトルの絶対値の比の分割された複数個の受信信号間の分散σ_abs（ｆ）と、上記２つのセンサで受信した２つの受信信号の周波数スペクトルの位相差の分割された複数個の受信信号間の分散σ_φ（ｆ）との積が予め決められた閾値以上になる周波数成分が０とするような周波数特性と、ＳＣＯＴフィルタの周波数特性とを重み付けし、これらを乗じた周波数特性を持つフィルタと、の内、いずれか一つあるいは２つ以上を含んでいる。
【０２６１】
さらに、下限周波数をｆｓとし、上限周波数をｆｅとして、式（５）で表わされる周波数特性Ｈ₅（ｆ）を持つバンドパスフィルタを含んでいる。
【０２６２】
上記前処理部７２ａが含んでいるフィルタ特性において、式（５）における下限周波数ｆｓおよび上限周波数ｆｅは、予備実験の結果に応じて決定され、制御部７４にこれら値が入力される。また、式（１）における実数ｉ、式（２）におけるコヒーレンシィの絶対値に関する閾値γ_l（ｆ）、式（３）における実数ｊ、式（４）における実数ｋ、および式（１７）における上記３つの超音波センサの内のある２つのセンサで受信した２つの受信信号の周波数スペクトルの振幅の絶対値の比の分割された複数個の受信信号間の分散σ_abs（ｆ）と、上記２つのセンサで受信した受信信号の周波数スペクトルの位相差の分割された複数個の受信信号間の分散σ_φ（ｆ）との積に関する閾値をσ_l（ｆ）も、同様の予備実験の結果に応じて決定される。
【０２６３】
この予備実験は、異常箇所２が存在する場合と、実質上、存在しない場合について、この実施の形態２に係る異常箇所検出装置と同じか、または、同様の異常箇所検出装置を用いて行われる。このような予備実験から得られた統計データから、検査の状況に応じて、異常箇所２の有無や異常箇所２の位置の検出を最も精度よく行なえるような、式（１）における実数ｉ、式（２）におけるコヒーレンシィの絶対値に関する閾値γ_l（ｆ）、式（３）における実数ｊ、式（４）における実数ｋ、および式（１７）における上記３つの超音波センサの内のある２つのセンサで受信した２つの受信信号の周波数スペクトルの振幅の絶対値の比の分割された複数個の受信信号間の分散σ_abs（ｆ）と、上記２つのセンサで受信した受信信号の周波数スペクトルの位相差の分割された複数個の受信信号間の分散σ_φ（ｆ）との積に関する閾値をσ_l（ｆ）も、同様の予備実験の結果に応じて決定される。
【０２６４】
また、実施の形態２と同様に、漏洩音が導管１を伝搬するときの伝搬速度が既知であれば、超音波センサ３ａ、３ｂ、３ｃの３つの内、２つの超音波センサの間が地表上に露出していてその間に異常箇所がないことが目視により判断できる領域等の異常箇所２がないことが既知である領域である２つの超音波センサの内の一つ、例えば、図１における超音波センサ３ｃを取り除いても、異常箇所２の有無、異常箇所２の位置の特定を行うことができる。このような場合であれば、超音波センサ３ａ、３ｂ、および３ｃの内の一つを取り除くことにより、検査を容易にし、検査システムを廉価にできるという効果がある。
【０２６５】
また、この実施の形態２に係る異常箇所検出装置は、実施の形態２と同様に、図２に示すフローチャートに従って動作し、同様な効果を奏する。なお、その詳細な説明は省略する。
また、この発明の実施の形態２における異常箇所検出装置の信号処理部の前処理部７２ａにおいては、式（１）における｜γ（ｆ）｜ⁱの周波数特性と、式（３）におけるｌｏｇ_j（|Ｃ（ｆ）|）、および式（４）における|Ｃ（ｆ）|^1/kの内、いずれか一方の周波数特性と、ＳＣＯＴフィルタの周波数特性とを重み付けし、これらを乗じた周波数特性を持つフィルタや、式（３）におけるｌｏｇ_j（|Ｃ（ｆ）|）、および式（４）における|Ｃ（ｆ）|^1/kの内、いずれか一方の周波数特性において、式（２）のように、コヒーレンシィγ（ｆ）の絶対値がある閾値以下になる周波数成分が０とするような周波数特性と、ＳＣＯＴフィルタの周波数特性とを重み付けし、これらを乗じた周波数特性を持つフィルタを用いても良い。
【０２６６】
このようなフィルタを用いることにより、式（１）や式（２）の周波数特性のフィルタが持つ効果と、式（３）や式（４）の周波数特性のフィルタが持つ効果とを兼ね備えた効果が得られる。なお、上記周波数特性に対する重みの係数については、上述したものと同様の予備実験からきめればよい。
【０２６７】
次に、ある周波数ｆにおける、２つの超音波センサで受信した受信信号の周波数スペクトルの絶対値の比Ａｎ／Ｂｎの分割されたＮ個のデータの間の分散をσ_abs（ｆ）、２つの超音波センサで受信した受信信号の周波数スペクトルの位相差φａｎ−φｂｎの分割されたＮ個にデータに関する分散をσ_φ（ｆ）とし、
【０２６８】
【数６６】

【０２６９】
で表される周波数特性を持つ関数σ'（ｆ）について説明する。この関数σ'（ｆ）は、２つの超音波センサで受信した受信信号において、振幅比と位相差が、Ｎ個に分割されたデータの間での分散が小さい、つまり、時間的なばらつきが小さい周波数成分ほど大きな値を持つ関数である。したがって、式（１８）の周波数特性は、２つの超音波センサで受信される受信信号の周波数スペクトルの各周波数成分において、漏洩音や、工場の機械音や停止している車のエンジン音のような、一定位置の発信源からの時間的に連続的な信号と、自動車の走行音、水道使用音や管内流水音といった雑音等の移動する発信源からの信号であるか、時間的に不連続的な信号であるかの、少なくともどちらかの性質を持つ信号とのレベル比を反映した関数である。つまり、式（１８）の関数σ'（ｆ）の物理的な意味は、コヒーレンシィの絶対値｜γ（ｆ）｜の物理的な意味と同じである。
【０２７０】
したがって、クロススペクトルＣ（ｆ）は、式（１）において、コヒーレンシィの絶対値｜γ（ｆ）｜に式（１８）の右辺を代入し、式（１６）の周波数特性を持つフィルタによりフィルタリングしても、式（１）の周波数特性を持つフィルタによりフィルタリングを行った場合と同様の効果が得られる。
【０２７１】
また、クロススペクトルＣ（ｆ）は、式（２）において、コヒーレンシィの絶対値｜γ（ｆ）｜に式（１８）の右辺を代入し、式（１７）の周波数特性を持つフィルタによりフィルタリングしても、式（２）の周波数特性を持つフィルタによりフィルタリングを行った場合と同様の効果が得られる。
【０２７２】
さらに、この発明の実施の形態２における異常箇所検出装置の信号処理部の前処理部７２ａにおいては、式（１６）における｛１／（σ_abs（ｆ）・σ_φ（ｆ））｝ⁱの周波数特性と、式（３）におけるｌｏｇ_j（|Ｃ（ｆ）|）、および式（４）における|Ｃ（ｆ）|^1/kの内、いずれか一方の周波数特性と、ＳＣＯＴフィルタの周波数特性とを重み付けし、これらを乗じた周波数特性を持つフィルタや、式（３）におけるｌｏｇ_j（|Ｃ（ｆ）|）、および式（４）における|Ｃ（ｆ）|^1/kの内、いずれか一方の周波数特性において、式（１７）のように、２つのセンサで受信した２つの受信信号の周波数スペクトルの絶対値の比の分割された複数個の受信信号間の分散σ_abs（ｆ）と、２つのセンサで受信した２つの受信信号の周波数スペクトルの位相差の分割された複数個の受信信号間の分散σ_φ（ｆ）との積に関する閾値σ_l（ｆ）が、ある閾値以上になる周波数成分が０とするような周波数特性と、ＳＣＯＴフィルタの周波数特性とを重み付けし、これらを乗じた周波数特性を持つフィルタを用いても良い。
【０２７３】
このようなフィルタを用いることにより、式（１６）や式（１７）の周波数特性のフィルタが持つ効果と、式（３）や式（４）の周波数特性のフィルタが持つ効果とを兼ね備えた効果が得られる。なお、上記周波数特性に対する重みの係数については、上述したものと同様の予備実験からきめればよい。
【０２７４】
この発明の実施の形態２においては、以上述べた前処理部７２ａにおけるフィルタリングにより、漏洩音の周波数帯域が未知な場合であっても、漏洩音を効率良く抽出できるので、式（５）の周波数特性を持つバンドパスフィルタは、特に用いなくても良い。したがって、前処理部７２ａにおいて、式（５）の周波数特性を持つバンドパスフィルタは取り除いても良い。これにより、前処理部７２ａの構成を簡易にし、装置全体を廉価にできるという効果がある。
なお、漏洩音の周波数帯域が既知であれば、フィルタリングされたクロススペクトルを、さらに式（５）のバンドパスフィルタによってフィルタリングする。これにより、漏洩音の周波数帯域の範囲外の周波数成分を持つ雑音を完全に除去することができる。
【０２７５】
この実施の形態２における異常箇所２の位置の特定方法に関して、その効果を確認する実験結果は、実施の形態１と同様に、図３から図３４に示される。
この実施の形態２において、前処理部におけるフィルタに、式（１）における｜γ（ｆ）｜ⁱ、および式（１６）における｛１／（σ_abs（ｆ）・σ_φ（ｆ））｝ⁱの内、いずれか一方の周波数特性と、式（３）におけるｌｏｇ_j（|Ｃ（ｆ）|）及び式（４）における|Ｃ（ｆ）|^1/kの内、いずれか一方の周波数特性と、ＳＣＯＴフィルタの周波数特性とを重み付けし、これらを乗じた周波数特性を持つフィルタや、式（３）におけるｌｏｇ_j（|Ｃ（ｆ）|）、および式（４）における|Ｃ（ｆ）|^1/kの内、いずれか一方の周波数特性において、式（２）のように、コヒーレンシィγ（ｆ）の絶対値が予め決めれれた閾値以下になる周波数成分が０とするような周波数特性、もしくは、式（１７）のように、上記３つの超音波センサの内、ある２つのセンサで受信した２つの受信信号の周波数スペクトルの絶対値の比の分割された複数個の受信信号間の分散σ_abs（ｆ）と、上記２つのセンサで受信した２つの受信信号の周波数スペクトルの位相差の分割された複数個の受信信号間の分散σ_φ（ｆ）との積が予め決めれれた閾値以上になる周波数成分が０とするような周波数特性と、ＳＣＯＴフィルタの周波数特性とを重み付けし、これらを乗じた周波数特性を持つフィルタの例として、式（１）における｜γ（ｆ）｜ⁱの周波数特性と、式（３）におけるｌｏｇ_j（|Ｃ（ｆ）|）の周波数特性と、ＳＣＯＴフィルタの周波数特性とを１：１：１で重み付けし、これらを乗じた式（１５）の周波数特性Ｈ₆（ｆ）を持つフィルタを用いてクロススペクトルをフィルタリングして求めたフィルタリングされたクロススペクトルの絶対値は、図２４に示す実施の形態１と同様なものを得ることができ、異常箇所２の位置を精度良く特定することができる。また、予め決められた繰り返し回数Ｍだけ受信を繰り返し、包絡線を平均化することにより、特定精度がさらに向上する効果が見られた。
【０２７６】
このように、この発明の実施の形態２においては、従来とは異なり、相関処理の前処理に、
式（１）の周波数特性を持つフィルタと、
式（２）の周波数特性を持つフィルタと、
式（３）の周波数特性を持つフィルタと、
式（４）の周波数特性を持つフィルタと、
式（１６）の周波数特性を持つフィルタと、
式（１７）の周波数特性を持つフィルタと、
式（１）における｜γ（ｆ）｜ⁱおよび式（１６）における｛１／（σ_abs（ｆ）・σ_φ（ｆ））｝ⁱの内、いずれか一つの周波数特性を持つフィルタと、式（３）におけるｌｏｇ_j（|Ｃ（ｆ）|）、および式（４）における|Ｃ（ｆ）|^1/kの内、いずれか一方の周波数特性と、ＳＣＯＴフィルタの周波数特性とを重み付けし、これらを乗じた周波数特性を持つフィルタや、式（３）におけるｌｏｇ_j（|Ｃ（ｆ）|）、および式（４）における|Ｃ（ｆ）|^1/kの内、いずれか一方の周波数特性において、式（２）もしくは式（１７）のように、コヒーレンシィγ（ｆ）の絶対値が予め決めれれた閾値以下になる周波数成分が０とするような周波数特性、もしくは、上記２つのセンサで受信した受信信号の周波数スペクトルの振幅の絶対値の比の、分割された複数個の受信信号間の分散σ_abs（ｆ）と、上記２つのセンサで受信した受信信号の周波数スペクトルの位相差の分割された複数個の受信信号間の分散σ_φ（ｆ）との積が予め決められたある閾値以上になる周波数成分が０とするような周波数特性と、ＳＣＯＴフィルタの周波数特性とを重み付けし、これらを乗じた周波数特性を持つフィルタとの内のいずれか一つ、または、二つ以上の周波数特性を持つフィルタを持ち合わせているので、漏洩音の周波数帯域が未知な場合であっても異常箇所２の位置を精度良く特定することができる。
【０２７７】
また、一定位置の発信源からの時間的に連続的な信号が漏洩音で、移動する発信源からの時間的に不連続な信号が雑音であれば、移動する発信源の移動速度が小さく、ＳＣＯＴフィルタのみでは、雑音が支配的な周波数成分を十分に抑圧できないような場合であっても、漏洩箇所の位置を精度良く特定することができる。
【０２７８】
また、一定位置の発信源からの時間的に連続的な信号が、漏洩音だけでなく、工場からの機械音や停止している車のエンジン音のような雑音であり、この雑音が支配的になっている周波数成分に関しても、精度良く漏洩箇所の特定を行うことができる。
【０２７９】
さらに、相互相関関数の包絡線を、複数回の繰り返し測定から平均化するという平均化処理をさらに行うことによって、漏洩音や雑音が、ランダムで周期性のない信号であるという問題を克服することができ、異常箇所２の位置の特定を、安定して行うことができる。
【０２８０】
上述した実施の形態２に係る異常箇所検出装置は、実施の形態１と同様な実施形態を採用することができ、要約すれば次の通りとなる。
１．被検査管に異常箇所が存在することにより発生する漏洩音を受信するための３つの超音波センサを備えると共に、これらの超音波センサによる漏洩音の受信信号をそれぞれ受信部を介して入力し信号処理する信号処理部に、ｆを受信信号の周波数、ｉを０以上のある実数、上記３つの超音波センサの内、ある２つのセンサで受信した受信信号から求めたコヒーレンシィをγ（ｆ）、上記２つのセンサの内の片方の超音波センサで受信した受信信号から求めたパワースペクトルをＣ₁₁（ｆ）、上記２つの超音波センサの他方で受信した受信信号から求めたパワースペクトルをＣ₂₂（ｆ）として、式（１）で表わされる周波数特性Ｈ₁（ｆ）を持ち、漏洩音の受信信号から求めたクロススペクトルをフィルタリングするためのフィルタＡと、上記コヒーレンシィγ（ｆ）の絶対値に関する予め決められた閾値をγ₁（ｆ）として、式（２）で表わされる周波数特性Ｈ₂（ｆ）を持つフィルタＢと、ｊを０以上のある実数、上記３つの超音波センサの内、２つのセンサで受信した受信信号から求めたクロススペクトルをＣ（ｆ）として、式（３）で表わされる周波数特性Ｈ₃（ｆ）を持つフィルタＣと、ｋを１以上のある実数として、式（４）で表わされる周波数特性Ｈ₄（ｆ）を持つフィルタＤと、σ_abs（ｆ）を上記３つの超音波センサの内、ある２つのセンサで受信した２つの受信信号の周波数スペクトルの絶対値の比の分割された複数個の受信信号間の分散とし、σ_φ（ｆ）を上記２つのセンサで受信した２つの受信信号の周波数スペクトルの位相差の分割された複数個の受信信号間の分散として、式（１６）で表わされる周波数特性Ｈ₇（ｆ）を持つフィルタＥと、上記３つの超音波センサの内、ある２つのセンサで受信した２つの受信信号の周波数スペクトルの絶対値の比の分割された複数個の受信信号間の分散σ_abs（ｆ）と、上記２つのセンサで受信した２つの受信信号の周波数スペクトルの位相差の分割された複数個の受信信号間の分散σ_φ（ｆ）との積に関する閾値をσ_l（ｆ）とし、式（１７）で表わされる周波数特性Ｈ₈（ｆ）を持つフィルタＦと、上記６つの式の内、少なくともｌｏｇ_j（|Ｃ（ｆ）|）と|Ｃ（ｆ）|^1/kのいずれか１つの周波数特性と、ＳＣＯＴ（Smoothed Coherence TrＡｎsform）フィルタの周波数特性とをそれぞれ重み付けし、重み付けしたこれらの周波数特性を乗じた周波数特性を持つフィルタＧとの内、一つまたは二つ以上のフィルタを含む前処理部と、上記前処理部により前処理された信号から相互相関関数を演算する相関処理部とを備える。
【０２８１】
２．上記フィルタＧは、｜γ（ｆ）｜ⁱおよび｛１／（σ_abs（ｆ）・σ_φ（ｆ））｝ⁱの内、いずれか一つの周波数特性と、ｌｏｇ_j（|Ｃ（ｆ）|）、および|Ｃ（ｆ）|^1/kの内、いずれか一つの周波数特性と、ＳＣＯＴフィルタの周波数特性とをそれぞれ重み付けし、重み付けしたこれらの周波数特性を乗じた周波数特性を持つフィルタ、または、ｌｏｇ_j（|Ｃ（ｆ）|）、および|Ｃ（ｆ）|^1/kの内、いずれか一つの周波数特性において、コヒーレンシィγ（ｆ）の絶対値が予め決められたある閾値以下になる周波数成分が０とするような周波数特性、もしくは、上記３つの超音波センサの内、ある２つのセンサで受信した２つの受信信号の周波数スペクトルの絶対値の比の分割された複数個の受信信号間の分散σ_abs（ｆ）と、上記２つのセンサで受信した２つの受信信号の周波数スペクトルの位相差の分割された複数個の受信信号間の分散σ_φ（ｆ）との積関する閾値σ_l（ｆ）がある閾値以上になる周波数成分が０とするような周波数特性と、ＳＣＯＴフィルタの周波数特性とをそれぞれ重み付けし、重み付けしたこれらの周波数特性を乗じた周波数特性を持つフィルタのいずれかである。
【０２８２】
３．上記前処理部は、下限周波数をｆｓ、上限周波数をｆｅとして、式（５）で表わされる周波数特性Ｈ₅（ｆ）を持つバンドパスフィルタをさらに含む。
【０２８３】
４．上記前処理部は、異常箇所が存在する場合としない場合の予備実験から得られる統計データに基づいて上記実数ｉ、ｊ、ｋおよび上記コヒーレンシィγ（ｆ）の絶対値に関する予め決められた閾値γ_l（ｆ）、上記３つの超音波センサの内、ある２つのセンサで受信した２つの受信信号の周波数スペクトルの絶対値の比の分割された複数個の受信信号間の分散σ_abs（ｆ）と、上記２つのセンサで受信した２つの受信信号の周波数スペクトルの位相差の分割された複数個の受信信号間の分散σ_φ（ｆ）との積に関する閾値σ_l（ｆ）を決定する。
【０２８４】
５．上記信号処理部は、上記相関処理部により演算された相互相関関数の包絡線を求め、さらに予め決められた繰り返し回数だけ求められた上記相互相関関数の包絡線を平均化する後処理部をさらに備える。
【０２８５】
６．上記後処理部は、上記平均化された相互相関関数の包絡線のピーク値と、予め決められたある閾値との大小関係から異常箇所の有無を判定する。
【０２８６】
７．上記後処理部は、３つの超音波センサの内、ある２つの超音波センサで受信した受信信号についての上記平均化された相互相関関数の包絡線がピークになる時間と、３つの超音波センサの内、上記２つの超音波センサの片方のセンサと上記２つの超音波センサ以外の超音波センサとで受信した受信信号についての上記平均化された相互相関関数の包絡線がピークになる時間と、上記３つの超音波センサの各離間距離とから異常箇所の位置を特定する。
【０２８７】
８．上記後処理部は、異常箇所が存在する場合としない場合の予備実験から得られる統計データに基づいて上記平均化された相互相関関数の包絡線のピーク値に関する閾値を決定する。
【０２８８】
９．被検査管に異常箇所が存在することにより発生する漏洩音を受信するための２つの超音波センサを備えると共に、これら超音波センサによる漏洩音の受信信号をそれぞれ受信部を介して入力し信号処理する信号処理部に、ｆを受信信号の周波数、ｉを０以上のある実数、上記２つの超音波センサで受信した受信信号から求めたコヒーレンシィをγ（ｆ）、上記２つのセンサの内の片方の超音波センサで受信した受信信号から求めたパワースペクトルをＣ₁₁（ｆ）とし、上記２つの超音波センサの他方で受信した受信信号から求めたパワースペクトルをＣ₂₂（ｆ）として、式（１）で表わされる周波数特性Ｈ₁（ｆ）を持ち、漏洩音の受信信号から求めたクロススペクトルをフィルタリングするためのフィルタＡと、上記コヒーレンシィγ（ｆ）の絶対値に関する予め決められた閾値をγ_l（ｆ）として、式（２）で表わされる周波数特性Ｈ₂（ｆ）を持つフィルタＢと、ｊを０以上のある実数、上記２つのセンサで受信した受信信号から求めたクロススペクトルをＣ（ｆ）として、式（３）で表わされる周波数特性Ｈ₃（ｆ）を持つフィルタＣと、ｋを１以上のある実数として、式（４）で表わされる周波数特性Ｈ₄（ｆ）を持つフィルタＤと、σ_abs（ｆ）を上記２つのセンサで受信した２つの受信信号の周波数スペクトルの絶対値の比の分割された複数個の受信信号間の分散とし、σ_φ（ｆ）を上記２つのセンサで受信した２つの受信信号の周波数スペクトルの位相差の分割された複数個の受信信号間の分散として、式（１６）で表わされる周波数特性Ｈ₇（ｆ）を持つフィルタＥと、上記２つのセンサで受信した２つの受信信号の周波数スペクトルの絶対値の比の分割された複数個の受信信号間の分散σ_abs（ｆ）と、上記２つのセンサで受信した２つの受信信号の周波数スペクトルの位相差の分割された複数個の受信信号間の分散σ_φ（ｆ）との積に関する閾値をσ_l（ｆ）とし、式（１７）で表わされる周波数特性Ｈ₈（ｆ）を持つフィルタＦと、上記６つの式の内、少なくともｌｏｇ_j（|Ｃ（ｆ）|）と|Ｃ（ｆ）|^1/kのいずれか１つの周波数特性と、ＳＣＯＴ（Smoothed Coherence TrＡｎsform）フィルタの周波数特性とをそれぞれ重み付けし、重み付けしたこれらの周波数特性を乗じた周波数特性を持つフィルタＧとの内、一つまたは二つ以上のフィルタを含む前処理部と、上記前処理部により前処理された信号から相互相関関数を演算する相関処理部とを備える。
【０２８９】
１０．上記フィルタＧは、｜γ（ｆ）｜ⁱおよび｛１／（σ_abs（ｆ）・σ_φ（ｆ））｝ⁱの内、いずれか一つの周波数特性と、ｌｏｇ_j（|Ｃ（ｆ）|）、および|Ｃ（ｆ）|^1/kの内、いずれか一つの周波数特性と、ＳＣＯＴフィルタの周波数特性とをそれぞれ重み付けし、重み付けしたこれらの周波数特性を乗じた周波数特性を持つフィルタ、またはｌｏｇ_j（|Ｃ（ｆ）|）、および|Ｃ（ｆ）|^1/kの内、いずれか一つの周波数特性において、コヒーレンシィγ（ｆ）の絶対値が予め決められたある閾値以下になる周波数成分が０となるような周波数特性、もしくは上記２つのセンサで受信した２つの受信信号の周波数スペクトルの絶対値の比の分割された複数個の受信信号間の分散σ_abs（ｆ）と、上記２つのセンサで受信した２つの受信信号の周波数スペクトルの位相差の分割された複数個の受信信号間の分散σ_φ（ｆ）との積に関する閾値σ_l（ｆ）が、予め決められたある閾値以上になる周波数成分が０とするような周波数特性と、ＳＣＯＴフィルタの周波数特性とをそれぞれ重み付けし、重み付けしたこれらの周波数特性を乗じた周波数特性を持つフィルタのいずれかである。
【０２９０】
１１．上記前処理部は、下限周波数をｆｓ、上限周波数をｆｅとして、式（５）で表わされる周波数特性Ｈ₅（ｆ）を持つバンドパスフィルタをさらに含む。
【０２９１】
１２．上記前処理部は、異常箇所が存在する場合としない場合の予備実験から得られる統計データに基づいて上記実数ｉ、ｊ、ｋ、上記コヒーレンシィγ（ｆ）の絶対値に関する予め決められた閾値γ_l（ｆ）、および上記２つのセンサで受信した２つの受信信号の周波数スペクトルの絶対値の比の分割された複数個の受信信号間の分散σ_abs（ｆ）と、上記２つのセンサで受信した２つの受信信号の周波数スペクトルの位相差の分割された複数個の受信信号間の分散σ_φ（ｆ）との積に関する閾値σ_l（ｆ）を決定する。
【０２９２】
１３．上記信号処理部は、上記相関処理部により演算された相互相関関数の包絡線を求め、さらに予め決められた繰り返し回数だけ求められた上記相互相関関数の包絡線を平均化する後処理部を備える。
【０２９３】
１４．上記後処理部は、上記平均化された相互相関関数の包絡線のピーク値と予め決められたある閾値との大小関係から異常箇所の有無を判定する。
【０２９４】
１５．上記後処理部は、２つの超音波センサで受信した受信信号についての上記平均化された相互相関関数の包絡線がピークになる時間と、被検査管を漏洩音が伝搬するときの伝搬速度と、上記２つの超音波センサの各離間距離とから異常箇所の位置を特定する。
【０２９５】
１６．上記後処理部は、異常箇所が存在する場合としない場合の予備実験から得られる統計データに基づいて上記平均化された相互相関関数の包絡線のピーク値に関する閾値を決定する。
【０２９６】
【発明の効果】
以上のように、この発明によれば、異常箇所からの漏洩音の特徴と、導管の異常箇所の検査の際にしばしば問題となるような雑音の特徴とを、十分に考慮した前処理を施し、その後に相関処理を行い、さらに後処理を行うようにすることにより、雑音中に含まれる導管からの漏洩音を、漏洩音の周波数帯域が未知であり、且つ、導管の漏洩箇所の検査においてしばしば問題となるような雑音が存在するような場合であっても、漏洩音を効率良く抽出し、導管に存在する異常箇所の存在の有無と位置を精度良く特定することができる。
【図面の簡単な説明】
【図１】 この発明の実施の形態１及び２による漏洩箇所検出装置の構成を示す模式図である。
【図２】 この発明の実施の形態１及び２における信号処理方法を説明するための説明図である。
【図３】 この発明の実施の形態１及び２における信号処理方法の効果を説明するための説明図である。
【図４】 この発明の実施の形態１及び２における信号処理方法の効果を説明するための説明図である。
【図５】 この発明の実施の形態１及び２における信号処理方法の効果を説明するための説明図である。
【図６】 この発明の実施の形態１及び２における信号処理方法を説明するための説明図である。
【図７】 この発明の実施の形態１及び２における信号処理方法の効果を説明するための説明図である。
【図８】 この発明の実施の形態１及び２における信号処理方法の効果を説明するための説明図である。
【図９】 この発明の実施の形態１及び２における信号処理方法の効果を説明するための説明図である。
【図１０】 この発明の実施の形態１及び２における信号処理方法を説明するための説明図である。
【図１１】 この発明の実施の形態１及び２における信号処理方法の効果を説明するための説明図である。
【図１２】 この発明の実施の形態１及び２における信号処理方法の効果を説明するための説明図である。
【図１３】 この発明の実施の形態１及び２における信号処理方法の効果を説明するための説明図である。
【図１４】 この発明の実施の形態１及び２における信号処理方法を説明するための説明図である。
【図１５】 この発明の実施の形態１及び２における信号処理方法の効果を説明するための説明図である。
【図１６】 この発明の実施の形態１及び２における信号処理方法の効果を説明するための説明図である。
【図１７】 この発明の実施の形態１及び２における信号処理方法の効果を説明するための説明図である。
【図１８】 この発明の実施の形態１及び２における信号処理方法を説明するための説明図である。
【図１９】 この発明の実施の形態１及び２における信号処理方法の効果を説明するための説明図である。
【図２０】 この発明の実施の形態１及び２における信号処理方法の効果を説明するための説明図である。
【図２１】 この発明の実施の形態１及び２における信号処理方法の効果を説明するための説明図である。
【図２２】 この発明の実施の形態１及び２における信号処理方法を説明するための説明図である。
【図２３】 この発明の実施の形態１及び２における信号処理方法の効果を説明するための説明図である。
【図２４】 この発明の実施の形態１及び２における信号処理方法の効果を説明するための説明図である。
【図２５】 この発明の実施の形態１及び２における信号処理方法の効果を説明するための説明図である。
【図２６】 この発明の実施の形態１及び２における信号処理方法を説明するための説明図である。
【図２７】 この発明の実施の形態１及び２における信号処理方法の効果を説明するための説明図である。
【図２８】 この発明の実施の形態１及び２における信号処理方法の効果を説明するための説明図である。
【図２９】 この発明の実施の形態１及び２における信号処理方法の効果を説明するための説明図である。
【図３０】 この発明の実施の形態１及び２における信号処理方法を説明するための説明図である。
【図３１】 この発明の実施の形態１及び２における信号処理方法の効果を説明するための説明図である。
【図３２】 この発明の実施の形態１及び２における信号処理方法の効果を説明するための説明図である。
【図３３】 この発明の実施の形態１及び２における信号処理方法の効果を説明するための説明図である。
【図３４】 この発明の実施の形態１及び２における信号処理方法の効果を説明するための説明図である。
【図３５】 従来例に係る漏洩箇所検出装置を説明するための構成図である。
【図３６】 従来例に係る漏洩箇所検出装置を説明するための構成図である。
【符号の説明】
１ 導管、２ 漏洩箇所、３ａ 超音波センサ、３ｂ 超音波センサ、３ｃ 超音波センサ、４ 音圧測定器、５ 相関器、６ 地中、７ 受信装置、７１ 受信部、７２信号処理部、７２ａ 前処理部、７２ｂ 相関処理部、７２ｃ 後処理部、７３ 表示部、７４ 制御部、８ 消火栓、９ レコーダー、１０ 計算機。[0001]
BACKGROUND OF THE INVENTION
The present invention relates to an abnormal point detection device for detecting, for example, an abnormal point of a conduit as an inspection tube through which gas such as water, oil, or other liquid or gas passes. It is related with the abnormal location detection apparatus which employ | adopted the system which detects the position of an abnormal location from receiving the leakage sound of this, and calculating the cross correlation function of a received signal.
[0002]
[Prior art]
As this type of conventional abnormal point detection device, as disclosed in Japanese Patent Laid-Open No. 5-87669, an abnormal point detection device that performs pre-processing and post-processing on received signals received by two sensors is known. ing.
FIG. 35 is a block diagram showing the above-described abnormal point detection apparatus.
In FIG. 35, 1 is a conduit, 2 is an abnormal location, 3a and 3b are ultrasonic sensors, 4 is a sound pressure measuring device, and 5 is a correlator.
[0003]
In the abnormal part detection apparatus described above, if the abnormal part 2 is present in the conduit 1, a leakage occurs thereby, and a leakage sound is generated accordingly. The leakage sound propagates through the conduit 1 together with noise and is received by the two ultrasonic sensors 3a and 3b.
The received signal is filtered by a bandpass filter having a predetermined band in order to remove noise. Thereafter, the cross-correlation function is calculated by the correlator 5 through the sound pressure measuring device 4, and the time at which the cross-correlation function takes a peak and the propagation speed at which the leaked sound propagates through the conduit 1 are calculated. Identify the location.
[0004]
As another conventional abnormal point detection device of this type, there is an abnormal point detection device that performs pre-processing and post-processing on received signals received by two sensors, as disclosed in JP-A-8-226865. Are known.
FIG. 36 is a block diagram showing the above-described abnormal point detection apparatus.
36, the same parts as those in FIG. 35 are denoted by the same reference numerals, and the description thereof is omitted. As a new code, 3c is an ultrasonic sensor.
[0005]
In the abnormal part detection apparatus described above, if the abnormal part 2 is present in the conduit 1, a leakage occurs thereby, and a leakage sound is generated accordingly. The leakage sound propagates along the conduit 1 together with noise and is received by the three ultrasonic sensors 3a, 3b, and 3c.
In order to remove noise in the received signal, the received signal is divided into a band-pass filter having a predetermined band and a SCOT (Smoothed Coherence Transform) filter (Carter et al. Proc. IEEE (Lett), 61, 10, pp 1497- 1498, 1973). Thereafter, using the signals obtained by the ultrasonic sensors 3a and 3b, a cross-correlation function is calculated in the correlation processing unit, and further, envelope detection is performed in the post-processing unit.
[0006]
From the time when the envelope of the cross-correlation function takes a peak and the distance between the two ultrasonic sensors 3a and 3b, the propagation speed of the abnormal part is obtained. Further, using the signals obtained by the ultrasonic sensors 3a and 3c, a cross-correlation function is calculated in the correlation processing unit, and further, envelope detection is performed in the post-processing unit. From the time when the envelope of the cross-correlation function takes a peak and the propagation speed obtained in the above process, the propagation speed of the abnormal part is obtained.
[0007]
[Problems to be solved by the invention]
As described above, the ultrasonic signal is received at two locations of the conduit, pre-processed into two received signals, and then the pre-processed signal is subjected to correlation processing and further subjected to post-processing. A technique for specifying the position is known.
[0008]
However, since the frequency spectrum of the leaking sound changes depending on the size and shape of the abnormal part and the pressure applied to the medium flowing in the conduit, the frequency band of the leaking sound cannot be accurately predicted in advance. Therefore, as a pre-processing of the correlation processing, even if filtering is performed with a band-pass filter whose band is determined in advance, the predetermined frequency band is not necessarily the frequency band possessed by the water leakage sound. Moreover, even if the predetermined frequency band is a frequency band that the water leaking sound has, the noise may also have the above frequency band.
[0009]
From the above, the bandpass filter alone is insufficient to remove noise. Therefore, since the correlation process is performed on the signal extracted by the bandpass filter, there is a problem in the accuracy of specifying the location of the abnormal part.
[0010]
In addition, as described in detail later, the SCOT filter continuously converts the absolute value of the filtered cross spectrum in time from a source at a fixed position regardless of the absolute value of the cross spectrum before being filtered. Filter that reflects the level ratio of a typical signal to a signal with at least one of the following characteristics: a signal from a moving source or a signal that is discontinuous in time .
[0011]
Therefore, a signal that is continuous in time from a source at a fixed position is a leaked sound and is a signal from a moving source or a signal that is discontinuous in time. If the characteristic signal is noise, the noise is dominant because the absolute value of the filtered cross spectrum is a function reflecting the signal-to-noise ratio regardless of the absolute value of the cross spectrum before filtering. It has the effect of suppressing frequency components.
[0012]
However, a signal that is continuous in time from a source at a fixed position is a leaked sound, that is, a signal from a moving source, or a signal that is discontinuous in time. Even if the characteristic signal is noise and this noise is dominant at a certain frequency component, for example, if the moving speed of the moving source is low, the absolute value of the filtered cross spectrum is Since the level is almost the same as the absolute value of the dominant frequency component, the function does not sufficiently reflect the SN ratio. That is, the frequency component dominated by the noise cannot be sufficiently suppressed.
[0013]
In addition, noise that is continuous in time from a certain position, such as mechanical noise from a factory or engine sound of a stopped car, cannot be distinguished from leakage sound.
Furthermore, from the characteristics of the SCOT filter, which will be described later, for example, the factory and the stopped vehicle are far away, and such frequency components dominated by temporally continuous noise from a certain position are filtered. In the previous cross spectrum, even if the level is small enough to be ignored, there is a problem that an adverse effect of amplifying this frequency component to almost the same level as the frequency component in which the leaking sound is dominant occurs.
[0014]
Therefore, there is a problem in the accuracy even if the position of the abnormal part is specified by using the SCOT filter for the preprocessing of the correlation process. Needless to say, the combination of the bandpass filter and the SCOT filter cannot solve the problems described above.
[0015]
In addition, as post-processing, the envelope of the cross-correlation function is calculated, and the location of the abnormal part is identified from the time when the envelope takes a peak, but for example, leakage noise from a water pipe, etc. Leaked sound is generally a random signal with no periodicity. Therefore, the position of the abnormal point 2 is identified from the time when the envelope of the cross-correlation function peaks in one limited measurement time. However, there is a problem that the stability of the inspection is poor.
[0016]
As described above, the pre-processing and post-processing methods of the correlation processing in the conventional abnormal point detection apparatus often cause problems in the characteristics of leakage sound from the abnormal point and the inspection of the abnormal point of the conduit. The characteristics of noise were not fully considered. Therefore, there is a problem in the accuracy of detecting the presence / absence of an abnormal location and the accuracy of specifying the position.
[0017]
The present invention has been made in view of the above circumstances, and pre-processing that sufficiently considers the characteristics of leaked sound from abnormal locations and the characteristics of noise that often becomes a problem when inspecting abnormal locations of conduits. After that, correlation processing is performed, and further post-processing is performed, thereby improving the accuracy and accuracy of the presence / absence of abnormal locations in the conduit and the location of abnormal locations and enabling stable detection. An abnormal location detecting device is provided.
[0018]
[Means for Solving the Problems]
The abnormal point detection device according to the present invention includes three ultrasonic sensors for receiving a leaked sound caused by the presence of an abnormal point in a tube to be inspected, and receives a reception signal of the leaked sound by these ultrasonic sensors. From the received signals received by two of the above three ultrasonic sensors, f is the frequency of the received signal, i is a real number greater than or equal to 0, The obtained coherency is γ (f), and the power spectrum obtained from the received signal received by one of the two sensors is C ₁₁ (F) The power spectrum obtained from the received signal received by the other of the two ultrasonic sensors is C _{twenty two} (F)
[0019]
[Expression 25]

[0020]
The frequency characteristic H expressed by ₁ A filter A having (f) for filtering the cross spectrum obtained from the received signal of the leaked sound;
A predetermined threshold value regarding the absolute value of the coherency γ (f) is expressed as γ _l (F)
[0021]
[Equation 26]

[0022]
The frequency characteristic H expressed by ₂ A filter B having (f);
j is a real number greater than or equal to 0, and among the above three ultrasonic sensors, a cross spectrum obtained from received signals received by two sensors is defined as C (f).
[0023]
[Expression 27]

[0024]
The frequency characteristic H expressed by _Three A filter C having (f);
Let k be a real number greater than or equal to 1,
[0025]
[Expression 28]

[0026]
The frequency characteristic H expressed by _Four A filter D having (f);
Of the above four expressions, at least log _j (| C (f) |) and | C (f) | ^{1 / k} And one of the frequency characteristics of the SCOT (Smoothed Coherence Transform) filter 1 / √ (C ₁₁ (F) ・ C _{twenty two} (F)), and a filter E having a frequency characteristic obtained by multiplying these weighted frequency characteristics,
A pre-processing unit including one or more filters,
And a correlation processing unit that calculates a cross-correlation function from the signal preprocessed by the preprocessing unit.
[0027]
In addition, the filter E is | γ (f) | ⁱ Frequency characteristics, log _j (| C (f) |) and | C (f) | ^{1 / k} A filter having a frequency characteristic obtained by weighting any one of the frequency characteristics and the frequency characteristic of the SCOT filter and multiplying the weighted frequency characteristics, or log _j (| C (f) |) and | C (f) | ^{1 / k} In any one of the frequency characteristics, the frequency characteristic in which the frequency component in which the absolute value of the coherency γ (f) is equal to or less than a certain threshold is 0 and the frequency characteristic of the SCOT filter are weighted, respectively. It is one of the filters having a frequency characteristic obtained by multiplying these frequency characteristics.
[0028]
In addition, the pre-processing unit sets the lower limit frequency to fs and the upper limit frequency to fe,
[0029]
[Expression 29]

[0030]
The frequency characteristic H expressed by _Five It further includes a bandpass filter having (f).
[0031]
In addition, the preprocessing unit is determined in advance with respect to the absolute values of the real numbers i, j, k and the coherency γ (f) based on statistical data obtained from a preliminary experiment with and without an abnormal location. Threshold γ _l (F) is determined.
[0032]
The signal processing unit obtains an envelope of the cross-correlation function calculated by the correlation processing unit, and further averages the envelope of the cross-correlation function obtained by a predetermined number of repetitions. Is further provided.
[0033]
The post-processing unit is characterized by determining the presence / absence of an abnormal portion from the magnitude relationship between the peak value of the averaged cross-correlation function envelope and a predetermined threshold value.
[0034]
In addition, the post-processing unit includes a time at which an envelope of the averaged cross-correlation function for the received signal received by two of the ultrasonic sensors reaches a peak, The envelope of the averaged cross-correlation function for the received signal received by one of the two ultrasonic sensors and the ultrasonic sensor other than the two ultrasonic sensors is a peak. The position of the abnormal part is specified from the time and the separation distances of the three ultrasonic sensors.
[0035]
Further, the post-processing unit determines a threshold value related to a peak value of an envelope of the averaged cross-correlation function based on statistical data obtained from a preliminary experiment in the presence or absence of an abnormal part. It is what.
[0036]
In addition, an abnormal point detection apparatus according to another invention includes two ultrasonic sensors for receiving leaked sound generated when an abnormal point exists in a tube to be inspected, and leaks sound leaked by these ultrasonic sensors. Coherency obtained from the received signals received by the above two ultrasonic sensors, where f is the frequency of the received signal, i is a real number greater than or equal to 0, and the received signal is input to the received signal through the receiving unit. Γ (f), the power spectrum obtained from the received signal received by one of the two sensors is C ₁₁ (F) The power spectrum obtained from the received signal received by the other of the two ultrasonic sensors is C _{twenty two} (F)
[0037]
[30]

[0038]
The frequency characteristic H expressed by ₁ (F), a filter A for filtering the cross spectrum obtained from the received signal of the leaked sound, and a predetermined threshold value regarding the absolute value of the coherency γ (f) as γ _l (F)
[0039]
[31]

[0040]
The frequency characteristic H expressed by ₂ C (f) is a cross spectrum obtained from a filter B having (f), j being a real number greater than or equal to 0, and received signals received by the two sensors.
[0041]
[Expression 32]

[0042]
The frequency characteristic H expressed by _Three Filter C with (f) and k as a real number greater than or equal to 1,
[0043]
[Expression 33]

[0044]
The frequency characteristic H expressed by _Four Filter D having (f) and at least log of the above four expressions _j (| C (f) |) and | C (f) | ^{1 / k} And the frequency characteristic of the SCOT filter 1 / √ (C ₁₁ (F) ・ C _{twenty two} (F)) and a pre-processing unit including one or two or more of the filters E having frequency characteristics obtained by multiplying these weighted frequency characteristics, and the pre-processing unit And a correlation processing unit that calculates a cross-correlation function from the processed signal.
[0045]
In addition, the filter E is | γ (f) | ⁱ Frequency characteristics, log _j (| C (f) |) and | C (f) | ^{1 / k} A filter having a frequency characteristic obtained by weighting any one of the frequency characteristics and the frequency characteristic of the SCOT filter and multiplying the weighted frequency characteristics, or log _j (| C (f) |) and | C (f) | ^{1 / k} In any one of the frequency characteristics, a frequency characteristic in which a frequency component in which the absolute value of the coherency γ (f) is equal to or less than a certain threshold is 0, and a frequency characteristic of a SCOT (Smoothed Coherence Transform) filter are obtained. Each of the filters has a frequency characteristic obtained by weighting and multiplying these weighted frequency characteristics.
[0046]
In addition, the pre-processing unit sets the lower limit frequency to fs and the upper limit frequency to fe,
[0047]
[Expression 34]

[0048]
The frequency characteristic H expressed by _Five It further includes a bandpass filter having (f).
[0049]
In addition, the preprocessing unit is determined in advance with respect to the absolute values of the real numbers i, j, k and the coherency γ (f) based on statistical data obtained from a preliminary experiment with and without an abnormal location. Threshold γ _l (F) is determined.
[0050]
The signal processing unit obtains an envelope of the cross-correlation function calculated by the correlation processing unit, and further averages the envelope of the cross-correlation function obtained by a predetermined number of repetitions. Is further provided.
[0051]
The post-processing unit is characterized by determining the presence / absence of an abnormal portion from the magnitude relationship between the peak value of the averaged cross-correlation function envelope and a predetermined threshold value.
[0052]
In addition, the post-processing unit has a time when the envelope of the averaged cross-correlation function for the received signals received by the two ultrasonic sensors reaches a peak, and a propagation when the leaked sound is propagated through the test tube. The position of the abnormal part is specified from the speed and the separation distances of the two ultrasonic sensors.
[0053]
Further, the post-processing unit determines a threshold value regarding the peak value of the envelope of the averaged cross-correlation function based on statistical data obtained from a preliminary experiment in the presence or absence of an abnormal location. It is what.
[0054]
In addition, an abnormal point detection apparatus according to another invention includes three ultrasonic sensors for receiving leaked sound generated due to the presence of an abnormal point in a tube to be inspected, and leaked sound by these ultrasonic sensors. The received signal is input to the signal processing unit via the receiving unit, and f is the frequency of the received signal, i is a real number greater than or equal to 0, and is received by two of the three ultrasonic sensors. The coherency obtained from the received signal is γ (f), and the power spectrum obtained from the received signal received by one of the two sensors is C ₁₁ (F) The power spectrum obtained from the received signal received by the other of the two ultrasonic sensors is C _{twenty two} (F)
[0055]
[Expression 35]

[0056]
The frequency characteristic H expressed by ₁ (F), a filter A for filtering the cross spectrum obtained from the received signal of the leaked sound, and a predetermined threshold value regarding the absolute value of the coherency γ (f) as γ ₁ (F)
[0057]
[Expression 36]

[0058]
The frequency characteristic H expressed by ₂ Filter B having (f), j being a real number greater than or equal to 0, and among the above three ultrasonic sensors, a cross spectrum obtained from received signals received by two sensors is defined as C (f),
[0059]
[Expression 37]

[0060]
The frequency characteristic H expressed by _Three Filter C with (f) and k as a real number greater than or equal to 1,
[0061]
[Formula 38]

[0062]
The frequency characteristic H expressed by _Four A filter D having (f) and σ _abs (F) is the variance between a plurality of received signals obtained by dividing the ratio of the absolute value of the frequency spectrum of two received signals received by two of the three ultrasonic sensors, and σ _φ (F) is the variance between a plurality of received signals obtained by dividing the phase difference of the frequency spectrum of the two received signals received by the two sensors.
[0063]
[39]

[0064]
The frequency characteristic H expressed by ₇ The variance E between the filter E having (f) and a plurality of received signals obtained by dividing the ratio of the absolute values of the frequency spectra of two received signals received by two of the three ultrasonic sensors. _abs (F) and the variance σ between a plurality of received signals obtained by dividing the phase difference of the frequency spectrum of the two received signals received by the two sensors. _φ The threshold for the product with (f) is σ _l (F)
[0065]
[Formula 40]

[0066]
The frequency characteristic H expressed by ₈ Filter F having (f) and at least log of the above six expressions _j (| C (f) |) and | C (f) | ^{1 / k} One or more of the filter G having a frequency characteristic obtained by weighting any one of the frequency characteristics and the frequency characteristic of the SCOT (Smoothed Coherence Transform) filter and multiplying these weighted frequency characteristics. And a correlation processing unit that calculates a cross-correlation function from the signal preprocessed by the preprocessing unit.
[0067]
The filter G is | γ (f) | ⁱ And {1 / (σ _abs (F) · σ _φ (F))} ⁱ Any one of the frequency characteristics and log _j (| C (f) |) and | C (f) | ^{1 / k} A filter having a frequency characteristic obtained by weighting any one of the frequency characteristics and the frequency characteristic of the SCOT filter and multiplying these weighted frequency characteristics, or log _j (| C (f) |) and | C (f) | ^{1 / k} In any one of the frequency characteristics, a frequency characteristic in which the frequency component at which the absolute value of the coherency γ (f) is equal to or less than a predetermined threshold is 0, or the above three ultrasonic sensors Among the received signals divided by the ratio of the absolute value of the frequency spectrum of the two received signals received by two sensors. _abs (F) and the variance σ between a plurality of received signals obtained by dividing the phase difference of the frequency spectrum of the two received signals received by the two sensors. _φ Threshold value σ related to product with (f) _l (F) A frequency characteristic in which a frequency component that exceeds a certain threshold is set to 0 and a frequency characteristic of the SCOT filter are each weighted, and any of the filters having a frequency characteristic obtained by multiplying these weighted frequency characteristics. It is characterized by being.
[0068]
In addition, the pre-processing unit sets the lower limit frequency to fs and the upper limit frequency to fe,
[0069]
[Expression 41]

[0070]
The frequency characteristic H expressed by _Five It further includes a bandpass filter having (f).
[0071]
In addition, the preprocessing unit is determined in advance with respect to the absolute values of the real numbers i, j, k and the coherency γ (f) based on statistical data obtained from a preliminary experiment with and without an abnormal location. Threshold γ _l (F) Dispersion σ between a plurality of received signals obtained by dividing the ratio of the absolute values of the frequency spectra of two received signals received by two of the three ultrasonic sensors. _abs (F) and the variance σ between a plurality of received signals obtained by dividing the phase difference of the frequency spectrum of the two received signals received by the two sensors. _φ Threshold value for product with (f) _l (F) is determined.
[0072]
The signal processing unit obtains an envelope of the cross-correlation function calculated by the correlation processing unit, and further averages the envelope of the cross-correlation function obtained by a predetermined number of repetitions. Is further provided.
[0073]
The post-processing unit is characterized by determining the presence / absence of an abnormal portion from the magnitude relationship between the peak value of the envelope of the averaged cross-correlation function and a predetermined threshold value. .
[0074]
In addition, the post-processing unit includes a time at which an envelope of the averaged cross-correlation function for the received signal received by two of the ultrasonic sensors reaches a peak, The envelope of the averaged cross-correlation function for the received signal received by one of the two ultrasonic sensors and the ultrasonic sensor other than the two ultrasonic sensors is a peak. The position of the abnormal part is specified from the time and the separation distances of the three ultrasonic sensors.
[0075]
Further, the post-processing unit determines a threshold value related to a peak value of an envelope of the averaged cross-correlation function based on statistical data obtained from a preliminary experiment in the presence or absence of an abnormal part. It is what.
[0076]
Furthermore, an abnormal point detection apparatus according to still another invention includes two ultrasonic sensors for receiving a leaked sound caused by the presence of an abnormal point in a tube to be inspected, and a leaked sound generated by these ultrasonic sensors. Are received through the receiving unit, and the signal processing unit that processes the received signal, f is the frequency of the received signal, i is a real number greater than or equal to 0, and the coherency obtained from the received signals received by the two ultrasonic sensors. Γ (f), and the power spectrum obtained from the received signal received by one of the two sensors is C ₁₁ (F), and the power spectrum obtained from the received signal received by the other of the two ultrasonic sensors is C _{twenty two} (F)
[0077]
[Expression 42]

The frequency characteristic H expressed by ₁ (F), a filter A for filtering the cross spectrum obtained from the received signal of the leaked sound, and a predetermined threshold value regarding the absolute value of the coherency γ (f) as γ _l (F)
[0078]
[Expression 43]

[0079]
The frequency characteristic H expressed by ₂ C (f) is a cross spectrum obtained from a filter B having (f), j being a real number greater than or equal to 0, and received signals received by the two sensors.
[0080]
(44)

[0081]
The frequency characteristic H expressed by _Three Filter C with (f) and k as a real number greater than or equal to 1,
[0082]
[Equation 45]

[0083]
The frequency characteristic H expressed by _Four A filter D having (f) and σ _abs (F) is the variance between a plurality of received signals divided by the ratio of the absolute values of the frequency spectra of the two received signals received by the two sensors, and σ _φ (F) is the variance between a plurality of received signals obtained by dividing the phase difference of the frequency spectrum of the two received signals received by the two sensors.
[0084]
[Equation 46]

[0085]
The frequency characteristic H expressed by ₇ Variance σ between a plurality of received signals obtained by dividing a ratio of absolute values of frequency spectra of a filter E having (f) and two received signals received by the two sensors. _abs (F) and the variance σ between a plurality of received signals obtained by dividing the phase difference of the frequency spectrum of the two received signals received by the two sensors. _φ The threshold for the product with (f) is σ _l (F)
[0086]
[Equation 47]

[0087]
The frequency characteristic H expressed by ₈ Filter F having (f) and at least log of the above six expressions _j (| C (f) |) and | C (f) | ^{1 / k} One or more of the filter G having a frequency characteristic obtained by weighting any one of the frequency characteristics and the frequency characteristic of the SCOT (Smoothed Coherence Transform) filter and multiplying these weighted frequency characteristics. And a correlation processing unit that calculates a cross-correlation function from the signal preprocessed by the preprocessing unit.
[0088]
The filter G is | γ (f) | ⁱ And {1 / (σ _abs (F) · σ _φ (F))} ⁱ Any one of the frequency characteristics and log _j (| C (f) |) and | C (f) | ^{1 / k} A filter having a frequency characteristic obtained by weighting any one of the frequency characteristics and the frequency characteristic of the SCOT filter and multiplying these weighted frequency characteristics, or log _j (| C (f) |) and | C (f) | ^{1 / k} In any one of the frequency characteristics, the frequency characteristic in which the frequency component at which the absolute value of the coherency γ (f) is equal to or less than a predetermined threshold becomes 0, or 2 received by the two sensors. The variance σ between the received signals divided by the ratio of the absolute value of the frequency spectrum of two received signals _abs (F) and the variance σ between a plurality of received signals obtained by dividing the phase difference of the frequency spectrum of the two received signals received by the two sensors. _φ Threshold value for product with (f) _l (F) has a frequency characteristic that weights the frequency characteristic such that the frequency component that exceeds a predetermined threshold value is 0 and the frequency characteristic of the SCOT filter, and multiplies these weighted frequency characteristics. It is one of the filters.
[0089]
In addition, the pre-processing unit sets the lower limit frequency to fs and the upper limit frequency to fe,
[0090]
[Formula 48]

[0091]
The frequency characteristic H expressed by _Five It further includes a bandpass filter having (f).
[0092]
Further, the preprocessing unit is determined in advance with respect to the absolute values of the real numbers i, j, k, and the coherency γ (f) based on statistical data obtained from a preliminary experiment with and without an abnormal location. Threshold γ _l (F) and the variance σ between a plurality of received signals divided by the ratio of the absolute values of the frequency spectra of the two received signals received by the two sensors. _abs (F) and the variance σ between a plurality of received signals obtained by dividing the phase difference of the frequency spectrum of the two received signals received by the two sensors. _φ Threshold value for product with (f) _l (F) is determined.
[0093]
The signal processing unit obtains an envelope of the cross-correlation function calculated by the correlation processing unit, and further averages the envelope of the cross-correlation function obtained by a predetermined number of repetitions. It is characterized by comprising.
[0094]
The post-processing unit is characterized by determining the presence / absence of an abnormal portion from the magnitude relationship between the peak value of the averaged cross-correlation function envelope and a predetermined threshold value.
[0095]
In addition, the post-processing unit transmits a time when the envelope of the averaged cross-correlation function for the received signals received by the two ultrasonic sensors reaches a peak, and a propagation when the leaked sound propagates through the test tube. The position of the abnormal part is specified from the speed and the separation distances of the two ultrasonic sensors.
[0096]
Further, the post-processing unit determines a threshold value regarding the peak value of the envelope of the averaged cross-correlation function based on statistical data obtained from a preliminary experiment in the presence or absence of an abnormal location. It is what.
[0097]
DETAILED DESCRIPTION OF THE INVENTION
Embodiment 1 FIG.
An abnormal point detection apparatus according to Embodiment 1 of the present invention will be described with reference to FIGS.
FIG. 1 is a block diagram showing an abnormal point detection apparatus according to Embodiment 1 of the present invention. In FIG. 1, 1 is a conduit as an inspection tube through which water, oil, or other liquid or gas such as gas passes, 2 is an abnormal portion of the conduit 1, and 6 is underground.
[0098]
1 includes ultrasonic sensors 3a, 3b, and 3c for receiving leaked sound generated when the abnormal location 2 is present, and a receiving device 7.
Although FIG. 1 shows a case where the conduit 1 is buried in the underground 6, the conduit 1 may be all or part of the conduit 1 even if the conduit 1 exists above the underground 6. I do not care. Moreover, although the case where the said abnormal part 2 is one place is described, the said abnormal part 2 may not be one place but may be multiple places.
[0099]
In FIG. 1, L1 is a distance along the conduit 1 between the ultrasonic sensor 3a and the ultrasonic sensor 3b, and L2 is a distance along the conduit 1 between the ultrasonic sensor 3b and the ultrasonic sensor 3c. x is a distance along the conduit 1 from the ultrasonic sensor 3a to the abnormal part 2;
[0100]
In FIG. 1, gas such as water, oil, or other liquid or gas in the conduit 1 may or may not flow. Moreover, when flowing, the direction of the flow may be either direction.
[0101]
In addition, although an ultrasonic wave is used as a term indicating a sound wave or an elastic wave having a frequency high enough to be inaudible to human ears, in the present invention, the frequency is not particularly defined. That is, the term “ultrasound” in the present invention is not limited to sound waves and elastic waves having a frequency higher than the upper limit of the frequency that can be heard by the human ear, but also includes sound waves and elastic waves having a frequency lower than the upper limit. Of course, it also includes the meaning of sound waves and elastic waves at frequencies lower than the lower limit of the frequency that can be heard by the human ear.
[0102]
FIG. 1 shows the case where the ultrasonic sensors 3a, 3b and 3c are placed against the conduit 1, but the ultrasonic sensors 3a, 3b and 3c are provided at three locations of the conduit 1. If the objective is to receive leaked sound and this objective can be achieved, the ultrasonic sensors 3a, 3b and 3c may not be in direct contact with the conduit 1. Further, if this object can be achieved, the ultrasonic sensors 3a, 3b and 3c may be arranged inside the conduit 1.
[0103]
Here, the positional relationship between the abnormal part 2 and the ultrasonic sensors 3a, 3b, and 3c will be described.
An area where two ultrasonic sensors among the three ultrasonic sensors, for example, the ultrasonic sensors 3b and 3c in FIG. 1, are exposed on the ground and can be visually determined that there is no abnormality between them. It is an area where it is known that there is no abnormal part 2. The abnormal part 2 is an ultrasonic sensor other than the two ultrasonic sensors of the three ultrasonic sensors, for example, the ultrasonic sensor 3a in FIG. 1 and the two ultrasonic sensors of the two ultrasonic sensors. It is located between the ultrasonic sensors closer to the ultrasonic sensors other than the ultrasonic sensor, for example, the ultrasonic sensor 3b in FIG.
[0104]
In FIG. 1, the reception device 7 includes three reception units 71, a signal processing unit 72, a display unit 73 as a notification unit, and a control unit 74.
The ultrasonic sensors 3 a, 3 b and 3 c are connected to the receiving unit 71. The receiving unit 71 is connected to the signal processing unit 72. The signal processing unit 72 is connected to the display unit 73.
[0105]
The control unit 74 is connected to the reception unit 71, the signal processing unit 72, and the display unit 73, and receives information and commands for performing an inspection, and receives the reception unit 71, the signal processing unit 72, and the display unit. The control signal for controlling these operations and the signal related to the information on the progress of the inspection are sequentially transmitted to 73 to control these operations.
[0106]
The receiving unit 71 includes an amplifier for amplifying the received signal and an A / D conversion unit (not shown).
The signal processing unit 72 is a preprocessing unit 72a for filtering the received signal, and a correlation for calculating a cross-correlation function of two signals among the three received signals received by the ultrasonic sensors 3a, 3b and 3c. It includes a processing unit 72b and a post-processing unit 72c for performing processing such as envelope detection on the cross-correlation function. The pre-processing unit 72a is connected to the correlation processing unit 72b, and the correlation processing unit 72b is connected to the post-processing unit 72c. The signal processing unit 72 has a memory therein although not shown. Various results of arithmetic processing are appropriately stored in this memory.
[0107]
Further, the preprocessing unit 72a in the signal processing unit 72 sets the frequency to f, sets i to a real number greater than or equal to 0, and receives signals received by two of the ultrasonic sensors 3a, 3b, and 3c. The obtained coherency is γ (f), and the power spectrum obtained from the received signal received by one of the two sensors is C ₁₁ (F) and the power spectrum obtained from the received signal received by the other of the two sensors is C _{twenty two} (F)
[0108]
[Formula 49]

[0109]
The frequency characteristic H expressed by ₁ A filter having (f);
The threshold for the absolute value of coherency γ (f) is γ _l (F)
[0110]
[Equation 50]

[0111]
The frequency characteristic H expressed by ₂ A filter having (f);
Let j be a real number greater than or equal to 0, and let C (f) be the cross spectrum obtained from the received signals received by the two sensors among the ultrasonic sensors 3a, 3b, and 3c.
[0112]
[Equation 51]

[0113]
The frequency characteristic H expressed by _Three A filter having (f);
Let k be a real number greater than or equal to 1,
[0114]
[Formula 52]

[0115]
The frequency characteristic H expressed by _Four A filter having (f);
| Γ (f) | in Equation (1) ⁱ Frequency characteristics and log in equation (3) _j (| C (f) |) and | C (f) | in equation (4) ^{1 / k} Among these, a filter having a frequency characteristic obtained by weighting and multiplying one of the frequency characteristics and the frequency characteristic of the SCOT filter, or log in the expression (3) _j (| C (f) |) and | C (f) | in equation (4) ^{1 / k} In any one of the frequency characteristics, a frequency characteristic in which the frequency component at which the absolute value of coherency γ (f) is equal to or smaller than a certain threshold value is 0 as shown in Expression (2), and the frequency of the SCOT filter Filter with frequency characteristics weighted and multiplied by these
Any one or two or more of them are included.
Further, the lower limit frequency is fs, the upper limit frequency is fe,
[0116]
[53]

[0117]
The frequency characteristic H expressed by _Five A bandpass filter with (f) is included.
[0118]
In the filter characteristics included in the preprocessing unit 72a, the lower limit frequency fs and the upper limit frequency fe in the equation (5) are determined according to the result of the preliminary experiment, and these values are input to the control unit 74. Further, a real number i in the equation (1) and a threshold value γ relating to the absolute value of coherency in the equation (2) _l (F), the real number j in the equation (3), and the real number k in the equation (4) are also determined according to the result of the same preliminary experiment.
[0119]
This preliminary experiment is performed using the same or similar abnormal point detection apparatus as the abnormal point detection apparatus according to the first embodiment when the abnormal part 2 exists and when it does not substantially exist. . From the statistical data obtained from such a preliminary experiment, the real number i and the formula (1) in Formula (1) that can detect the presence / absence of the abnormal location 2 and the position of the abnormal location 2 with the highest accuracy according to the state of the inspection. Threshold value γ for the absolute value of coherency in 2) _l (F), the real number j in the expression (3), the real number k in the expression (4), and the lower limit frequency fs and the upper limit frequency fe in the expression (5) are determined.
[0120]
As will be described later, if the propagation speed when the leaked sound propagates through the conduit 1 is known, the space between the two ultrasonic sensors 3a, 3b, and 3c is exposed on the ground surface. One of the two ultrasonic sensors that are known to have no abnormal part 2, such as an area where it can be visually determined that there is no abnormal part between them, such as the ultrasonic sensor in FIG. Even if 3c is removed, the presence / absence of the abnormal part 2 and the position of the abnormal part 2 can be specified. In such a case, removing one of the ultrasonic sensors 3a, 3b, and 3c has an effect of facilitating the inspection and reducing the cost of the inspection system.
[0121]
Here, the characteristics of the leaked sound and the characteristics of the noise will be described.
The leakage sound has a feature that it is a continuous signal in time from a transmission source at a fixed position.
On the other hand, for example, noises such as automobile running sounds and temporary water usage sounds that are often received in water pipe leak inspections are signals from moving sources or signals that are discontinuous in time. It has at least one of the characteristics.
[0122]
There are also noises that are continuous in time from a fixed source, such as factory machine noise and engine sound of a car that is often stopped in water pipe leak inspections. To do. Leaked sound and such noise cannot be distinguished from characteristics such as signal continuity and whether or not the transmission source moves in time.
[0123]
However, with regard to the transmission source of signals that are continuous in time from the transmission source at a certain position, such as the mechanical sound of the factory and the engine sound of the stopped car, the transmission source exists nearby. Therefore, prior to the inspection, it is easy to know the location of the transmission source because the map near the inspection site or the vicinity of the inspection site is observed. It is possible to prevent noise from such a transmission source from being generated by requesting not to generate it or by requesting the driver of the car not to generate engine sound temporarily.
[0124]
In addition, there is a case where such a source is far away and it is not easy to know its presence prior to the inspection. In such a case, the signal level is weak. Have.
As described above, leakage sound and noise have characteristics in signal continuity, temporal stability of the transmission source position, and signal level.
[0125]
Next, the operation of the abnormal point detection apparatus shown in FIG. 1 will be described using the flowchart of FIG.
First, the frequency band of the leaked sound, the separation distance between the ultrasonic sensor 3a and the ultrasonic sensor 3b, the separation distance between the ultrasonic sensor 3b and the ultrasonic sensor 3c, and the time length of data captured by one reception in the control unit 74. Determine the number N of divisions on the time axis for data received in a single reception, the number M of repeated data acquisitions to stabilize the accuracy of the position of an abnormal location 2 to be described later, and the presence or absence of leakage A predetermined threshold value is input (step S1). When the frequency band of the leaked sound is unknown, information that the frequency band is unknown is input.
[0126]
Further, the time length of the data to be captured by the one reception and the number of repetitions M of the data capture are set by the allowable stability of the inspection and the allowable time for the inspection, and are captured by the one reception. By increasing the data length and increasing the number of repetitions M of the data acquisition, the stability of the inspection can be increased. If the inspection time allowed is short, the length of the data to be captured may be reduced and the number of repetitions M of the data capture may be reduced.
[0127]
The control unit 74 also stores information such as the date of the inspection, the time of the inspection, the types and serial numbers of the ultrasonic sensors 3a, 3b, and 3c, and the position of the conduit on the map that has been inspected. All of the information, or any one or more of the above information can be input, the input information can be displayed on the display unit 73, and the input information is recorded. If stored, this not only increases the stability of the inspection, but also helps to construct a database relating to the occurrence tendency of the abnormal part 2. Furthermore, there is an effect that is useful as reference data in a periodic inspection after a certain period. Further, when a second inspection is performed after a certain period of time, it can be used for confirming the reproducibility of inspection data and investigating changes with time.
[0128]
Next, the inspection start timing is input by the inspector. The timing input may be performed by turning on / off the switch, or may be performed, for example, by picking up the sound of the inspector or the sound of the inspector clapping with a microphone. Further, by providing the control unit 74 with a timer, the inspection may be started when a certain time has elapsed after the timing is input.
[0129]
In FIG. 1, when the abnormal location 2 exists in the conduit | pipe 1, a leak will arise from the said abnormal location 2, and a leak sound will generate | occur | produce in connection with it. The generated leakage sound propagates through the conduit 1 and is received together with noise by the ultrasonic sensors 3a, 3b and 3c.
[0130]
Next, a signal for starting reception by the ultrasonic sensors 3a, 3b and 3c is transmitted from the control unit 74 to the reception unit 71, and reception signals received by the ultrasonic sensors 3a, 3b and 3c are received by the reception unit 71. (Step S2). In the reception unit 71, the received signal is A / D converted after being amplified and sent to a preprocessing unit 72 a included in the signal processing unit 72.
[0131]
In the preprocessing unit 72a, the three received signals are each divided into N, which is a predetermined number on the time axis. Next, each of the divided N pieces of data is Fourier transformed, and a frequency spectrum for each of the N pieces of data is calculated.
[0132]
With respect to the signal received by the ultrasonic sensor 3a, the time length of the data captured in one reception is T, and the frequency spectrum in the component of a certain frequency f in the nth data of the N divided data An is the absolute value of, the phase is φan, the symbol representing the conjugate complex number is *,
[0133]
[Formula 54]

[0134]
The power spectrum C _aa (F) is calculated. Further, regarding the signal received by the ultrasonic sensor 3b, the absolute value of the frequency spectrum in a certain frequency f component in the nth data of the N divided data is Bn, and its phase is φbn.
[0135]
[Expression 55]

[0136]
The power spectrum C _bb (F) is calculated. Further, regarding the signal received by the ultrasonic sensor 3c, the absolute value of the frequency spectrum in a component of a certain frequency f in the nth data of the N divided data is Cn, and its phase is φcn.
[0137]
[Expression 56]

[0138]
The power spectrum C _CC (F) is calculated (step S3).
[0139]
Next, it is known that the region between the two ultrasonic sensors 3b and 3c is exposed on the ground surface, and there is no abnormal portion 2 such as a region where it can be visually determined that there is no abnormal portion between them. A combination of one of the two ultrasonic sensors as a region and another ultrasonic sensor 3a different from the two ultrasonic sensors, for example, a combination of the ultrasonic sensor 3a and the ultrasonic sensor 3b in FIG. Is selected.
[0140]
Further, a region where the region between the two ultrasonic sensors 3b and 3c is exposed on the ground surface, and is known to have no abnormal part 2, such as a region where it can be visually determined that there is no abnormal part between them. Of the two ultrasonic sensors, the ultrasonic sensor 3c that is not selected in the combination of the ultrasonic sensors described above and the region between the two ultrasonic sensors are exposed on the ground surface, and there is an abnormal point between them. Another combination of two ultrasonic sensors that are areas where it is known that there is no abnormal part 2, such as an area where it can be visually determined that there is no abnormality, such as the ultrasonic sensor 3a and the ultrasonic sensor 3c in FIG. A combination is selected (step S4).
[0141]
Here, a signal processing method for the combination of the ultrasonic sensors 3a and 3b among the combination of the two ultrasonic sensors, the combination of the ultrasonic sensors 3a and 3b, and the combination of the ultrasonic sensors 3a and 3c will be described.
Cross spectrum C of received signals received by two ultrasonic sensors 3a and 3b _ab (F)
[0142]
[Equation 57]

[0143]
(Step S5). The cross spectrum C (f) is expressed as C in Equation (1) ₁₁ (F) = C _aa (F), C _{twenty two} (F) = C _bb Further filtering is performed by a filter having a frequency characteristic of Formula (1).
[0144]
Here, the effect of the filter having the frequency characteristic of Expression (1) will be described in detail.
Expression (1) is obtained by multiplying the frequency characteristic of the SCOT filter by a value obtained by multiplying the absolute value of coherency γ (f) to the real number i.
[0145]
First, the characteristics of coherency γ (f) will be described.
Coherency γ (f) is
[0146]
[Formula 58]

[0147]
(Step S6). The coherency γ (f) represented by the equation (10) can be understood from substituting the equations (6), (7), and (9) into the equation (10).
[0148]
[Formula 59]

[0149]
If the condition is satisfied, the absolute value is 1. If the condition of Expression (11) is not satisfied, the absolute value is 1 or less. A and φ _a Indicates the amplitude and phase of the frequency spectrum of the received signal received by the ultrasonic sensor 3a. Similarly, B and φ _b Indicates the amplitude and phase of the frequency spectrum of the received signal received by the ultrasonic sensor 3b (subscript is time data).
[0150]
The fact that the condition of Expression (11) is satisfied means that, at the frequency f, the amplitude ratio and the phase difference of the frequency spectrum of the received signals received by the two ultrasonic sensors 3a and 3b are divided on the time axis. It means that it is constant among N data.
This means that the component of the frequency f is a temporally continuous signal from a transmission source at a fixed position.
This means that the signal of frequency f includes noise such as leakage sound, for example, mechanical sound of a factory or engine sound of a stopped car.
[0151]
In addition, the fact that the condition of Expression (11) is not satisfied means that, at the frequency f, the amplitude ratio and the phase difference of the frequency spectrum of the received signals received by the two ultrasonic sensors 3a and 3b are divided on the time axis. It means that it is not constant among the N pieces of data. In other words, the signal of frequency f means that the signal has at least one of the characteristics of a signal from a moving transmission source or a signal that is discontinuous in time.
Further, the component of the frequency f includes leakage sound, noise such as mechanical sound of a factory and engine sound of a stopped car, driving sound of a car, water use sound, pipe running water sound, etc. This means that the noise is also included.
[0152]
Further, at a certain frequency f, if a temporally continuous signal from a source at a fixed position is dominant, such as leakage sound, machine noise of a factory, or engine sound of a stopped car, Equation (11) is approximately established and the absolute value of coherency γ (f) is approximately 1. However, a moving source of noise such as automobile running sound, water use sound, pipe running water sound, etc. As the signal having at least one of the following characteristics becomes dominant, the equation (11) does not hold approximately, and the coherency The absolute value of γ (f) takes a small value accordingly.
[0153]
As described above, the coherency γ (f) of the equation (10) is expressed by the frequency spectrum of the received signal received by the two ultrasonic sensors, here, the ultrasonic sensors 3a and 3b at the frequency f. In terms of frequency components, signals such as leaking noise, factory machine noise, and engine sound of a stopped car, such as a continuous signal from a source at a fixed position, driving noise of a car, It is a function reflecting a level ratio with a signal having at least one of the characteristics of a signal from a moving transmission source such as noise such as a running water sound in a pipe or a signal discontinuous in time.
[0154]
Next, the characteristic determination of the SCOT filter will be described (step S7).
The SCOT filter makes the absolute value of the filtered cross spectrum C ′ (f) continuous in time from a fixed source regardless of the absolute value of the cross spectrum C (f) before being filtered. It is a filter that makes the same value as the absolute value of coherency γ (f), which is a function reflecting the level ratio of a simple signal to a temporally discontinuous signal from a moving source.
[0155]
Therefore, if a temporally continuous signal from a fixed source is a leaky sound and a temporally discontinuous signal from a moving source is noise, the filtered cross spectrum C ′ (f) Is a function that reflects the noise-to-noise ratio of the leaked sound regardless of the absolute value of the cross spectrum C (f) before being filtered. Have.
[0156]
Also, if the frequency spectrum of the leaked sound is broadband compared to that of noise, the filtered cross spectrum C ′ (f) is broadened accordingly, reducing the side lobe of the cross-correlation function, This has the effect of improving the resolution for specifying the position of the abnormal part 2.
Therefore, the filter having the frequency characteristic of Expression (1) has the characteristic effect of the above-described SCOT filter.
[0157]
However, the SCOT filter has the following problems.
First, a temporally continuous signal from a source at a fixed position is a leaking sound, a signal from a moving source, or a signal that is discontinuous in time. Even if the characteristic signal is noise and this noise is dominant at a certain frequency component, for example, if the moving speed of the moving source is low, the absolute value C ′ (f) of the filtered cross spectrum is small. Since the leaked sound is at the same level as the absolute value of the dominant frequency component, the function does not sufficiently reflect the noise-to-noise ratio of the leaked sound.
[0158]
That is, the frequency component dominated by the noise cannot be sufficiently suppressed. In addition, the SCOT filter converts the absolute value of the filtered cross spectrum C ′ (f) into a temporally discontinuous signal from a moving source of a temporally continuous signal from a source at a fixed position. It is the same as the absolute value of coherency γ (f), which is a function that reflects the level ratio to the time, so that the time from a certain position, such as mechanical sound from a factory or engine sound of a stopped car, As for continuous noise, the factory and the stopped vehicle are far away, and the frequency component in which temporally continuous noise from such a fixed position is dominant is the cross spectrum C before filtering. In (f), even if the level is small enough to be ignored, an adverse effect of amplifying this frequency component to substantially the same level as the frequency component in which the leaking sound is dominant occurs.
[0159]
In order to solve the above-mentioned problems of the SCOT filter, the filter having the frequency characteristic of the formula (1) multiplies the frequency characteristic of the SCOT filter by further multiplying the absolute value of the coherency γ (f) to the real number i. Have the characteristics. Multiplying the frequency characteristic of the SCOT filter by the real value i of the absolute value of coherency γ (f) further emphasizes the frequency component having a large absolute value of coherency γ (f). There is an effect of extracting and further suppressing the frequency component having a small absolute value of the coherency γ (f).
[0160]
Therefore, in the filter having the frequency characteristic of the equation (1), a temporally continuous signal from a transmission source at a fixed position is a leakage sound, and a temporally discontinuous signal from a moving transmission source is noise. For example, even if the moving speed of the moving transmission source is low and the frequency component dominated by noise cannot be sufficiently suppressed by using only the SCOT filter, the noise is generated without impairing the effect of the SCOT filter. The dominant frequency component can be sufficiently suppressed.
[0161]
Note that a certain threshold value γ with respect to the absolute value of coherency γ (f) _l (F) is determined in advance, and C in formula (2) ₁₁ (F) = C _aa (F), C _{twenty two} (F) = C _bb As (f), a filter having the frequency characteristic of Expression (2) may be used. In the filter having the frequency characteristic of Expression (2), the absolute value of the coherency γ (f) is the threshold γ related to the coherency γ (f). _l (F) Since the frequency components below are completely removed, a temporally continuous signal from a fixed source is a leaking sound, and a temporally discontinuous signal from a moving source is noise. If so, even if the moving speed of the moving transmission source is low and the frequency component dominated by noise cannot be sufficiently suppressed by using only the SCOT filter, the noise is not impaired without impairing the effect of the SCOT filter. Can sufficiently suppress the dominant frequency component. Therefore, the same effect as that obtained when the filter having the frequency characteristic of the expression (1) is used can be obtained.
[0162]
Further, the cross spectrum C (f) is represented by C in the formula (3). ₁₁ (F) = C _aa (F), C _{twenty two} (F) = C _bb As (f), it may be filtered by a filter having the frequency characteristic of Expression (3) (step S8).
[0163]
Here, the effect of the filter having the frequency characteristic of Expression (3) will be described in detail.
Expression (3) is obtained by multiplying the frequency characteristic of the SCOT filter by the logarithm of the absolute value of the cross spectrum C (f).
Therefore, the filter having the frequency characteristic of Expression (3) has the effect of the above-described SCOT filter.
[0164]
Furthermore, by multiplying the logarithm of the absolute value of the cross spectrum C (f), the function according to the magnitude of the absolute value of the cross spectrum C (f), that is, the level of the received signal is multiplied. For example, in the frequency component in which the absolute value of the coherency γ (f) takes a large value, it is not a leakage sound but a time from a certain position such as a mechanical sound from a factory or a stationary engine sound of a car. Even if continuous noise is dominant, if the level is low, the level of the frequency component will not be amplified and passed to the same level as the level of the frequency component where the leaked sound is dominant. Furthermore, since the absolute value itself is multiplied by the logarithm of the absolute value, a broadband spectrum is obtained compared to the cross spectrum C (f), so the effect of the SCOT filter is not impaired and the side lobe is small. A correlation waveform can be obtained.
[0165]
In formula (4), C ₁₁ (F) = C _aa (F), C _{twenty two} (F) = C _bb As (f), a filter having the frequency characteristic of Expression (4) may be used. As a result, since the function according to the magnitude of the absolute value C (f) of the cross spectrum, that is, the level of the received signal is multiplied, for example, a value having a large absolute value of the coherency γ (f). Even if the continuous noise from a certain position such as mechanical sound from the factory or engine sound of a stopped car is dominant in the frequency component taking If it is low, the level of the frequency component will not be amplified and passed to the same level as the level of the frequency component in which the leaking sound is dominant.
[0166]
Further, since the absolute value of the cross spectrum C (f) is multiplied by the 1 / kth power of the absolute value, a broadband spectrum is obtained compared to the cross spectrum C (f). Thus, a correlation waveform having a small side lobe can be obtained. Therefore, the same effect as that obtained when the filter having the frequency characteristic of Expression (3) is used can be obtained.
In the above, the effect of the filter having the frequency characteristics of Expression (1), Expression (2), Expression (3), and Expression (4) has been described.
[0167]
Further, in the preprocessing unit 72a of the signal processing unit of the abnormal point detection device according to Embodiment 1 of the present invention, | γ (f) | ⁱ Frequency characteristics and log in equation (3) _j (| C (f) |) and | C (f) | in equation (4) ^{1 / k} Among these, a filter having a frequency characteristic obtained by weighting and multiplying one of the frequency characteristics and the frequency characteristic of the SCOT filter, or log in the expression (3) _j (| C (f) |) and | C (f) | in equation (4) ^{1 / k} In any one of the frequency characteristics, a frequency characteristic in which the frequency component that the absolute value of coherency γ (f) is equal to or smaller than a certain threshold value is 0 as shown in Expression (2), and the frequency of the SCOT filter A filter having a frequency characteristic obtained by weighting the characteristics and multiplying them may be used.
[0168]
By using such a filter, an effect having both the effect of the frequency characteristic filter of the formula (1) and the formula (2) and the effect of the frequency characteristic filter of the formula (3) and the formula (4). Is obtained. The weighting coefficient for the frequency characteristic may be determined from the same preliminary experiment as described above.
[0169]
In Embodiment 1 of the present invention, the leakage sound can be efficiently extracted even when the frequency band of the leakage sound is unknown by the filtering in the preprocessing unit 72a described above. A bandpass filter having characteristics may not be particularly used. Therefore, the bandpass filter having the frequency characteristic of Expression (5) may be removed from the preprocessing unit 72a. Thereby, there is an effect that the configuration of the preprocessing unit 72a can be simplified and the entire apparatus can be made inexpensive.
If the frequency band of the leaking sound is known, the filtered cross spectrum is further filtered by the bandpass filter of Expression (5). Thereby, noise having a frequency component outside the range of the frequency band of the leaked sound can be completely removed.
[0170]
The filtered cross spectrum C ′ (f) is sent to the correlation processing unit 72b, and a symbol F indicating inverse Fourier transform. ^-1 As a result of the inverse Fourier transform, the delay time with respect to the reception time at the ultrasonic sensor 3a is set to τ, and the cross-correlation function φ (τ) of the equation (12) is calculated (step S9).
The calculation result of the cross-correlation function φ (τ) calculated by the correlation processing unit 72b is output to the post-processing unit 72c.
[0171]
[Expression 60]

[0172]
Next, the operation of the post-processing unit 72c will be described.
In the post-processing unit 72c, the cross-correlation function φ (τ) is subjected to envelope detection (step S10).
The envelope is stored in the memory each time. At the same time, the number of times the envelope is stored in the memory is counted. This number of times is the same as the number of times that a control signal for capturing data is transmitted from the control unit 74 to the reception unit 71. If the number of times the envelope has been stored in the memory is smaller than the predetermined number M of data fetching repetitions, a signal requesting the control unit 74 to fetch data again from the post-processing unit 72c Send. Accordingly, the control unit 74 transmits a control signal for capturing data to the receiving unit 71.
[0173]
The above-described repetition is performed until the number of times that the envelope is stored in the memory becomes equal to a predetermined number M of data acquisition repetitions. If the number of times the envelope has been stored in the memory is equal to a predetermined number M of data fetching repetitions, this repetition is terminated (step S11).
[0174]
Next, after the envelopes stored in the memory are averaged, the peak value Aab of the averaged envelope and the ultrasonic sensor 3a when the averaged envelope peaks A delay time τab with respect to the reception time is obtained (step S12).
[0175]
The signal processing described above is performed with respect to the combination of the ultrasonic sensor 3a and the ultrasonic sensor 3b, but in parallel with being performed with respect to the combination of the ultrasonic sensors 3a and 3b, the ultrasonic sensor 3a and the ultrasonic sensor 3c. It is performed also about the combination of.
Thereby, also regarding the combination of the ultrasonic sensor 3a and the ultrasonic sensor 3c, when the peak value Aac of the averaged cross-correlation function φ (τ) and the averaged envelope become peaks The delay time τac with respect to the reception time at the ultrasonic sensor 3a is obtained.
[0176]
Next, the peak value Aab of the envelope of the averaged cross-correlation function between the ultrasonic sensor 3a and the ultrasonic sensor 3b and the averaged cross-correlation function between the ultrasonic sensor 3a and the ultrasonic sensor 3c It is determined whether one or both of the peak values Aac of the envelope are larger or smaller than a predetermined threshold value for determining the presence or absence of the abnormal part 2. Thereby, the presence or absence of the abnormal location 2 is determined (steps S13 and S14).
[0177]
The threshold value for determining the presence or absence of the abnormal part 2 may be a threshold value related to the peak value Aab of the envelope of the averaged cross-correlation function between the ultrasonic sensor 3a and the ultrasonic sensor 3b. It may be a threshold value regarding the peak value Aac of the envelope of the averaged cross-correlation function between the ultrasonic sensor 3a and the ultrasonic sensor 3c. The threshold value is averaged between the peak value Aab of the envelope of the averaged cross-correlation function between the ultrasonic sensor 3a and the ultrasonic sensor 3b, and between the ultrasonic sensor 3a and the ultrasonic sensor 3c. It may be a product with the peak value Aac of the envelope of the cross-correlation function or a threshold related to the sum.
[0178]
Further, the threshold value for determining the presence / absence of the abnormal part 2 may not be a threshold value related to the peak value of the envelope of the averaged cross-correlation function, but may be a threshold value related to the level of the received signal. It may be a threshold value related to the absolute value of the frequency spectrum or a threshold value related to the absolute value of the filtered cross spectrum C (f). In that case, the presence or absence of the abnormal part 2 is determined at the stage of the preprocessing unit 72.
[0179]
Moreover, the determination regarding the presence / absence of the abnormal part 2 may be made by any one of all the thresholds, or may be made by combining two or more pieces of information.
[0180]
The threshold value is determined by a preliminary experiment. This preliminary experiment is performed using the same or similar abnormal point detection apparatus as the abnormal point detection apparatus according to the first embodiment when the abnormal part 2 exists and when it does not substantially exist. The threshold data for determining the presence / absence of the abnormal part 2 is determined in advance from statistical data obtained from such preliminary experiments.
[0181]
Further, the threshold value may be used for classifying the shape and size of the abnormal location 2 as well as the determination of the presence or absence of the abnormal location 2. In this case, the preliminary experiment is the same as or similar to the abnormal point detection device according to the first embodiment for each case where the classification and the size of the abnormal point 2 are classified. Done.
[0182]
In addition, if the presence / absence of an abnormal part is determined by combining more information among the above-described information on the magnitude relationship, an effect of performing determination with higher accuracy can be obtained. In particular, if each determination result is weighted differently and is determined using the logic of weighted majority, when the above three determination results are disjoint determination results, a determination result with higher accuracy can be obtained. If the weighting coefficient used for the determination of the weighted majority is determined from the same preliminary experiment as described above, the accuracy of the determination can be further increased.
[0183]
When it is determined that there is an abnormal location 2, the distance between the ultrasonic sensors 3a and 3b, the distance between the ultrasonic sensors 3b and 3c, and the distance between the ultrasonic sensors 3a and 3b. Of the averaged cross-correlation function between the ultrasonic sensors 3a and 3c and the delay time τab with respect to the reception time at the ultrasonic sensor 3a when the envelope of the averaged cross-correlation function reaches a peak The distance x along the conduit 1 from the ultrasonic sensor 3a to the abnormal point 2 is specified by the equation (13) from the delay time τac with respect to the reception time at the ultrasonic sensor 3a when the envelope becomes a peak. The
[0184]
[Equation 61]

[0185]
If the propagation speed at which the leaking sound propagates through the conduit 1 is known, the space between the two ultrasonic sensors among the ultrasonic sensors 3a, 3b, and 3c is exposed on the ground surface. One of the two ultrasonic sensors that are known to have no abnormal part 2, such as an area where it can be visually determined that there is no abnormal part, for example, the ultrasonic sensor 3c in FIG. 1 is removed. In addition, the presence / absence of the abnormal part 2 and the position of the abnormal part 2 can be specified. For example, when the ultrasonic sensor 3c is removed, the distance x from the ultrasonic sensor 3a to the abnormal location 2 is given by Expression (14), where v is the propagation speed when the leaked sound propagates through the conduit 1. Thereby, there exists an effect that inspection can be facilitated and the inspection system can be made inexpensive.
[0186]
[62]

[0187]
The presence / absence of the abnormal location 2 thus obtained and the position of the specified abnormal location 2 are output to the display unit 73 (step S15).
The display unit 73 displays the presence / absence of the abnormal location 2 and the position of the identified abnormal location 2. Not only can this information be displayed, but if it is recorded and stored as a record of inspection results, it not only helps to build a database on the tendency of abnormalities to occur, but also during periodic inspections after a certain period of time. There is an effect that is useful as reference data.
[0188]
Further, as described above, information regarding the presence / absence of the abnormal part 2 is input to the display unit 73. This information is binary information. Therefore, in addition to the display unit, other informing means are provided so that this can be informed to the inspector in a form that reacts to the inspector's five senses, such as turning on / off the light and turning on / off the alarm sound in addition to the display. You may make it tell. Further, when the allowable time required for the inspection is large, the inspector may determine the presence / absence of the abnormal part 2 and the determination regarding the position of the abnormal part 2 by visually checking the averaged envelope. . In such a case, needless to say, the function of displaying the presence / absence of the abnormal location 2 or the specified water leakage location on the display unit 2 may be removed. As a result, it goes without saying that the effect of reducing the cost of the apparatus can be obtained.
[0189]
In the display unit 73, the averaged envelope, the peak value of the averaged envelope, the time when the averaged envelope peaks, and the reception obtained by the ultrasonic sensors 3a, 3b, and 3c Displays all or all of the signal waveform, frequency spectrum of the received signal, coherency obtained from the received signal, etc., or two or more of the above information. Further recording and storage not only increases the stability of the inspection, but also helps to build a database on the tendency of abnormalities to occur. Furthermore, there is an effect that is useful as reference data in a periodic inspection after a certain period. Further, when a second inspection is performed after a certain period of time, it can be used for confirming the reproducibility of inspection data and investigating changes with time.
[0190]
Next, experimental results for confirming the effect of the above-described method for specifying the position of the abnormal part 2 will be described with reference to FIGS.
FIG. 3 shows an experimental system of the above-mentioned experiment, where 8 is a fire hydrant, 9 is a recorder, and 10 is a computer. In addition, the conduit 1 is a water pipe here, and the region between the ultrasonic sensor 3b and the ultrasonic sensor 3c is an area where it can be visually determined that the conduit 1 is on the ground and there is no abnormal part 2. The separation distance L1 along the conduit 1 between the ultrasonic sensor 3a and the ultrasonic sensor 3b is 19.8 m, and the separation distance L2 along the conduit 1 between the ultrasonic sensor 3b and the ultrasonic sensor 3c is 10 m. Further, when the separation distance x along the conduit 1 from the ultrasonic sensor 3a to the abnormal part 2 was measured, this value was 9 m.
[0191]
In FIG. 3, three ultrasonic sensors 3 a, 3 b, and 3 c receive the leakage sound from the abnormal location 2 and record the reception signal on the recorder 9. The received signal received by the recorder 9 is taken into the computer 10 and the same processing as the processing in the preprocessing unit 72a, the correlation processing unit 72b, and the postprocessing unit 72c shown in FIG. Identified the location. Note that the frequency band of the leaked sound is unknown, the time length of the data acquired by one reception is 1 second, and the division number N for dividing the data acquired by one reception on the time axis is 12. The number M of repetitions of data loading is set to 10. Moreover, since this experiment was performed on the assumption that the abnormal part 2 exists, the threshold value for determining the presence or absence of the abnormal part was not set.
[0192]
4 to 34 show the received signal waveforms obtained by the ultrasonic sensors 3a, 3b, and 3c, the absolute value of the frequency spectrum of the received signal, the absolute value of the power spectrum of the received signal in the experimental system of FIG. The absolute value of the cross spectrum obtained from the received signals received by two of the three ultrasonic sensors, the absolute value of the filtered cross spectrum, the envelope of the cross-correlation function, and the averaged cross-correlation It is a figure which shows the envelope of a function.
[0193]
In the above experiment, the threshold value for determining the presence / absence of the abnormal location was not set because the abnormal location 2 was present, so the received signal waveform and the frequency of the received signal in FIGS. Absolute value of spectrum, absolute value of power spectrum of received signal, absolute value of cross spectrum obtained from received signals received by two of the three ultrasonic sensors, and absolute value of filtered cross spectrum The values of the envelope of the cross-correlation function and the averaged envelope of the cross-correlation function are not compared in magnitude relation with the threshold value for determining the presence / absence of an abnormal part.
[0194]
Therefore, from the received signal waveforms in FIGS. 4 to 34, the absolute value of the frequency spectrum of the received signal, the absolute value of the power spectrum of the received signal, and the received signals received by two of the three ultrasonic sensors. The calculated cross-spectrum absolute value, filtered cross-spectrum absolute value, cross-correlation function envelope, and averaged cross-correlation function envelope values should be set to appropriate values to make the figure easier to read. Normalized as 1 and shown as relative amplitude.
[0195]
FIG. 4 shows received signal waveforms received by the ultrasonic sensors 3a, 3b and 3c in one reception. 4A shows a reception signal waveform received by the ultrasonic sensor 3a, FIG. 4B shows a reception signal waveform received by the ultrasonic sensor 3b, and FIG. 4C shows a reception signal waveform received by the ultrasonic sensor 3c. The received signal waveform.
[0196]
FIG. 5 shows the frequency spectrum obtained with respect to the first data obtained by dividing the received signal received by the ultrasonic sensors 3a, 3b and 3c into N pieces on the time axis. FIG. 5A relates to a received signal received by the ultrasonic sensor 3a once, and FIG. 5B shows a received signal received by the ultrasonic sensor 3b once. FIG. 5C relates to a received signal received by the ultrasonic sensor 3c in a single reception.
[0197]
FIG. 6 shows the absolute value of the power spectrum obtained by using the frequency spectrum of the received signals received by the ultrasonic sensors 3a, 3b and 3c and the equations (6) to (8). 6A shows the absolute value of the power spectrum obtained from the received signal received by the ultrasonic sensor 3a, and FIG. 6B shows the absolute value of the power spectrum obtained from the received signal received by the ultrasonic sensor 3b. FIG. 6C shows the absolute value of the power spectrum obtained from the received signal received by the ultrasonic sensor 3c.
[0198]
FIG. 7 shows the absolute value of the cross spectrum for the combination of the ultrasonic sensor 3a and the ultrasonic wave 3b, and the combination of the ultrasonic sensors 3a and 3c. FIG. The absolute value of the cross spectrum obtained from the combination, FIG. 7B is the absolute value of the cross spectrum obtained from the combination of the ultrasonic sensor 3a and the ultrasonic sensor 3c.
[0199]
FIG. 8 shows the absolute value of the filtered cross spectrum obtained by filtering the cross spectrum using the filter having the frequency characteristic of Expression (1) as the filter in the preprocessing unit 72a. FIG. 8A relates to the combination of the ultrasonic sensor 3a and the ultrasonic sensor 3b, and FIG. 8B relates to the combination of the ultrasonic sensor 3a and the ultrasonic sensor 3c. The real number i in the formula (1) is 1.
[0200]
FIG. 9 shows a cross-correlation function obtained by inverse Fourier transform of the filtered cross spectrum of FIG. FIG. 9A relates to the combination of the ultrasonic sensor 3a and the ultrasonic sensor 3b, and FIG. 9B relates to the combination of the ultrasonic sensor 3a and the ultrasonic sensor 3c.
[0201]
An envelope curve obtained by envelope detection of the cross-correlation function of FIG. 9 is shown in FIG. FIG. 10A relates to the combination of the ultrasonic sensor 3a and the ultrasonic sensor 3b, and FIG. 10B relates to the combination of the ultrasonic sensor 3a and the ultrasonic sensor 3c.
[0202]
10A, the delay time τab with respect to the reception time at the ultrasonic sensor 3a when the envelope becomes a peak is 0.25 ms, and the ultrasonic wave when the envelope becomes a peak in FIG. 10B. The delay time τac with respect to the reception time at the sensor 3a is 6.54 ms. From these delay times, when the position of the abnormal part 2 in FIG. 3 is specified by the equation (13), the distance x along the conduit 1 from the ultrasonic sensor 3a of the abnormal part 2 is 9.70 m.
[0203]
FIG. 11 shows an averaged envelope obtained by repeating the process of obtaining the envelope in FIG. 10 by a predetermined number M of data acquisition repetitions and averaging them. FIG. 11A relates to the combination of the ultrasonic sensor 3a and the ultrasonic sensor 3b, and FIG. 11B relates to the combination of the ultrasonic sensor 3a and the ultrasonic sensor 3c.
[0204]
In FIG. 11A, the delay time τab with respect to the reception time at the ultrasonic sensor 3a when the averaged envelope becomes a peak is 0.54 ms, and in FIG. 11B, the averaged envelope is The delay time τac relative to the reception time at the ultrasonic sensor 3a at the time of the peak is 6.29 ms. From these delay times, when the position of the abnormal part 2 in FIG. 3 is specified by the equation (13), the distance x along the conduit 1 from the ultrasonic sensor 3a of the abnormal part 2 is 9.43 m.
[0205]
Thus, by filtering the cross spectrum using the filter having the frequency characteristic of Expression (1), the position of the abnormal part 2 could be specified with high accuracy. In addition, by repeating reception for a predetermined number of repetitions M and averaging the envelope, the effect of further improving the specific accuracy was seen.
[0206]
FIG. 12 shows the absolute value of the filtered cross spectrum obtained by filtering the cross spectrum using a filter having the frequency characteristic of Expression (2) as a filter in the preprocessing unit 72a. FIG. 12A relates to the combination of the ultrasonic sensor 3a and the ultrasonic sensor 3b, and FIG. 12B relates to the combination of the ultrasonic sensor 3a and the ultrasonic sensor 3c. Threshold value γ for the absolute value of coherency γ (f) in equation (2) _l (F) was set to 0.5.
[0207]
FIG. 13 shows a cross-correlation function obtained by inverse Fourier transform of the filtered cross spectrum of FIG. FIG. 13A relates to the combination of the ultrasonic sensor 3a and the ultrasonic sensor 3b, and FIG. 13B relates to the combination of the ultrasonic sensor 3a and the ultrasonic sensor 3c.
[0208]
FIG. 14 shows an envelope obtained by envelope detection of the cross-correlation function shown in FIG. FIG. 14A relates to the combination of the ultrasonic sensor 3a and the ultrasonic sensor 3b, and FIG. 14B relates to the combination of the ultrasonic sensor 3a and the ultrasonic sensor 3c.
[0209]
In FIG. 14A, the delay time τab with respect to the reception time at the ultrasonic sensor 3a when the envelope becomes a peak is 0.54 ms. In FIG. 14B, the ultrasonic sensor when the envelope becomes a peak. The delay time τac for the reception time at 3a is 6.17 ms. From these delay times, when the position of the abnormal part 2 in FIG. 3 is specified by Expression (13), the distance x along the conduit 1 from the ultrasonic sensor 3a of the abnormal part 2 is 9.42 m.
[0210]
FIG. 15 shows an averaged envelope obtained by repeating the process of obtaining the envelope in FIG. 14 by a predetermined number of repetitions M and averaging them. FIG. 15A relates to the combination of the ultrasonic sensor 3a and the ultrasonic sensor 3b, and FIG. 15B relates to the combination of the ultrasonic sensor 3a and the ultrasonic sensor 3c.
[0211]
In FIG. 15A, the delay time τab with respect to the reception time at the ultrasonic sensor 3a when the averaged envelope becomes a peak is 0.54 ms, and in FIG. 15B, the averaged envelope is The delay time τac with respect to the reception time at the ultrasonic sensor 3a at the time of the peak is 6.17 ms. From these delay times, when the position of the abnormal part 2 in FIG. 3 is specified by Expression (13), the distance x along the conduit 1 from the ultrasonic sensor 3a of the abnormal part 2 is 9.42 m.
[0212]
Thus, by filtering the cross spectrum using the filter having the frequency characteristic of Expression (2), the position of the abnormal part 2 could be specified with high accuracy.
[0213]
FIG. 16 shows the absolute value of the filtered cross spectrum obtained by filtering the cross spectrum using a filter having the frequency characteristic of Expression (3) as a filter in the preprocessing unit 72a. FIG. 16A relates to the combination of the ultrasonic sensor 3a and the ultrasonic sensor 3b, and FIG. 16B relates to the combination of the ultrasonic sensor 3a and the ultrasonic sensor 3c. The real number j in the equation (3) is set to e = 2.
[0214]
FIG. 17 shows a cross-correlation function obtained by inverse Fourier transform of the filtered cross spectrum of FIG. FIG. 17A relates to the combination of the ultrasonic sensor 3a and the ultrasonic sensor 3b, and FIG. 17B relates to the combination of the ultrasonic sensor 3a and the ultrasonic sensor 3c.
[0215]
FIG. 18 shows an envelope obtained by envelope detection of the cross-correlation function shown in FIG. FIG. 18A relates to the combination of the ultrasonic sensor 3a and the ultrasonic sensor 3b, and FIG. 18B relates to the combination of the ultrasonic sensor 3a and the ultrasonic sensor 3c.
[0216]
In FIG. 18A, the delay time τab with respect to the reception time at the ultrasonic sensor 3a when the envelope becomes a peak is 0.50 ms. In FIG. 18B, the ultrasonic sensor when the envelope becomes a peak. The delay time τac relative to the reception time at 3a is 6.63 ms. From these delay times, when the position of the abnormal part 2 in FIG. 3 is specified by the equation (13), the distance x along the conduit 1 from the ultrasonic sensor 3a of the abnormal part 2 is 9.49 m.
[0217]
FIG. 19 shows an averaged envelope obtained by repeating the process of obtaining the envelope in FIG. 18 by a predetermined number of repetitions M and averaging them. FIG. 19A relates to the combination of the ultrasonic sensor 3a and the ultrasonic sensor 3b, and FIG. 19B relates to the combination of the ultrasonic sensor 3a and the ultrasonic sensor 3c.
[0218]
In FIG. 19A, the delay time τab with respect to the reception time at the ultrasonic sensor 3a when the averaged envelope becomes a peak is 0.54 ms, and in FIG. 19B, the averaged envelope is The delay time τac with respect to the reception time at the ultrasonic sensor 3a when the peak is reached is 6.46 ms. From these delay times, when the position of the abnormal part 2 in FIG. 3 is specified by the equation (13), the distance x along the conduit 1 from the ultrasonic sensor 3a of the abnormal part 2 is 9.44 m.
[0219]
Thus, by filtering the cross spectrum using the filter having the frequency characteristic of Expression (3), the position of the abnormal part 2 could be specified with high accuracy. In addition, by repeating reception for a predetermined number of repetitions M and averaging the envelope, the effect of further improving the specific accuracy was seen.
[0220]
FIG. 20 shows the absolute value of the filtered cross spectrum obtained by filtering the cross spectrum using a filter having the frequency characteristic of Expression (4) as a filter in the preprocessing unit 72a. FIG. 20A relates to the combination of the ultrasonic sensor 3a and the ultrasonic sensor 3b, and FIG. 20B relates to the combination of the ultrasonic sensor 3a and the ultrasonic sensor 3c. The real number k in the formula (4) is 2.
[0221]
FIG. 21 shows a cross-correlation function obtained by inverse Fourier transform of the filtered cross spectrum of FIG. FIG. 21A relates to the combination of the ultrasonic sensor 3a and the ultrasonic sensor 3b, and FIG. 21B relates to the combination of the ultrasonic sensor 3a and the ultrasonic sensor 3c.
[0222]
An envelope curve obtained by envelope detection of the cross-correlation function of FIG. 21 is shown in FIG. FIG. 22A relates to the combination of the ultrasonic sensor 3a and the ultrasonic sensor 3b, and FIG. 22B relates to the combination of the ultrasonic sensor 3a and the ultrasonic sensor 3c.
[0223]
22A, the delay time τab with respect to the reception time at the ultrasonic sensor 3a when the envelope becomes a peak is 0.63 ms. In FIG. 22B, the ultrasonic sensor when the envelope becomes a peak. The delay time τac with respect to the reception time at 3a is 6.29 ms. From these delay times, when the position of the abnormal part 2 in FIG. 3 is specified by the equation (13), the distance x along the conduit 1 from the ultrasonic sensor 3a of the abnormal part 2 is 9.34 m.
[0224]
FIG. 23 shows an averaged envelope obtained by repeating the process of obtaining the envelope in FIG. 22 by a predetermined number of repetitions M and averaging them. FIG. 23 (a) relates to the combination of the ultrasonic sensor 3a and the ultrasonic sensor 3b, and FIG. 23 (b) relates to the combination of the ultrasonic sensor 3a and the ultrasonic sensor 3c.
[0225]
In FIG. 23A, the delay time τab with respect to the reception time at the ultrasonic sensor 3a when the averaged envelope becomes a peak is 0.71 ms, and in FIG. 23B, the averaged envelope is The delay time τac relative to the reception time at the ultrasonic sensor 3a at the time of the peak is 6.29 ms. From these delay times, when the position of the abnormal part 2 in FIG. 3 is specified by the equation (13), the distance x along the conduit 1 from the ultrasonic sensor 3a of the abnormal part 2 is 9.26 m.
[0226]
Thus, by filtering the cross spectrum using the filter having the frequency characteristic of Expression (4), the position of the abnormal part 2 could be specified with high accuracy. In addition, by repeating reception for a predetermined number of repetitions M and averaging the envelope, the effect of further improving the specific accuracy was seen.
[0227]
Next, the filter in the pre-processing unit 72a is | γ (f) | ⁱ Frequency characteristics and log in equation (3) _j (| C (f) |) and | C (f) | in equation (4) ^{1 / k} Among these, a filter having a frequency characteristic obtained by weighting and multiplying one of the frequency characteristics and the frequency characteristic of the SCOT filter, or log in the expression (3) _j (| C (f) |) and | C (f) | in equation (4) ^{1 / k} In any one of the frequency characteristics, a frequency characteristic in which the frequency component that the absolute value of coherency γ (f) is equal to or smaller than a certain threshold value is 0 as shown in Expression (2), and the frequency of the SCOT filter As an example of a filter having a frequency characteristic obtained by weighting and multiplying the characteristics, | γ (f) | in Expression (1) ⁱ Frequency characteristics and log in equation (3) _j (| C (f) |) and the frequency characteristic of the SCOT filter are weighted by 1: 1: 1 and multiplied by the frequency characteristic H of Expression (15). ₆ FIG. 24 shows the absolute value of the filtered cross spectrum obtained by filtering the cross spectrum using the filter having (f).
[0228]
FIG. 24A relates to the combination of the ultrasonic sensor 3a and the ultrasonic sensor 3b, and FIG. 24B relates to the combination of the ultrasonic sensor 3a and the ultrasonic sensor 3c. In equation (15), the real number i is 1, and the real number j is e = 2.71881818, which is the base of the natural logarithm.
[0229]
[Equation 63]

[0230]
FIG. 25 shows a cross-correlation function obtained by inverse Fourier transform of the filtered cross spectrum of FIG. FIG. 25A relates to the combination of the ultrasonic sensor 3a and the ultrasonic sensor 3b, and FIG. 25B relates to the combination of the ultrasonic sensor 3a and the ultrasonic sensor 3c.
[0231]
FIG. 26 shows an envelope obtained by envelope detection of the cross-correlation function of FIG. FIG. 26 (a) relates to the combination of the ultrasonic sensor 3a and the ultrasonic sensor 3b, and FIG. 26 (b) relates to the combination of the ultrasonic sensor 3a and the ultrasonic sensor 3c.
[0232]
In FIG. 26A, the delay time τab with respect to the reception time at the ultrasonic sensor 3a when the envelope becomes a peak is 0.54 ms. In FIG. 26B, the ultrasonic sensor when the envelope becomes a peak. The delay time τac with respect to the reception time at 3a is 6.42 ms. From these delay times, when the position of the abnormal part 2 in FIG. 3 is specified by the equation (13), the distance x along the conduit 1 from the ultrasonic sensor 3a of the abnormal part 2 is 9.44 m.
[0233]
FIG. 27 shows an averaged envelope obtained by repeating the process of obtaining the envelope in FIG. 26 by a predetermined number of repetitions M and averaging them. FIG. 27A relates to the combination of the ultrasonic sensor 3a and the ultrasonic sensor 3b, and FIG. 27B relates to the combination of the ultrasonic sensor 3a and the ultrasonic sensor 3c.
[0234]
In FIG. 27A, the delay time τab with respect to the reception time at the ultrasonic sensor 3a when the averaged envelope becomes a peak is 0.58 ms, and in FIG. 27B, the averaged envelope is The delay time τac with respect to the reception time at the ultrasonic sensor 3a when the peak is reached is 6.25 ms. From these delay times, when the position of the abnormal part 2 in FIG. 3 is specified by the equation (13), the distance x along the conduit 1 from the ultrasonic sensor 3a of the abnormal part 2 is 9.39 m.
[0235]
Thus, | γ (f) | in equation (1) ⁱ Frequency characteristics and log in equation (3) _j (| C (f) |) and | C (f) | in equation (4) ^{1 / k} Among these, a filter having a frequency characteristic obtained by weighting and multiplying one of the frequency characteristics and the frequency characteristic of the SCOT filter, or log in the expression (3) _j (| C (f) |) and | C (f) | in equation (4) ^{1 / k} In any one of the frequency characteristics, the frequency characteristic in which the frequency component that the absolute value of the coherency γ (f) is equal to or smaller than a certain threshold value is 0 as shown in Expression (2), and the frequency of the SCOT filter As an example of a filter having a frequency characteristic obtained by weighting and multiplying the characteristics, | γ (f) | in Expression (1) ⁱ Frequency characteristics and log in equation (3) _j (| C (f) |) and the frequency characteristic of the SCOT filter are weighted by 1: 1: 1 and multiplied by the frequency characteristic H of Expression (15). ₆ By filtering the cross spectrum using the filter having (f), the position of the abnormal part 2 could be specified with high accuracy. In addition, by repeating reception for a predetermined number of repetitions M and averaging the envelope, the effect of further improving the specific accuracy was seen.
[0236]
For comparison, the result when the preprocessing unit 72a performs processing without using a filter for filtering the cross spectrum is shown below.
FIG. 28 shows a cross-correlation function obtained by performing inverse Fourier transform on the cross spectrum without using a filter for filtering the cross spectrum in the preprocessing unit 72a. FIG. 28A relates to the combination of the ultrasonic sensor 3a and the ultrasonic sensor 3b, and FIG. 28B relates to the combination of the ultrasonic sensor 3a and the ultrasonic sensor 3c.
[0237]
FIG. 29 shows an envelope obtained by envelope detection of the cross-correlation function of FIG. FIG. 29A relates to the combination of the ultrasonic sensor 3a and the ultrasonic sensor 3b, and FIG. 29B relates to the combination of the ultrasonic sensor 3a and the ultrasonic sensor 3c.
[0238]
In FIG. 29A, the delay time τab with respect to the reception time at the ultrasonic sensor 3a when the envelope becomes a peak is 4.42 ms, and the ultrasonic wave when the envelope becomes a peak in FIG. 29B. The delay time τac with respect to the reception time at the sensor 3a is 6.29 ms. From these delay times, when the position of the abnormal part 2 in FIG. 3 is specified by the equation (13), the distance x along the conduit 1 from the ultrasonic sensor 3a of the abnormal part 2 is −1.91 m.
[0239]
FIG. 30 shows an averaged envelope obtained by repeating the process of obtaining the envelope of FIG. 29 by a predetermined number of repetitions M and averaging them. FIG. 30A relates to the combination of the ultrasonic sensor 3a and the ultrasonic sensor 3b, and FIG. 30B relates to the combination of the ultrasonic sensor 3a and the ultrasonic sensor 3c.
[0240]
30A, the delay time τab with respect to the reception time at the ultrasonic sensor 3a when the averaged envelope becomes a peak is 4.42 ms. In FIG. 30B, the averaged envelope is obtained. The delay time τac with respect to the reception time at the ultrasonic sensor 3a when becomes a peak is 6.21 ms. From these delay times, when the position of the abnormal part 2 in FIG. 3 is specified by the equation (13), the distance x along the conduit 1 from the ultrasonic sensor 3a of the abnormal part 2 is −2.46 m.
[0241]
As described above, in the preprocessing unit 72a, when the processing is performed without using the filter for filtering the cross spectrum, the envelope is not averaged, and the position of the abnormal portion 2 is accurately determined both when the processing is performed. I couldn't identify well.
[0242]
Furthermore, for comparison, the results when processing is performed using only the SCOT filter as the filter in the preprocessing unit 72a are shown below.
FIG. 31 shows the absolute value of the filtered cross spectrum obtained by filtering the cross spectrum using only the SCOT filter as the filter in the preprocessing unit 72a. FIG. 31A relates to the combination of the ultrasonic sensor 3a and the ultrasonic sensor 3b, and FIG. 31B relates to the combination of the ultrasonic sensor 3a and the ultrasonic sensor 3c.
[0243]
FIG. 32 shows a cross-correlation function obtained by inverse Fourier transform of the cross spectrum of FIG. FIG. 32A relates to the combination of the ultrasonic sensor 3a and the ultrasonic sensor 3b, and FIG. 32B relates to the combination of the ultrasonic sensor 3a and the ultrasonic sensor 3c.
[0244]
FIG. 33 shows an envelope obtained by envelope detection of the cross-correlation function of FIG. FIG. 33A relates to the combination of the ultrasonic sensor 3a and the ultrasonic sensor 3b, and FIG. 33B relates to the combination of the ultrasonic sensor 3a and the ultrasonic sensor 3c.
[0245]
In FIG. 33A, the delay time τab with respect to the reception time at the ultrasonic sensor 3a when the envelope becomes a peak is 0.63 ms, and in FIG. 33B, the ultrasonic sensor when the envelope becomes a peak. The delay time τac relative to the reception time at 3a is −0.04 ms. From these delay times, when the position of the abnormal part 2 in FIG. 3 is specified by the equation (13), the distance x along the conduit 1 from the ultrasonic sensor 3a of the abnormal part 2 is 14.6 m.
[0246]
FIG. 34 shows an averaged envelope obtained by repeating the process of obtaining the envelope in FIG. 33 by a predetermined number of repetitions M and averaging them. FIG. 34 (a) relates to the combination of the ultrasonic sensor 3a and the ultrasonic sensor 3b, and FIG. 34 (b) relates to the combination of the ultrasonic sensor 3a and the ultrasonic sensor 3c.
[0247]
In FIG. 34A, the delay time τab with respect to the reception time at the ultrasonic sensor 3a when the averaged envelope becomes a peak is 0.50 ms, and in FIG. 34B, the averaged envelope is The delay time τac with respect to the reception time at the ultrasonic sensor 3a when the peak is reached is −0.08 ms. From these delay times, when the position of the abnormal location 2 in FIG. 3 is specified by the equation (13), the distance x along the conduit 1 from the ultrasonic sensor 3a of the abnormal location 2 is 14.21 m.
[0248]
As described above, when processing is performed using only the SCOT filter as the filter in the pre-processing unit 72a, it is possible to accurately identify the position of the abnormal part 2 even when the envelope is not averaged. could not.
[0249]
From the above experimental results, it was experimentally confirmed that the abnormal part of the conduit can be identified with higher accuracy than before by the signal processing in the preprocessing unit 72a of the signal processing unit 72 in the first embodiment of the present invention. Furthermore, it has been experimentally confirmed that the abnormal part of the conduit can be identified more stably than in the past by the signal processing in the post-processing unit 72c.
[0250]
Further, from the experimental results described above, it is experimentally shown that the position of the abnormal location 2 can be accurately identified even when the frequency band of the leaked sound is unknown by the signal processing method according to the first embodiment of the present invention. Was also confirmed.
[0251]
In Embodiment 1 of the present invention, unlike the prior art, a filter having a frequency characteristic of Expression (1), a filter having a frequency characteristic of Expression (2), and Expression (3) are pre-processed for correlation processing. A filter having a frequency characteristic of (4), a filter having a frequency characteristic of Equation (4), and | γ (f) | ⁱ Frequency characteristics and log in equation (3) _j (| C (f) |) and | C (f) | in equation (4) ^{1 / k} Among these, a filter having a frequency characteristic obtained by weighting and multiplying one of the frequency characteristics and the frequency characteristic of the SCOT filter, or log in the expression (3) _j (| C (f) |) and | C (f) | in equation (4) ^{1 / k} In any one of the frequency characteristics, a frequency characteristic in which the frequency component at which coherency γ (f) is equal to or less than a certain threshold is 0 as shown in Equation (2), and the frequency characteristic of the SCOT filter are: This is a case where the frequency band of the leaked sound is unknown because one of the filters having frequency characteristics obtained by weighting and multiplying these, or a filter having two or more frequency characteristics is included. Also, the position of the abnormal part 2 can be specified with high accuracy.
[0252]
In addition, if the temporally continuous signal from the transmission source at a certain position is a leaking sound and the temporally discontinuous signal from the moving transmission source is noise, the moving speed of the moving transmission source is small, Even with the SCOT filter alone, it is possible to accurately identify the position of the leakage location even in the case where the frequency component dominated by noise cannot be sufficiently suppressed.
[0253]
In addition, a continuous signal from a transmission source at a fixed position is not only a leaking sound but also a noise such as a mechanical sound from a factory and a car engine sound from a stopped vehicle. With respect to the frequency components which are, it is possible to specify the leak location with high accuracy.
[0254]
Furthermore, the problem of leaked sound and noise being random and non-periodic signals can be overcome by further averaging the cross-correlation function envelope from multiple repeated measurements. The position of the abnormal part 2 can be identified stably.
[0255]
Embodiment 2. FIG.
Next, an abnormal point detection apparatus according to Embodiment 2 of the present invention will be described.
The abnormal point detection apparatus according to the second embodiment has the same configuration as that of the second embodiment shown in FIG. 1, but the preprocessing unit 72a in the signal processing unit 72 has a frequency represented by the equation (1). Characteristic H ₁ A filter having (f) and a frequency characteristic H expressed by the equation (2) ₂ A filter having (f) and a frequency characteristic H expressed by the equation (3) _Three The filter having (f) and the frequency characteristic H expressed by the equation (4) _Four In addition to the filter having (f), any one or two or more of filters having frequency characteristics to be described later are included.
[0256]
That is, the pre-processing unit 72a in the signal processing unit 72 has a frequency characteristic H ₁ A filter having (f) and a frequency characteristic H ₂ A filter having (f) and a frequency characteristic H _Three A filter having (f) and a frequency characteristic H _Four A filter having (f) and σ _abs (F) is the dispersion between a plurality of received signals obtained by dividing the ratio of the absolute value of the frequency spectrum of two received signals received by a certain two of the three ultrasonic sensors 3a, 3b and 3c. , Σ _φ (F) is the variance between a plurality of received signals obtained by dividing the phase difference of the frequency spectrum of the two received signals received by the two sensors.
[0257]
[Expression 64]

[0258]
The frequency characteristic H expressed by ₇ A filter having (f);
Among the three ultrasonic sensors 3a, 3b and 3c, the variance σ between a plurality of received signals divided by the ratio of the absolute values of the amplitudes of the frequency spectra of two received signals received by two sensors. _abs (F) and variance σ between a plurality of received signals obtained by dividing the phase difference of the frequency spectrum of the received signals received by the two sensors. _φ The threshold for the product with (f) is σ _l (F)
[0259]
[Equation 65]

[0260]
The frequency characteristic H expressed by ₈ A filter having (f);
| Γ (f) | in Equation (1) ⁱ , And {1 / (σ in equation (16) _abs (F) · σ _φ (F))} ⁱ Any one of the frequency characteristics and log in equation (3) _j (| C (f) |) and | C (f) | in Equation (4) ^{1 / k} Among these, a filter having a frequency characteristic obtained by weighting and multiplying one of the frequency characteristics and the frequency characteristic of the SCOT filter, or log in the expression (3) _j (| C (f) |) and | C (f) | in equation (4) ^{1 / k} In any one of the frequency characteristics, a frequency characteristic in which the frequency component at which the absolute value of the coherency γ (f) is equal to or less than a predetermined threshold is set to 0 as shown in Expression (2), or As shown in the equation (17), among the above three ultrasonic sensors, the variance σ between a plurality of received signals obtained by dividing the ratio of the absolute values of the frequency spectra of two received signals received by two sensors. _abs (F) and the variance σ between a plurality of received signals obtained by dividing the phase difference of the frequency spectrum of the two received signals received by the two sensors. _φ Of the frequency characteristics in which the frequency component with which the product of (f) is equal to or greater than a predetermined threshold is 0 and the frequency characteristics of the SCOT filter are weighted and multiplied by these, , Any one or two or more.
[0261]
Further, the frequency characteristic H represented by the equation (5) is set such that the lower limit frequency is fs and the upper limit frequency is fe. _Five A bandpass filter with (f) is included.
[0262]
In the filter characteristics included in the preprocessing unit 72a, the lower limit frequency fs and the upper limit frequency fe in the equation (5) are determined according to the result of the preliminary experiment, and these values are input to the control unit 74. Further, a real number i in the equation (1) and a threshold value γ relating to the absolute value of coherency in the equation (2) _l (F), real number j in equation (3), real number k in equation (4), and frequency spectra of two received signals received by two of the three ultrasonic sensors in equation (17). Variance σ between multiple received signals divided by the ratio of absolute amplitude values _abs (F) and variance σ between a plurality of received signals obtained by dividing the phase difference of the frequency spectrum of the received signals received by the two sensors. _φ The threshold for the product with (f) is σ _l (F) is also determined according to the result of a similar preliminary experiment.
[0263]
This preliminary experiment is performed using the same or similar abnormal point detection device according to the second embodiment when the abnormal point 2 exists and when it does not substantially exist. . From the statistical data obtained from such a preliminary experiment, the real number i in the equation (1) that can detect the presence / absence of the abnormal part 2 and the position of the abnormal part 2 with the highest accuracy according to the state of the examination, Threshold value γ for the absolute value of coherency in equation (2) _l (F), real number j in equation (3), real number k in equation (4), and frequency spectra of two received signals received by two of the three ultrasonic sensors in equation (17). Dispersion σ between multiple received signals divided by the ratio of absolute amplitude values _abs (F) and variance σ between a plurality of received signals obtained by dividing the phase difference of the frequency spectrum of the received signals received by the two sensors. _φ The threshold for the product with (f) is σ _l (F) is also determined according to the result of a similar preliminary experiment.
[0264]
Similarly to the second embodiment, if the propagation speed when leaking sound propagates through the conduit 1 is known, the space between the two ultrasonic sensors 3a, 3b, and 3c is between the two ultrasonic sensors. One of the two ultrasonic sensors that are known to have no abnormal part 2 such as an area that can be visually determined that there is no abnormal part between them, for example, in FIG. Even if the ultrasonic sensor 3c is removed, the presence / absence of the abnormal part 2 and the position of the abnormal part 2 can be specified. In such a case, removing one of the ultrasonic sensors 3a, 3b, and 3c has an effect of facilitating the inspection and reducing the cost of the inspection system.
[0265]
Further, similarly to the second embodiment, the abnormal point detection apparatus according to the second embodiment operates according to the flowchart shown in FIG. Detailed description thereof is omitted.
Further, in the preprocessing unit 72a of the signal processing unit of the abnormal point detection device according to Embodiment 2 of the present invention, | γ (f) | ⁱ Frequency characteristics and log in equation (3) _j (| C (f) |) and | C (f) | in equation (4) ^{1 / k} Among these, a filter having a frequency characteristic obtained by weighting and multiplying one of the frequency characteristics and the frequency characteristic of the SCOT filter, or log in the expression (3) _j (| C (f) |) and | C (f) | in equation (4) ^{1 / k} In any one of the frequency characteristics, a frequency characteristic in which the frequency component that the absolute value of coherency γ (f) is equal to or smaller than a certain threshold value is 0 as shown in Expression (2), and the frequency of the SCOT filter A filter having a frequency characteristic obtained by weighting the characteristics and multiplying them may be used.
[0266]
By using such a filter, an effect having both the effect of the frequency characteristic filter of the formula (1) and the formula (2) and the effect of the frequency characteristic filter of the formula (3) and the formula (4). Is obtained. The weighting coefficient for the frequency characteristic may be determined from the same preliminary experiment as described above.
[0267]
Next, the variance between the divided N data of the ratio An / Bn of the absolute value of the frequency spectrum of the received signal received by the two ultrasonic sensors at a certain frequency f is expressed as σ. _abs (F) The variance relating to the data is divided into N divided by the phase difference φan−φbn of the frequency spectrum of the received signals received by the two ultrasonic sensors. _φ (F)
[0268]
[Equation 66]

[0269]
A function σ ′ (f) having a frequency characteristic represented by This function σ ′ (f) indicates that the amplitude ratio and the phase difference in the received signals received by the two ultrasonic sensors have a small variance among the data divided into N pieces, that is, there are temporal variations. The smaller frequency component is a function having a larger value. Therefore, the frequency characteristic of the equation (18) is similar to the leakage sound, the mechanical sound of the factory, and the engine sound of the stopped car in each frequency component of the frequency spectrum of the reception signal received by the two ultrasonic sensors. In addition, a signal that is continuous in time from a source at a fixed position and a signal from a moving source such as noise from driving sounds of automobiles, water usage sounds, pipe running water, etc. It is a function that reflects a level ratio with a signal that is at least one of the characteristics of a typical signal. That is, the physical meaning of the function σ ′ (f) in Expression (18) is the same as the physical meaning of the absolute value of coherency | γ (f) |.
[0270]
Therefore, the cross spectrum C (f) is filtered by a filter having the frequency characteristic of Expression (16) by substituting the right side of Expression (18) into the absolute value | γ (f) | of coherency in Expression (1). Even in this case, the same effect as that obtained when filtering is performed using a filter having the frequency characteristic of Expression (1) can be obtained.
[0271]
Further, the cross spectrum C (f) is filtered by a filter having the frequency characteristic of Expression (17) by substituting the right side of Expression (18) into the absolute value | γ (f) | of coherency in Expression (2). Even in this case, the same effect as that obtained when filtering is performed using a filter having the frequency characteristic of Expression (2) can be obtained.
[0272]
Furthermore, in the pre-processing unit 72a of the signal processing unit of the abnormal point detection apparatus according to Embodiment 2 of the present invention, {1 / (σ _abs (F) · σ _φ (F))} ⁱ Frequency characteristics and log in equation (3) _j (| C (f) |) and | C (f) | in equation (4) ^{1 / k} Among these, a filter having a frequency characteristic obtained by weighting and multiplying one of the frequency characteristics and the frequency characteristic of the SCOT filter, or log in the expression (3) _j (| C (f) |) and | C (f) | in equation (4) ^{1 / k} In any one of the frequency characteristics, the variance σ between a plurality of received signals divided by the ratio of the absolute values of the frequency spectra of the two received signals received by the two sensors as shown in Expression (17) _abs (F) and variance σ between a plurality of received signals obtained by dividing the phase difference of the frequency spectrum of two received signals received by two sensors _φ Threshold value for product with (f) _l A filter having a frequency characteristic in which (f) is weighted with a frequency characteristic such that a frequency component equal to or greater than a certain threshold is 0 and a frequency characteristic of the SCOT filter are multiplied by these may be used.
[0273]
By using such a filter, the effect which has the effect which the filter of the frequency characteristic of Formula (16) and Formula (17) has, and the effect which the filter of the frequency characteristic of Formula (3) and Formula (4) has Is obtained. The weighting coefficient for the frequency characteristic may be determined from the same preliminary experiment as described above.
[0274]
In the second embodiment of the present invention, the leakage sound can be extracted efficiently even when the frequency band of the leakage sound is unknown by the filtering in the preprocessing unit 72a described above. A bandpass filter having characteristics may not be particularly used. Therefore, the bandpass filter having the frequency characteristic of Expression (5) may be removed from the preprocessing unit 72a. Thereby, there is an effect that the configuration of the preprocessing unit 72a can be simplified and the entire apparatus can be made inexpensive.
If the frequency band of the leaking sound is known, the filtered cross spectrum is further filtered by the bandpass filter of Expression (5). Thereby, noise having a frequency component outside the range of the frequency band of the leaked sound can be completely removed.
[0275]
The experimental results for confirming the effect of the method for specifying the position of the abnormal part 2 in the second embodiment are shown in FIGS. 3 to 34 as in the first embodiment.
In the second embodiment, the filter in the pre-processing unit includes | γ (f) | ⁱ , And {1 / (σ in equation (16) _abs (F) · σ _φ (F))} ⁱ Any one of the frequency characteristics and log in equation (3) _j (| C (f) |) and | C (f) | in Equation (4) ^{1 / k} Among these, a filter having a frequency characteristic obtained by weighting and multiplying one of the frequency characteristics and the frequency characteristic of the SCOT filter, or log in the expression (3) _j (| C (f) |) and | C (f) | in equation (4) ^{1 / k} In any one of the frequency characteristics, a frequency characteristic in which the frequency component at which the absolute value of the coherency γ (f) is equal to or less than a predetermined threshold is set to 0, as in Expression (2), or As shown in the equation (17), among the above three ultrasonic sensors, the variance σ between a plurality of received signals obtained by dividing the ratio of the absolute values of the frequency spectra of two received signals received by two sensors. _abs (F) and the variance σ between a plurality of received signals obtained by dividing the phase difference of the frequency spectrum of the two received signals received by the two sensors. _φ As an example of a filter having a frequency characteristic in which a frequency characteristic in which the product of (f) is equal to or greater than a predetermined threshold is 0 and the frequency characteristic of the SCOT filter are weighted and multiplied, | Γ (f) | in equation (1) ⁱ Frequency characteristics and log in equation (3) _j (| C (f) |) and the frequency characteristic of the SCOT filter are weighted by 1: 1: 1 and multiplied by the frequency characteristic H of Expression (15). ₆ The absolute value of the filtered cross spectrum obtained by filtering the cross spectrum using the filter having (f) can be the same as that of the first embodiment shown in FIG. Can be specified with high accuracy. In addition, by repeating reception for a predetermined number of repetitions M and averaging the envelope, the effect of further improving the specific accuracy was seen.
[0276]
As described above, in the second embodiment of the present invention, unlike the conventional case, the pre-processing of the correlation processing is performed.
A filter having the frequency characteristic of equation (1);
A filter having the frequency characteristic of equation (2);
A filter having the frequency characteristic of Equation (3);
A filter having the frequency characteristic of equation (4);
A filter having the frequency characteristic of equation (16);
A filter having the frequency characteristic of Expression (17);
| Γ (f) | in Equation (1) ⁱ And {1 / (σ in equation (16) _abs (F) · σ _φ (F))} ⁱ Filter having any one frequency characteristic, and log in Expression (3) _j (| C (f) |) and | C (f) | in equation (4) ^{1 / k} Among these, a filter having a frequency characteristic obtained by weighting and multiplying one of the frequency characteristics and the frequency characteristic of the SCOT filter, or log in the expression (3) _j (| C (f) |) and | C (f) | in equation (4) ^{1 / k} In any one of the frequency characteristics, the frequency component at which the absolute value of the coherency γ (f) is equal to or less than a predetermined threshold is set to 0 as shown in the equation (2) or the equation (17). Dispersion σ between a plurality of divided received signals of the frequency characteristics or the ratio of the absolute values of the amplitudes of the frequency spectra of the received signals received by the above two sensors _abs (F) and variance σ between a plurality of received signals obtained by dividing the phase difference of the frequency spectrum of the received signals received by the two sensors. _φ Of the filters having a frequency characteristic in which the frequency characteristic that the product of (f) is equal to or greater than a predetermined threshold value is set to 0 and the frequency characteristic of the SCOT filter are weighted and multiplied. Therefore, even if the frequency band of the leaked sound is unknown, the position of the abnormal part 2 can be accurately identified.
[0277]
In addition, if the temporally continuous signal from the transmission source at a certain position is a leaking sound and the temporally discontinuous signal from the moving transmission source is noise, the moving speed of the moving transmission source is small, Even with the SCOT filter alone, it is possible to accurately identify the position of the leakage location even in the case where the frequency component dominated by noise cannot be sufficiently suppressed.
[0278]
In addition, a continuous signal from a transmission source at a fixed position is not only a leaking sound but also a noise such as a mechanical sound from a factory and a car engine sound from a stopped vehicle. With respect to the frequency components which are, it is possible to specify the leak location with high accuracy.
[0279]
Furthermore, the problem of leaked sound and noise being random and non-periodic signals can be overcome by further averaging the cross-correlation function envelope from multiple repeated measurements. And the position of the abnormal part 2 can be identified stably.
[0280]
The above-described abnormal part detection device according to the second embodiment can adopt the same embodiment as that of the first embodiment, and is summarized as follows.
1. Provided with three ultrasonic sensors for receiving leaked sound generated due to the presence of an abnormal location in the tube to be inspected, and receiving signals of leaked sound from these ultrasonic sensors via the receiving unit, respectively In the signal processing unit to be processed, f is the frequency of the received signal, i is a real number greater than or equal to 0, and the coherency obtained from the received signals received by two of the three ultrasonic sensors is γ (f) The power spectrum obtained from the received signal received by one of the two sensors is C ₁₁ (F) The power spectrum obtained from the received signal received by the other of the two ultrasonic sensors is C _{twenty two} As (f), the frequency characteristic H expressed by the equation (1) ₁ (F), a filter A for filtering the cross spectrum obtained from the received signal of the leaked sound, and a predetermined threshold value regarding the absolute value of the coherency γ (f) as γ ₁ As (f), the frequency characteristic H represented by the equation (2) ₂ The filter B having (f), j being a real number greater than or equal to 0, and among the above three ultrasonic sensors, the cross spectrum obtained from the received signals received by two sensors is represented by C (f). The frequency characteristic H expressed by _Three The filter C having (f) and the frequency characteristic H expressed by the equation (4) where k is a real number of 1 or more. _Four A filter D having (f) and σ _abs (F) is the variance between a plurality of received signals obtained by dividing the ratio of the absolute value of the frequency spectrum of two received signals received by two of the three ultrasonic sensors, and σ _φ (F) is the frequency characteristic H expressed by the equation (16) as the variance between the plurality of received signals obtained by dividing the phase difference of the frequency spectrum of the two received signals received by the two sensors. ₇ The variance E between the filter E having (f) and a plurality of received signals obtained by dividing the ratio of the absolute values of the frequency spectra of two received signals received by two of the three ultrasonic sensors. _abs (F) and the variance σ between a plurality of received signals obtained by dividing the phase difference of the frequency spectrum of the two received signals received by the two sensors. _φ The threshold for the product with (f) is σ _l (F) and the frequency characteristic H expressed by the equation (17) ₈ Filter F having (f) and at least log of the above six expressions _j (| C (f) |) and | C (f) | ^{1 / k} Any one or two or more of the frequency characteristics of the SCOT (Smoothed Coherence Tr Ansform) filter and the filter G having the frequency characteristics obtained by multiplying the weighted frequency characteristics. And a correlation processing unit that calculates a cross-correlation function from the signal preprocessed by the preprocessing unit.
[0281]
2. The filter G is | γ (f) | ⁱ And {1 / (σ _abs (F) · σ _φ (F))} ⁱ Any one of the frequency characteristics and log _j (| C (f) |) and | C (f) | ^{1 / k} A filter having a frequency characteristic obtained by weighting any one of the frequency characteristics and the frequency characteristic of the SCOT filter and multiplying these weighted frequency characteristics, or log _j (| C (f) |) and | C (f) | ^{1 / k} In any one of the frequency characteristics, a frequency characteristic in which the frequency component at which the absolute value of the coherency γ (f) is equal to or less than a predetermined threshold is 0, or the above three ultrasonic sensors Among the received signals divided by the ratio of the absolute value of the frequency spectrum of the two received signals received by two sensors. _abs (F) and the variance σ between a plurality of received signals obtained by dividing the phase difference of the frequency spectrum of the two received signals received by the two sensors. _φ Threshold value σ related to product with (f) _l (F) A frequency characteristic in which a frequency component that exceeds a certain threshold is set to 0 and a frequency characteristic of the SCOT filter are each weighted, and any of the filters having a frequency characteristic obtained by multiplying these weighted frequency characteristics. is there.
[0282]
3. The pre-processing unit has a frequency characteristic H expressed by equation (5), where fs is the lower limit frequency and fe is the upper limit frequency. _Five It further includes a bandpass filter having (f).
[0283]
4). The pre-processing unit is configured to determine a predetermined threshold relating to the absolute values of the real numbers i, j, k and the coherency γ (f) based on statistical data obtained from a preliminary experiment in the presence or absence of an abnormal location. γ _l (F) Dispersion σ between a plurality of received signals obtained by dividing the ratio of the absolute values of the frequency spectra of two received signals received by two of the three ultrasonic sensors. _abs (F) and the variance σ between a plurality of received signals obtained by dividing the phase difference of the frequency spectrum of the two received signals received by the two sensors. _φ Threshold value for product with (f) _l Determine (f).
[0284]
5. The signal processing unit further includes a post-processing unit that obtains an envelope of the cross-correlation function calculated by the correlation processing unit, and further averages the envelope of the cross-correlation function obtained by a predetermined number of repetitions. Prepare.
[0285]
6). The post-processing unit determines the presence / absence of an abnormal portion from the magnitude relationship between the peak value of the averaged envelope of the cross-correlation function and a predetermined threshold value.
[0286]
7). The post-processing unit includes a time at which an envelope of the averaged cross-correlation function for the received signal received by two of the three ultrasonic sensors reaches a peak, and three ultrasonic sensors. And the time when the envelope of the averaged cross-correlation function peaks for the received signals received by one of the two ultrasonic sensors and the ultrasonic sensor other than the two ultrasonic sensors. The position of the abnormal part is specified from the separation distances of the three ultrasonic sensors.
[0287]
8). The said post-processing part determines the threshold value regarding the peak value of the envelope of the said averaged cross correlation function based on the statistical data obtained from the preliminary experiment in the case where an abnormal location exists and it does not exist.
[0288]
9. Provided with two ultrasonic sensors for receiving leaked sound generated due to the presence of an abnormal location in the tube to be inspected, and receiving the leaked sound received by these ultrasonic sensors via the receiving unit, respectively. In the signal processing unit, f is the frequency of the received signal, i is a real number greater than or equal to 0, the coherency obtained from the received signals received by the two ultrasonic sensors is γ (f), The power spectrum obtained from the received signal received by one ultrasonic sensor is C ₁₁ (F), and the power spectrum obtained from the received signal received by the other of the two ultrasonic sensors is C _{twenty two} As (f), the frequency characteristic H expressed by the equation (1) ₁ (F), a filter A for filtering the cross spectrum obtained from the received signal of the leaked sound, and a predetermined threshold value regarding the absolute value of the coherency γ (f) as γ _l As (f), the frequency characteristic H expressed by the equation (2) ₂ A frequency characteristic H represented by the equation (3), where a filter B having (f), j is a real number greater than or equal to 0, and a cross spectrum obtained from the received signals received by the two sensors is C (f). _Three The filter C having (f) and the frequency characteristic H expressed by the equation (4) where k is a real number of 1 or more. _Four A filter D having (f) and σ _abs (F) is the variance between a plurality of received signals divided by the ratio of the absolute values of the frequency spectra of the two received signals received by the two sensors, and σ _φ (F) is the frequency characteristic H expressed by the equation (16) as the variance between the plurality of received signals obtained by dividing the phase difference of the frequency spectrum of the two received signals received by the two sensors. ₇ Variance σ between a plurality of received signals obtained by dividing a ratio of absolute values of frequency spectra of a filter E having (f) and two received signals received by the two sensors. _abs (F) and the variance σ between a plurality of received signals obtained by dividing the phase difference of the frequency spectrum of the two received signals received by the two sensors. _φ The threshold for the product with (f) is σ _l (F) and the frequency characteristic H expressed by the equation (17) ₈ Filter F having (f) and at least log of the above six expressions _j (| C (f) |) and | C (f) | ^{1 / k} 1 or two or more of the filter G having a frequency characteristic obtained by weighting each of the frequency characteristics and the frequency characteristic of the SCOT (Smoothed Coherence Tr Ansform) filter and multiplying these weighted frequency characteristics. And a correlation processing unit that calculates a cross-correlation function from the signal preprocessed by the preprocessing unit.
[0289]
10. The filter G is | γ (f) | ⁱ And {1 / (σ _abs (F) · σ _φ (F))} ⁱ Any one of the frequency characteristics and log _j (| C (f) |) and | C (f) | ^{1 / k} A filter having a frequency characteristic obtained by weighting any one of the frequency characteristics and the frequency characteristic of the SCOT filter and multiplying these weighted frequency characteristics, or log _j (| C (f) |) and | C (f) | ^{1 / k} In any one of the frequency characteristics, the frequency characteristic in which the frequency component at which the absolute value of the coherency γ (f) is equal to or less than a predetermined threshold becomes 0, or 2 received by the two sensors. The variance σ between the received signals divided by the ratio of the absolute value of the frequency spectrum of two received signals _abs (F) and the variance σ between a plurality of received signals obtained by dividing the phase difference of the frequency spectrum of the two received signals received by the two sensors. _φ Threshold value for product with (f) _l (F) has a frequency characteristic that weights the frequency characteristic such that the frequency component that exceeds a predetermined threshold value is 0 and the frequency characteristic of the SCOT filter, and multiplies these weighted frequency characteristics. One of the filters.
[0290]
11. The pre-processing unit has a frequency characteristic H expressed by equation (5), where fs is the lower limit frequency and fe is the upper limit frequency. _Five It further includes a bandpass filter having (f).
[0291]
12 The pre-processing unit is configured to determine a predetermined threshold regarding the absolute values of the real numbers i, j, k and the coherency γ (f) based on statistical data obtained from a preliminary experiment in the presence or absence of an abnormal location. γ _l (F) and the variance σ between a plurality of received signals divided by the ratio of the absolute values of the frequency spectra of the two received signals received by the two sensors. _abs (F) and the variance σ between a plurality of received signals obtained by dividing the phase difference of the frequency spectrum of the two received signals received by the two sensors. _φ Threshold value for product with (f) _l Determine (f).
[0292]
13. The signal processing unit includes a post-processing unit that obtains an envelope of the cross-correlation function calculated by the correlation processing unit and averages the envelope of the cross-correlation function obtained by a predetermined number of repetitions. .
[0293]
14 The post-processing unit determines the presence / absence of an abnormal location from the magnitude relationship between the peak value of the averaged envelope of the cross-correlation function and a predetermined threshold value.
[0294]
15. The post-processing unit includes a time when the envelope of the averaged cross-correlation function for the received signals received by the two ultrasonic sensors reaches a peak, and a propagation speed when the leaked sound propagates through the test tube. The position of the abnormal part is specified from the distance between the two ultrasonic sensors.
[0295]
16. The said post-processing part determines the threshold value regarding the peak value of the envelope of the said averaged cross correlation function based on the statistical data obtained from the preliminary experiment in the case where an abnormal location exists and it does not exist.
[0296]
【The invention's effect】
As described above, according to the present invention, preprocessing is performed in which the characteristics of leakage sound from an abnormal location and the noise characteristics that often cause problems when inspecting an abnormal location of a conduit are sufficiently considered. Then, the correlation process is performed, and the post-processing is further performed, so that the leaked sound from the conduit contained in the noise is unknown in the frequency band of the leaked sound and the leaked portion of the conduit is inspected. Even in the presence of noise that often causes problems, it is possible to efficiently extract the leaking sound and accurately identify the presence and location of an abnormal location in the conduit.
[Brief description of the drawings]
FIG. 1 is a schematic diagram showing the configuration of a leak location detection apparatus according to Embodiments 1 and 2 of the present invention.
FIG. 2 is an explanatory diagram for explaining a signal processing method according to Embodiments 1 and 2 of the present invention;
FIG. 3 is an explanatory diagram for explaining the effect of the signal processing method according to the first and second embodiments of the present invention.
FIG. 4 is an explanatory diagram for explaining the effect of the signal processing method according to the first and second embodiments of the present invention.
FIG. 5 is an explanatory diagram for explaining the effect of the signal processing method according to the first and second embodiments of the present invention.
FIG. 6 is an explanatory diagram for explaining a signal processing method according to the first and second embodiments of the present invention.
FIG. 7 is an explanatory diagram for explaining the effect of the signal processing method according to the first and second embodiments of the present invention.
FIG. 8 is an explanatory diagram for explaining the effect of the signal processing method according to the first and second embodiments of the present invention.
FIG. 9 is an explanatory diagram for explaining the effect of the signal processing method according to the first and second embodiments of the present invention.
FIG. 10 is an explanatory diagram for explaining a signal processing method according to Embodiments 1 and 2 of the present invention;
FIG. 11 is an explanatory diagram for explaining the effect of the signal processing method according to the first and second embodiments of the present invention.
FIG. 12 is an explanatory diagram for explaining the effect of the signal processing method according to the first and second embodiments of the present invention.
FIG. 13 is an explanatory diagram for explaining the effect of the signal processing method according to the first and second embodiments of the present invention.
FIG. 14 is an explanatory diagram for explaining a signal processing method according to the first and second embodiments of the present invention;
FIG. 15 is an explanatory diagram for explaining the effect of the signal processing method according to the first and second embodiments of the present invention;
FIG. 16 is an explanatory diagram for explaining the effect of the signal processing method according to the first and second embodiments of the present invention;
FIG. 17 is an explanatory diagram for explaining the effect of the signal processing method according to the first and second embodiments of the present invention;
FIG. 18 is an explanatory diagram for explaining a signal processing method according to Embodiments 1 and 2 of the present invention;
FIG. 19 is an explanatory diagram for explaining the effect of the signal processing method according to the first and second embodiments of the present invention.
FIG. 20 is an explanatory diagram for explaining the effect of the signal processing method according to the first and second embodiments of the present invention.
FIG. 21 is an explanatory diagram for explaining the effect of the signal processing method according to the first and second embodiments of the present invention;
FIG. 22 is an explanatory diagram for explaining a signal processing method according to the first and second embodiments of the present invention.
FIG. 23 is an explanatory diagram for explaining the effect of the signal processing method according to the first and second embodiments of the present invention.
FIG. 24 is an explanatory diagram for explaining the effect of the signal processing method according to the first and second embodiments of the present invention.
FIG. 25 is an explanatory diagram for explaining the effect of the signal processing method according to the first and second embodiments of the present invention;
FIG. 26 is an explanatory diagram for explaining a signal processing method according to the first and second embodiments of the present invention.
FIG. 27 is an explanatory diagram for explaining the effect of the signal processing method according to the first and second embodiments of the present invention;
FIG. 28 is an explanatory diagram for explaining the effect of the signal processing method according to the first and second embodiments of the present invention.
FIG. 29 is an explanatory diagram for explaining the effect of the signal processing method according to the first and second embodiments of the present invention.
FIG. 30 is an explanatory diagram for explaining a signal processing method according to the first and second embodiments of the present invention.
FIG. 31 is an explanatory diagram for explaining the effect of the signal processing method according to the first and second embodiments of the present invention;
FIG. 32 is an explanatory diagram for explaining the effect of the signal processing method according to the first and second embodiments of the present invention.
FIG. 33 is an explanatory diagram for explaining the effect of the signal processing method according to the first and second embodiments of the present invention.
FIG. 34 is an explanatory diagram for explaining the effect of the signal processing method according to the first and second embodiments of the present invention.
FIG. 35 is a configuration diagram for explaining a leakage spot detection device according to a conventional example.
FIG. 36 is a configuration diagram for explaining a leakage spot detection device according to a conventional example.
[Explanation of symbols]
1 Conduit, 2 Leakage location, 3a Ultrasonic sensor, 3b Ultrasonic sensor, 3c Ultrasonic sensor, 4 Sound pressure measuring device, 5 Correlator, 6 Underground, 7 Receiver, 71 Receiver, 72 Signal processor, 72a Pre-processing unit, 72b Correlation processing unit, 72c Post-processing unit, 73 Display unit, 74 Control unit, 8 Fire hydrant, 9 Recorder, 10 Computer.

Claims (32)

While equipped with three ultrasonic sensors for receiving leaked sound generated by the presence of an abnormal location in the tube to be inspected,
In the signal processing unit that inputs the signal of leakage sound from these ultrasonic sensors via the receiving unit and performs signal processing,
f is the frequency of the received signal, i is a real number greater than or equal to 0, and the coherency obtained from the received signals received by two of the three ultrasonic sensors is γ (f), C ₁₁ (f) is a power spectrum obtained from the received signal received by one of the ultrasonic sensors, and C ₂₂ (f) is a power spectrum obtained from the received signal received by the other of the two ultrasonic sensors.

A filter A having a frequency characteristic H ₁ (f) expressed by: for filtering a cross spectrum obtained from a received signal of leaked sound;
A predetermined threshold value regarding the absolute value of the coherency γ (f) is γ _l (f),

A filter B having a frequency characteristic H ₂ (f) represented by:
j is a real number greater than or equal to 0, and among the above three ultrasonic sensors, a cross spectrum obtained from received signals received by two sensors is defined as C (f).

A filter C having a frequency characteristic H ₃ (f) represented by:
Let k be a real number greater than or equal to 1,

A filter D having a frequency characteristic H ₄ (f) represented by:
Of the above four expressions, at least one of the frequency characteristics of log _j (| C (f) |) and | C (f) | ^{1 / k} and the frequency characteristic 1 / √ of a SCOT (Smoothed Coherence Transform) filter. (C ₁₁ (f) · C ₂₂ (f)) is weighted, and pre-processing including one or more filters among the filter E having frequency characteristics obtained by multiplying these weighted frequency characteristics And
And a correlation processing unit that calculates a cross-correlation function from the signal preprocessed by the preprocessing unit. The filter E includes a frequency characteristic of | γ (f) | ^i, a frequency characteristic of log _j (| C (f) |), and | C (f) | ^{1 / k} , and a SCOT filter. Filter having frequency characteristics obtained by multiplying these frequency characteristics by weight, or log _j (| C (f) |) and | C (f) | ^{1 / k.} In such frequency characteristics, the frequency characteristics in which the frequency component in which the absolute value of coherency γ (f) is equal to or less than a certain threshold value is set to 0 and the frequency characteristics of the SCOT filter are weighted, and these weighted frequencies. The abnormal point detection apparatus according to claim 1, wherein the abnormality point detection apparatus is one of filters having frequency characteristics multiplied by the characteristics. The pre-processing unit has a lower limit frequency fs and an upper limit frequency fe,

The abnormal point detection device according to claim 1, further comprising a bandpass filter having a frequency characteristic H ₅ (f) represented by: The pre-processing unit is configured to determine a predetermined threshold relating to the absolute values of the real numbers i, j, k and the coherency γ (f) based on statistical data obtained from a preliminary experiment in the presence or absence of an abnormal location. The abnormal point detection apparatus according to claim 1, wherein γ _l (f) is determined. The signal processing unit further includes a post-processing unit that obtains an envelope of the cross-correlation function calculated by the correlation processing unit, and further averages the envelope of the cross-correlation function obtained by a predetermined number of repetitions. The abnormality location detection apparatus according to claim 1, further comprising: The said post-processing part determines the presence or absence of an abnormal location from the magnitude relationship between the peak value of the envelope of the averaged cross-correlation function, and a predetermined threshold value. Abnormal point detection device. The post-processing unit includes a time at which an envelope of the averaged cross-correlation function for a received signal received by two of the three ultrasonic sensors reaches a peak, and three ultrasonic sensors. And the time when the envelope of the averaged cross-correlation function peaks for the received signals received by one of the two ultrasonic sensors and the ultrasonic sensor other than the two ultrasonic sensors. The abnormal point detection device according to claim 5 or 6, wherein the position of the abnormal point is specified from the separation distances of the three ultrasonic sensors. The post-processing unit determines a threshold value regarding a peak value of an envelope of the averaged cross-correlation function based on statistical data obtained from a preliminary experiment in the presence or absence of an abnormal part. The abnormal point detection apparatus according to claim 5. In addition to providing two ultrasonic sensors for receiving leakage sound that occurs due to the presence of an abnormal location in the tube to be inspected,
In the signal processing unit that inputs the signal of leakage sound from these ultrasonic sensors via the receiving unit and performs signal processing,
f is the frequency of the received signal, i is a real number greater than or equal to 0, the coherency obtained from the received signals received by the two ultrasonic sensors is γ (f), and one of the two sensors is the ultrasonic sensor. The power spectrum obtained from the received signal received is C ₁₁ (f), and the power spectrum obtained from the received signal received by the other of the two ultrasonic sensors is C ₂₂ (f).

A filter A having a frequency characteristic H ₁ (f) expressed by: for filtering a cross spectrum obtained from a received signal of leaked sound;
A predetermined threshold value regarding the absolute value of the coherency γ (f) is γ _l (f),

A filter B having a frequency characteristic H ₂ (f) represented by:
Let j be a real number greater than or equal to 0, and C (f) be the cross spectrum obtained from the received signals received by the two sensors.

A filter C having a frequency characteristic H ₃ (f) represented by:
Let k be a real number greater than or equal to 1,

A filter D having a frequency characteristic H ₄ (f) represented by:
Of the above four expressions, at least one of the frequency characteristics of log _j (| C (f) |) and | C (f) | ^{1 / k} and the frequency characteristic of the SCOT filter 1 / √ (C ₁₁ (f ) · C ₂₂ (f)), and a pre-processing unit including one or more filters among the filters E having frequency characteristics obtained by multiplying these weighted frequency characteristics,
And a correlation processing unit that calculates a cross-correlation function from the signal preprocessed by the preprocessing unit. The filter E includes a frequency characteristic of | γ (f) | ^i, a frequency characteristic of log _j (| C (f) |), and | C (f) | ^{1 / k} , a SCOT filter Filter having frequency characteristics obtained by multiplying these frequency characteristics by weight, or log _j (| C (f) |) and | C (f) | ^{1 / k.} In one frequency characteristic, the frequency characteristic in which the frequency component in which the absolute value of coherency γ (f) is equal to or less than a certain threshold is 0 and the frequency characteristic of a SCOT (Smoothed Coherence Transform) filter are respectively weighted. The abnormal point detection device according to claim 9, wherein the abnormal point detection device is any one of filters having frequency characteristics obtained by multiplying these weighted frequency characteristics. The pre-processing unit has a lower limit frequency fs and an upper limit frequency fe,

The abnormal point detection apparatus according to claim 9, further comprising a bandpass filter having a frequency characteristic H ₅ (f) represented by: The pre-processing unit is configured to determine a predetermined threshold relating to the absolute values of the real numbers i, j, k and the coherency γ (f) based on statistical data obtained from a preliminary experiment in the presence or absence of an abnormal location. The abnormal point detection apparatus according to claim 9, wherein γ _l (f) is determined. The signal processing unit further includes a post-processing unit that obtains an envelope of the cross-correlation function calculated by the correlation processing unit, and further averages the envelope of the cross-correlation function obtained by a predetermined number of repetitions. The abnormality location detection device according to claim 9, further comprising: The said post-processing part determines the presence or absence of an abnormal location from the magnitude relationship between the peak value of the envelope of the averaged cross-correlation function and a predetermined threshold value. Abnormal point detection device. The post-processing unit includes a time when the envelope of the averaged cross-correlation function for the received signals received by two ultrasonic sensors reaches a peak, a propagation speed when propagating leaked sound through the test tube, and The abnormal point detection device according to claim 13 or 14, wherein the position of the abnormal point is specified from the distance between the two ultrasonic sensors. The post-processing unit determines a threshold value regarding a peak value of an envelope of the averaged cross-correlation function based on statistical data obtained from a preliminary experiment in the presence or absence of an abnormal part. The abnormal point detection apparatus according to any one of claims 13 to 15. While equipped with three ultrasonic sensors for receiving leaked sound generated by the presence of an abnormal location in the tube to be inspected,
In the signal processing unit that receives and processes the received signal of the leakage sound by these ultrasonic sensors via the receiving unit,
f is the frequency of the received signal, i is a real number greater than or equal to 0, and the coherency obtained from the received signals received by two of the three ultrasonic sensors is γ (f), C ₁₁ (f) is a power spectrum obtained from the received signal received by one of the ultrasonic sensors, and C ₂₂ (f) is a power spectrum obtained from the received signal received by the other of the two ultrasonic sensors.

A filter A having a frequency characteristic H ₁ (f) expressed by: for filtering a cross spectrum obtained from a received signal of leaked sound;
A predetermined threshold regarding the absolute value of the coherency γ (f) is γ ₁ (f),

A filter B having a frequency characteristic H ₂ (f) represented by:
j is a real number greater than or equal to 0, and among the above three ultrasonic sensors, a cross spectrum obtained from received signals received by two sensors is defined as C (f).

A filter C having a frequency characteristic H ₃ (f) represented by:
Let k be a real number greater than or equal to 1,

A filter D having a frequency characteristic H ₄ (f) represented by:
σ _abs (f) is a variance between a plurality of received signals obtained by dividing the ratio of absolute values of frequency spectra of two received signals received by two of the three ultrasonic sensors, and σ _φ (F) is the variance between a plurality of received signals obtained by dividing the phase difference of the frequency spectrum of the two received signals received by the two sensors.

A filter E having a frequency characteristic H ₇ (f) represented by:
Among the three ultrasonic sensors, a variance σ _abs (f) between a plurality of reception signals divided by a ratio of absolute values of frequency spectra of two reception signals received by two sensors, and the two Let σ _l (f) be a threshold value related to the product of the variance σ _φ (f) between a plurality of received signals obtained by dividing the phase difference of the frequency spectrum of two received signals received by the sensor,

A filter F having a frequency characteristic H ₈ (f) represented by:
Of the above six expressions, at least one of the frequency characteristics of log _j (| C (f) |) and | C (f) | ^{1 / k} and the frequency characteristic of the SCOT (Smoothed Coherence Transform) filter are respectively shown. A pre-processing unit including one or two or more of the filters G having frequency characteristics obtained by weighting and multiplying these weighted frequency characteristics;
And a correlation processing unit that calculates a cross-correlation function from the signal preprocessed by the preprocessing unit. The filter G has a frequency characteristic of | γ (f) | ⁱ and {1 / (σ _abs (f) · σ _φ (f))} ⁱ and log _j (| C (f) |), And | C (f) | ^{1 / k} , each of the frequency characteristics and the frequency characteristic of the SCOT filter are weighted, and a filter having a frequency characteristic obtained by multiplying these weighted frequency characteristics, Alternatively, a threshold value in which the absolute value of coherency γ (f) is predetermined in any one of the frequency characteristics of log _j (| C (f) |) and | C (f) | ^{1 / k.} Frequency characteristics such that the frequency component to be zero is 0, or a plurality of divided frequency spectrum absolute value ratios of two received signals received by two of the three ultrasonic sensors. variance σ and _abs (f) between the received signals, with the two sensors The product relates threshold σ _l (f) frequency components equal to or higher than a certain threshold of the variance σ φ _(f) between the divided plurality of received signals having a phase difference of the frequency spectrum of the signal by the two received signals is zero The abnormal part according to claim 17, wherein the frequency characteristic and the frequency characteristic of the SCOT filter are weighted respectively, and the filter has frequency characteristics obtained by multiplying the weighted frequency characteristics. Detection device. The pre-processing unit has a lower limit frequency fs and an upper limit frequency fe,

The abnormal point detection device according to claim 17 or 18, further comprising a bandpass filter having a frequency characteristic H ₅ (f) represented by: The pre-processing unit is configured to determine a predetermined threshold relating to the absolute values of the real numbers i, j, k and the coherency γ (f) based on statistical data obtained from a preliminary experiment in the presence or absence of an abnormal location. γ _l (f), variance σ _abs (f) between a plurality of received signals divided by the ratio of absolute values of frequency spectra of two received signals received by two of the three ultrasonic sensors ) And the variance σ _φ (f) between a plurality of received signals obtained by dividing the phase difference of the frequency spectrum of the two received signals received by the two sensors, a threshold σ _l (f) is determined. 20. An abnormal point detection apparatus according to claim 17, The signal processing unit further includes a post-processing unit that obtains an envelope of the cross-correlation function calculated by the correlation processing unit, and further averages the envelope of the cross-correlation function obtained by a predetermined number of repetitions. 21. The abnormal point detection device according to claim 17, further comprising: The said post-processing part determines the presence or absence of an abnormal location from the magnitude relationship between the peak value of the envelope of the averaged cross-correlation function, and a predetermined threshold value. Abnormal point detection device. The post-processing unit includes a time at which an envelope of the averaged cross-correlation function for a received signal received by two of the three ultrasonic sensors reaches a peak, and three ultrasonic sensors. And the time when the envelope of the averaged cross-correlation function peaks for the received signals received by one of the two ultrasonic sensors and the ultrasonic sensor other than the two ultrasonic sensors. The abnormal point detection device according to claim 21 or 22, wherein the position of the abnormal point is specified from the separation distances of the three ultrasonic sensors. The post-processing unit determines a threshold value regarding a peak value of an envelope of the averaged cross-correlation function based on statistical data obtained from a preliminary experiment in the presence or absence of an abnormal part. The abnormal point detection apparatus according to any one of claims 21 to 23. In addition to providing two ultrasonic sensors for receiving leakage sound that occurs due to the presence of an abnormal location in the tube to be inspected,
In the signal processing unit that inputs the signal of leakage sound from these ultrasonic sensors via the receiving unit and performs signal processing,
f is the frequency of the received signal, i is a real number greater than or equal to 0, the coherency obtained from the received signals received by the two ultrasonic sensors is γ (f), and one of the two sensors is the ultrasonic sensor. The power spectrum obtained from the received signal received is C ₁₁ (f), and the power spectrum obtained from the received signal received by the other of the two ultrasonic sensors is C ₂₂ (f).

A filter A having a frequency characteristic H ₁ (f) expressed by: for filtering a cross spectrum obtained from a received signal of leaked sound;
A predetermined threshold value regarding the absolute value of the coherency γ (f) is γ _l (f),

A filter B having a frequency characteristic H ₂ (f) represented by:
Let j be a real number greater than or equal to 0, and C (f) be the cross spectrum obtained from the received signals received by the two sensors.

A filter C having a frequency characteristic H ₃ (f) represented by:
Let k be a real number greater than or equal to 1,

A filter D having a frequency characteristic H ₄ (f) represented by:
σ _abs (f) is a variance between a plurality of received signals obtained by dividing the ratio of the absolute values of the frequency spectra of the two received signals received by the two sensors, and σ _φ (f) is determined by the two sensors. As a variance between a plurality of received signals obtained by dividing the phase difference of the frequency spectrum of two received signals,

A filter E having a frequency characteristic H ₇ (f) represented by:
The variance σ _abs (f) between a plurality of received signals divided by the ratio of the absolute value of the frequency spectrum of the two received signals received by the two sensors, and the two received signals received by the two sensors Let σ _l (f) be a threshold for the product of the variance σ _φ (f) between a plurality of received signals divided in the phase difference of the frequency spectrum,

A filter F having a frequency characteristic H ₈ (f) represented by:
Of the above six expressions, at least one of the frequency characteristics of log _j (| C (f) |) and | C (f) | ^{1 / k} and the frequency characteristic of the SCOT (Smoothed Coherence Transform) filter are respectively shown. A pre-processing unit including one or two or more of the filters G having frequency characteristics obtained by weighting and multiplying these weighted frequency characteristics;
And a correlation processing unit that calculates a cross-correlation function from the signal preprocessed by the preprocessing unit. The filter G has a frequency characteristic of | γ (f) | ⁱ and {1 / (σ _abs (f) · σ _φ (f))} ⁱ and log _j (| C (f) |) And | C (f) | ^{1 / k} , one of the frequency characteristics and the frequency characteristic of the SCOT filter are respectively weighted, and a filter having a frequency characteristic obtained by multiplying these weighted frequency characteristics, Alternatively, in any one of the frequency characteristics of log _j (| C (f) |) and | C (f) | ^{1 / k} , the absolute value of coherency γ (f) is less than a predetermined threshold value. The frequency characteristic such that the frequency component becomes 0, or the variance σ _abs (f) between the plurality of received signals divided by the ratio of the absolute values of the frequency spectra of the two received signals received by the two sensors And the frequency scan of the two received signals received by the two sensors. Distributed between the divided plurality of received signals of the phase difference vector sigma _phi threshold for the product of (f) σ _l (f) is, as previously frequency component equal to or higher than a certain threshold value which is determined is the 0 26. The abnormal point detection device according to claim 25, wherein the frequency characteristic and the frequency characteristic of the SCOT filter are weighted respectively, and the filter has frequency characteristics obtained by multiplying the weighted frequency characteristics. The pre-processing unit has a lower limit frequency fs and an upper limit frequency fe,

27. The abnormal point detection apparatus according to claim 25 or 26, further comprising a bandpass filter having a frequency characteristic H ₅ (f) represented by: The pre-processing unit is configured to determine a predetermined threshold regarding the absolute values of the real numbers i, j, k and the coherency γ (f) based on statistical data obtained from a preliminary experiment in the presence or absence of an abnormal location. γ _l (f) and the variance σ _abs (f) between the plurality of received signals divided by the ratio of the absolute values of the frequency spectra of the two received signals received by the two sensors, and the two sensors 26. A threshold value σ _l (f) relating to a product of variance σ _φ (f) between a plurality of received signals obtained by dividing a phase difference of a frequency spectrum of two received signals is determined. The abnormal part detection apparatus in any one of thru | or 27. The signal processing unit includes a post-processing unit that obtains an envelope of the cross-correlation function calculated by the correlation processing unit and averages the envelope of the cross-correlation function obtained by a predetermined number of repetitions. 29. The abnormal point detection apparatus according to claim 25, wherein The said post-processing part determines the presence or absence of an abnormal location from the magnitude relationship between the peak value of the envelope of the averaged cross-correlation function and a predetermined threshold value. Abnormal point detection device. The post-processing unit includes a time at which an envelope of the averaged cross-correlation function for received signals received by two ultrasonic sensors reaches a peak, a propagation speed at which leaked sound propagates through the test tube, and The abnormal location detection device according to claim 29 or 30, wherein the location of the abnormal location is specified from each of the separation distances of the two ultrasonic sensors. The post-processing unit determines a threshold value regarding a peak value of an envelope of the averaged cross-correlation function based on statistical data obtained from a preliminary experiment in the presence or absence of an abnormal part. 32. The abnormal point detection apparatus according to claim 29.

Images