JP2004012322A - Device for diagnosing deterioration in semiconductor detector - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To patternize change in an output noise waveform of a semiconductor detector to carry out automatic discrimination. <P>SOLUTION: A computer 26 computes wavelet transformation using the Gabor function of the output noise waveform observed by the semiconductor detector 1 as a basic function, based on a time-serial data Xi output from a sampling means 24, an average value in every passing center frequency computed by the wavelet transformation using the Gabor function as the basic function is compared with an average value in the every passing center frequency when the output noise waveform of the semiconductor detector 1 is normal, and a condition of the output noise waveform is output based on a comparison result therein. <P>COPYRIGHT: (C)2004,JPO

Description

【００２９】
【数２】

【００３０】
次に、連続時間形式の基本ウェーブレット関数ψ（ｔ），−Ｄ≦ｔ＜Ｄを、入力信号Ｘ（ｔ）と同じデータ数Ｌ（点）に引き伸ばし、式（３）のような連続ウェーブレット変換用の基本ウェーブレット関数（解像度＝０の基底関数）Ｗ（ｔ）を生成する。
【００３１】
【数３】

【００３２】
次に、上記基本ウェーブレット関数（解像度＝０の基底関数）Ｗ（ｔ）を相似変換し、式（４）のように各解像度ごとに基底関数Ｐｎ（ｔ）を計算する。これら基本ウェーブレット関数の相似変換の様子は図３に表わしている。ここで、Ｗ（ｔ）は離散的な時刻ｔでしか値を持っていないので、線形補間により中間値を求める。
【００３３】
【数４】

【００３４】
次に、解像度ｎごとに並んだ基底関数の配列Ｐｎ（ｔ）とデータＸ（ｔ）を各時刻毎に連続ウェーブレット変換（コンボリューション計算）を行うことにより、式（５）に示す解像度ｎ，時刻ｔにおけるウェーブレット係数値Ｙｎ（ｔ）を求める。これら連続ウェーブレット変換の様子は図４に表わしている。
【００３５】
【数５】

【００３６】
以上のようにして、Ｇａｂｏｒ関数を基底関数とするウェーブレット変換結果が通過中心周波数０，１，．．．，ｊ−１として求められると、診断器２５における平均演算器（０）３１、平均演算器（１）３２、平均演算器（ｊ−１）３３は、下記の式（６）に基づいて、通過中心周波数０，１，．．．，ｊの各平均μ_１ ^０，μ_１ ^１，．．．，μ_１ ^ｊ−１を演算する。
【００３７】
【数６】

【００３８】
そして、平均演算器（０）３１、平均演算器（１）３２、平均演算器（ｊ−１）３３により、各平均μ_１ ^０，μ_１ ^１，．．．，μ_１ ^ｊ−１が演算されると、比較器（０）３５、比較器（１）３７、比較器（ｊ−１）３９が、出力雑音波形が正常であるときの、Ｇａｂｏｒ関数を基底関数とするウェーブレット変換の通過中心周波数０，１，．．．，ｊ−１の平均を設定器（０）３４、設定器（１）３６、設定器（ｊ−１）３８から入力し、その演算された平均μ_１ ^０，μ_１ ^１，．．．，μ_１ ^ｊ−１と正常時の各通過中心周波数０，１，．．．，ｊ−１の平均と比較する。なお、ここで、設定器（０）３４、設定器（１）３６、設定器（ｊ−１）３８には、出力雑音波形が正常であるときの、Ｇａｂｏｒ関数を基底関数とするウェーブレット変換の通過中心周波数０，１，．．．，ｊ−１の平均が記憶されているが、外部の装置（図示せず）が演算した正常時の平均を記憶するようにしてもよい。
【００３９】
そして、比較器（０）３５、比較器（１）３７、比較器（ｊ−１）３９は演算された平均μ_１ ^０，μ_１ ^１，．．．，μ_１ ^ｊ−１と正常時の平均を比較した結果、例えば、演算された平均μ_１ ^０，μ_１ ^１，．．．，μ_１ ^ｊ−１が正常時の平均の３倍を超えた時、異常と判定し、比較器Ｍ（４１）に出力する。
【００４０】
ここで、出力雑音波形のＧａｂｏｒ関数を基底関数とするウェーブレット変換の通過中心周波数０，１，．．．，ｊ−１の平均により、出力雑音波形の正常・異常が区別できる理由を簡単に説明すると、出力雑音波形のＧａｂｏｒ関数を基底関数とするウェーブレット変換の通過中心周波数０，１，．．．，ｊ−１は相当通過中心周波数ごとのフィルタになっており、フィルタをかけた時系列データの平均は、正常の場合に所定値のまわりに分布する。異常になると、特定の通過中心周波数あるいは、全部の通過中心周波数の平均が正常の場合の平均からずれる。このずれ方は、異常な出力雑音波形に依存するので、平均を比較することで正常・異常の判別することができる。
【００４１】
そして、比較器Ｍ４１は、比較器（０）３５、比較器（１）３７、比較器（ｊ−１）３９の判定結果と正常時にあらかじめ定められた各通過中心周波数ごとの正常・異常パターンと比較して最終的な判定を決定する。この判定は、対象に応じて、設定器Ｍ４０に格納しておく。例えば、一番単純な設定基準は、各通過中心周波数が一つでも異常なら、最終判断は異常であるという設定である。
【００４２】
そして、最後に、比較器Ｍ４１が異常と判断すると、警報発生器（警報発生手段）４２が、異常な出力雑音波形が発生されたことを明らかにすべく、表示装置（図示せず）に異常が発生した旨を表示し、あるいは、プラントを監視する監視装置（図示せず）等に異常が発生した旨を示す信号等を出力し、一連の処理を終了する。
【００４３】
以上のように、この発明の実施の形態１によれば正常・異常を判断する指標と成り得る出力雑音波形のＧａｂｏｒ関数を基底関数とするウェーブレット変換の通過中心周波数０，１，．．．，ｊ−１の平均に基づいて出力雑音波形の正常・異常を判断するように構成したので、従来と比較して短時間かつ高精度に半導体検出器の劣化を判定することができる。
【００４４】
実施の形態２．
以下、この発明の実施の形態２を図について説明する。図１はこの発明の実施の形態２による半導体検出器の劣化診断システムを示す全体図であり、その構成は実施の形態１と同様であるので説明を省略する。
【００４５】
図５はこの発明の実施の形態２に係る診断器２５の詳細な構成を示すブロック図である。図において、２６はＡ／Ｄ変換器２４から出力された時系列データＸｉに基づいて出力雑音波形のＧａｂｏｒ関数を基底関数とするウェーブレット変換を演算するウェーブレット変換演算器（演算手段）、２７はウェーブレット変換演算器２６の基底関数であるＧａｂｏｒ関数を与える基底関数発生器（演算手段）、２８、２９、３０は、それぞれＧａｂｏｒ関数を基底関数とするウェーブレット変換演算器２６から出力された結果であり、０，１，２，．．．，ｊ−１個の通過中心周波数と０，１，２，．．．，Ｌ−１の時間で表現される。これらをそれぞれ通過中心周波数（０）２８、通過中心周波数（１）２９、通過中心周波数（ｊ−１）３０と呼ぶ。
【００４６】
３１０はＧａｂｏｒ関数を基底関数とするウェーブレット変換演算器２６から出力された通過中心周波数（０）２８の分散を演算する分散演算器（０）（演算手段）、３２０はＧａｂｏｒ関数を基底関数とするウェーブレット変換演算器２６から出力された通過中心周波数（１）２９の分散を演算する分散演算器（１）（演算手段）、３３０はＧａｂｏｒ関数を基底関数とするウェーブレット変換演算器２６から出力された通過中心周波数（ｊ−１）３０の分散を演算する分散演算器（ｊ−１）（演算手段）である。なおこの分散演算器はｊ個設定しておく。
【００４７】
３４０は半導体検出器１の正常出力雑音波形の場合のＧａｂｏｒ関数を基底関数とするウェーブレット変換の通過中心周波数（０）２８の分散を記憶する設定器（０）、３５０は分散演算器（０）３１０により演算された出力雑音波形のＧａｂｏｒ関数を基底関数とするウェーブレット変換の通過中心周波数（０）２８の分散と、設定器（０）３４０に記憶された正常時のＧａｂｏｒ関数を基底関数とするウェーブレット変換の通過中心周波数（０）２８の分散を比較する比較器（０）（比較手段）である。
【００４８】
３６０は半導体検出器１の正常出力雑音波形の場合のＧａｂｏｒ関数を基底関数とするウェーブレット変換の通過中心周波数（１）２９の分散を記憶する設定器（１）、３７０は分散演算器（１）３２０により演算された出力雑音波形のＧａｂｏｒ関数を基底関数とするウェーブレット変換の通過中心周波数（１）２９の分散と、設定器（１）３６０に記憶された正常時のＧａｂｏｒ関数を基底関数とするウェーブレット変換の通過中心周波数（１）２９の分散を比較する比較器（１）（比較手段）である。
【００４９】
３８０は半導体検出器１の正常出力雑音波形の場合のＧａｂｏｒ関数を基底関数とするウェーブレット変換の通過中心周波数（ｊ−１）３０の分散を記憶する設定器（ｊ−１）、３９は分散演算器（ｊ−１）３３０により演算された出力雑音波形のＧａｂｏｒ関数を基底関数とするウェーブレット変換の通過中心周波数（ｊ−１）３０の分散と、設定器（ｊ−１）３８０に記憶された正常時のＧａｂｏｒ関数を基底関数とするウェーブレット変換の通過中心周波数（ｊ−１）３０の分散を比較する比較器（ｊ−１）（比較手段）である。
【００５０】
なお、図では通過中心周波数（０）２８、通過中心周波数（１）２９、通過中心周波数（ｊ−１）３０という具合に３個で代表して表記しているが、実際にはｊ個の通過中心周波数が得られる。同様に、分散演算器（０）３１０、分散演算器（１）３２０、分散演算器（ｊ−１）３３０，比較器（０）３５０、比較器（１）３７０、比較器（ｊ−１）３９０，設定器（０）３４０、設定器（１）３６０、設定器（ｊ−１）３８０という具合にそれぞれ３個で代表して表記しているが、実際にはそれぞれｊ個必要である。
【００５１】
４００は比較器（０）３５０、比較器（１）３７０，．．．，比較器（ｊ−１）３９０において正常出力雑音波形の場合の許容パターンを記憶する設定器Ｍである。４１０は比較器（０）３５０、比較器（１）３７０，．．．，比較器（ｊ−１）３９０の比較結果と設定器Ｍ４００に記憶された正常時の許容パターンを比較する比較器Ｍ（比較手段）、４２０は比較器Ｍ４１０の比較結果に基づいて警報を出力する警報発生器（警報出力手段）である。
【００５２】
次に、この発明の実施の形態２による半導体検出器の劣化診断動作について説明する。まず、測定された半導体検出器１の出力雑音波形をＡ／Ｄ変換器２４が所定時間ごとにサンプリングしてアナログ・ディジタル変換し、複数個の時系列データＸｉ（ｉ＝０，１，．．．，Ｌ−１）を出力する。
【００５３】
そして、複数個の時系列データＸｉが診断器２５に入力されると、診断器２５のウェーブレット変換演算器２６は、実施の形態１に説明した手法と同様の手法で、Ｇａｂｏｒ関数を基底関数とするウェーブレット変換を演算する。
【００５４】
Ｇａｂｏｒ関数を基底関数とするウェーブレット変換結果が通過中心周波数１，２，．．．，ｊとして求められると、診断器２５における分散演算器（０）３１０、分散演算器（１）３２０、分散演算器（ｊ−１）３３０は、下記の式（７）に基づいて、通過中心周波数０，１，．．．，ｊ−１の各分散μ_２ ^０，μ_２ ^１，．．．，μ_２ ^ｊ−１を演算する。
【００５５】
【数７】

【００５６】
そして、分散演算器（０）３１０、分散演算器（１）３２０、分散演算器（ｊ−１）３３０により各分散μ_２ ^０，μ_２ ^１，．．．，μ_２ ^ｊ−１が演算されると、比較器（０）３５０、比較器（１）３７０、比較器（ｊ−１）３９０が、出力雑音波形が正常であるときの、Ｇａｂｏｒ関数を基底関数とするウェーブレット変換の通過中心周波数０，１，．．．，ｊ−１の分散を設定器（０）３４０、設定器（１）３６０、設定器（ｊ−１）３８０から入力し、その演算された分散μ_２ ^０，μ_２ ^１，．．．，μ_２ ^ｊ−１と正常時の各通過中心周波数０，１，．．．，ｊ−１の分散と比較する。なお、ここで、設定器（０）３４０、設定器（１）３６０、設定器（ｊ−１）３８０には、出力雑音波形が正常であるときの、Ｇａｂｏｒ関数を基底関数とするウェーブレット変換の通過中心周波数０，１，．．．，ｊ−１の分散が記憶されているが、外部の装置（図示せず）が演算した正常時の分散を記憶するようにしてもよい。
【００５７】
そして、比較器（０）３５０、比較器（１）３７０、比較器（ｊ−１）３９０は演算された分散μ_２ ^０，μ_２ ^１，．．．，μ_２ ^ｊ−１と正常時の分散を比較した結果、例えば、演算された分散μ_２ ^０，μ_２ ^１，．．．，μ_２ ^ｊ−１が正常時の分散の３倍を超えた時、異常と判定し、比較器Ｍ４１０に出力する。
【００５８】
ここで、出力雑音波形のＧａｂｏｒ関数を基底関数とするウェーブレット変換の通過中心周波数０，１，．．．，ｊ−１の分散により、出力雑音波形の正常・異常が区別できる理由を簡単に説明すると、出力雑音波形のＧａｂｏｒ関数を基底関数とするウェーブレット変換の通過中心周波数０，１，．．．，ｊ−１は相当通過中心周波数ごとのフィルタになっており、フィルタをかけた時系列データの分散は、正常の場合ある値のまわりに分布する。異常になると、特定の通過中心周波数あるいは、全部の通過中心周波数の分散が正常の場合の分散からずれる。このずれ方は、異常な出力雑音波形に依存するので、分散を比較することで正常・異常の判別することができる。特に、分散は確率密度分布関数のばらつきを表わす指標であるから、異常が確率密度分布関数のばらつきに表われる場合に顕著に検出できる。
【００５９】
そして、比較器Ｍ４１０は、比較器（０）３５０、比較器（１）３７０、比較器（ｊ−１）３９０の判定結果と正常時にあらかじめ定められた各通過中心周波数ごとの正常・異常パターンと比較して最終的な判定を決定する。この判定は、対象に応じて、設定器Ｍ４００に格納しておく。例えば、一番単純な設定基準は、各通過中心周波数が一つでも異常なら、最終判断は異常であるという設定である。
【００６０】
そして、最後に、比較器Ｍ４１０が異常と判断すると、警報発生器（警報発生手段）４２０が、異常な出力雑音波形が発生されたことを明らかにすべく、表示装置（図示せず）に異常が発生した旨を表示し、あるいは、プラントを監視する監視装置（図示せず）等に異常が発生した旨を示す信号等を出力し、一連の処理を終了する。
【００６１】
以上のように、この発明の実施の形態２によれば正常・異常を判断する指標と成り得る出力雑音波形のＧａｂｏｒ関数を基底関数とするウェーブレット変換の通過中心周波数０，１，．．．，ｊ−１の分散に基づいて出力雑音波形の正常・異常を判断するように構成したので、従来と比較して短時間かつ高精度に半導体検出器の劣化を判定することができる。
【００６２】
実施の形態３．
以下、この発明の実施の形態３を図について説明する。図１はこの発明の実施の形態３による半導体検出器の劣化診断システムを示す全体図であり、その構成は実施の形態１と同様であるので説明を省略する。
【００６３】
図６はこの発明の実施の形態３に係る診断器２５の詳細な構成を示すブロック図である。図において、２６はＡ／Ｄ変換器２４から出力された時系列データＸｉに基づいて出力雑音波形のＧａｂｏｒ関数を基底関数とするウェーブレット変換を演算するウェーブレット変換演算器（演算手段）、２７はウェーブレット変換演算器２６の基底関数であるＧａｂｏｒ関数を与える基底関数発生器（演算手段）、２８、２９、３０は、それぞれＧａｂｏｒ関数を基底関数とするウェーブレット変換演算器２６から出力された結果であり、０，１，２，．．．，ｊ−１個の通過中心周波数と０，１，２，．．．，Ｌ−１の時間で表現される。これらをそれぞれ通過中心周波数（０）２８、通過中心周波数（１）２９、通過中心周波数（ｊ−１）３０と呼ぶ。
【００６４】
３１１はＧａｂｏｒ関数を基底関数とするウェーブレット変換演算器２６から出力された通過中心周波数（０）２８の歪度を演算する歪度演算器（０）（演算手段）、３２１はＧａｂｏｒ関数を基底関数とするウェーブレット変換演算器２６から出力された通過中心周波数（１）２９の歪度を演算する歪度演算器（１）（演算手段）、３３１はＧａｂｏｒ関数を基底関数とするウェーブレット変換演算器２６から出力された通過中心周波数（ｊ−１）３０の歪度を演算する歪度演算器（ｊ−１）（演算手段）である。なおこの歪度演算器はｊ個設定しておく。
【００６５】
３４１は半導体検出器１の正常出力雑音波形の場合のＧａｂｏｒ関数を基底関数とするウェーブレット変換の通過中心周波数（０）２８の歪度を記憶する設定器（０）、３５１は歪度演算器（０）３１１により演算された出力雑音波形のＧａｂｏｒ関数を基底関数とするウェーブレット変換の通過中心周波数（０）２８の歪度と、設定器（０）３４１に記憶された正常時のＧａｂｏｒ関数を基底関数とするウェーブレット変換の通過中心周波数（０）２８の歪度を比較する比較器（０）（比較手段）である。
【００６６】
３６１は半導体検出器１の正常出力雑音波形の場合のＧａｂｏｒ関数を基底関数とするウェーブレット変換の通過中心周波数（１）２９の歪度を記憶する設定器（１）、３７１は歪度演算器（１）３２１により演算された出力雑音波形のＧａｂｏｒ関数を基底関数とするウェーブレット変換の通過中心周波数（１）２９の歪度と、設定器（１）３６１に記憶された正常時のＧａｂｏｒ関数を基底関数とするウェーブレット変換の通過中心周波数（１）２９の歪度を比較する比較器（１）（比較手段）である。
【００６７】
３８１は半導体検出器１の正常出力雑音波形の場合のＧａｂｏｒ関数を基底関数とするウェーブレット変換の通過中心周波数（ｊ−１）３０の歪度を記憶する設定器（ｊ−１）、３９は歪度演算器（ｊ−１）３３１により演算された出力雑音波形のＧａｂｏｒ関数を基底関数とするウェーブレット変換の通過中心周波数（ｊ−１）３０の歪度と、設定器（ｊ−１）３８１に記憶された正常時のＧａｂｏｒ関数を基底関数とするウェーブレット変換の通過中心周波数（ｊ−１）３０の歪度を比較する比較器（ｊ−１）（比較手段）である。
【００６８】
なお、図では通過中心周波数（０）２８、通過中心周波数（１）２９、通過中心周波数（ｊ−１）３０という具合に３個で代表して表記しているが、実際にはｊ個の通過中心周波数が得られる。同様に、歪度演算器（０）３１１、歪度演算器（１）３２１、歪度演算器（ｊ−１）３３１，比較器（０）３５１、比較器（１）３７１、比較器（ｊ−１）３９１，設定器（０）３４１、設定器（１）３６１、設定器（ｊ−１）３８１という具合にそれぞれ３個で代表して表記しているが、実際にはそれぞれｊ個必要である。
【００６９】
４０１は比較器（０）３５１、比較器（１）３７１，．．．，比較器（ｊ−１）３９１において正常出力雑音波形の場合の許容パターンを記憶する設定器Ｍである。４１１は比較器（０）３５１、比較器（１）３７１，．．．，比較器（ｊ−１）３９１の比較結果と設定器Ｍ４０１に記憶された正常時の許容パターンを比較する比較器Ｍ（比較手段）、４２１は比較器Ｍ４１１の比較結果に基づいて警報を出力する警報発生器（警報出力手段）である。
【００７０】
次に、この発明の実施の形態３による半導体検出器の劣化診断動作について説明する。まず、測定された半導体検出器１の出力雑音波形をＡ／Ｄ変換器２４が所定時間ごとにサンプリングしてアナログ・ディジタル変換し、複数個の時系列データＸｉ（ｉ＝０，１，．．．，Ｌ−１）を出力する。
【００７１】
そして、複数個の時系列データＸｉが診断器２５に入力されると、診断器２５のウェーブレット変換演算器２６は、実施の形態１に説明した手法と同様の手法で、Ｇａｂｏｒ関数を基底関数とするウェーブレット変換を演算する。
【００７２】
Ｇａｂｏｒ関数を基底関数とするウェーブレット変換結果が通過中心周波数０，１，．．．，ｊ−１として求められると、診断器２５における歪度演算器（０）３１１、歪度演算器（１）３２１、歪度演算器（ｊ−１）３３１は、下記の式（８）に基づいて、通過中心周波数０，１，．．．，ｊ−１のそれぞれ歪度μ_３ ^０，μ_３ ^１，．．．，μ_３ ^ｊ−１を演算する。式（８）において、歪度μ_３は平均値μ_１と分散値μ_２から求められ、この歪度は各数値の算術平均からの偏差の３乗に基づくもので、統計データの処理によく用いられる。
【００７３】
【数８】

【００７４】
そして、歪度演算器（０）３１１、歪度演算器（１）３２１、歪度演算器（ｊ−１）３３１により各歪度μ_３ ^０，μ_３ ^１，．．．，μ_３ ^ｊ−１が演算されると、比較器（０）３５１、比較器（１）３７１、比較器（ｊ−１）３９１が、出力雑音波形が正常であるときの、Ｇａｂｏｒ関数を基底関数とするウェーブレット変換の通過中心周波数０，１，．．．，ｊ−１の歪度を設定器（０）３４１、設定器（１）３６１、設定器（ｊ−１）３８１から入力し、その演算された歪度μ_３ ^０，μ_３ ^１，．．．，μ_３ ^ｊ−１と正常時の各通過中心周波数０，１，．．．，ｊ−１の歪度と比較する。なお、ここで、設定器（０）３４１、設定器（１）３６１、設定器（ｊ−１）３８１には、出力雑音波形が正常であるときの、Ｇａｂｏｒ関数を基底関数とするウェーブレット変換の通過中心周波数０，１，．．．，ｊ−１の歪度が記憶されているが、外部の装置（図示せず）が演算した正常時の歪度を記憶するようにしてもよい。
【００７５】
そして、比較器（０）３５１、比較器（１）３７１、比較器（ｊ−１）３９１は演算された歪度μ_３ ^０，μ_３ ^１，．．．，μ_３ ^ｊ−１と正常時の歪度を比較した結果、例えば、演算された歪度μ_３ ^０，μ_３ ^１，．．．，μ_３ ^ｊ−１が正常時の歪度の３倍を超えた時、異常と判定し、比較器Ｍ４１１に出力する。
【００７６】
ここで、出力雑音波形のＧａｂｏｒ関数を基底関数とするウェーブレット変換の通過中心周波数０，１，．．．，ｊ−１の歪度により、出力雑音波形の正常・異常が区別できる理由を簡単に説明すると、出力雑音波形のＧａｂｏｒ関数を基底関数とするウェーブレット変換の通過中心周波数０，１，．．．，ｊ−１は相当通過中心周波数ごとのフィルタになっており、フィルタをかけた時系列データの歪度は、正常の場合ある値のまわりに分布する。異常になると、特定の通過中心周波数あるいは、全部の通過中心周波数の歪度が正常の場合の歪度からずれる。このずれ方は、異常な出力雑音波形に依存するので、歪度を比較することで正常・異常の判別することができる。特に、歪度は確率密度分布関数の偏りを表わす指標であるから、異常が確率密度分布関数の偏りに表われる場合に顕著に検出できる。
【００７７】
そして、比較器Ｍ４１１は、比較器（０）３５１、比較器（１）３７１、比較器（ｊ−１）３９１の判定結果と正常時にあらかじめ定められた各通過中心周波数ごとの正常・異常パターンと比較して最終的な判定を決定する。この判定は、対象に応じて、設定器Ｍ４０１に格納しておく。例えば、一番単純な設定基準は、各通過中心周波数が一つでも異常なら、最終判断は異常であるという設定である。
【００７８】
そして、最後に、比較器Ｍ４１１が異常と判断すると、警報発生器（警報発生手段）４２１が、異常な出力雑音波形が発生されたことを明らかにすべく、表示装置（図示せず）に異常が発生した旨を表示し、あるいは、プラントを監視する監視装置（図示せず）等に異常が発生した旨を示す信号等を出力し、一連の処理を終了する。
【００７９】
以上のように、この発明の実施の形態３によれば正常・異常を判断する指標と成り得る出力雑音波形のＧａｂｏｒ関数を基底関数とするウェーブレット変換の通過中心周波数０，１，．．．，ｊ−１の歪度に基づいて出力雑音波形の正常・異常を判断するように構成したので、従来と比較して短時間かつ高精度に半導体検出器の劣化を判定することができる。
【００８０】
実施の形態４．
以下、この発明の実施の形態４を図について説明する。図１はこの発明の実施の形態４による半導体検出器の劣化診断システムを示す全体図であり、その構成は実施の形態１と同様であるので説明を省略する。
【００８１】
図７はこの発明の実施の形態４に係る診断器２５の詳細な構成を示すブロック図である。図において、２６はＡ／Ｄ変換器２４から出力された時系列データＸｉに基づいて出力雑音波形のＧａｂｏｒ関数を基底関数とするウェーブレット変換を演算するウェーブレット変換演算器（演算手段）、２７はウェーブレット変換演算器２６の基底関数であるＧａｂｏｒ関数を与える基底関数発生器（演算手段）、２８、２９、３０は、それぞれＧａｂｏｒ関数を基底関数とするウェーブレット変換演算器２６から出力された結果であり、０，１，２，．．．，ｊ−１の通過中心周波数と０，１，２，．．．，Ｌ−１の時間で表現される。これらをそれぞれ通過中心周波数（０）２８、通過中心周波数（１）２９、通過中心周波数（ｊ−１）３０と呼ぶ。
【００８２】
３１２はＧａｂｏｒ関数を基底関数とするウェーブレット変換演算器２６から出力された通過中心周波数（０）２８の尖度を演算する尖度演算器（０）（演算手段）、３２２はＧａｂｏｒ関数を基底関数とするウェーブレット変換演算器２６から出力された通過中心周波数（１）２９の尖度を演算する尖度演算器（１）（演算手段）、３３２はＧａｂｏｒ関数を基底関数とするウェーブレット変換演算器２６から出力された通過中心周波数（ｊ−１）３０の尖度を演算する尖度演算器（ｊ−１）（演算手段）である。なおこの尖度演算器はｊ個設定しておく。
【００８３】
３４２は半導体検出器１の正常出力雑音波形の場合のＧａｂｏｒ関数を基底関数とするウェーブレット変換の通過中心周波数（０）２８の尖度を記憶する設定器（０）、３５２は尖度演算器（０）３１２により演算された出力雑音波形のＧａｂｏｒ関数を基底関数とするウェーブレット変換の通過中心周波数（０）２８の尖度と、設定器（０）３４２に記憶された正常時のＧａｂｏｒ関数を基底関数とするウェーブレット変換の通過中心周波数（０）２８の尖度を比較する比較器（０）（比較手段）である。
【００８４】
３６２は半導体検出器１の正常出力雑音波形の場合のＧａｂｏｒ関数を基底関数とするウェーブレット変換の通過中心周波数（１）２９の尖度を記憶する設定器（１）、３７２は尖度演算器（１）３２２により演算された出力雑音波形のＧａｂｏｒ関数を基底関数とするウェーブレット変換の通過中心周波数（１）２９の尖度と、設定器（１）３６２に記憶された正常時のＧａｂｏｒ関数を基底関数とするウェーブレット変換の通過中心周波数（１）２９の尖度を比較する比較器（１）（比較手段）である。
【００８５】
３８２は半導体検出器１の正常出力雑音波形の場合のＧａｂｏｒ関数を基底関数とするウェーブレット変換の通過中心周波数（ｊ−１）３０の尖度を記憶する設定器（ｊ−１）、３９は尖度演算器（ｊ−１）３３２により演算された出力雑音波形のＧａｂｏｒ関数を基底関数とするウェーブレット変換の通過中心周波数（ｊ−１）３０の尖度と、設定器（ｊ−１）３８２に記憶された正常時のＧａｂｏｒ関数を基底関数とするウェーブレット変換の通過中心周波数（ｊ−１）３０の尖度を比較する比較器（ｊ−１）（比較手段）である。
【００８６】
なお、図では通過中心周波数（０）２８、通過中心周波数（１）２９、通過中心周波数（ｊ−１）３０という具合に３個で代表して表記しているが、実際にはｊ個の通過中心周波数が得られる。同様に、尖度演算器（０）３１２、尖度演算器（１）３２２、尖度演算器（ｊ−１）３３２，比較器（０）３５２、比較器（１）３７２、比較器（ｊ−１）３９２，設定器（０）３４２、設定器（１）３６２、設定器（ｊ−１）３８２という具合にそれぞれ３個で代表して表記しているが、実際にはそれぞれｊ個必要である。
【００８７】
４０２は比較器（０）３５２、比較器（１）３７２，．．．，比較器（ｊ−１）３９２において正常出力雑音波形の場合の許容パターンを記憶する設定器Ｍである。４１２は比較器（０）３５２、比較器（１）３７２，．．．，比較器（ｊ−１）３９２の比較結果と設定器Ｍ４０２に記憶された正常時の許容パターンを比較する比較器Ｍ（比較手段）、４２２は比較器Ｍ４１２の比較結果に基づいて警報を出力する警報発生器（警報出力手段）である。
【００８８】
次に、この発明の実施の形態４による半導体検出器の劣化診断動作について説明する。まず、測定された半導体検出器１の出力雑音波形をＡ／Ｄ変換器２４が所定時間ごとにサンプリングしてアナログ・ディジタル変換し、複数個の時系列データＸｉ（ｉ＝０，１，２，．．．，Ｌ−１）を出力する。
【００８９】
そして、複数個の時系列データＸｉが診断器２５に入力されると、診断器２５のウェーブレット変換演算器２６は、実施の形態１に説明した手法と同様の手法で、Ｇａｂｏｒ関数を基底関数とするウェーブレット変換を演算する。
【００９０】
Ｇａｂｏｒ関数を基底関数とするウェーブレット変換結果が通過中心周波数０，１，．．．，ｊ−１として求められると、診断器２５における尖度演算器（０）３１２、尖度演算器（１）３２２、尖度演算器（ｊ−１）３３２は、下記の式（９）に基づいて、通過中心周波数０，１，．．．，ｊ−１の各々の尖度μ_４ ^０，μ_４ ^１，．．．，μ_４ ^ｊ−１を演算する。
【００９１】
【数９】

【００９２】
そして、尖度演算器（０）３１２、尖度演算器（１）３２２、尖度演算器（ｊ−１）３３２により各尖度μ_４ ^０，μ_４ ^１，．．．，μ_４ ^ｊ−１が演算されると、比較器（０）３５２、比較器（１）３７２、比較器（ｊ−１）３９２が、出力雑音波形が正常であるときの、Ｇａｂｏｒ関数を基底関数とするウェーブレット変換の通過中心周波数０，１，．．．，ｊ−１の尖度を設定器（０）３４２、設定器（１）３６２、設定器（ｊ−１）３８２から入力し、その演算された尖度μ_４ ^０，μ_４ ^１，．．．，μ_４ ^ｊ−１と正常時の各通過中心周波数０，１，．．．，ｊ−１の尖度と比較する。なお、ここで、設定器（０）３４２、設定器（１）３６２、設定器（ｊ−１）３８２には、出力雑音波形が正常であるときの、Ｇａｂｏｒ関数を基底関数とするウェーブレット変換の通過中心周波数０，１，．．．，ｊ−１の尖度が記憶されているが、外部の装置（図示せず）が演算した正常時の尖度を記憶するようにしてもよい。
【００９３】
そして、比較器（０）３５２、比較器（１）３７２、比較器（ｊ−１）３９２は演算された尖度μ_４ ^０，μ_４ ^１，．．．，μ_４ ^ｊ−１と正常時の尖度を比較した結果、例えば、演算された尖度μ_４ ^０，μ_４ ^１，．．．，μ_４ ^ｊ−１が正常時の尖度の３倍を超えた時、異常と判定し、比較器Ｍ４１２に出力する。
【００９４】
ここで、出力雑音波形のＧａｂｏｒ関数を基底関数とするウェーブレット変換の通過中心周波数０，１，．．．，ｊ−１の尖度により、出力雑音波形の正常・異常が区別できる理由を簡単に説明すると、出力雑音波形のＧａｂｏｒ関数を基底関数とするウェーブレット変換の通過中心周波数０，１，．．．，ｊ−１は相当通過中心周波数ごとのフィルタになっており、フィルタをかけた時系列データの尖度は、正常の場合ある値のまわりに分布する。異常になると、特定の通過中心周波数あるいは、全部の通過中心周波数の尖度が正常の場合の尖度からずれる。このずれ方は、異常な出力雑音波形に依存するので、尖度を比較することで正常・異常の判別することができる。特に、尖度は確率密度分布関数の広がり具合を表わす指標であるから、異常が確率密度分布関数の広がりに表われる場合に顕著に検出できる。
【００９５】
そして、比較器Ｍ４１２は、比較器（０）３５２、比較器（１）３７２、比較器（ｊ−１）３９２の判定結果と正常時にあらかじめ定められた各通過中心周波数ごとの正常・異常パターンと比較して最終的な判定を決定する。この判定は、対象に応じて、設定器Ｍ４０２に格納しておく。例えば、一番単純な設定基準は、各通過中心周波数が一つでも異常なら、最終判断は異常であるという設定である。
【００９６】
そして、最後に、比較器Ｍ４１２が異常と判断すると、警報発生器（警報発生手段）４２２が、異常な出力雑音波形が発生されたことを明らかにすべく、表示装置（図示せず）に異常が発生した旨を表示し、あるいは、プラントを監視する監視装置（図示せず）等に異常が発生した旨を示す信号等を出力し、一連の処理を終了する。
【００９７】
以上のように、この発明の実施の形態４によれば正常・異常を判断する指標と成り得る出力雑音波形のＧａｂｏｒ関数を基底関数とするウェーブレット変換の通過中心周波数０，１，．．．，ｊ−１の尖度に基づいて出力雑音波形の正常・異常を判断するように構成したので、従来と比較して短時間かつ高精度に半導体検出器の劣化を判定することができる。
【００９８】
【発明の効果】
以上のようにこの発明によれば、下記のような効果を達成することができる。
【００９９】
請求項１の発明によれば、サンプリング手段から出力された時系列データに基づく出力雑音波形のＧａｂｏｒ関数を基底関数とするウェーブレット変換の通過中心周波数０，１，．．．，ｊ−１の平均値を演算し、出力雑音波形の正常時の平均値と比較するよう構成したので、従来のものと比較して高速に状況を判断できるように成り、人の感度に即した診断をすることが可能となった。特に、平均値を採用することで、計算時間を短縮することもできる。
【０１００】
請求項２の発明によれば、サンプリング手段から出力された時系列データに基づく出力雑音波形のＧａｂｏｒ関数を基底関数とするウェーブレット変換の通過中心周波数０，１，．．．，ｊ−１の分散値を演算し、出力雑音波形の正常時の分散値と比較するよう構成したので、従来のものと比較して高速に状況を判断できるように成り、人の感度に即した診断をすることが可能となった。特に、分散値を採用することで、ばらつきに異常が現われる現象の感度向上が図れる。
【０１０１】
請求項３の発明によれば、サンプリング手段から出力された時系列データに基づく出力雑音波形のＧａｂｏｒ関数を基底関数とするウェーブレット変換の通過中心周波数０，１，．．．，ｊ−１の歪度を演算し、出力雑音波形の正常時の歪度と比較するよう構成したので、従来のものと比較して高速に状況を判断できるように成り、人の感度に即した診断をすることが可能となった。特に、歪度を採用することで、偏りに異常が現われる現象の感度向上が図れる。
【０１０２】
請求項４の発明によれば、サンプリング手段から出力された時系列データに基づく出力雑音波形のＧａｂｏｒ関数を基底関数とするウェーブレット変換の通過中心周波数０，１，．．．，ｊ−１の尖度を演算し、出力雑音波形の正常時の尖度と比較するよう構成したので、従来のものと比較して高速に状況を判断できるように成り、人の感度に即した診断をすることが可能となった。特に、尖度を採用することで、ひろがりに異常が現われる現象の感度向上が図れる。
【０１０３】
請求項５の発明によれば、請求項１から請求項４の発明の比較結果に基づいて警報を出力するようにしたので、半導体検出装置が劣化した場合、直ちにその異常を外部に知らせることができる。
【図面の簡単な説明】
【図１】この発明による半導体検出器の劣化診断システムを示す全体構成図である。
【図２】この発明の実施の形態１による診断器の詳細な構成を示すブロック図である。
【図３】この発明による基本ウェーブレット関数の相似変換の様子を表わす図である。
【図４】この発明による連続ウェーブレット変換の様子を表わす図である。
【図５】この発明の実施の形態２による診断器の詳細な構成を示すブロック図である。
【図６】この発明の実施の形態３による診断器の詳細な構成を示すブロック図である。
【図７】この発明の実施の形態４による診断器の詳細な構成を示すブロック図である。
【図８】従来の半導体検出器の診断手法を示す構成図である。
【符号の説明】
１　半導体検出器、２　電池、４　負荷抵抗、５　カップリングコンデンサ、２０　前置増幅器、２１　出力雑音波形、２２　フィルタ、２３　増幅器、２４Ａ／Ｄ変換器、２５　診断器、２６　ウェーブレット変換演算器、２７　基底関数、２８　通過中心周波数（０）、２９　通過中心周波数（１）、３０　通過中心周波数（ｊ−１）、３１　平均演算器（０）、３２　平均演算器（１）、３３　平均演算器（ｊ−１）、３４，３４０，３４１，３４２　設定器（０）、３５，３５０，３５１，３５２　比較器（０）、３６，３６０，３６１，３６２　設定器（１）、３７，３７０，３７１，３７２　比較器（１）、３８，３８０，３８１，３８２　設定器（ｊ−１）、３９，３９０，３９１，３９２　比較器（ｊ−１）、４０，４００，４０１，４０２　設定器Ｍ、４１，４１０，４１１，４１２　比較器Ｍ、４２，４２０，４２１，４２２　警報発生器、３１０　分散演算器（０）、３２０　分散演算器（１）、３３０　分散演算器（ｊ−１）、３１１　歪度演算器（０）、３２１　歪度演算器（１）、３３１　歪度演算器（ｊ−１）、３１２　尖度演算器（０）、３２２　尖度演算器（１）、３３２　尖度演算器（ｊ−１）。[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention relates to an apparatus for diagnosing deterioration of a semiconductor detector for detecting radiation. The present invention relates to a semiconductor detector based on an effective value of a continuous wavelet transform using a Gabor function (Gabor function) of an output noise waveform of the semiconductor detector as a basis function. Is to evaluate the state of
[0002]
[Prior art]
Conventionally, as a technique for determining the state of a semiconductor detector, it has been necessary to observe the absolute value of the current flowing through the semiconductor detector and the absolute value of the voltage at the output terminal of the semiconductor detector and evaluate the measured value. I have been. This is a method in which the measured current value or voltage value is compared with a predetermined value and determined by a computer or the like.
[0003]
A conventional method for diagnosing deterioration of a semiconductor detector will be described. FIG. 8 is a diagram showing a configuration of a diagnostic method of a semiconductor detector shown in, for example, "Radiation Measurement Handbook" (2nd edition, published by Nikkan Kogyo Shimbun), 1991.
[0004]
In the figure, 1 is a semiconductor detector for detecting radiation, 2 and 3 are batteries 1 and 2 for applying a high voltage, 4 is a load resistor, 5 is a coupling capacitor, 6 is a connector, and 13 is radiation. The counter 14 is a display. The description of the preamplifier and the like is omitted.
[0005]
In the above configuration, a voltmeter 7 is connected to the connector 6, and a voltage signal waveform 8 output from the voltmeter 7 is input to an effective value calculator 9 to calculate an effective value of the voltage signal waveform. The effective value of the voltage signal waveform is input to the comparator voltage 11, where it is compared with a predetermined value set by the setter voltage 10. When the effective value of the voltage signal waveform exceeds a predetermined range, an alarm a1 of the voltage signal is issued. Be emitted. On the other hand, an ammeter 15 is connected to the semiconductor detector 1, and a current signal waveform 16 output from the ammeter 15 is input to an effective value calculator 17 to calculate an effective value of the current signal waveform. The effective value of the current signal waveform is input to the comparator current 18, where it is compared with a preset value set by the setter current 19, and when the effective value of the voltage signal waveform exceeds a predetermined range, an alarm a2 of the current signal is issued. Be emitted.
[0006]
As shown in FIG. 8, the conventional semiconductor detector deterioration diagnosis apparatus measures and evaluates the absolute value of the current flowing through the semiconductor detector and the absolute value of the voltage observed at the output terminal. The determination condition is that the measured current value and voltage value are compared with predetermined values by a computer or the like.
[0007]
[Problems to be solved by the invention]
Since the conventional deterioration diagnostic device for a semiconductor detector is configured as described above, information on the output noise waveform included in the output voltage of the semiconductor detector is out of the determination condition. For this reason, there is a problem that information on a minute voltage change is not used, and the detection sensitivity due to the minute deterioration is poor.
[0008]
SUMMARY OF THE INVENTION The present invention has been made to solve the above-described problems, and has as its object to obtain a semiconductor detector deterioration diagnosis apparatus that performs automatic discrimination by patterning a detected output noise waveform change. I do.
[0009]
[Means for Solving the Problems]
The degradation diagnostic apparatus for a semiconductor detector according to the first aspect of the present invention calculates a wavelet transform using a Gabor function of an output noise waveform as a basis function based on the time-series data output from the sampling means, and uses the Gabor function as a basis. The average value of each pass center frequency of the wavelet transform as a function is calculated and compared with the average value of the output noise waveform in a normal state. This makes it possible to distinguish whether the output noise waveform observed by the semiconductor detector is normal or not. That is, the average value is an index indicating the average value of the probability density distribution function.The probability density distribution function has a fixed distribution when the output noise waveform is normal, but differs from the fixed distribution when the output noise waveform is not normal. The above distinction can be made by comparing the values.
[0010]
The deterioration diagnostic apparatus for a semiconductor detector according to the second aspect of the present invention calculates a wavelet transform using a Gabor function of an output noise waveform as a basis function based on the time series data output from the sampling means, and uses the Gabor function as a basis. The variance of each pass center frequency of the wavelet transform as a function is calculated and compared with the variance of the output noise waveform in a normal state. This makes it possible to distinguish whether the output noise waveform observed by the semiconductor detector is normal or not. That is, the variance value is an index indicating the variance value of the probability density distribution function.The probability density distribution function has a fixed distribution when the output noise waveform is normal, but differs from the fixed distribution when the output noise waveform is not normal. The above distinction can be made by comparing the values.
[0011]
According to the third aspect of the present invention, there is provided a semiconductor detector deterioration diagnosis apparatus that performs a wavelet transform using a Gabor function of an output noise waveform as a basis function based on the time series data output from the sampling means, and uses the Gabor function as a basis. The skewness of each pass center frequency of the wavelet transform as a function is calculated and compared with the skewness of the output noise waveform in a normal state. This makes it possible to distinguish whether the output noise waveform observed by the semiconductor detector is normal or not. That is, the skewness is an index indicating the skewness of the probability density distribution function.The probability density distribution function has a fixed distribution when the output noise waveform is normal, but differs from the fixed distribution when the output noise waveform is not normal. The distinction can be made by comparing the degrees.
[0012]
According to a fourth aspect of the present invention, there is provided a deterioration diagnostic apparatus for a semiconductor detector, which calculates a wavelet transform using a Gabor function of an output noise waveform as a basis function based on the time series data output from the sampling means, and uses the Gabor function as a basis. The kurtosis of each pass center frequency of the wavelet transform as a function is calculated and compared with the kurtosis of the output noise waveform in a normal state. This makes it possible to distinguish whether the output noise waveform observed by the semiconductor detector is normal or not. That is, the kurtosis is an index indicating the kurtosis of the probability density distribution function, and the probability density distribution function has a fixed distribution when the output noise waveform is normal, but differs from the fixed distribution when the output noise waveform is not normal. The above distinction can be made by comparing.
[0013]
According to a fifth aspect of the present invention, there is provided a semiconductor detector deterioration diagnostic apparatus which outputs an alarm based on a comparison result according to the first to fourth aspects of the present invention. Thus, when the semiconductor detection device is deteriorated, the abnormality can be immediately notified to the outside.
[0014]
BEST MODE FOR CARRYING OUT THE INVENTION
Embodiment 1 FIG.
Hereinafter, a first embodiment of the present invention will be described with reference to the drawings. FIG. 1 is an overall configuration diagram showing a semiconductor detector deterioration diagnosis system according to Embodiment 1 of the present invention.
[0015]
In the figure, a semiconductor detector 1 includes a semiconductor element (not shown) made of, for example, a compound semiconductor, and upper and lower electrodes (not shown) arranged above and below the semiconductor element. When incident on a semiconductor element, pairs of electrons and holes are generated in an amount proportional to the energy applied thereto. A voltage is applied to the semiconductor detector 1 by the power supply 2 via the load resistor 4 to generate an electric field in the semiconductor element, and electrons and holes of the semiconductor element flow toward the upper and lower electrodes due to the electric field. The electrons and holes flow in the semiconductor device, so that induced charges are generated in the upper and lower electrodes of the semiconductor device. A coupling capacitor 5 is connected to one end of the semiconductor detector 1. When the capacitance of the coupling capacitor 5 is sufficiently larger than the capacitance of the semiconductor element, the induced charge generated in the semiconductor detector 1 is reduced. Most are transmitted to the coupling capacitor 5 and further input to the preamplifier 20. The preamplifier 20 amplifies the input electric charge and outputs it as a voltage pulse. Since the wave height of this voltage pulse is proportional to the radiation energy applied to the semiconductor detector 1, the energy spectrum of the radiation can be measured by measuring the wave height of the voltage pulse with a measuring device (not shown). . Further, the dose rate can be measured by measuring the number of pulses per unit time using a counter (not shown).
[0016]
In the system for diagnosing deterioration of a semiconductor detector according to the present invention, an output noise waveform 21 of the semiconductor detector 1 output from the preamplifier 20 is input to a filter 22. Pass band components. Then, the output noise waveform output from the filter 22 is input to the amplifier 23, and the amplifier 23 amplifies the output noise waveform at an appropriate amplification factor. Further, the amplified output noise waveform is input to an A / D converter (sampling means A) 24, sampled at predetermined time intervals, and is converted from analog to digital, and a plurality of time-series data Xi (i = 0, 1) , 2, ..., L-1). The diagnostic device 25 receives a plurality of time-series data Xi from the A / D converter 24 and determines deterioration of the semiconductor detector 1 based on an effective value of continuous wavelet transform using a Gabor function (Gabor function) as a basis function. Diagnose, and output an alarm if any abnormality is found.
[0017]
FIG. 2 is a block diagram showing a detailed configuration of the diagnostic device 25 according to Embodiment 1 of the present invention. In the figure, reference numeral 26 denotes a wavelet transform calculator (calculating means) for calculating a wavelet transform using the Gabor function of the output noise waveform as a basis function based on the time series data Xi output from the A / D converter 24, and 27 denotes a wavelet. Basis function generators (computing means) 28, 29, and 30 that provide Gabor functions as basis functions of the transformation computing unit 26 are results output from the wavelet transformation computing unit 26 using the Gabor function as a basis function, respectively. .., J−1 passing center frequencies and 0, 1, 2,..., L−1 times. These are referred to as pass center frequency (0) 28, pass center frequency (1) 29,..., Pass center frequency (j-1) 30, respectively. Here, when calculating a convolution to be described later, the basis function of the wavelet transform can be expanded and contracted and treated as if it were a frequency. This is the same as expanding and contracting a sine wave or the like by Fourier transform and projecting it to the frequency domain. However, unlike the sine wave, the basis function of the Gabor function reacts strongly to a specific frequency. This is called a pass center frequency.
[0018]
31 is an average calculator (0) (calculating means) for calculating the average of the passing center frequency (0) 28 output from the wavelet transform calculator 26 using the Gabor function as a basis function, and 32 is a Gabor function as a basis function. An average calculator (1) (calculating means) for calculating the average of the passing center frequency (1) 29 output from the wavelet transform calculator 26, and 33 output from the wavelet transform calculator 26 using the Gabor function as a basis function. An average calculator (j-1) (calculating means) for calculating the average of the passing center frequency (j-1) 30. Note that j averaging units are set in advance.
[0019]
Numeral 34 denotes a setting unit (0) for storing the average of the passing center frequencies (0) 28 of the wavelet transform using the Gabor function as a basis function in the case of the normal output noise waveform of the semiconductor detector 1, and numeral 35 denotes an averaging unit (0). The average of the passing center frequency (0) 28 of the wavelet transform using the Gabor function of the output noise waveform calculated by 31 as the basis function, and the normal-time Gabor function stored in the setting unit (0) 34 as the basis function. The comparator (0) (comparing means) compares the average of the pass center frequencies (0) 28 of the wavelet transform.
[0020]
36 is a setting unit (1) for storing the average of the passing center frequencies (1) 29 of the wavelet transform using the Gabor function as the basis function in the case of the normal output noise waveform of the semiconductor detector 1, and 37 is an averaging unit (1) The average of the pass center frequency (1) 29 of the wavelet transform using the Gabor function of the output noise waveform calculated by the base function as the basis function, and the normal-time Gabor function stored in the setting unit (1) 36 as the basis function. The comparator (1) (comparing means) compares the average of the pass center frequencies (1) 29 of the wavelet transform.
[0021]
38 is a setter (j-1) that stores the average of the pass center frequency (j-1) 30 of the wavelet transform using the Gabor function as the basis function in the case of the normal output noise waveform of the semiconductor detector 1, and 39 is the average calculation The average of the pass center frequencies (j-1) 30 of the wavelet transform using the Gabor function of the output noise waveform calculated by the unit (j-1) 33 as a basis function, and stored in the setting unit (j-1) 38. A comparator (j-1) (comparing means) for comparing the average of the passing center frequencies (j-1) 30 of the wavelet transform using the Gabor function in a normal state as a basis function.
[0022]
In the figure, three pass center frequencies (0) 28, pass center frequencies (1) 29, and pass center frequencies (j-1) 30 are represented by three, but in actuality, j The pass center frequency is obtained. Similarly, the average calculator (0) 31, the average calculator (1) 32, the average calculator (j-1) 33, the comparator (0) 35, the comparator (1) 37, and the comparator (j-1). 39, a setting device (0) 34, a setting device (1) 36, and a setting device (j-1) 38, each of which is represented by three, but in actuality j is required.
[0023]
40 are comparators (0) 35, comparators (1) 37,. . . , Comparator (j-1) 39 is a setter M for storing an allowable pattern for a normal output noise waveform. 41 are comparators (0) 35, comparators (1) 37,. . . , A comparator M (comparing means) for comparing the comparison result of the comparator (j-1) 39 with the normal allowable pattern stored in the setting device M40, and outputs an alarm based on the comparison result of the comparator M41. Alarm generator (alarm output means).
[0024]
Next, an operation of diagnosing deterioration of the semiconductor detector according to the first embodiment of the present invention will be described. First, the A / D converter 24 samples the measured output noise waveform of the semiconductor detector 1 at predetermined time intervals, and performs analog-to-digital conversion to obtain a plurality of time-series data Xi (i = 0, 1,. , L-1).
[0025]
When a plurality of time-series data Xi are input to the diagnostic device 25, the wavelet transform calculator 26 of the diagnostic device 25 performs a wavelet process using the Gabor function as a basis function in the following steps (1) to (3). Compute the transformation.
(1) Generate a basic wavelet function.
(2) The basic wavelet function is subjected to similarity conversion for each resolution (pass center frequency) to generate a conversion basis function.
(3) Continuous wavelet transform is performed using the basis functions created in (2).
Note that the result of the wavelet transform is that of the pass center frequencies 0, 1,. . . , J−1.
[0026]
Next, the wavelet transform of the present embodiment using the Gabor function as a basis function will be described in detail.
[0027]
First, an input signal is expressed as Expression (1), and a Gabor function basic wavelet function is generated as in Expression (2).
[0028]
(Equation 1)

[0029]
(Equation 2)

[0030]
Next, the continuous wavelet basic wavelet function ψ (t), −D ≦ t <D is extended to the same number of data L (points) as the input signal X (t), and the continuous wavelet transform as shown in the equation (3) is performed. A basic wavelet function (basis function of resolution = 0) W (t) is generated.
[0031]
(Equation 3)

[0032]
Next, the above basic wavelet function (basis function with resolution = 0) W (t) is subjected to similarity conversion, and the basis function Pn (t) is calculated for each resolution as in equation (4). FIG. 3 shows the similarity transformation of these basic wavelet functions. Here, since W (t) has a value only at discrete times t, an intermediate value is obtained by linear interpolation.
[0033]
(Equation 4)

[0034]
Next, an array Pn (t) of basis functions and data X (t) arranged for each resolution n are subjected to a continuous wavelet transform (convolution calculation) at each time, so that the resolutions n, The wavelet coefficient value Yn (t) at time t is obtained. The state of these continuous wavelet transforms is shown in FIG.
[0035]
(Equation 5)

[0036]
As described above, the result of the wavelet transform using the Gabor function as the basis function is obtained when the pass center frequencies 0, 1,. . . , J−1, the average calculator (0) 31, the average calculator (1) 32, and the average calculator (j−1) 33 in the diagnostic device 25 are based on the following equation (6). Pass center frequencies 0, 1,. . . , J for each mean μ₁ ⁰, Μ₁ ¹,. . . , Μ₁ ^j-1Is calculated.
[0037]
(Equation 6)

[0038]
The average calculator (0) 31, the average calculator (1) 32, and the average calculator (j-1) 33 calculate each average μ.₁ ⁰, Μ₁ ¹,. . . , Μ₁ ^j-1Is calculated, the comparator (0) 35, the comparator (1) 37, and the comparator (j-1) 39 perform the wavelet transform using the Gabor function as a basis function when the output noise waveform is normal. Pass center frequencies 0, 1,. . . , J−1 from the setting device (0) 34, the setting device (1) 36, and the setting device (j−1) 38, and calculates the calculated average μ.₁ ⁰, Μ₁ ¹,. . . , Μ₁ ^j-1And the respective pass center frequencies 0, 1,. . . , J−1. Here, the setter (0) 34, the setter (1) 36, and the setter (j-1) 38 have a wavelet transform with a Gabor function as a basis function when the output noise waveform is normal. Pass center frequencies 0, 1,. . . , J−1 are stored, but an average in a normal state calculated by an external device (not shown) may be stored.
[0039]
The comparator (0) 35, the comparator (1) 37, and the comparator (j−1) 39 calculate the calculated average μ.₁ ⁰, Μ₁ ¹,. . . , Μ₁ ^j-1As a result of comparing the average with the normal state, for example, the calculated average μ₁ ⁰, Μ₁ ¹,. . . , Μ₁ ^j-1Is greater than three times the average of normal, it is determined to be abnormal and output to the comparator M (41).
[0040]
Here, the pass center frequencies 0, 1,... Of the wavelet transform using the Gabor function of the output noise waveform as a basis function. . . , J−1, the reason why the output noise waveform can be distinguished between normal and abnormal is briefly described. The center frequencies 0, 1,... Of the wavelet transform using the Gabor function of the output noise waveform as a basis function are described. . . , J-1 are filters for each corresponding pass center frequency, and the average of the filtered time-series data is distributed around a predetermined value in a normal case. When an abnormality occurs, the specific pass center frequency or the average of all pass center frequencies deviates from the average in the case of normal. Since this shift depends on an abnormal output noise waveform, it is possible to determine normal or abnormal by comparing the averages.
[0041]
Then, the comparator M41 determines the determination results of the comparator (0) 35, the comparator (1) 37, and the comparator (j-1) 39, and a normal / abnormal pattern for each pass center frequency which is predetermined in a normal state. The final judgment is determined by comparison. This determination is stored in the setting device M40 according to the target. For example, the simplest setting criterion is a setting in which the final judgment is abnormal if at least one pass center frequency is abnormal.
[0042]
Finally, when the comparator M41 determines that an abnormality has occurred, the alarm generator (alarm generation means) 42 displays an abnormality on a display device (not shown) to clarify that an abnormal output noise waveform has been generated. Is displayed, or a signal or the like indicating that an abnormality has occurred is output to a monitoring device (not shown) or the like that monitors the plant, and the series of processes is terminated.
[0043]
As described above, according to the first embodiment of the present invention, the pass center frequencies 0, 1,... Of the wavelet transform using the Gabor function of the output noise waveform, which can be an index for determining normality / abnormality, as a basis function. . . , J−1, the output noise waveform is determined to be normal or abnormal, so that the deterioration of the semiconductor detector can be determined in a shorter time and more accurately than in the conventional case.
[0044]
Embodiment 2 FIG.
Hereinafter, a second embodiment of the present invention will be described with reference to the drawings. FIG. 1 is an overall view showing a deterioration diagnostic system for a semiconductor detector according to a second embodiment of the present invention.
[0045]
FIG. 5 is a block diagram showing a detailed configuration of the diagnostic device 25 according to Embodiment 2 of the present invention. In the figure, reference numeral 26 denotes a wavelet transform calculator (calculating means) for calculating a wavelet transform using a Gabor function of an output noise waveform as a basis function based on the time series data Xi output from the A / D converter 24, and 27 denotes a wavelet. Basis function generators (computing means) 28, 29, and 30 that provide Gabor functions as basis functions of the transformation computing unit 26 are results output from the wavelet transformation computing unit 26 using the Gabor function as a basis function, respectively. 0, 1, 2,. . . , J−1 pass center frequencies and 0, 1, 2,. . . , L−1. These are referred to as a pass center frequency (0) 28, a pass center frequency (1) 29, and a pass center frequency (j-1) 30, respectively.
[0046]
Reference numeral 310 denotes a variance calculator (0) (calculation means) for calculating the variance of the pass center frequency (0) 28 output from the wavelet transform calculator 26 using the Gabor function as a basis function, and 320 denotes a Gabor function as a basis function. A variance calculator (1) (calculation means) for calculating the variance of the pass center frequency (1) 29 output from the wavelet transform calculator 26, 330 is output from the wavelet transform calculator 26 using the Gabor function as a basis function. It is a dispersion calculator (j-1) (calculation means) for calculating the variance of the passing center frequency (j-1) 30. Note that j distributed arithmetic units are set in advance.
[0047]
340 is a setter (0) for storing the variance of the pass center frequency (0) 28 of the wavelet transform using the Gabor function as the basis function in the case of the normal output noise waveform of the semiconductor detector 1, and 350 is the variance calculator (0) The variance of the passing center frequency (0) 28 of the wavelet transform using the Gabor function of the output noise waveform calculated by 310 as the basis function, and the normal-time Gabor function stored in the setting unit (0) 340 as the basis function. The comparator (0) (comparing means) compares the variance of the pass center frequency (0) 28 of the wavelet transform.
[0048]
360 is a setter (1) that stores the variance of the pass center frequency (1) 29 of the wavelet transform using the Gabor function as a basis function in the case of the normal output noise waveform of the semiconductor detector 1, and 370 is a variance calculator (1) The variance of the passing center frequency (1) 29 of the wavelet transform using the Gabor function of the output noise waveform calculated by 320 as the basis function, and the normal-time Gabor function stored in the setting unit (1) 360 as the basis function. A comparator (1) (comparing means) for comparing the variance of the pass center frequency (1) 29 of the wavelet transform.
[0049]
380 is a setting unit (j-1) for storing the variance of the pass center frequency (j-1) 30 of the wavelet transform using the Gabor function as the basis function in the case of the normal output noise waveform of the semiconductor detector 1, and 39 is the variance operation Variance of the pass center frequency (j-1) 30 of the wavelet transform using the Gabor function of the output noise waveform calculated by the unit (j-1) 330 as a basis function, and stored in the setting unit (j-1) 380. This is a comparator (j-1) (comparing means) for comparing the variance of the pass center frequency (j-1) 30 of the wavelet transform using the Gabor function at the normal time as a basis function.
[0050]
In the figure, three pass center frequencies (0) 28, pass center frequencies (1) 29, and pass center frequencies (j-1) 30 are represented by three, but in actuality, j The pass center frequency is obtained. Similarly, distributed computing unit (0) 310, distributed computing unit (1) 320, distributed computing unit (j-1) 330, comparator (0) 350, comparator (1) 370, comparator (j-1) Although 390, setting device (0) 340, setting device (1) 360, and setting device (j-1) 380 are represented by three each, in actuality j are required respectively.
[0051]
400 is a comparator (0) 350, a comparator (1) 370,. . . , Comparator (j-1) 390 is a setter M for storing an allowable pattern for a normal output noise waveform. 410 are comparators (0) 350, comparators (1) 370,. . . , A comparator (comparing means) 420 for comparing the comparison result of the comparator (j-1) 390 with the allowable pattern stored in the setting unit M400 in the normal state, and outputs an alarm based on the comparison result of the comparator M410. Alarm generator (alarm output means).
[0052]
Next, the operation of diagnosing deterioration of a semiconductor detector according to the second embodiment of the present invention will be described. First, the A / D converter 24 samples the measured output noise waveform of the semiconductor detector 1 at predetermined time intervals, and performs analog-to-digital conversion to obtain a plurality of time-series data Xi (i = 0, 1,. , L-1).
[0053]
When a plurality of time-series data Xi are input to the diagnostic device 25, the wavelet transform calculator 26 of the diagnostic device 25 converts the Gabor function into the basis function by the same method as that described in the first embodiment. Calculate the wavelet transform to be performed.
[0054]
The result of the wavelet transform using the Gabor function as a basis function is the passing center frequencies 1, 2,. . . , J, the distributed computing unit (0) 310, the distributed computing unit (1) 320, and the distributed computing unit (j-1) 330 in the diagnostic device 25 determine the passing center based on the following equation (7). The frequencies 0, 1,. . . , J−1₂ ⁰, Μ₂ ¹,. . . , Μ₂ ^j-1Is calculated.
[0055]
(Equation 7)

[0056]
Then, the distributed arithmetic unit (0) 310, the distributed arithmetic unit (1) 320, and the distributed arithmetic unit (j-1) 330₂ ⁰, Μ₂ ¹,. . . , Μ₂ ^j-1Is calculated, the comparator (0) 350, the comparator (1) 370, and the comparator (j-1) 390 perform the wavelet transform using the Gabor function as a basis function when the output noise waveform is normal. Pass center frequencies 0, 1,. . . , J−1 are input from the setter (0) 340, the setter (1) 360, and the setter (j−1) 380, and the calculated variance μ is input.₂ ⁰, Μ₂ ¹,. . . , Μ₂ ^j-1And the respective pass center frequencies 0, 1,. . . , J−1. Here, the setter (0) 340, the setter (1) 360, and the setter (j-1) 380 have a wavelet transform with a Gabor function as a basis function when the output noise waveform is normal. Pass center frequencies 0, 1,. . . , J−1 are stored, but the normal distribution calculated by an external device (not shown) may be stored.
[0057]
The comparator (0) 350, the comparator (1) 370, and the comparator (j−1) 390 calculate the calculated variance μ.₂ ⁰, Μ₂ ¹,. . . , Μ₂ ^j-1And the variance in the normal state, for example, the calculated variance μ₂ ⁰, Μ₂ ¹,. . . , Μ₂ ^j-1Is larger than three times the normal dispersion, it is determined to be abnormal and output to the comparator M410.
[0058]
Here, the pass center frequencies 0, 1,... Of the wavelet transform using the Gabor function of the output noise waveform as a basis function. . . , J−1, the reason why the output noise waveform can be distinguished between normal and abnormal is briefly described below. The center frequencies 0, 1,. . . , J-1 are filters for each corresponding pass center frequency, and the variance of the filtered time-series data is distributed around a certain value in a normal case. When an abnormality occurs, the variance of a specific pass center frequency or all the pass center frequencies deviates from the variance in a normal case. Since this shift depends on an abnormal output noise waveform, it is possible to determine normal or abnormal by comparing the variances. In particular, since the variance is an index indicating the variation of the probability density distribution function, it can be remarkably detected when the abnormality appears in the variation of the probability density distribution function.
[0059]
Then, the comparator M410 outputs the judgment results of the comparator (0) 350, the comparator (1) 370, and the comparator (j-1) 390 and the normal / abnormal pattern for each pass center frequency which is predetermined in the normal state. The final judgment is determined by comparison. This determination is stored in the setting device M400 according to the target. For example, the simplest setting criterion is a setting in which the final judgment is abnormal if at least one pass center frequency is abnormal.
[0060]
Finally, when the comparator M410 determines that an abnormality has occurred, the alarm generator (alarm generation means) 420 displays an abnormality on a display device (not shown) in order to clarify that an abnormal output noise waveform has been generated. Is displayed, or a signal or the like indicating that an abnormality has occurred is output to a monitoring device (not shown) or the like that monitors the plant, and the series of processes is terminated.
[0061]
As described above, according to the second embodiment of the present invention, the pass center frequencies 0, 1,... Of the wavelet transform using the Gabor function of the output noise waveform, which can be an index for determining normality / abnormality, as a basis function. . . , J−1, the output noise waveform is determined to be normal or abnormal, so that the deterioration of the semiconductor detector can be determined in a shorter time and more accurately than in the conventional case.
[0062]
Embodiment 3 FIG.
Hereinafter, a third embodiment of the present invention will be described with reference to the drawings. FIG. 1 is an overall view showing a deterioration detection system for a semiconductor detector according to a third embodiment of the present invention. The configuration is the same as that of the first embodiment and will not be described.
[0063]
FIG. 6 is a block diagram showing a detailed configuration of the diagnostic device 25 according to Embodiment 3 of the present invention. In the figure, reference numeral 26 denotes a wavelet transform calculator (calculating means) for calculating a wavelet transform using a Gabor function of an output noise waveform as a basis function based on the time series data Xi output from the A / D converter 24, and 27 denotes a wavelet. Basis function generators (computing means) 28, 29, and 30 that provide Gabor functions as basis functions of the transformation computing unit 26 are results output from the wavelet transformation computing unit 26 using the Gabor function as a basis function, respectively. 0, 1, 2,. . . , J−1 pass center frequencies and 0, 1, 2,. . . , L−1. These are referred to as a pass center frequency (0) 28, a pass center frequency (1) 29, and a pass center frequency (j-1) 30, respectively.
[0064]
Reference numeral 311 denotes a skewness calculator (0) (calculation means) for calculating the skewness of the pass center frequency (0) 28 output from the wavelet transform calculator 26 using the Gabor function as a basis function, and 321 denotes a Gabor function as a basis function. A skewness calculator (1) (calculating means) for calculating the skewness of the pass center frequency (1) 29 output from the wavelet transform calculator 26, and a wavelet transform calculator 26 using a Gabor function as a basis function Is a skewness calculator (j-1) (arithmetic means) for calculating the skewness of the pass center frequency (j-1) 30 output from. Note that j skewness calculators are set in advance.
[0065]
341 is a setter (0) that stores the skewness of the pass center frequency (0) 28 of the wavelet transform using the Gabor function as a basis function in the case of the normal output noise waveform of the semiconductor detector 1, and 351 is a skewness calculator ( 0) The skewness of the pass center frequency (0) 28 of the wavelet transform using the Gabor function of the output noise waveform calculated by 311 as the basis function, and the Gabor function in the normal state stored in the setting unit (0) 341 The comparator (0) (comparing means) compares the skewness of the pass center frequency (0) 28 of the wavelet transform as a function.
[0066]
361 is a setting unit (1) for storing the skewness of the pass center frequency (1) 29 of the wavelet transform using the Gabor function as a basis function in the case of the normal output noise waveform of the semiconductor detector 1, and 371 is a skewness calculator ( 1) The skewness of the pass center frequency (1) 29 of the wavelet transform using the Gabor function of the output noise waveform calculated by 321 as a basis function, and the Gabor function at normal time stored in the setting unit (1) 361 The comparator (1) (comparing means) compares the skewness of the pass center frequency (1) 29 of the wavelet transform as a function.
[0067]
381 is a setter (j-1) that stores the skewness of the pass center frequency (j-1) 30 of the wavelet transform using the Gabor function as the basis function in the case of the normal output noise waveform of the semiconductor detector 1; The skewness of the pass center frequency (j-1) 30 of the wavelet transform using the Gabor function of the output noise waveform calculated by the degree calculator (j-1) 331 as a basis function and the setter (j-1) 381 A comparator (j-1) (comparing means) for comparing the skewness of the pass center frequency (j-1) 30 of the wavelet transform using the stored normal Gabor function as a basis function.
[0068]
In the figure, three pass center frequencies (0) 28, pass center frequencies (1) 29, and pass center frequencies (j-1) 30 are represented by three, but in actuality, j The pass center frequency is obtained. Similarly, the skewness calculator (0) 311, the skewness calculator (1) 321, the skewness calculator (j-1) 331, the comparator (0) 351, the comparator (1) 371, and the comparator (j -1) 391, setting device (0) 341, setting device (1) 361, setting device (j-1) 381, each of which is represented by three. It is.
[0069]
Reference numeral 401 denotes a comparator (0) 351, a comparator (1) 371,. . . , Comparator (j-1) 391 is a setting unit M for storing an allowable pattern in the case of a normal output noise waveform. 411 are comparators (0) 351, comparators (1) 371,. . . , A comparator M (comparing means) for comparing the comparison result of the comparator (j-1) 391 with the normal allowable pattern stored in the setting unit M401, and 421 outputs an alarm based on the comparison result of the comparator M411. Alarm generator (alarm output means).
[0070]
Next, an operation of diagnosing deterioration of a semiconductor detector according to Embodiment 3 of the present invention will be described. First, the A / D converter 24 samples the measured output noise waveform of the semiconductor detector 1 at predetermined time intervals, and performs analog-to-digital conversion to obtain a plurality of time-series data Xi (i = 0, 1,. , L-1).
[0071]
When a plurality of time-series data Xi are input to the diagnostic device 25, the wavelet transform calculator 26 of the diagnostic device 25 converts the Gabor function into the basis function by the same method as that described in the first embodiment. Calculate the wavelet transform to be performed.
[0072]
The result of the wavelet transform using the Gabor function as a basis function is the passing center frequencies 0, 1,. . . , J−1, the skewness calculator (0) 311, the skewness calculator (1) 321 and the skewness calculator (j−1) 331 in the diagnostic device 25 are expressed by the following equation (8). Based on the passing center frequencies 0, 1,. . . , J−1 skewness μ₃ ⁰, Μ₃ ¹,. . . , Μ₃ ^j-1Is calculated. In equation (8), the skewness μ₃Is the average μ₁And the variance μ₂This skewness is based on the cube of the deviation of each numerical value from the arithmetic mean, and is often used for processing statistical data.
[0073]
(Equation 8)

[0074]
Then, the skewness calculator (0) 311, the skewness calculator (1) 321 and the skewness calculator (j-1) 331 calculate each skewness μ.₃ ⁰, Μ₃ ¹,. . . , Μ₃ ^j-1Is calculated, the comparator (0) 351, the comparator (1) 371, and the comparator (j-1) 391 perform the wavelet transform using the Gabor function as a basis function when the output noise waveform is normal. Pass center frequencies 0, 1,. . . , J−1 are input from the setting device (0) 341, the setting device (1) 361, and the setting device (j−1) 381, and the calculated skewness μ₃ ⁰, Μ₃ ¹,. . . , Μ₃ ^j-1And the respective pass center frequencies 0, 1,. . . , J−1. Here, the setter (0) 341, the setter (1) 361, and the setter (j-1) 381 have a wavelet transform with a Gabor function as a basis function when the output noise waveform is normal. Pass center frequencies 0, 1,. . . , J−1 are stored, but the normal state skewness calculated by an external device (not shown) may be stored.
[0075]
The comparator (0) 351, the comparator (1) 371, and the comparator (j−1) 391 calculate the calculated skewness μ.₃ ⁰, Μ₃ ¹,. . . , Μ₃ ^j-1As a result of comparing the skewness with the normal state, for example, the calculated skewness μ₃ ⁰, Μ₃ ¹,. . . , Μ₃ ^j-1Exceeds three times the normal skewness, it is determined to be abnormal and output to the comparator M411.
[0076]
Here, the pass center frequencies 0, 1,... Of the wavelet transform using the Gabor function of the output noise waveform as a basis function. . . , J−1, the reason why the output noise waveform can be distinguished between normal and abnormal is briefly described below. The center frequencies 0, 1,. . . , J-1 are filters for each corresponding pass center frequency, and the skewness of the filtered time-series data is distributed around a certain value in a normal case. When an abnormality occurs, the skewness of a specific pass center frequency or all pass center frequencies deviates from the skewness in the case where the skew is normal. Since this shift depends on an abnormal output noise waveform, normality / abnormality can be determined by comparing the skewness. In particular, since the skewness is an index indicating the bias of the probability density distribution function, it can be remarkably detected when the abnormality appears in the bias of the probability density distribution function.
[0077]
Then, the comparator M411 determines the determination results of the comparator (0) 351, the comparator (1) 371, and the comparator (j-1) 391 and a normal / abnormal pattern for each passing center frequency which is predetermined in a normal state. The final judgment is determined by comparison. This determination is stored in the setting device M401 according to the target. For example, the simplest setting criterion is a setting in which the final judgment is abnormal if at least one pass center frequency is abnormal.
[0078]
Finally, when the comparator M411 determines that there is an abnormality, the alarm generator (alarm generation means) 421 displays an abnormality on a display device (not shown) in order to clarify that an abnormal output noise waveform has been generated. Is displayed, or a signal or the like indicating that an abnormality has occurred is output to a monitoring device (not shown) or the like that monitors the plant, and the series of processes is terminated.
[0079]
As described above, according to the third embodiment of the present invention, the pass center frequencies 0, 1,... Of the wavelet transform using the Gabor function of the output noise waveform, which can be an index for determining normality / abnormality, as a basis function. . . , J−1, the output noise waveform is determined to be normal or abnormal, so that the deterioration of the semiconductor detector can be determined in a shorter time and more accurately than in the conventional case.
[0080]
Embodiment 4 FIG.
Hereinafter, a fourth embodiment of the present invention will be described with reference to the drawings. FIG. 1 is a general view showing a semiconductor detector deterioration diagnosis system according to Embodiment 4 of the present invention. The configuration is the same as that of Embodiment 1 and will not be described.
[0081]
FIG. 7 is a block diagram showing a detailed configuration of a diagnostic device 25 according to Embodiment 4 of the present invention. In the figure, reference numeral 26 denotes a wavelet transform calculator (calculating means) for calculating a wavelet transform using a Gabor function of an output noise waveform as a basis function based on the time series data Xi output from the A / D converter 24, and 27 denotes a wavelet. Basis function generators (computing means) 28, 29, and 30 that provide Gabor functions as basis functions of the transformation computing unit 26 are results output from the wavelet transformation computing unit 26 using the Gabor function as a basis function, respectively. 0, 1, 2,. . . , J−1 and 0, 1, 2,. . . , L−1. These are referred to as a pass center frequency (0) 28, a pass center frequency (1) 29, and a pass center frequency (j-1) 30, respectively.
[0082]
Reference numeral 312 denotes a kurtosis calculator (0) (calculating means) for calculating the kurtosis of the passing center frequency (0) 28 output from the wavelet transform calculator 26 using the Gabor function as a basis function. A kurtosis calculator (1) (calculating means) for calculating the kurtosis of the passing center frequency (1) 29 output from the wavelet transform calculator 26, and a wavelet transform calculator 26 using a Gabor function as a basis function Is a kurtosis calculator (j-1) (calculation means) for calculating the kurtosis of the pass center frequency (j-1) 30 output from. Note that j kurtosis calculators are set in advance.
[0083]
342 is a setter (0) that stores the kurtosis of the pass center frequency (0) 28 of the wavelet transform using the Gabor function as a basis function in the case of the normal output noise waveform of the semiconductor detector 1, and 352 is a kurtosis calculator ( 0) The kurtosis of the pass center frequency (0) 28 of the wavelet transform using the Gabor function of the output noise waveform calculated by 312 as the basis function, and the normal-time Gabor function stored in the setting unit (0) 342 A comparator (0) (comparing means) for comparing the kurtosis of the pass center frequency (0) 28 of the wavelet transform as a function.
[0084]
362 is a setter (1) that stores the kurtosis of the pass center frequency (1) 29 of the wavelet transform using the Gabor function as a basis function in the case of the normal output noise waveform of the semiconductor detector 1, and 372 is a kurtosis calculator ( 1) The kurtosis of the pass center frequency (1) 29 of the wavelet transform using the Gabor function of the output noise waveform calculated by 322 as a basis function, and the Gabor function in the normal state stored in the setting unit (1) 362 It is a comparator (1) (comparing means) for comparing the kurtosis of the pass center frequency (1) 29 of the wavelet transform as a function.
[0085]
382 is a setter (j-1) that stores the kurtosis of the passing center frequency (j-1) 30 of the wavelet transform using the Gabor function as the basis function in the case of the normal output noise waveform of the semiconductor detector 1; The kurtosis of the pass center frequency (j-1) 30 of the wavelet transform using the Gabor function of the output noise waveform calculated by the degree calculator (j-1) 332 as a basis function and the setter (j-1) 382 The comparator (j-1) (comparing means) compares the kurtosis of the pass center frequency (j-1) 30 of the wavelet transform using the stored normal Gabor function as a basis function.
[0086]
In the figure, three pass center frequencies (0) 28, pass center frequencies (1) 29, and pass center frequencies (j-1) 30 are represented by three, but in actuality, j The pass center frequency is obtained. Similarly, a kurtosis calculator (0) 312, a kurtosis calculator (1) 322, a kurtosis calculator (j-1) 332, a comparator (0) 352, a comparator (1) 372, and a comparator (j -1) 392, setting device (0) 342, setting device (1) 362, setting device (j-1) 382, and so on. It is.
[0087]
402 is a comparator (0) 352, a comparator (1) 372,. . . , Comparator (j−1) 392 is a setter M that stores an allowable pattern in the case of a normal output noise waveform. 412 are comparators (0) 352, comparators (1) 372,. . . , A comparator M (comparing means) for comparing the comparison result of the comparator (j-1) 392 with the allowable pattern stored in the setting unit M402 in the normal state, and outputs an alarm based on the comparison result of the comparator M412. Alarm generator (alarm output means).
[0088]
Next, an operation of diagnosing deterioration of a semiconductor detector according to Embodiment 4 of the present invention will be described. First, the A / D converter 24 samples the measured output noise waveform of the semiconductor detector 1 at predetermined time intervals and performs analog-to-digital conversion to obtain a plurality of time-series data Xi (i = 0, 1, 2, 2). , L-1) are output.
[0089]
When a plurality of time-series data Xi are input to the diagnostic device 25, the wavelet transform calculator 26 of the diagnostic device 25 converts the Gabor function into the basis function by the same method as that described in the first embodiment. Calculate the wavelet transform to be performed.
[0090]
The result of the wavelet transform using the Gabor function as a basis function is the passing center frequencies 0, 1,. . . , J−1, the kurtosis calculator (0) 312, the kurtosis calculator (1) 322, and the kurtosis calculator (j−1) 332 in the diagnostic device 25 are calculated by the following equation (9). Based on the passing center frequencies 0, 1,. . . , J−1 of each kurtosis μ₄ ⁰, Μ₄ ¹,. . . , Μ₄ ^j-1Is calculated.
[0091]
(Equation 9)

[0092]
Then, the kurtosis calculator (0) 312, the kurtosis calculator (1) 322, and the kurtosis calculator (j-1) 332 calculate each kurtosis μ.₄ ⁰, Μ₄ ¹,. . . , Μ₄ ^j-1Is calculated, the comparator (0) 352, the comparator (1) 372, and the comparator (j-1) 392 perform the wavelet transform using the Gabor function as a basis function when the output noise waveform is normal. Pass center frequencies 0, 1,. . . , J-1 are input from the setting device (0) 342, the setting device (1) 362, and the setting device (j-1) 382, and the calculated kurtosis μ₄ ⁰, Μ₄ ¹,. . . , Μ₄ ^j-1And the respective pass center frequencies 0, 1,. . . , J−1. Here, the setter (0) 342, the setter (1) 362, and the setter (j-1) 382 have a wavelet transform with a Gabor function as a basis function when the output noise waveform is normal. Pass center frequencies 0, 1,. . . , J−1 are stored, but the kurtosis in a normal state calculated by an external device (not shown) may be stored.
[0093]
Then, the comparator (0) 352, the comparator (1) 372, and the comparator (j-1) 392 calculate the kurtosis μ₄ ⁰, Μ₄ ¹,. . . , Μ₄ ^j-1And the kurtosis of the normal state, for example, the calculated kurtosis μ₄ ⁰, Μ₄ ¹,. . . , Μ₄ ^j-1Exceeds three times the normal kurtosis, it is determined to be abnormal and output to the comparator M412.
[0094]
Here, the pass center frequencies 0, 1,... Of the wavelet transform using the Gabor function of the output noise waveform as a basis function. . . , J−1, the reason why the output noise waveform can be distinguished between normal and abnormal is briefly described as follows: Pass center frequencies 0, 1,... Of the wavelet transform using the Gabor function of the output noise waveform as a basis function. . . , J-1 are filters for each corresponding pass center frequency, and the kurtosis of the filtered time-series data is distributed around a certain value in a normal case. When an abnormality occurs, the kurtosis of a specific pass center frequency or all the pass center frequencies deviates from the kurtosis in a normal case. Since this deviation depends on an abnormal output noise waveform, normality / abnormality can be determined by comparing the kurtosis. In particular, since the kurtosis is an index indicating the degree of spread of the probability density distribution function, it can be remarkably detected when an abnormality appears in the spread of the probability density distribution function.
[0095]
Then, the comparator M412 determines the determination results of the comparator (0) 352, the comparator (1) 372, and the comparator (j-1) 392, and a normal / abnormal pattern for each passing center frequency that is predetermined in a normal state. The final judgment is determined by comparison. This determination is stored in the setting device M402 according to the target. For example, the simplest setting criterion is a setting in which the final judgment is abnormal if at least one pass center frequency is abnormal.
[0096]
Finally, when the comparator M412 determines that there is an abnormality, the alarm generator (alarm generation means) 422 displays an abnormality on a display device (not shown) in order to clarify that an abnormal output noise waveform has been generated. Is displayed, or a signal or the like indicating that an abnormality has occurred is output to a monitoring device (not shown) or the like that monitors the plant, and the series of processes is terminated.
[0097]
As described above, according to the fourth embodiment of the present invention, the passing center frequencies 0, 1,... Of the wavelet transform using the Gabor function of the output noise waveform, which can be an index for determining normality / abnormality, as a basis function. . . , J−1, the normality / abnormality of the output noise waveform is determined, so that the deterioration of the semiconductor detector can be determined in a shorter time and more accurately than in the conventional case.
[0098]
【The invention's effect】
As described above, according to the present invention, the following effects can be achieved.
[0099]
According to the first aspect of the present invention, the pass center frequencies 0, 1,... Of the wavelet transform using the Gabor function of the output noise waveform based on the time series data output from the sampling means as the basis function. . . , J-1 are calculated and compared with the average value of the output noise waveform in the normal state, so that the situation can be determined at a higher speed than in the conventional case, and the sensitivity of human beings can be determined immediately. It is now possible to make a diagnosis. In particular, by using the average value, the calculation time can be reduced.
[0100]
According to the invention of claim 2, the pass center frequencies 0, 1,... Of the wavelet transform using the Gabor function of the output noise waveform based on the time-series data output from the sampling means as a basis function. . . , J-1 are calculated and compared with the variance of the output noise waveform in the normal state, so that the situation can be judged at a higher speed than in the conventional case, and the sensitivity of human beings can be immediately determined. It is now possible to make a diagnosis. In particular, by using a variance value, the sensitivity of a phenomenon in which an abnormality appears in the variation can be improved.
[0101]
According to the third aspect of the present invention, the pass center frequencies 0, 1,... Of the wavelet transform using the Gabor function of the output noise waveform based on the time series data output from the sampling means as a basis function. . . , J-1 is calculated and compared with the skewness of the output noise waveform in the normal state, so that the situation can be determined at a higher speed than in the conventional case, and the sensitivity of human beings can be instantly determined. It is now possible to make a diagnosis. In particular, by adopting the skewness, it is possible to improve the sensitivity of the phenomenon in which abnormalities appear in the bias.
[0102]
According to the invention of claim 4, the passing center frequencies 0, 1,... Of the wavelet transform using the Gabor function of the output noise waveform based on the time series data output from the sampling means as a basis function. . . , J-1 are calculated and compared with the normal kurtosis of the output noise waveform, so that the situation can be determined at a higher speed as compared with the conventional one, and the sensitivity of human beings can be immediately determined. It is now possible to make a diagnosis. In particular, by adopting the kurtosis, the sensitivity of a phenomenon in which an abnormality appears in the spread can be improved.
[0103]
According to the fifth aspect of the present invention, the alarm is output based on the comparison result of the first to fourth aspects of the present invention. Therefore, when the semiconductor detection device is deteriorated, the abnormality can be immediately notified to the outside. it can.
[Brief description of the drawings]
FIG. 1 is an overall configuration diagram showing a system for diagnosing deterioration of a semiconductor detector according to the present invention.
FIG. 2 is a block diagram showing a detailed configuration of a diagnostic device according to the first embodiment of the present invention.
FIG. 3 is a diagram showing a state of similarity conversion of a basic wavelet function according to the present invention.
FIG. 4 is a diagram illustrating a state of a continuous wavelet transform according to the present invention.
FIG. 5 is a block diagram showing a detailed configuration of a diagnostic device according to a second embodiment of the present invention.
FIG. 6 is a block diagram showing a detailed configuration of a diagnostic device according to Embodiment 3 of the present invention.
FIG. 7 is a block diagram showing a detailed configuration of a diagnostic device according to Embodiment 4 of the present invention.
FIG. 8 is a configuration diagram illustrating a conventional method of diagnosing a semiconductor detector.
[Explanation of symbols]
1 semiconductor detector, 2 battery, 4 load resistance, 5 coupling capacitor, 20 preamplifier, 21 output noise waveform, 22 filter, 23 amplifier, 24 A / D converter, 25 diagnostic device, 26 wavelet transform calculator, 27 Basis function, 28 ° pass center frequency (0), 29 ° pass center frequency (1), 30 ° pass center frequency (j−1), 31 ° average operator (0), 32 ° average operator (1), 33 ° average operator ( j-1), 34, 340, 341, 342 {setter (0), 35, 350, 351, 352} comparator (0), 36, 360, 361, 362 {setter (1), 37, 370, 371, 372 {comparator (1), 38, 380, 381, 382} setter (j-1), 39, 390, 391, 392} comparator (j-1), 40, 400 401, 402 {setting device M, 41, 410, 411, 412} comparator M, 42, 420, 421, 422 {alarm generator, 310} distributed computing device (0), 320} distributed computing device (1), 330} distributed computing device ( j-1), 311 skewness calculator (0), 321 skewness calculator (1), 331 skewness calculator (j-1), 312 kurtosis calculator (0), 322 kurtosis calculator (1 ), 332 ° kurtosis calculator (j−1).

Claims (5)

Sampling means for sampling the output noise waveform signal of the semiconductor detector every predetermined time and outputting a plurality of time-series data, and wavelet transform using a Gabor function of the output noise waveform output from the sampling means as a basis function. A wavelet transform calculating means for calculating, an average calculating means for calculating an average value for each pass center frequency calculated by the wavelet transform using the Gabor function as a basis function, and a pass center frequency calculated by the average calculating means First comparing means for comparing the average value of each of the noise levels with the average value of each of the passing center frequencies when the output noise waveform of the semiconductor detector is normal, and the output noise waveform based on the comparison result of the comparing means. A degradation diagnostic device for a semiconductor detector, comprising: a second comparing unit that outputs a state. Sampling means for sampling the output noise waveform signal of the semiconductor detector every predetermined time and outputting a plurality of time-series data, and wavelet transform using a Gabor function of the output noise waveform output from the sampling means as a basis function. A wavelet transform calculating means for calculating, a dispersion calculating means for calculating a variance value for each pass center frequency calculated by the wavelet transform using the Gabor function as a basis function, and a pass center frequency calculated by the spread calculating means Comparing means for comparing the variance value of each of the semiconductor detectors with the variance value of each pass center frequency when the output noise waveform of the semiconductor detector is normal; and outputting the output based on the comparison result of the first comparing means. A degradation diagnostic device for a semiconductor detector, comprising: a second comparing unit that outputs a state of a noise waveform. Sampling means for sampling the output noise waveform signal of the semiconductor detector every predetermined time and outputting a plurality of time-series data, and wavelet transform using a Gabor function of the output noise waveform output from the sampling means as a basis function. A wavelet transform calculating means for calculating, a skewness calculating means for calculating a skewness for each pass center frequency calculated by the wavelet transform using the Gabor function as a basis function, and a skewness calculated for each skewness calculating means. First comparing means for comparing the skewness for each center frequency with the skewness for each pass center frequency when the output noise waveform of the semiconductor detector is normal; and a comparison result of the first comparing means. And a second comparing means for outputting the state of the output noise waveform. Sampling means for sampling the output noise waveform signal of the semiconductor detector every predetermined time and outputting a plurality of time-series data, and wavelet transform using a Gabor function of the output noise waveform output from the sampling means as a basis function. A wavelet transform calculating means for calculating, a kurtosis calculating means for calculating a kurtosis for each pass center frequency calculated by the wavelet transform using the Gabor function as a basis function, and a pass kurtosis calculated by the kurtosis calculating means. First comparing means for comparing the kurtosis for each center frequency and the kurtosis for each passing center frequency when the output noise waveform of the semiconductor detector is normal; and a comparison result of the first comparing means. And a second comparing means for outputting the state of the output noise waveform. 5. The apparatus according to claim 1, further comprising: an alarm output unit configured to output an alarm based on a comparison result of the second comparison unit. 6.

Images