JP3651693B2 - Plant monitoring diagnosis apparatus and method - Google Patents

Description

の最大値を類似度とし、一定の値以上の高い類似度を与える事例の異常原因を現在の異常原因の候補として出力することを特徴とするプラント監視診断装置を提供する。
【００３７】
また、本発明は、上述の目的を達成するために、請求項７に記載したように、因果表には、プラントに対して人為的、機械的に行われる計画的操作に対して想定される各監視指標の徴候パターンを含むことを特徴とする請求項１または２記載のプラント監視診断装置を提供する。
【００３８】
また、本発明は、上述の目的を達成するために、請求項８に記載したように、事例データには、プラントに対して人為的、機械的に行われる計画的操作が行われた事例データを含むことを特徴とする請求項６記載のプラント監視診断装置を提供する。
【００３９】
また、本発明は、上述の目的を達成するために、請求項９に記載したように、プラントで観測されるプロセス信号より算出される監視指標の正常範囲からの逸脱を異常徴候として検出するに際し、プラントに対して人為的、機械的に行われる計画的操作毎にその開始と終了に伴い生じる監視指標の変化と、操作により最初に変化するアナログ監視指標Ｘとその影響を受けて変化する可能性のあるその他のアナログ監視指標｛Ｙ i ｝と、変化の現れる期間を予め登録しておき、操作の開始に伴う変化を検出した時点から一定期間あるいは終了に伴う変化を検出するまでの間、前記アナログ監視指標｛Ｙ i ｝については相対変化率｛Ｙ i ／Ｘ｝を新たな指標としてその正常範囲からの逸脱を異常徴候として検出することを特徴とするプラント監視診断方法を提供する。
【００４０】
また、本発明は、上述の目的を達成するために、請求項１０に記載したように、請求項１，２または６記載のプラント監視診断装置において、異常伝播経路同定部では、監視指標間の影響伝播特性にフィードバック制御によって一定に保持される関係を導入し、制御系のフィードバック効果によって過渡変化が抑制されたことにより、監視指標の観測値に影響が見かけ上現れない場合でも、伝播経路が成立したと見なすことを特徴とするプラント監視診断装置を提供する。
【００４１】
また、本発明は、上述の目的を達成するために、請求項１１に記載したように、請求項１，２または６記載のプラント監視診断装置において、異常伝播経路同定部では、監視指標間の影響伝播特性に時間遅れを与えておき、監視処理部において原因側の監視指標に異常徴候が検出されていて結果側の監視指標には検出されていない場合、この時間遅れを基に結果側の監視指標の変化の特徴を抽出する時間範囲と正常値の範囲を限定して異常徴候を再評価した後、異常伝播経路を同定することを特徴とするプラント監視診断装置を提供する。
【００４２】
また、本発明は、上述の目的を達成するために、請求項１２に記載したように、請求項１，２または６記載のプラント監視診断装置において、異常伝播経路同定部では、監視指標間の影響伝播特性に時間遅れを与えておき、同定された伝播経路が閉ループを構成した場合には最初に異常徴候の検出された閉ループ内の監視指標を当該閉ループの起点とし、２つの監視指標間に異なる伝播経路が見かけ上成立した場合には原因側監視指標の異常徴候検出時刻に伝播時間遅れを加えた時刻を結果側の監視指標の異常徴候検出時刻と比較することにより真の伝播経路を推定することを特徴とするプラント監視診断装置を提供する。
【００４３】
また、本発明は、上述の目的を達成するために、請求項１３に記載したように、請求項１，２または６記載のプラント監視診断装置において、異常伝播経路同定部では、プラントの異常に対して設けられている保護シーケンスの作動が監視指標に与える影響伝播特性を定性モデルに含めておき、シーケンスの作動を検出した時点から観測される監視指標の異常徴候を当該定性モデルに基づき予測したその後の監視指標の徴候と比較することによりプラントの応答を監視することを特徴とするプラント監視診断装置を提供する。
【００４４】
また、本発明は、上述の目的を達成するために、請求項１４に記載したように、請求項１，２，６または１０記載のプラント監視診断装置において、異常伝播経路同定部による診断結果の出力表示では、ゲインが正の伝播特性、ゲインが負の伝播特性、フィードバック効果を示す伝播特性の３種類の有向線分で監視指標に該当するプロセス信号の名称を結んだ図を描き、監視指標の観測された変化挙動を、増加、減少、変化なしの３種類に分類した結果をプロセス信号の名称と併せて表示し、変化の起点と同定されたプロセス信号の名称を強調表示すると共に、同定された影響伝播経路に該当する有向線分の色を変えて表示するようにしたことを特徴とするプラント監視診断装置を提供する。
【００４５】
また、本発明は、上述の目的を達成するために、請求項１５に記載したように、請求項１，２または６記載のプラント監視診断装置において、異常伝播経路同定部による診断結果の出力表示では、プラントの系統図の中にプロセス信号の観測点と信号名称を明示し、当該プロセス信号より算出した監視指標の変化挙動を、増加、減少、および変化なしに分類した結果をプロセス信号名称と併せて表示して、変化の起点と同定された監視指標に該当するプロセス信号とその信号に直接影響を与える上流側の機器を強調表示するとともに、同定された影響伝播経路に存在する、制御回路、配管、弁、ポンプなどの色を変えて表示するようにしたことを特徴とするプラント監視診断装置を提供する。
【００４６】
【作用】
請求項１記載の本発明のプラント監視診断装置では、監視指標に異常徴候が検出された場合、監視指標間の影響伝播特性を記述したネットワークモデルによって変化の起点となった監視指標とその影響伝播経路が同定されることから、全ての異常徴候が１つの起点により説明可能か否かを容易に把握することができる。
【００４７】
また、起点の監視指標の算出に用いたプロセス信号が得られたプラントの系統あるいはサブシステムが異常源であることが明かとなり、異常原因同定において観測される異常徴候パターンとの照合に用いる因果表の規模を系統あるいはサブシステム単位に小さく限定することが可能となる。
【００４８】
これにより、短時間の中にプラントの広い範囲のプロセス信号に影響が伝播するような異常が発生した場合にも、それが想定可能な異常であればその原因まで診断し、想定外の異常であっても少なくとも異常源と影響伝播経路を提示することができる。
【００４９】
請求項２記載の本発明のプラント監視診断装置では、請求項１記載のプラント監視診断装置によって想定外の異常と診断された場合に、観測された徴候パターンを直ちに因果表に登録することにより想定事象の１つとして学習し、以後の診断に役立つ知識として提供することができる。
【００５０】
請求項３記載の本発明のプラント監視診断方法は、プラントで観測されるプロセス信号より算出される監視指標の正常範囲からの逸脱を異常徴候として検出し、得られた各監視指標の徴候パターンを、種々の異常原因に対して想定される各監視指標の徴候パターンを記述した因果表と照合することにより診断を行った結果を、徴候の一致する監視指標の多い原因を可能性の高い原因候補として可能性の高い順に並べ替え、また監視指標については異常徴候の現れた時間の早い順に並べ替えて因果表の形式で表示するので、より一層精度の高い診断を行うことができ、同定の根拠となった各監視指標の異常徴候の関連を把握することができる。
【００５１】
請求項４，５記載の本発明のプラント監視診断装置では、特定の異常に固有の監視指標に基づく第１診断手段を、多くのプロセス信号に一律的に適用可能な処理によって算出し得る監視指標の異常徴候パターンと因果表との照合に基づく第２診断手段と分離して設けたことにより、短絡的な診断が可能な場合の処理時間を短縮することができる。逆に、固有の監視指標の算出に長時間を要する場合には、このような分離を行うことにより一律的な監視指標を用いた診断処理時間への影響を排除することができる。
【００５２】
請求項６記載の本発明のプラント監視診断装置では、定性モデルによって異常変化の起点と同定された監視指標についてのみ、事例データとの波形の比較照合を行うことにより、容易に類似事例を検索し異常原因の推定に資することができる。定性モデルに基づく診断を先に行うことにより、異常徴候を示す全ての監視指標についての波形照合の必要性を排除できる。
【００５３】
請求項７，８記載の本発明のプラント監視診断装置では、計画的操作が行われる場合にも監視処理は実行し、操作に起因して監視指標に異常徴候が検出された場合には観測された徴候パターンを因果表の徴候パターンと照合することにより、あるいは定性モデルにより異常変化の起点と同定された監視指標の波形を事例データと照合することにより、操作の識別を行うことができる。これにより、計画的操作が行われる毎にプラント監視診断装置の動作が正常か否かを確認することができる。
【００５４】
請求項９記載の本発明のプラント監視診断方法では、監視対象から除外した監視指標の中、計画的操作によってプラントに与えられた外乱の大きさを評価し得るアナログ監視指標を入力、その影響を受けて変化する各アナログ監視指標を出力と見なし、外乱発生期間中の入出力間の伝達ゲインを監視することであたかも能動的な外乱印加試験を行った場合のようにプラントの挙動の監視を行うことができる。
【００５５】
請求項１０記載の本発明のプラント監視診断装置では、定性モデルに導入した、フィードバック効果によって一定に保持され得る、と言う関係を考慮して監視指標の徴候パターンを照合することにより、伝播経路の途中にフィードバック効果の影響を受けて変化が観測されない監視指標があった場合にも伝播経路として同定することができる。
【００５６】
請求項１１記載の本発明のプラント監視診断装置では、定性モデルに基づく診断に必要な異常徴候を確認する範囲を、原因側の信号が変化した時刻から与えられた時間遅れだけ経過した時刻の前後の限定した時間帯に設定することにより、感度よく変化の徴候を検出することができる。
【００５７】
請求項１２記載の本発明のプラント監視診断装置では、定性モデルによって同定された異常の影響伝播経路がループ状になった場合にも各監視指標の異常検出時刻を基にループの起点を同定することができる。また、監視指標間に複数の伝播経路が成立し得る場合にも、各監視指標の異常徴候が検出された時刻と、伝播特性に与えられた時間遅れを照合して真の伝播経路を同定することができる。
【００５８】
請求項１３記載の本発明のプラント監視診断装置では、プラント異常時に備えて設けられた保護シーケンスの作動を起点にしてプロセス信号の応答を予測し、観測される挙動が予測された動きと一致するかどうかを逐次確認することにより、時々刻々変化するプラント状態をリアルタイムで監視することができる。
【００５９】
請求項１４，１５記載の本発明のプラント監視診断装置では、フィードバック効果を含む符号付き有向グラフで表現した請求項２０記載の定性モデル、あるいはプラント系統図の上に、異常徴候の検出された監視指標と同定された伝播経路を色を変えて示すことにより、各監視指標の異常徴候の関連が容易に把握できるように支援することができる。
【００６０】
【実施例】
以下、本発明に係るプラント監視診断装置および方法の実施例を図面に基づいて説明する。
【００６１】
図１は本発明に係るプラント監視診断装置の第１の実施例の構成を示すブロック図、図２は第１の実施例の処理の流れを示すフローチャート図である。図１において、制御部１は装置全体の実行を制御する機能を有し、定期的に診断実行指令を出力する。信号入力部２はこの実行指令を受けて予め設定された信号IＤ、サンプリング周期などの入力条件に従ってプラントよりプロセス信号を入力する。
【００６２】
監視処理部３は、監視指標毎に信号処理方式、正常範囲を示す上下限閾値などが予め設定された監視条件に従って入力されたプロセス信号より監視指標を算出して上下限閾値と比較し、上限閾値を越えた指標に対しては「増加」、下限閾値を下回った指標に対しては「減少」、正常範囲内であれば「変化なし」とする３種の徴候を判定する。上記「増加」または「減少」が異常な変化である。
【００６３】
監視指標としては信号そのものあるいは検波処理して得られる標準偏差相当の信号の現在値から、例えばその過去の一定期間内の平均値、あるいはその信号を平滑化して得られる値、あるいは他の１あるいは複数のプロセス信号の現在および／または過去の値の関数として与えられる値を予測値として差し引いた残差を用いる。このような信号処理は特定の異常事象に着目したものではないことから、多くの信号に対して一律的に適用可能なものである。
【００６４】
また、監視処理部３では、監視指標毎に初めに異常徴候が検出されたときの変化の方向を徴候として保存する。すなわち、一度「増加」と判定された後、「変化なし」あるいは「減少」に変わったとしても初めの徴候である「増加」をその徴候として後の診断に用いる。このようにして、監視処理部３は各監視指標の徴候判定結果の集合である徴候パターンを出力する。
【００６５】
異常伝播経路同定部４は、定性モデルデータベース６に登録された各監視指標間の影響伝播特性を符号付き有向グラフで表現した定性モデルと、観測された徴候パターンとを照合し、異常徴候の影響伝播経路を同定する。この経路の起点に位置する監視指標が最初に変化を示したものとして同定され、当該指標の算出に用いたプロセス信号の観測されたプラントの系統あるいはサブシステムが異常源と判断される。なお、変化の起点となった監視指標の同定までを異常源同定部の機能とし、図２に示したように、異常源系統またはサブシステムを同定する機能は次に述べる異常原因同定部５に含まれるとしてもよい。
【００６６】
異常原因同定部５は、プラントの各系統あるいはサブシステム毎に作成されるとともに、因果表データベース７に登録されている複数の因果表の中から、異常伝播経路同定部４により異常源と同定された系統あるいはサブシステムに対する因果表を読み出し、監視処理部３で作成された徴候パターンとの照合を行う。照合においては例えば異常徴候に矛盾がなく、且つ徴候の一致する監視指標の数が多い想定原因を現在の異常原因の候補として同定する。このとき、計画的な操作によって観測される徴候パターンを因果表に含めておくことにより、操作の行われたことを識別することができる。
【００６７】
以下、図１において破線で囲んだ監視処理部３、異常伝播経路同定部４、異常原因同定部５、および定性モデルデータベース６と因果表データベース７の部分を監視診断部８ということにする。また、監視診断部８から監視処理部３を除いた部分を診断処理部９という。
【００６８】
制御部１は、異常伝播経路同定部４の診断結果を出力表示部１０に送るとともに、異常原因同定部５において原因候補が同定された場合、その結果も出力表示部１０に送る。異常原因同定部５において該当する想定原因が同定されなかった場合は、想定外の異常が発生したものと判断し、データベース管理部１１に現在の徴候パターンを送る。
【００６９】
データベース管理部１１では、制御部１より徴候パターンが送られた場合、未登録のパターンとして因果表データベース７に追加登録する。このようにして学習された徴候パターンに対しては、後刻原因が判明した時点で対話処理部１２により事象名などの必要な情報が追加される。
【００７０】
出力表示部１０は、制御部１より送られる種々の情報を分かり易い形式で表示する。
【００７１】
以下、定性モデルに基づく異常伝播経路同定、因果表に基づく異常原因同定、および出力表示部１０の特徴についてさらに詳細を述べる。
【００７２】
先ず、前記定性モデルに基づく異常伝播経路同定において従来方法の課題を解決するために本実施例が提供する手段の特徴について図面を用いて詳細に説明する。具体例における監視指標名は、簡略化するためプロセス信号名そのものを用いている。
【００７３】
始めに、図３に基づいて定性モデルへのフィードバック特性の導入について説明する。例えば、沸騰水型原子炉の圧力制御系では、主蒸気の圧力が一定になるように加減弁の開度を制御系が調節している。原子炉圧力が減少すると主蒸気圧力も減少するので、制御系は加減弁の開度を絞って圧力を上げようとする。したがって、これらの観測信号の間には、図中に実線の有向線分で表現した「助長」の関係が存在することになる。
【００７４】
ところが、図３に示すように、原子炉圧力が下り勾配で表現した通り、減少したとしても、過渡変化の速度が遅い場合には制御系の応答の方が速くなり、結果として主蒸気圧力には→印で表現した通り変化が見えず、従来の符号付き有向グラフの方法では図中に破線の矢印の有向線分で表現したように過渡変化の伝達経路がその前後で切れてしまう。
【００７５】
そこで、本実施例では、加減弁の開度が増加あるいは減少することによって主蒸気圧力を一定に保つというフィードバック特性を表現した定性モデルを導入することにより、見かけ上変化が現れない場合でも正しく伝播経路を同定することができる。
【００７６】
次に、図４を用いて監視指標間の影響伝播特性への時間遅れの導入について説明する。図４に示したように、監視処理部３において監視指標の挙動から「増加」、「減少」、「変化なし」といった変化の徴候を検出する時には、例えば通常時に観測される監視指標の最大振幅の２倍を正常範囲と定めることにより、上下限閾値が設定されているが、異常によって生じた過渡変化の程度が小さい場合には、信号Ｃの例のように通常の変動に隠れて正確に検出されないことがある。
【００７７】
そこで、本実施例では、監視指標間の因果関係に伝達遅れを導入し、原因側の監視指標の変化が検出されていて且つ結果側の監視指標に変化が検出されていない場合、原因側の監視指標が変化してから伝達遅れ時間だけ後の限定した範囲の結果側の監視指標の変化を正常範囲についても限定した上で調べることにより、精度よく検出することができる。これにより、過渡変化の程度が小さい場合でも、正しく経路を推定できるようになる。
【００７８】
また、同様に図４に示すように、同定された影響伝播経路がループ状になる場合でも、各監視指標の異常検出時刻を基にループの起点を同定するとともに、監視指標間に複数の伝播経路が成立し得る場合にも、各監視指標の異常徴候が検出された時刻と、導入した時間遅れを照合して真の伝播経路を同定することができる。
【００７９】
さらに、図５に基づいてプラントの保護シーケンス作動後の診断方法について説明する。従来の符号付き有向グラフに基づく診断方法は、正常な運転状態のフェーズIから異常の徴候が検出されたフェーズIIにかけてのプラント信号間の影響伝播特性を基に、異常による観測信号の初期変化から過渡変化の発生した信号を推定するものである。したがって、異常が進展して機器や系統の保護シーケンスが作動し、ポンプがトリップしたり、プラントがスクラムしたことによってそれまでとは異なる影響伝播特性を示すフェーズIIIの段階に移った後の状態を診断することは困難であった。
【００８０】
そこで、本実施例では、図５に示すように保護シーケンスの動作を知らせるデジタル信号を取り込むことにより、例えばポンプがトリップすると、それを起点として信号の変化を予測し、プラントの応答が正常か否かを診断することができる。
【００８１】
以下、図６に示すフローチャート図を用いて異常伝播経路同定部４の動作を説明する。
【００８２】
監視処理部３で信号変化の徴候が検知されると、異常伝播経路同定部４が起動する。先ず、監視処理部３で抽出された観測信号の増加・減少を入力し、個々の信号について以下のような検査を行う。着目している信号が増加または減少している場合には、定性モデルデータベース６に記述された通常の伝達モデル（増減の伝わり方と時間遅れ）を参照し、原因側の信号の変化がモデルと合致していれば両信号間を変化が伝わったとして伝達経路を保存し、合致しなかった場合には起点の信号とみなす。起点の信号が保護シーケンスの動作を表すものであればシーケンス動作確認として、通常のプロセス信号であれば変化の起点としてその上流の機器を異常箇所として保存する。着目している信号に変化がなかった場合には、フィードバックモデルを参照して、原因側の信号の動きがモデルと合致していれば、フィードバックループを伝達経路として保存し、合致しなければ、あるいはフィードバックループが存在しなければ、着目している信号は事象に無関係として無視する。
【００８３】
以上の処理を全ての信号に施すことにより、異常の発生箇所とその影響が伝わった経路の同定、または保護シーケンス動作とその応答の確認することができる。
【００８４】
次に、因果表に基づく異常原因同定について説明する。図１および図２を用いて説明した第１の実施例では、異常変化の生じた系統あるいはサブシステムを、異常変化の起点と同定された監視指標から求め、系統またはサブシステム別に与えられた因果表と観測された異常徴候パターンを照合することにより異常原因を同定している。
【００８５】
これに対して、第２の実施例の監視診断部の構成、監視指標の因果表をそれぞれ図７（Ａ），（Ｂ）に、その処理の流れを図８にそれぞれ示す。なお、前記第１の実施例と同一または対応する部分には同一の符号を用いて説明する。以下の実施例でも同様である。
【００８６】
図７に示すように、異常伝播経路同定部４では、異常変化の起点となった監視指標が同定されていることから、各監視指標毎に当該指標に最初に影響が現れることが想定される種々の異常原因に対して徴候パターンを与えた因果表をデータベース７ａに登録しておき、異常原因同定部５においては該当する監視指標別因果表に基づく原因同定を行うようにすることも可能である。
【００８７】
このように第２の実施例によれば、徴候パターンとの照合の対象となる因果表を系統あるいはサブシステム単位から監視指標単位に限定することにより、計算機による診断処理時間の短縮、実行時の所要メモリの縮小、診断精度の向上が期待される。
【００８８】
本発明に係るプラント監視診断装置の第３の実施例における監視診断部の構成、サブシステムの因果表をそれぞれ図９（Ａ），（Ｂ）に、また同じ監視診断部の処理の流れを図１０にそれぞれ示す。サブシステム因果表データベース７ｂには、プラントの系統あるいはサブシステム毎に当該系統あるいはサブシステムが異常源となることが想定される種々の原因に対して徴候パターンを登録するばかりでなく、他の系統あるいはサブシステムで生じた異常の影響が伝播することが想定される場合にはこれを含めて登録するとともに、上流側の原因として他の系統あるいはサブシステムの異常原因を併せて登録しておく。
【００８９】
異常原因同定部５では、監視条件として与えられている各監視指標とその対象としている系統またはサブシステムとの関係を用いて、異常徴候を示している監視指標から異常源系統またはサブシステムの候補群｛Ｓi｝を求める。
【００９０】
次に、監視処理部３で作成された異常徴候パターンを系統あるいはサブシステム毎に作成された因果表と照合することにより、各異常源候補Ｓi毎に異常原因候補｛Ａｊi｝と、それぞれの候補に対して図９（Ｂ）のように与えられた上流側要因｛Ｆｊi｝を同定し、全ての異常源候補について求めた異常原因候補を集めた確認対象原因リストＬａ、上両側要因を集めたリストＬｅを求める。
【００９１】
そして、確認対象原因リストＬａに含まれる異常原因候補の中から、それ自身がＬｅに含まれ、且つその上流側要因がＬａに含まれていないものだけを最終的な異常原因候補と判定する。但し、異常源候補が１つだけの場合には確認対象のＬａがそのまま異常原因候補のリストとなる。
【００９２】
これにより、本実施例では異常徴候が観測されている他の系統あるいはサブシステムの影響を受けていない系統あるいはサブシステムを異常源と判定し、当該系統あるいはサブシステムに対する因果表を基に同定された異常原因候補の中で他の系統あるいはサブシステムの異常徴候を説明し得るものを現在の異常原因として同定することができる。
【００９３】
また、図１１に示す例では、第２サブシステムおよび第３サブシステムの異常徴候はそれぞれ原因２−１と原因３−２で説明されるが、何れの原因も上流側の原因１−２によって説明されることから、第１サブシステムが異常源として判定される。さらに、第１サブシステムの因果表による診断結果では、異常原因候補として１−１と１−２の２つが同定されているが、原因１−１は他のサブシステムの異常徴候を説明できないことから、最終的に異常原因１−２が同定されることとなる。
【００９４】
図１２は本発明に係るプラント監視診断装置の第４の実施例における監視診断部の構成を示すブロック図、図１３はその監視診断処理の流れを示すフローチャート図である。本実施例においては、図１、図２で示した第１の実施例、図７、図８で示した第２の実施例の場合と同様に、異常伝播経路同定部４では監視処理部３で作成された各監視指標の異常徴候パターンを定性モデルと照合することにより、変化の起点となった監視指標とその影響伝播経路を同定する。
【００９５】
事例データ抽出部１３は、事例データベース１４に登録されている異常発生時のプロセス信号の時系列データの中から、変化の起点と同定された監視指標が同じ事例データを検索し、監視条件として与えられた信号処理方式に従い、各該当事例についてプロセス信号から当該監視指標の時系列データを算出する。変化の起点として観測された監視指標の時系列データの波形と前記抽出された事例データの波形の類似度を比較照合部１５において評価し、一定の値以上の高い類似度を示す事例の異常原因を現在の異常原因と診断する。
【００９６】
類似度の評価方法を図１４を用いて説明する。変化の起点と同定された監視指標の現在のデータと１事例データの波形が同図に示すようであったとき、現在のデータで異常が検出された時刻前後の一定期間ｔ0を取り出した時系列データを基本関数ｆ（ｔ）とし、事例データｘ（ｔ）との間で次の評価関数
【数３】

を算出する。そして遅れ時間τの関数としての上記評価関数の最大値を類似度とするものである。上式において事例データを同じ方法で期間ｔ0だけ取り出して評価したものは両データの相互相関関数となり、これを評価関数として用いることも１つの方法である。
【００９７】
因果表による原因同定の場合と同様、事例データベースに計画的操作が行われたときのデータを含めておくことにより、操作が行われたことを識別することができる。
【００９８】
このように、計画的操作が行われている間も通常と同じ監視を継続し、操作に伴うプロセス信号の変化を異常と見なしてその原因を判定するようにすることは、プラント監視診断装置の機能と動作の健全性を確認する手段として有効なものと考えることができる。
【００９９】
しかしながら、計画的操作による影響はあくまでも異常とは見なさず、正確に識別すべきという考え方もある。このような考えに添って、本発明に係るプラント監視診断装置の第５の実施例の場合の監視処理部の構成を図１５に示す。第５の実施例では、図１、図７、図９、あるいは図１２に示した各実施例の監視診断部８の監視処理部３の前段に操作識別部１６が設けられ、図１５の第１監視処理手段３ａは前記各図に示した実施例の監視処理部３と同一のものである。
【０１００】
操作情報データベース１７には、図１６に示すように、計画的操作毎にその名称、操作の開始を判定する条件、終了を判定する条件、最長操作期間、操作iにより最初に変化するアナログ監視指標をＸi、操作の影響を受けて変化するアナログ監視指標｛Ｙji｝、新たな指標である相対変化率｛Ｙji／Ｘi｝を、Ｘiを入力変数、Ｙjiを出力変数と見なしたときの伝達ゲインと考え、その正常ゲインの範囲などが登録されている。
【０１０１】
次に、図１７に示す処理の流れに従って前記第５の実施例における監視処理部３ａ，３ｂの動作を説明する。
【０１０２】
図１５に示す操作識別部１６は、開始および終了判定に必要な指標を算出してそれぞれの判定条件を確認し、操作iの開始が確認された時点から最長操作期間だけ、あるいは終了条件が成立するまでの間、操作の影響によって変化する監視指標であるＸiおよび｛Ｙji｝を監視バイパス対象指標として異常検出の対象から除外する指令を第１監視処理手段３ａに送る一方、第２監視処理手段３ｂに対して実行指令を送る。第２監視処理手段３ｂでは、図１８に示すように前記相対変化率を新たな監視指標として算出し、正常ゲインとして与えられた上限閾値Ｈi、下限閾値Ｌiと比較することにより監視を行う。
【０１０３】
以上のような監視処理部の構成を採ることにより、計画的操作が行われている間も、操作の影響を受けない監視指標に対しては通常通りの第１監視処理手段３ａによる監視処理を継続しながら、操作によって変化を生じる監視指標については、その変化の仕方が正常か否かを第２監視処理手段３ｂにより判定することができる。
【０１０４】
図１９は本発明に係るプラント監視診断装置の第６の実施例の構成を示すブロック図である。同図において、第２診断手段８ｂは図１、図７、図９、あるいは図１２に示した監視診断部８と同様の機能を有し、一律的な監視処理によって得られる異常徴候パターンを想定可能な異常原因に対して例えば因果表の形式で与えられた徴候パターンと照合することにより診断を行う。一方、第１診断手段８ａは特定の異常のみに反応するプロセス信号から当該異常に固有の監視指標を算出して正常範囲を示す閾値と比較することにより、当該異常の発生を検出する。例えば、流量信号に現れる特定周期の振動成分の振幅を監視することにより不安定流動現象を検出するといった場合がこれに該当する。
【０１０５】
また、図１９の構成において、図１５に示した操作情報データベース１７を含む操作識別部１６と、第２監視処理手段３ｂを全て第１診断手段８ａに含め、第１監視処理手段３ａを第２診断手段８ｂに含め、操作識別部１６の有する実行制御機能を制御部１に含めることにより、図１５の構成の監視処理部の有する機能を実現することが可能である。
【０１０６】
次に、本発明のプラント監視診断装置の出力表示方法について説明する。
【０１０７】
図２０は定性モデルによる異常の影響伝播経路同定結果の表示の一例を示すものであり、各監視指標には簡略化するためプロセス信号名そのものを用いてある。伝達特性のゲインの正負とフィードバック効果の種別のわかる実線、破線、一点鎖線などで表現した有向線分でプロセス信号の名称を結び、信号間の因果関係を符号付き有向グラフで表現したのと同等の情報を表わしている。
【０１０８】
これに観測された信号の定性的な挙動を「増加」、「減少」、「変化なし」に分類した結果を信号名称と併せてそれぞれ矢印上り勾配、下り勾配、→で表示し、過渡変化の起点の信号名称を強調表示するととともに、変化が伝わったと同定された経路に相当する有向線分の色および／または濃淡を変えることにより、過渡変化の発生箇所と波及経路が容易に認識できるように表示されている。
【０１０９】
定性モデルによる診断結果の異なる表示例を図２１に示す。この場合、プラントの系統図の中にプロセス信号の観測点と信号名称が明示され、観測された信号の定性的な挙動を「増加」、「減少」、「変化なし」に分類した結果が信号名称と併せて表示されている。過渡変化の起点の信号とその上流の機器を強調表示するとともに、変化が伝わったと同定された信号間に存在する、制御回路、配管、弁、ポンプなどの色を変えることにより、過渡変化の発生箇所と波及経路が容易に認識できるように表示されている。
【０１１０】
因果表に基づく診断結果を表示する場合には、監視指標毎の異常徴候の関連を把握し易いように考慮することが必要である。図２２に示す例は、観測される徴候パターンと矛盾せず、且つ徴候の一致する監視指標の数の多い原因を可能性の高い原因候補として上位に並べ換え、さらに監視指標についても異常な変化の発生時刻が早い順に並べ換えて表示したものである。
【０１１１】
また、図２３は因果表作成時に行われるイベントツリー解析の結果に監視指標の変化を含めてツリー状に表現することで、各監視指標の徴候の相違が診断結果に与える影響を明示した例を示すものである。
【０１１２】
図２４は本発明のプラント監視診断装置の第７の実施例の構成を示すブロック図である。同図の操作識別部１６と操作情報データベース１７は図１５および図１６に示したものと同じである。また、図２４の第１監視処理部１８ａは図１、図７、図９、および図１２に示した各実施例の監視診断部８の監視処理部３と同じであり、一律的な処理による監視を行う部分である。
【０１１３】
図２４の異常伝播経路同定部４および定性モデルデータベース６は、図１、図７、および図１２に示した各実施例の監視診断部８の異常伝播経路同定部４および定性モデルデータベース６と同じである。図２４の第１異常原因同定部５ａおよび因果表データベース７は、図１および図７に示した各実施例の監視診断部８の異常原因同定部５および因果表データベース７と同じである。
【０１１４】
また、図２４の第２異常原因同定部５ｂは、図１２に示した実施例の事例データ抽出部１３と比較照合部１５を合わせた機能を有し、事例データベース１４は両図とも同じものである。さらに、図２４の第２監視処理部１８ｂは特定の異常に反応する固有の監視指標を算出して当該異常の発生を検出する部分であり、図１９の第１診断手段８ａと同じものである。したがって、この第２監視処理部１８ｂには、図１９の説明でも述べた通り、図１５の第２監視処理手段３ｂの機能を含めたものとすることができる。
【０１１５】
【発明の効果】
以上説明したように、請求項１記載の本発明のプラント監視診断装置によれば、監視指標に異常徴候が検出された場合、監視指標間の影響伝播特性を記述したネットワークモデルによって変化の起点となった監視指標とその影響伝播経路が同定されることから、全ての異常徴候が１つの起点により説明可能か否かを容易に把握することができる。
【０１１６】
また、起点の監視指標の算出に用いたプロセス信号が得られたプラントの系統あるいはサブシステムが異常源であることが明かとなり、異常原因同定において観測される異常徴候パターンとの照合に用いる因果表の規模を系統あるいはサブシステム単位に小さく限定することが可能となる。
【０１１７】
これにより、短時間の中にプラントの広い範囲のプロセス信号に影響が伝播するような異常が発生した場合にも、それが想定可能な異常であればその原因まで診断し、想定外の異常であっても少なくとも異常源と影響伝播経路を提示することができる。
【０１１８】
このように、大規模プラントで生じる異常を計画的操作の影響と識別しながら徴候段階で検出し診断することが可能となる。また、異常の影響が短時間の内に系統間を伝播し広範なプロセス信号に変化が現れる場合にも速やかに異常源を同定することが可能となる。何れの場合にも、想定可能かあるいは経験済みの異常であれば、その原因あるいは機器単位の異常原因同定が可能であり、想定外の異常の場合にも少なくとも信号単位での異常源の同定が可能となる。したがって、異常時の状況把握が容易となり、徴候段階での早期対応と迅速な正常復帰が期待される。
【図面の簡単な説明】
【図１】 本発明のプラント監視診断装置の第１の実施例の構成を示すブロック図。
【図２】 第１の実施例の処理の流れを示すフローチャート図。
【図３】 フィードバック特性を導入した定性モデルの効果を示す図。
【図４】 監視指標間の影響伝播特性に導入した時間遅れの効果を示す図。
【図５】 定性モデルによるリアルタイム診断の例を示す図。
【図６】 異常伝播経路同定処理の流れを示すフローチャート図。
【図７】 （Ａ），（Ｂ）本発明に係るプラント監視診断装置の第２の実施例の監視診断部の構成を示すブロック図，監視指標の因果表を示す図。
【図８】 第２の実施例の監視診断処理の流れを示すフローチャート図。
【図９】 本発明のプラント監視診断装置の第３の実施例の監視診断部の構成を示すブロック図，サブシステムの因果表を示す図。
【図１０】 第３の実施例の監視診断処理の流れを示すフローチャート図。
【図１１】 第３の実施例における原因同定の方法を示す図。
【図１２】 本発明に係るプラント監視診断装置の第４の実施例の監視診断部の構成を示すブロック図。
【図１３】 第４の実施例の監視診断処理の流れを示すフローチャート図。
【図１４】 事例データとの比較照合方法を示す図。
【図１５】 本発明に係るプラント監視診断装置の第５の実施例の監視処理部の構成を示すブロック図。
【図１６】 操作情報データベースの内容を示す図。
【図１７】 第５の実施例の監視処理の流れを示すフローチャート図。
【図１８】 第５の実施例の第２監視処理手段の監視方法を示す図。
【図１９】 本発明に係るプラント監視診断装置の第６の実施例の構成を示すブロック図。
【図２０】 信号間のネットワークによる異常伝播経路の表示例を示す説明図。
【図２１】 プラント系統図による異常伝播経路の表示例を示す説明図。
【図２２】 因果表に基づく診断結果の表示例を示す説明図。
【図２３】 因果表に基づく診断結果の異なる表示例示す説明図。
【図２４】 本発明に係るプラント監視診断装置の第７の実施例の構成を示すブロック図。
【符号の説明】
１ 制御部
２ 信号入力部
３ 監視処理部
４ 異常伝播経路同定部
５ 異常原因同定部
６ 定性モデルデータベース
７ 因果表データベース
７ａ 監視指標別因果表データベース
７ｂ サブシステム因果表データベース
８ 監視診断部
８ａ 第１診断手段
８ｂ 第２診断手段
９ 診断処理部
１０ 出力表示部
１１ データベース管理部
１２ 対話処理部
１３ 事例データ抽出部
１４ 事例データベース
１５ 比較照合部
１６ 操作識別部
１７ 操作情報データベース
１８ａ 第１監視処理部
１８ｂ 第２監視処理部[0001]
[Industrial application fields]
  The present invention relates to a plant monitoring diagnostic apparatus and method for monitoring an operation state based on a process signal of a large-scale plant such as a nuclear power plant, a thermal power plant, and a chemical plant to detect and diagnose an abnormality at an early stage.
[0002]
[Prior art]
  In general, there are two types of abnormalities that occur in a plant: mechanical abnormalities in the equipment that constitutes the plant, and abnormalities in the operating state of mechanically sound components such as unstable vibrations of the fluid flow rate. There is. In addition, in the case of a pump, process signals used to detect these abnormalities include main effect parameters indicating performance and functions such as discharge pressure and flow rate, and secondary effects such as vibration, sound, and temperature that accompany operation. There are parameters.
[0003]
  The mechanical abnormality is likely to be detected if a detector for measuring the secondary effect parameter is provided for each device and monitored. However, in a large-scale plant, a large cost is required. For this reason, only the main devices are subject to such direct monitoring, and for many devices, a detection method is adopted only when the influence of the abnormality appears as a change in the main effect parameter.
[0004]
  However, the change that appears in the secondary effect parameter is almost always associated with the abnormality of the specific device, whereas the main effect parameter is affected by the abnormality of many devices in addition to the abnormality as the operation state. May give. In addition, when a change in operating conditions, a device operation check test, or the like is performed, just because a change appears in the main effect parameter, it cannot always be determined that it is caused by an abnormality.
[0005]
  From the above background, a technique for detecting an abnormality from a main effect parameter of a plant and diagnosing which of the systems or subsystems constituting the abnormality when the entire plant is viewed as one system is an abnormality source. Has been developed. Also, by using the secondary effect parameters, including the secondary effect parameters, what kind of abnormality is the cause, that is, if the abnormality is in the operating state, and which equipment is abnormal if the equipment is abnormal Techniques for diagnosing this are also being developed.
[0006]
  These techniques are introduced in, for example, “Equipment Diagnosis Predictive Maintenance Encyclopedia” (published in 1988) supervised by Koji Oshima. Typical examples of techniques for diagnosing abnormal sources include so-called model-based methods. There is a method based on a causal table as a typical example of a technique for diagnosing the problem.
[0007]
  In the model-based method, the behavior in the normal operation state of the main effect parameter governed by the physical law is described by a mathematical model for each subsystem. Then, using this model, an output parameter is calculated from the observed value of the input parameter to the subsystem, and the result is compared with the actually observed output parameter value using the result as a reference value. In other words, the subsystem that shows the change in the output parameter that cannot be explained by the change in the input parameter is used as the abnormality source. However, the model-based method requires a highly accurate model in order to detect the occurrence of an abnormality at an early stage, and it is difficult to construct a highly accurate behavior model for the entire large-scale plant.
[0008]
  For this reason, a diagnostic method using a signed directed graph has been developed in which only information indicating whether the phase delay of the transmission behavior between main effect parameters is less than 180 degrees (in phase) or 180 degrees or more (reverse phase) is described as a network model. This method detects changes in the main effect parameters based on their normal values, and changes the direction of change by comparing it with the qualitative behavior model of the entire plant system described by a signed directed graph. This is to identify the parameter that is the starting point and the transmission path of the influence.
[0009]
  On the other hand, the method based on the causal table can be preliminarily assumed as a symptom of whether the process signal value has increased (or abnormally high), decreased (abnormally low), or normal compared to the normal value. A table in which a symptom pattern that is a set of symptoms of each signal is associated in advance for each abnormal cause is given in advance, and the observed symptom pattern is compared with this table for diagnosis. The causal table is generally created by predicting a symptom pattern by an event tree analysis that evaluates a ripple effect by covering the abnormal causes to be assumed by failure mode / effect evaluation of equipment constituting the plant.
[0010]
  In addition, for the influence of planned operation that appears in the process signal by being performed artificially and mechanically, when the status signal of the operation switch is measured, it is detected that the operation is in progress. A method of not performing monitoring diagnosis is adopted.
[0011]
[Problems to be solved by the invention]
  By the way, in order to maintain the soundness of the plant and perform stable operation, the process signal is monitored continuously or periodically, the change that appears due to the occurrence of abnormality is detected quickly, the cause is investigated, and the appropriate It is required to deal with. However, it has been difficult to satisfy the demands of the conventionally developed technologies due to the following problems. That is,
  (1) The diagnostic method using a qualitative model identifies abnormal sources in units of signals even when the effects of abnormalities occurring in the plant propagate between systems or subsystems within a short period of time, and changes appear in a wide range of process signals. Although it is possible, it is generally impossible to identify the cause, except in the special case where individual signals react one-on-one to a specific anomaly cause.
[0012]
  (2) Although the method based on the causal table can be diagnosed with respect to possible causes, the scale of the table exponentially increases depending on the number of signals to be handled, and thus cannot be simply applied to a large-scale plant.
[0013]
  In order to deal with this problem, it is possible to divide the entire plant into several systems or subsystems and create a causal table for each of them. Diagnosis is difficult when propagating between subsystems.
[0014]
  (3) The method based on the causal table cannot cope with an abnormality that was not assumed at the start of operation of the plant monitoring and diagnosis apparatus.
[0015]
  (4) The symptom pattern defined in the causal table does not always completely match the pattern actually observed depending on the degree of abnormality. In addition, when the causal table is divided for each system or subsystem, signal changes that are not included in the given symptom pattern for the identified cause may be observed, and all signs are identified. It is difficult to grasp whether or not explanation is possible due to the cause.
[0016]
  (5) The causal table can include not only the symptoms of the measured values of each signal itself, but also the symptoms when various characteristic parameters obtained by processing the signals are used as monitoring indicators. Because it does not contain causal information, it is difficult to understand the reason for diagnosis.
[0017]
  The above (2) is a problem to be solved in order to achieve the processing time required for practical use as a monitoring / diagnosis apparatus. In addition, the following problems can be solved from the viewpoint of practicality. It is done.
[0018]
  (6) It is effective to monitor important abnormal events unique to the target plant or abnormalities actually experienced, if there is a unique index suitable for each detection. As described in (5) above, it is possible to diagnose by including the symptoms of such a special index in the causal table, but it is not efficient when the abnormality of the special index directly indicates the occurrence of a specific abnormality .
[0019]
  (7) In addition to simple symptoms such as increase or decrease in process signal, the cause of abnormality may be estimated by comparing the way of change, that is, the waveform with the waveform of past case data. Diagnosis is not always possible because it relied on the flashes of engaged humans.
[0020]
  (8) Signals that have not been positively identified in the past because they have been artificially and mechanically performed, so that the contents of planned operations have not been positively identified based on changes appearing in the process signal. There has been no meticulous response such as continuing monitoring. In addition, when a human operation is performed for the purpose of confirming the operation of the device, the operation of the device to be confirmed is naturally monitored, but whether or not the change is appropriate for other affected process states It is usually difficult to constantly monitor this.
[0021]
  In addition, it is not always possible to identify an abnormal source even in the diagnosis method using the qualitative model described in (1) above, and some problems to be solved in practice have been pointed out as follows. That is,
  (9) As described in the related art, the qualitative model expressed by the signed directed graph expresses the propagation characteristics of the signal change only by the propagation direction and phase. In other words, if the cause signal increases, the result signal also increases, and if the cause signal decreases, it decreases, and the “promotion” relationship has a positive transfer gain, and if the cause signal increases, the result signal decreases. There are only two types of “inhibition” relationships that have a negative transfer gain that increases when decreased. Therefore, it is impossible to express a relationship in which the result-side signal is kept constant by increasing or decreasing the cause-side signal, and the feedback effect cannot be handled.
[0022]
  (10) The qualitative model does not include the concept of time in the propagation characteristics of signal changes. Therefore, it is not possible to diagnose the propagation state of the transient change in consideration of the passage of time. Further, when the influence propagation path of the monitoring index indicating the identified abnormal sign becomes a closed loop, the signal that is the starting point of the change cannot be identified.
[0023]
  (11) The diagnosis method based on the qualitative model is to identify the signal and the influence propagation path that are the starting point of the change based on the initial change of the observation signal. Changing plant conditions cannot be diagnosed in real time.
[0024]
  The present invention has been made in view of the above circumstances, and its first object is to solve the problems (1), (2), (3), and (4), and to produce anomalies that occurred in a large-scale plant. It is an object of the present invention to provide a plant monitoring diagnosis apparatus and method that can promptly diagnose the cause of an abnormality even when the influence of the problem propagates between systems or subsystems within a short time and changes occur in a wide range of process signals.
[0025]
  A second object of the present invention is to provide a plant monitoring diagnosis apparatus and method that can solve the above-mentioned problem (5) and provide easy-to-understand diagnosis results.
[0026]
  The third object of the present invention is to solve the above-mentioned problem (6) and provide a plant monitoring diagnosis apparatus having a function of diagnosing a widely assumed abnormality and a function of efficiently performing a diagnosis focusing on a specific event There is to do.
[0027]
  The fourth object of the present invention is to solve the above-mentioned problem (7) and to make a diagnosis by comparing the waveform of the observed process signal with the waveform in an abnormal case experienced in the past. And to provide a method.
[0028]
  A fifth object of the present invention is to provide a plant monitoring diagnostic apparatus capable of detecting the occurrence of a true abnormality by identifying the influence of a planned operation appearing in a process signal with respect to the problem (8). And to provide a method.
[0029]
  The sixth object of the present invention is that abnormal signs that should appear in the process signal are suppressed by the feedback effect and do not appear apparently, or because the signal change is small, abnormal signs cannot be detected hidden behind normal fluctuations, and influence propagation A plant monitoring diagnosis device that solves the problems (9), (10), and (11) of the qualitative model diagnosis method such that the starting point of the change cannot be identified when the path becomes a closed loop, without complicated numerical processing. And to provide a method.
[0030]
  Hereinafter, in the present invention, all monitoring targets for which abnormality detection is performed by comparison with a threshold value indicating a normal range including the process signal observation value itself are referred to as monitoring indices.
[0031]
[Means for Solving the Problems]
  The present inventionAboveTo achieve the purpose,As described in claim 1,Describes the deviation from the normal range of the monitoring index calculated from the process signal observed at the plant as an abnormal sign, and describes the effect propagation characteristics between each monitoring index and the monitoring processing unit that creates the symptom pattern of each monitoring index A qualitative model database in which the network model is registered, an abnormal propagation path identification unit for identifying the influence propagation path due to the monitoring index and the monitoring index that is the starting point of the abnormal change by comparing the observed symptom pattern with the network model,It is assumed that the influence of abnormalities will first appear in the monitoring index that started the change.Various abnormal causesEveryCausal table database that registers causal tables describing symptom patterns, and monitoring indicators identified as the origin of abnormal changesCorresponding toAn abnormal cause identification unit that identifies the cause of the abnormality by comparing the causal table with the observed symptom pattern, and outputs at least the identified effect propagation path. A plant monitoring diagnostic apparatus characterized by comprising an output display unit for outputting the above together.
[0032]
  Further, in order to achieve the above-mentioned object, the present invention provides a plant monitoring diagnostic apparatus according to claim 1, in which the symptom pattern corresponding to the causal table is not included, as described in claim 2. Provided is a plant monitoring diagnosis apparatus characterized by including a database management unit for performing learning by adding this.
[0033]
  Further, in order to achieve the above-mentioned object, the present invention detects a deviation from a normal range of a monitoring index calculated from a process signal observed in a plant as an abnormal sign as described in claim 3, The result of diagnosis by comparing the obtained symptom pattern of each monitoring index with the causal table describing the symptom pattern of each monitoring index assumed for various abnormal causes The plant monitoring is characterized by rearranging the most common causes as the most likely cause candidates in order from the most likely cause, and displaying the monitoring indices in the form of a causal table by sorting in ascending order of time of appearance of abnormal signs. Provide a diagnostic method.
[0034]
  Further, in order to achieve the above-mentioned object, the present invention calculates a monitoring index specific to an abnormality from a process signal that reacts only to a specific abnormality occurring in the plant, as described in claim 4, The first diagnostic means for identifying the abnormality as the cause of the abnormality by detecting the deviation from the range, the process signal and the current value of the detected signal, the average value in the past fixed period, the signal is smoothed Deviation from the normal range is detected as an abnormal sign by using the residual obtained by subtracting the value obtained as a function of the current value and / or the past value of at least one other process signal as a predicted value, as a monitoring index. The second diagnostic means for identifying the cause of abnormality by collating the sign pattern of each monitoring index with a causal table describing sign patterns assumed for various cause of abnormality A plant monitoring diagnosis apparatus characterized in that the first diagnosis means and the second diagnosis means are executed in parallel and the result of the first diagnosis means is output in preference to the result of the second diagnosis means. To do.
[0035]
  Further, in order to achieve the above-mentioned object, the present invention calculates a monitoring index specific to an abnormality from a process signal that reacts only to a specific abnormality occurring in the plant, as described in claim 5, The first diagnostic means for identifying the abnormality as the cause of the abnormality by detecting the deviation from the range, the process signal and the current value of the detected signal, the average value in the past fixed period, the signal is smoothed Deviation from the normal range is detected as an abnormal sign by using the residual obtained by subtracting the value obtained as a function of the current value and / or the past value of at least one other process signal as a predicted value, as a monitoring index. The second diagnostic means for identifying the cause of abnormality by collating the sign pattern of each monitoring index with a causal table describing sign patterns assumed for various cause of abnormality When the cause of abnormality is identified by executing the first diagnosis means, the result is output and the second diagnosis means is not executed, and only when the corresponding cause is not identified by the first diagnosis means A plant monitoring diagnostic apparatus characterized by executing second diagnostic means is provided.
[0036]
  Further, in order to achieve the above-mentioned object, the present invention detects a deviation from a normal range of a monitoring index calculated from a process signal observed in a plant as an abnormal sign, as described in claim 6, A monitoring processing unit that creates a symptom pattern for each monitoring index, a network model that describes the effect propagation characteristics between each monitoring index, and the observed symptom pattern, and the monitoring index that started the abnormal change and its effect An abnormal propagation path identification unit that identifies a propagation path, a case data extraction unit that searches for registered cases where the monitoring index is the starting point of change and calculates the monitoring index, and a monitoring index that is the starting point of the change A period of time t before and after detection of the change of the monitoring index that is the starting point of the change, comprising a comparison and collation unit that evaluates the similarity between the waveform of the time change and the waveform of the time change of the extracted case data 0 Function obtained from the case data x (t) using the time-series data obtained from the data as the basic function f (t)
[Expression 2]

The plant monitoring and diagnosis apparatus is characterized by outputting the cause of abnormality of a case giving a high similarity of a certain value or more as a candidate for the current abnormality cause.
[0037]
  Further, in order to achieve the above-mentioned object, the present invention is envisaged for the planned operation that is artificially and mechanically performed on the plant in the causal table as described in claim 7. The plant monitoring diagnostic apparatus according to claim 1 or 2, characterized by including a symptom pattern of each monitoring index.
[0038]
  Further, in order to achieve the above-mentioned object, the present invention provides, as described in claim 8, case data in which case-by-case operation that is artificially and mechanically performed on a plant is performed. The plant monitoring diagnostic apparatus according to claim 6 is provided.
[0039]
  In order to achieve the above-mentioned object, the present invention detects a deviation from a normal range of a monitoring index calculated from a process signal observed in a plant as an abnormal sign. Changes in the monitoring index that occur at the beginning and end of each planned and artificially performed operation on the plant, and the analog monitoring index X that changes first by the operation and the influence of the change Other analog monitoring indicators {Y i } And a period in which the change appears in advance, the analog monitoring index {Y from the time when the change accompanying the start of the operation is detected until the change accompanying the fixed period or the end is detected. i }, Relative change rate {Y i / X} is used as a new index, and a deviation from the normal range is detected as an abnormal sign.
[0040]
  In order to achieve the above object, according to the present invention, as described in claim 10, in the plant monitoring diagnosis apparatus according to claim 1, 2, or 6, the abnormal propagation path identification unit includes a monitoring index between the monitoring indices. Even if there is no apparent influence on the observed value of the monitoring index due to the introduction of a relationship that is kept constant by feedback control in the influence propagation characteristic and the transient change is suppressed by the feedback effect of the control system, the propagation path is Provided is a plant monitoring diagnostic apparatus characterized by being regarded as established.
[0041]
  In order to achieve the above-mentioned object, according to the present invention, as described in claim 11, in the plant monitoring diagnostic apparatus according to claim 1, 2, or 6, the abnormal propagation path identification unit includes a monitoring index between If a time delay is given to the effect propagation characteristics and an abnormal sign is detected in the monitoring indicator on the cause side in the monitoring processing unit and not detected in the monitoring indicator on the result side, the result side Provided is a plant monitoring diagnostic apparatus characterized by identifying an abnormal propagation path after re-evaluating abnormal signs by limiting a time range for extracting features of changes in monitoring indices and a range of normal values.
[0042]
  In order to achieve the above object, according to the present invention, as described in claim 12, in the plant monitoring diagnosis apparatus according to claim 1, 2, or 6, the abnormal propagation path identification unit includes a monitoring index. If a delay is given to the effect propagation characteristics and the identified propagation path forms a closed loop, the monitoring index in the closed loop where the abnormal sign is first detected is the starting point of the closed loop, and the two monitoring indices are If a different propagation path is apparently established, the true propagation path is estimated by comparing the time when the abnormal sign detection time of the cause side monitoring index is added with the propagation time delay with the time of detecting the abnormal sign of the monitoring index on the result side Provided is a plant monitoring diagnosis apparatus characterized by:
[0043]
  In order to achieve the above object, according to the present invention, as described in claim 13, in the plant monitoring diagnosis apparatus according to claim 1, 2, or 6, the abnormal propagation path identification unit is configured to detect abnormalities in the plant. The qualitative model includes the effect propagation characteristics that the operation of the protection sequence provided for the monitoring index has on the qualitative model, and the abnormal sign of the monitoring index observed from the time when the sequence operation is detected is predicted based on the qualitative model. There is provided a plant monitoring diagnostic apparatus characterized by monitoring a response of a plant by comparing with a sign of a subsequent monitoring index.
[0044]
  In order to achieve the above object, according to the present invention, as described in claim 14, in the plant monitoring diagnosis apparatus according to claim 1, 2, 6, or 10, the diagnosis result of the abnormal propagation path identification unit In the output display, draw a diagram that connects the name of the process signal corresponding to the monitoring index with three types of directed segments: propagation characteristics with positive gain, propagation characteristics with negative gain, and propagation characteristics showing the feedback effect. The result of classifying the observed change behavior of the indicator into three types of increase, decrease and no change is displayed together with the name of the process signal, and the name of the process signal identified as the starting point of the change is highlighted. Provided is a plant monitoring diagnostic apparatus characterized in that a directed line corresponding to an identified influence propagation path is displayed in a different color.
[0045]
  In order to achieve the above object, according to the present invention, as described in claim 15, in the plant monitoring diagnosis apparatus according to claim 1, 2, or 6, an output display of a diagnosis result by the abnormal propagation path identification unit In the plant system diagram, the process signal observation point and the signal name are clearly indicated, and the change behavior of the monitoring index calculated from the process signal is classified as increase, decrease, and no change. In addition to displaying the process signal corresponding to the starting point of change and the identified monitoring index and the upstream equipment that directly affects the signal, the control circuit that exists in the identified influence propagation path Provided is a plant monitoring / diagnosis device characterized in that the colors of piping, valves, pumps, etc. are changed and displayed.
[0046]
[Action]
  In the plant monitoring diagnosis apparatus of the present invention according to claim 1, when an abnormal sign is detected in the monitoring index, the monitoring index that is the starting point of the change by the network model that describes the effect propagation characteristics between the monitoring indices and its influence propagation Since the route is identified, it can be easily grasped whether or not all abnormal signs can be explained by one starting point.
[0047]
  In addition, it becomes clear that the plant system or subsystem from which the process signal used to calculate the monitoring index of the starting point was obtained is an abnormal source, and a causal table used for collation with the abnormal sign pattern observed in the abnormal cause identification Can be limited to a system or subsystem unit.
[0048]
  As a result, even if an abnormality that affects the process signal in a wide range of the plant occurs in a short time, if the abnormality can be assumed, the cause is diagnosed and an unexpected abnormality is detected. Even if it exists, at least the abnormal source and the influence propagation path can be presented.
[0049]
  In the plant monitoring diagnostic apparatus according to the present invention described in claim 2, when an unexpected abnormality is diagnosed by the plant monitoring diagnostic apparatus according to claim 1, the observed sign pattern is immediately registered in the causal table. Learning as one of the events and providing it as useful knowledge for subsequent diagnosisit can.
[0050]
  The plant monitoring diagnosis method of the present invention according to claim 3 detects a deviation from a normal range of a monitoring index calculated from a process signal observed in the plant as an abnormal sign, and displays a sign pattern of each monitoring index obtained. The results of diagnosis by collating with the causal table describing the symptom pattern of each monitoring index assumed for various abnormal causes, the cause of the possibility that there are many monitoring indices with the same signs Since the monitoring indicators are displayed in the form of causal tables in the order of the time of appearance of abnormal signs, they are displayed in the form of a causal table. It is possible to grasp the relationship between abnormal signs of each monitoring index.
[0051]
  Claims 4 and 5In the plant monitoring diagnostic apparatus of the present invention described, the first diagnostic means based on the monitoring index unique to the specific abnormality is an abnormality sign pattern of the monitoring index that can be calculated by processing that can be uniformly applied to many process signals. By providing it separately from the second diagnostic means based on collation with the causal table, the processing time when a short-circuit diagnosis is possible can be shortened. On the other hand, when it takes a long time to calculate the unique monitoring index, it is possible to eliminate the influence on the diagnosis processing time using the uniform monitoring index by performing such separation.
[0052]
  In the plant monitoring diagnostic apparatus of the present invention according to claim 6Only the monitoring index identified as the starting point of the abnormal change by the qualitative model can be used to easily search for similar cases and contribute to the estimation of the cause of the abnormality by comparing and comparing waveforms with the case data. By conducting a diagnosis based on a qualitative model first, the need for waveform matching for all monitoring indicators that show abnormal signsCan be eliminated.
[0053]
  Claims 7 and 8In the plant monitoring and diagnosis apparatus of the present invention described above, monitoring processing is executed even when a planned operation is performed, and when an abnormal sign is detected in the monitoring index due to the operation, the observed sign pattern is causal. The operation can be identified by collating with the symptom pattern of the table or by collating the waveform of the monitoring index identified as the starting point of the abnormal change by the qualitative model with the case data. Thereby, it is possible to confirm whether or not the operation of the plant monitoring diagnostic apparatus is normal each time a planned operation is performed.
[0054]
  Claim 9The plant monitoring diagnosis method of the present invention describedIn the monitoring targetThe monitoring indicators excluded from the above are input as analog monitoring indicators that can evaluate the magnitude of disturbance given to the plant by planned operation, and each analog monitoring indicator that changes under the influence is regarded as output, and the disturbance occurrence period By monitoring the transfer gain between the input and output, the behavior of the plant can be monitored as if an active disturbance application test was performed.
[0055]
  Claim 10In the plant monitoring diagnostic apparatus of the present invention described, feedback is provided in the middle of the propagation path by collating the symptom pattern of the monitoring index in consideration of the relation that it can be held constant by the feedback effect introduced into the qualitative model. Even if there is a monitoring index whose change is not observed due to the effect, it can be identified as a propagation path.
[0056]
  Claim 11In the plant monitoring diagnostic apparatus of the present invention described, the range for confirming the abnormal sign necessary for diagnosis based on the qualitative model is limited to before and after the time when a given time delay has elapsed from the time when the signal on the cause side changes. By setting the time zone, signs of changes can be detected with high sensitivity.
[0057]
  Claim 12The plant monitoring diagnosis apparatus of the present invention described can identify the starting point of the loop based on the abnormality detection time of each monitoring index even when the influence propagation path of the abnormality identified by the qualitative model becomes a loop. . Even when multiple propagation paths can be established between monitoring indices, the true propagation path is identified by comparing the time at which an abnormal sign of each monitoring index is detected with the time delay given to the propagation characteristics. be able to.
[0058]
  Claim 13In the plant monitoring diagnostic apparatus of the present invention described, the response of the process signal is predicted based on the operation of a protection sequence provided in case of a plant abnormality, and it is determined whether or not the observed behavior matches the predicted motion. By confirming sequentially, the plant state which changes every moment can be monitored in real time.
[0059]
  Claims 14 and 15In the plant monitoring and diagnosis apparatus of the present invention described above, the propagation identified as the monitoring index detected as an abnormal sign on the qualitative model according to claim 20 expressed by a signed directed graph including a feedback effect or the plant system diagram. By showing the route in different colors, it is possible to assist in easily grasping the relationship between abnormal signs of each monitoring index.it can.
[0060]
【Example】
  Embodiments of a plant monitoring diagnosis apparatus and method according to the present invention will be described below with reference to the drawings.
[0061]
  FIG. 1 is a block diagram showing a configuration of a first embodiment of a plant monitoring and diagnosing apparatus according to the present invention, and FIG. 2 is a flowchart showing a processing flow of the first embodiment. In FIG. 1, the control unit 1 has a function of controlling the execution of the entire apparatus, and periodically outputs a diagnosis execution command. The signal input unit 2 receives this execution command and inputs a process signal from the plant in accordance with input conditions such as a preset signal ID and sampling period.
[0062]
  The monitoring processing unit 3 calculates a monitoring index from a process signal input in accordance with a monitoring condition in which a signal processing method, an upper and lower limit threshold value indicating a normal range, etc. are set in advance for each monitoring index, and compares it with an upper and lower threshold value. Three types of signs are determined: “increase” for an index exceeding the threshold, “decrease” for an index below the lower threshold, and “no change” if within the normal range. The above “increase” or “decrease” is an abnormal change.
[0063]
  As the monitoring index, from the current value of the signal itself or the signal corresponding to the standard deviation obtained by the detection process, for example, an average value in the past certain period, a value obtained by smoothing the signal, or another 1 or A residual obtained by subtracting a value given as a function of a current value and / or a past value of a plurality of process signals as a predicted value is used. Since such signal processing does not focus on specific abnormal events, it can be applied uniformly to many signals.
[0064]
  The monitoring processing unit 3 stores the direction of change when an abnormal sign is first detected for each monitoring index as a sign. That is, once it is determined as “increase”, even if it changes to “no change” or “decrease”, the first sign “increase” is used as a sign for subsequent diagnosis. In this way, the monitoring processing unit 3 outputs a symptom pattern that is a set of symptom determination results for each monitoring index.
[0065]
  The anomalous propagation path identification unit 4 collates the qualitative model expressing the influence propagation characteristics between the respective monitoring indices registered in the qualitative model database 6 with a signed directed graph with the observed symptom pattern, and propagates the influence of the abnormal symptom. Identify the pathway. The monitoring index located at the starting point of this route is identified as having first changed, and the system or subsystem of the plant in which the process signal used to calculate the index is observed is determined to be an abnormal source. The function up to the identification of the monitoring index that is the starting point of the change is the function of the abnormality source identification unit. As shown in FIG. 2, the function of identifying the abnormality source system or subsystem is the function of the abnormality cause identification unit 5 described below. It may be included.
[0066]
  The abnormality cause identifying unit 5 is created for each system or subsystem of the plant, and is identified as an abnormal source by the abnormality propagation path identifying unit 4 from a plurality of causal tables registered in the causal table database 7. The causal table for the selected system or subsystem is read out and collated with the symptom pattern created by the monitoring processor 3. In the collation, for example, an assumed cause in which there is no contradiction in abnormal signs and the number of monitoring indexes with the same signs is large is identified as a candidate for the current abnormal cause. At this time, by including the symptom pattern observed by the planned operation in the causal table, it is possible to identify that the operation has been performed.
[0067]
  Hereinafter, the monitoring processing unit 3, the abnormal propagation path identification unit 4, the abnormal cause identification unit 5, and the qualitative model database 6 and the causal table database 7 surrounded by a broken line in FIG. Further, a portion obtained by removing the monitoring processing unit 3 from the monitoring diagnostic unit 8 is referred to as a diagnostic processing unit 9.
[0068]
  The control unit 1 sends the diagnosis result of the abnormal propagation path identification unit 4 to the output display unit 10, and also sends the result to the output display unit 10 when the cause candidate is identified in the abnormality cause identification unit 5. If the corresponding cause is not identified by the abnormality cause identifying unit 5, it is determined that an unexpected abnormality has occurred, and the current symptom pattern is sent to the database management unit 11.
[0069]
  In the database management unit 11, when a symptom pattern is sent from the control unit 1, it is additionally registered in the causal table database 7 as an unregistered pattern. Necessary information such as an event name is added to the symptom pattern learned in this way by the dialogue processing unit 12 when the cause is determined later.
[0070]
  The output display unit 10 displays various information sent from the control unit 1 in an easy-to-understand format.
[0071]
  In the following, further details will be described regarding abnormal propagation path identification based on a qualitative model, abnormal cause identification based on a causal table, and characteristics of the output display unit 10.
[0072]
  First, the features of the means provided by the present embodiment in order to solve the problems of the conventional method in the abnormal propagation path identification based on the qualitative model will be described in detail with reference to the drawings. The monitor signal name in the specific example uses the process signal name itself for simplification.
[0073]
  First, the introduction of feedback characteristics to the qualitative model will be described with reference to FIG. For example, in the pressure control system of a boiling water reactor, the control system adjusts the opening of the control valve so that the main steam pressure is constant. When the reactor pressure decreases, the main steam pressure also decreases, so the control system tries to increase the pressure by reducing the opening of the control valve. Therefore, between these observation signals, there exists a “promotion” relationship expressed by a solid directed line segment in the figure.
[0074]
  However, as shown in FIG. 3, even if the reactor pressure decreases as expressed by a downward gradient, the response of the control system becomes faster when the speed of the transient change is slow, resulting in the main steam pressure being reduced. No change can be seen as expressed by the → symbol, and in the conventional signed directed graph method, the transmission path of the transient change is cut before and after, as represented by the directed line segment of the dashed arrow in the figure.
[0075]
  Therefore, in this embodiment, by introducing a qualitative model that expresses a feedback characteristic that keeps the main steam pressure constant by increasing or decreasing the opening of the control valve, it propagates correctly even if no apparent change appears. Pathways can be identified.
[0076]
  Next, the introduction of a time delay to the effect propagation characteristics between monitoring indices will be described with reference to FIG. As shown in FIG. 4, when detecting signs of changes such as “increase”, “decrease”, and “no change” from the behavior of the monitoring index in the monitoring processing unit 3, for example, the maximum amplitude of the monitoring index observed in the normal time The upper and lower thresholds are set by setting the normal range to be twice the normal range. However, when the degree of transient change caused by an abnormality is small, it is accurately hidden behind normal fluctuations as in the example of signal C. May not be detected.
[0077]
  Therefore, in this embodiment, a transmission delay is introduced into the causal relationship between the monitoring indexes, and when a change in the monitoring indicator on the cause side is detected and no change is detected in the monitoring indicator on the result side, It is possible to detect with high accuracy by examining the change in the monitoring index on the result side in the limited range after the transmission delay time after the monitoring index changes, after limiting the normal range as well. As a result, the path can be correctly estimated even when the degree of transient change is small.
[0078]
  Similarly, as shown in FIG. 4, even when the identified influence propagation path is in a loop shape, the starting point of the loop is identified based on the abnormality detection time of each monitoring index, and a plurality of propagations are transmitted between the monitoring indices. Even when the path can be established, the true propagation path can be identified by comparing the time when the abnormal sign of each monitoring index is detected with the introduced time delay.
[0079]
  Further, a diagnostic method after the plant protection sequence is activated will be described with reference to FIG. The conventional diagnostic method based on a signed directed graph is based on the effect propagation characteristics between plant signals from phase I in normal operating state to phase II in which signs of abnormality are detected. This is to estimate a signal in which a change has occurred. Therefore, after the abnormality progresses, the protection sequence of the equipment or system is activated, the pump trips or the plant is scrammed, and the state after moving to the phase III stage where the influence propagation characteristics are different from the previous ones is displayed. It was difficult to diagnose.
[0080]
  Therefore, in this embodiment, as shown in FIG. 5, a digital signal that informs the operation of the protection sequence is taken in, for example, when the pump trips, the change in the signal is predicted from that point, and whether the plant response is normal or not. Can be diagnosed.
[0081]
  Hereinafter, the operation of the abnormal propagation path identification unit 4 will be described with reference to the flowchart shown in FIG.
[0082]
  When the monitoring processor 3 detects a sign of a signal change, the abnormal propagation path identification unit 4 is activated. First, the increase / decrease of the observation signal extracted by the monitoring processing unit 3 is input, and the following inspection is performed for each signal. When the signal of interest is increasing or decreasing, the normal transmission model (how to increase and decrease and the time delay) described in the qualitative model database 6 is referred to, and the change in the cause signal is If they match, the transmission path is saved assuming that a change has been transmitted between both signals, and if they do not match, the signal is regarded as a starting signal. If the starting signal represents the operation of the protection sequence, the sequence operation is confirmed, and if it is a normal process signal, the upstream device is stored as an abnormal point as the starting point of the change. If there is no change in the signal of interest, referring to the feedback model, if the movement of the causal signal matches the model, the feedback loop is saved as a transmission path, and if it does not match, Alternatively, if there is no feedback loop, the signal of interest is ignored as being irrelevant to the event.
[0083]
  By applying the above processing to all the signals, it is possible to identify the location where the abnormality has occurred and the route through which the influence is transmitted, or to confirm the protection sequence operation and its response.
[0084]
  Next, abnormality cause identification based on a causal table will be described. In the first embodiment described with reference to FIGS. 1 and 2, the system or subsystem in which the abnormal change has occurred is obtained from the monitoring index identified as the starting point of the abnormal change, and the causality given for each system or subsystem. The cause of the abnormality is identified by comparing the table with the observed abnormal sign pattern.
[0085]
  On the other hand, FIGS. 7A and 7B show the configuration of the monitoring diagnosis unit of the second embodiment and the causal table of the monitoring index, respectively, and FIG. 8 shows the processing flow. The same or corresponding parts as those in the first embodiment are described using the same reference numerals. The same applies to the following embodiments.
[0086]
  As shown in FIG. 7, in the abnormal propagation path identification unit 4, since the monitoring index that is the starting point of the abnormal change has been identified, it is assumed that an influence will appear on the index first for each monitoring index. It is also possible to register a causal table giving symptom patterns for various abnormal causes in the database 7a, and to perform cause identification based on the corresponding causal table for each monitoring index in the abnormal cause identifying unit 5. is there.
[0087]
  As described above, according to the second embodiment, by limiting the causal table to be collated with the symptom pattern from the system or subsystem unit to the monitoring index unit, it is possible to shorten the diagnostic processing time by the computer, Expected to reduce memory requirements and improve diagnostic accuracy.
[0088]
  FIGS. 9A and 9B show the configuration of the monitoring diagnostic unit and the causal table of the subsystem in the third embodiment of the plant monitoring diagnostic apparatus according to the present invention, respectively, and FIG. 9B shows the processing flow of the same monitoring diagnostic unit. 10 respectively. In the subsystem causality table database 7b, not only the symptom pattern is registered for various causes for which the system or subsystem is assumed to be an abnormal source for each system or subsystem of the plant, but also other systems Alternatively, when it is assumed that the influence of an abnormality that has occurred in the subsystem is propagated, it is registered including this, and the cause of the abnormality of another system or subsystem is also registered as an upstream cause.
[0089]
  The abnormality cause identifying unit 5 uses the relationship between each monitoring index given as a monitoring condition and the target system or subsystem, and from the monitoring index indicating an abnormal sign, a candidate for an abnormal source system or subsystem Find the group {Si}.
[0090]
  Next, by comparing the abnormality sign pattern created by the monitoring processing unit 3 with the causal table created for each system or subsystem, the abnormality cause candidate {Aji} for each abnormality source candidate Si and each candidate As shown in FIG. 9B, the upstream factor {Fji} given is identified, and the cause list La to be confirmed, which collects the abnormal cause candidates obtained for all the abnormal source candidates, is collected. The list Le is obtained.
[0091]
  Then, only abnormal cause candidates included in the confirmation target cause list La that are included in Le and whose upstream factors are not included in La are determined as final abnormal cause candidates. However, when there is only one abnormality source candidate, La to be confirmed becomes a list of abnormality cause candidates as it is.
[0092]
  As a result, in this embodiment, a system or subsystem that is not affected by other systems or subsystems in which abnormal signs are observed is determined as an abnormal source, and is identified based on the causal table for the system or subsystem. Among the abnormal cause candidates, those that can explain the abnormal signs of other systems or subsystems can be identified as current abnormal causes.
[0093]
  Further, in the example shown in FIG. 11, the abnormal signs of the second subsystem and the third subsystem are explained by cause 2-1 and cause 3-2, respectively, but both causes are caused by upstream cause 1-2. Since it will be described, the first subsystem is determined as the abnormal source. Furthermore, in the diagnosis result from the causal table of the first subsystem, two abnormal causes, 1-1 and 1-2, have been identified, but cause 1-1 cannot explain the abnormal signs of other subsystems Therefore, the cause of abnormality 1-2 is finally identified.
[0094]
  FIG. 12 is a block diagram showing the configuration of the monitoring diagnosis unit in the fourth embodiment of the plant monitoring diagnosis apparatus according to the present invention, and FIG. 13 is a flowchart showing the flow of the monitoring diagnosis process. In this embodiment, as in the case of the first embodiment shown in FIG. 1 and FIG. 2 and the second embodiment shown in FIG. 7 and FIG. By comparing the abnormal symptom pattern of each monitoring index created in step 1 with a qualitative model, the monitoring index that has started the change and its influence propagation path are identified.
[0095]
  The case data extraction unit 13 searches the time series data of the process signal at the time of occurrence of abnormality registered in the case database 14 for the case data having the same monitoring index identified as the starting point of change, and gives it as a monitoring condition. In accordance with the signal processing method, the time series data of the monitoring index is calculated from the process signal for each corresponding case. The comparison / matching unit 15 evaluates the similarity between the waveform of the time-series data of the monitoring index observed as the starting point of the change and the waveform of the extracted case data, and causes the abnormality of the case showing a high similarity of a certain value or more. Is diagnosed as the current cause of abnormality.
[0096]
  A method for evaluating the similarity will be described with reference to FIG. When the current data of the monitoring index identified as the starting point of the change and the waveform of one case data are as shown in the figure, a time series obtained by taking a fixed period t0 around the time when the abnormality was detected in the current data The data is a basic function f (t), and the following evaluation function with the case data x (t)
[Equation 3]

Is calculated. The maximum value of the evaluation function as a function of the delay time τ is used as the similarity. In the above formula, the case data extracted by the same method for the period t0 and evaluated becomes a cross-correlation function of both data, and this can be used as an evaluation function.
[0097]
  As in the case of cause identification using the causal table, it is possible to identify that the operation has been performed by including data when the planned operation is performed in the case database.
[0098]
  As described above, the same monitoring as usual is continued while the planned operation is performed, and the change of the process signal accompanying the operation is regarded as abnormal and the cause is determined. It can be considered as an effective means for confirming the soundness of functions and operations.
[0099]
  However, there is an idea that the influence of planned operations should not be regarded as abnormal and should be accurately identified. Based on this idea, FIG. 15 shows the configuration of the monitoring processing unit in the case of the fifth embodiment of the plant monitoring diagnostic apparatus according to the present invention. In the fifth embodiment, an operation identification unit 16 is provided in the preceding stage of the monitoring processing unit 3 of the monitoring diagnosis unit 8 of each embodiment shown in FIG. 1, FIG. 7, FIG. 9, or FIG. The one monitoring processing means 3a is the same as the monitoring processing unit 3 of the embodiment shown in the above figures.
[0100]
  In the operation information database 17, as shown in FIG. 16, for each planned operation, the name, the condition for determining the start of the operation, the condition for determining the end, the longest operation period, and the analog monitoring index that changes first depending on the operation i Is the analog monitoring index {Yji} that changes under the influence of the operation, the relative change rate {Yji / Xi} as a new index, the transfer gain when Xi is regarded as an input variable, and Yji as an output variable The normal gain range is registered.
[0101]
  Next, the operation of the monitoring processing units 3a and 3b in the fifth embodiment will be described in accordance with the processing flow shown in FIG.
[0102]
  The operation identification unit 16 shown in FIG. 15 calculates indices necessary for start and end determinations and confirms the respective determination conditions, and the end condition is satisfied only for the longest operation period from the time when the start of the operation i is confirmed. Until the first monitoring processing means 3a is sent a command to exclude Xi and {Yji}, which are monitoring indices that change due to the influence of the operation, from monitoring targets as monitoring bypass target indices, while the second monitoring processing means 3a An execution command is sent to 3b. In the second monitoring processing means 3b, as shown in FIG. 18, the relative change rate is calculated as a new monitoring index, and monitoring is performed by comparing with an upper limit threshold Hi and a lower limit threshold Li given as normal gains.
[0103]
  By adopting the configuration of the monitoring processing unit as described above, the monitoring processing by the first monitoring processing unit 3a as usual is performed for the monitoring index that is not affected by the operation even during the planned operation. Continuing, about the monitoring index which changes by operation, the 2nd monitoring process means 3b can determine whether the way of the change is normal.
[0104]
  FIG. 19 is a block diagram showing the configuration of the sixth embodiment of the plant monitoring diagnostic apparatus according to the present invention. In the figure, the second diagnosis means 8b has the same function as the monitoring diagnosis unit 8 shown in FIG. 1, FIG. 7, FIG. 9, or FIG. 12, and assumes an abnormal sign pattern obtained by uniform monitoring processing. Diagnosis is made by checking against possible anomalies, for example, with symptom patterns given in the form of a causal table. On the other hand, the first diagnosis means 8a detects the occurrence of the abnormality by calculating a monitoring index specific to the abnormality from a process signal that reacts only to the specific abnormality and comparing it with a threshold value indicating a normal range. For example, this is the case when an unstable flow phenomenon is detected by monitoring the amplitude of a vibration component of a specific period appearing in the flow signal.
[0105]
  19, the operation identifying unit 16 including the operation information database 17 shown in FIG. 15 and the second monitoring processing unit 3b are all included in the first diagnosis unit 8a, and the first monitoring processing unit 3a is the second monitoring unit. By including the execution control function of the operation identification unit 16 included in the diagnosis unit 8b in the control unit 1, the function of the monitoring processing unit having the configuration of FIG. 15 can be realized.
[0106]
  Next, an output display method of the plant monitoring diagnostic apparatus of the present invention will be described.
[0107]
  FIG. 20 shows an example of the display of the abnormal influence propagation path identification result by the qualitative model, and the process signal name itself is used for each monitoring index for simplification. It is equivalent to expressing the causal relationship between signals with a directed directed graph by connecting the names of the process signals with directed lines represented by solid lines, broken lines, and alternate long and short dash lines that show the positive and negative of the transfer characteristics and the type of feedback effect. Information.
[0108]
  The results of categorizing the observed signal qualitative behavior into “increase”, “decrease”, and “no change” are displayed together with the signal name in the direction of arrow upslope, downslope, and →, respectively. By highlighting the signal name of the starting point and changing the color and / or shading of the directed line corresponding to the path identified that the change has been transmitted, the location of the transient change and the propagation path can be easily recognized. Is displayed.
[0109]
  FIG. 21 shows display examples of different diagnosis results based on the qualitative model. In this case, the process signal observation points and signal names are clearly shown in the plant system diagram, and the qualitative behavior of the observed signals is classified as “increase”, “decrease”, and “no change”. It is displayed together with the name. Transient changes occur by highlighting the signal at the beginning of the transition and the equipment upstream of it, and changing the color of the control circuit, piping, valves, pumps, etc. that exist between the signals identified as having undergone the change. The location and the propagation path are displayed so that they can be easily recognized.
[0110]
  When displaying the diagnosis result based on the causal table, it is necessary to consider so as to easily grasp the relationship of the abnormal signs for each monitoring index. The example shown in FIG. 22 rearranges the causes with a high number of monitoring indicators that are consistent with the observed symptom patterns and have the same signs as the most likely cause candidates, and the monitoring indicators also show abnormal changes. They are rearranged and displayed in order of their occurrence time.
[0111]
  FIG. 23 shows an example in which the effect of the difference in the sign of each monitoring index on the diagnostic result is expressed by expressing the result of the event tree analysis performed when creating the causal table in a tree shape including the change of the monitoring index. It is shown.
[0112]
  FIG. 24 is a block diagram showing the configuration of the seventh embodiment of the plant monitoring diagnostic apparatus of the present invention. The operation identification unit 16 and the operation information database 17 shown in the figure are the same as those shown in FIGS. 24 is the same as the monitoring processing unit 3 of the monitoring diagnosis unit 8 of each embodiment shown in FIGS. 1, 7, 9, and 12, and is based on uniform processing. This is the part that monitors.
[0113]
  The abnormal propagation path identification unit 4 and the qualitative model database 6 in FIG. 24 are the same as the abnormal propagation path identification unit 4 and the qualitative model database 6 in the monitoring diagnosis unit 8 of each embodiment shown in FIGS. 1, 7, and 12. It is. The first abnormality cause identifying unit 5a and the causal table database 7 in FIG. 24 are the same as the abnormality cause identifying unit 5 and the causal table database 7 in the monitoring diagnosis unit 8 of each embodiment shown in FIGS.
[0114]
  Further, the second abnormality cause identifying unit 5b in FIG. 24 has a function of combining the case data extracting unit 13 and the comparison collating unit 15 of the embodiment shown in FIG. 12, and the case database 14 is the same in both figures. is there. Furthermore, the second monitoring processing unit 18b in FIG. 24 is a part that calculates a unique monitoring index that reacts to a specific abnormality and detects the occurrence of the abnormality, and is the same as the first diagnosis unit 8a in FIG. . Therefore, the second monitoring processing unit 18b can include the function of the second monitoring processing unit 3b of FIG. 15 as described in the description of FIG.
[0115]
【The invention's effect】
  As described above, according to the plant monitoring diagnostic apparatus of the present invention described in claim 1, when an abnormal sign is detected in the monitoring index, the network model describing the influence propagation characteristics between the monitoring indexes Since the monitoring index and its influence propagation path are identified, it is possible to easily grasp whether or not all abnormal signs can be explained by one starting point.
[0116]
  In addition, it becomes clear that the plant system or subsystem from which the process signal used to calculate the monitoring index of the starting point was obtained is an abnormal source, and a causal table used for collation with the abnormal sign pattern observed in the abnormal cause identification Can be limited to a system or subsystem unit.
[0117]
  As a result, even if an abnormality that affects the process signal in a wide range of the plant occurs in a short time, if the abnormality can be assumed, the cause is diagnosed and an unexpected abnormality is detected. Even if it exists, at least the abnormal source and the influence propagation path can be presented.
[0118]
  In this way, it is possible to detect and diagnose an abnormality occurring in a large-scale plant at the symptom stage while distinguishing it from the influence of planned operation. Further, even when the influence of an abnormality propagates between systems within a short period of time and changes appear in a wide range of process signals, it is possible to quickly identify the abnormality source. In any case, if it is possible to assume or has experienced an abnormality, it is possible to identify the cause or the cause of the abnormality in units of equipment. It becomes possible. Therefore, it is easy to grasp the situation at the time of abnormality, and early response at the symptom stage and quick return to normal are expected.Is done.
[Brief description of the drawings]
FIG. 1 is a block diagram showing the configuration of a first embodiment of a plant monitoring and diagnosing apparatus according to the present invention.
FIG. 2 is a flowchart showing a process flow of the first embodiment.
FIG. 3 is a diagram showing the effect of a qualitative model in which feedback characteristics are introduced.
FIG. 4 is a diagram showing the effect of time delay introduced in the effect propagation characteristics between monitoring indices.
FIG. 5 is a diagram showing an example of real-time diagnosis using a qualitative model.
FIG. 6 is a flowchart showing the flow of an abnormal propagation path identification process.
FIGS. 7A and 7B are a block diagram showing a configuration of a monitoring diagnostic unit of a second embodiment of the plant monitoring diagnostic apparatus according to the present invention, and a diagram showing a causal table of monitoring indices.
FIG. 8 is a flowchart showing a flow of monitoring diagnosis processing of the second embodiment.
FIG. 9 is a block diagram showing a configuration of a monitoring diagnosis unit of a third embodiment of the plant monitoring diagnosis apparatus of the present invention, and a diagram showing a subsystem causality table.
FIG. 10 is a flowchart showing a flow of monitoring diagnosis processing of the third embodiment.
FIG. 11 is a diagram illustrating a cause identification method according to a third embodiment.
FIG. 12 is a block diagram showing a configuration of a monitoring diagnosis unit of the fourth embodiment of the plant monitoring diagnosis apparatus according to the present invention.
FIG. 13 is a flowchart showing the flow of monitoring diagnosis processing of the fourth embodiment.
FIG. 14 is a diagram showing a method for comparing and collating with case data.
FIG. 15 is a block diagram showing a configuration of a monitoring processing unit of a fifth embodiment of the plant monitoring diagnostic apparatus according to the present invention.
FIG. 16 is a view showing the contents of an operation information database.
FIG. 17 is a flowchart showing the flow of monitoring processing of the fifth embodiment.
FIG. 18 is a diagram showing a monitoring method of second monitoring processing means of the fifth embodiment.
FIG. 19 is a block diagram showing a configuration of a sixth embodiment of the plant monitoring diagnostic apparatus according to the present invention.
FIG. 20 is an explanatory diagram showing a display example of an abnormal propagation path by a network between signals.
FIG. 21 is an explanatory diagram showing a display example of an abnormal propagation path by a plant system diagram.
FIG. 22 is an explanatory diagram showing a display example of a diagnosis result based on a causal table.
FIG. 23 is an explanatory diagram showing different display examples of diagnosis results based on a causal table.
FIG. 24 is a block diagram showing a configuration of a seventh embodiment of the plant monitoring diagnostic apparatus according to the present invention.
[Explanation of symbols]
1 Control unit
2 Signal input section
3 Monitoring processor
4 Abnormal propagation path identification part
5 Abnormal cause identification part
6 Qualitative model database
7 Causal table database
7a Causal table database by monitoring index
7b Subsystem causal table database
8 Monitoring and diagnosis department
8a First diagnostic means
8b Second diagnostic means
9 Diagnosis processing section
10 Output display
11 Database management department
12 Dialogue processing part
13 Case data extraction unit
14 Case database
15 Comparison and verification part
16 Operation identification part
17 Operation information database
18a First monitoring processing unit
18b Second monitoring processor

Claims (15)

Describes the deviation from the normal range of the monitoring index calculated from the process signal observed at the plant as an abnormal sign, and describes the effect propagation characteristics between each monitoring index and the monitoring processing unit that creates the symptom pattern of each monitoring index A qualitative model database in which the network model is registered, an observed propagation pattern that is compared with the network model, an abnormal propagation path identification unit that identifies an influence propagation path due to the monitoring index and anomaly change starting point , A causal table database in which a causal table describing symptom patterns for each of the various abnormal causes that are expected to have the first effect of abnormalities on the starting monitoring index, and the monitoring index identified as the starting point of abnormal changes and abnormality cause identifying unit to identify the cause of abnormality by collating the causal table and observed symptoms pattern corresponding to at least the identified affected Outputs 播経 path, plant monitoring diagnostic apparatus characterized by comprising an output display unit for outputting together abnormality cause if it contains signs pattern corresponding to a causal table.   2. The plant monitoring and diagnosing apparatus according to claim 1, further comprising a database management unit for performing learning by adding a causal table when a symptom pattern corresponding to the causal table is not included. .   Deviations from the normal range of monitoring indices calculated from process signals observed at the plant are detected as abnormal signs, and the resulting monitoring index signs are monitored for various abnormal causes. The results of diagnosis by collating with the causal table describing the symptom pattern of the symptom are rearranged in order of the most likely cause as the most likely cause candidate, and the monitoring index A plant monitoring diagnosis method, characterized by rearranging in order of time of appearance of abnormal signs and displaying them in the form of a causal table.   A first diagnostic means for calculating a monitoring index specific to the abnormality from a process signal that reacts only to a specific abnormality occurring in the plant and detecting a deviation from the normal range, and identifying the abnormality as a cause of the abnormality; and a process Given as a function of the current value of the signal and its detection signal, as a function of the average value over a certain period in the past, the value obtained by smoothing the signal, and the current and / or past value of at least one other process signal Causal in which the deviation from the normal range is detected as an abnormal sign using the residual obtained by subtracting the estimated value as the predicted value as the monitoring index, and the sign pattern of each monitoring index is described for various abnormal causes A second diagnostic means for identifying the cause of the abnormality by collating with the table, the first diagnostic means and the second diagnostic means are executed in parallel, and the first diagnostic means The results plant monitoring diagnostic apparatus and outputs in preference to the results of the second diagnostic means.   A first diagnostic means for calculating a monitoring index specific to the abnormality from a process signal that reacts only to a specific abnormality occurring in the plant and detecting a deviation from the normal range, and identifying the abnormality as a cause of the abnormality; and a process Given as a function of the current value of the signal and its detection signal, as a function of the average value over a certain period in the past, the value obtained by smoothing the signal, and the current and / or past value of at least one other process signal Causal in which the deviation from the normal range is detected as an abnormal sign using the residual obtained by subtracting the estimated value as the predicted value as the monitoring index, and the sign pattern of each monitoring index is described for various abnormal causes A second diagnostic means for identifying the cause of the abnormality by collating with the table, and when the cause of the abnormality is identified by executing the first diagnostic means, the result is output. To not execute the second diagnostic means, the plant monitoring and diagnosing apparatus cause applicable in the first diagnostic means and executes a second diagnosis means only if it is not identified. Describes the deviation from the normal range of the monitoring index calculated from the process signal observed at the plant as an abnormal sign, and describes the effect propagation characteristics between each monitoring index and the monitoring processing unit that creates the symptom pattern of each monitoring index An abnormal propagation path identification unit that identifies the monitoring index and its effect propagation path that started the abnormal change by comparing the network model and the observed symptom pattern, and registered that the monitoring index is the starting point of the change A case data extraction unit that searches for cases and calculates the monitoring index, and evaluates the similarity between the waveform of the temporal change of the monitoring index that is the starting point of the change and the waveform of the temporal change of the extracted case data and a comparison unit, the time-series data obtained by extracting duration t 0 of the change detection before and after the monitoring indicators which became the starting point for changing the basic function f (t), is calculated with the case data x (t) Evaluation function

A plant monitoring diagnosis apparatus characterized in that the abnormality cause of a case giving a high similarity not less than a certain value is output as a candidate for the current abnormality cause. 3. The plant monitoring diagnosis apparatus according to claim 1 , wherein the causal table includes a symptom pattern of each monitoring index assumed for a planned operation that is artificially and mechanically performed on the plant. . 7. The plant monitoring and diagnosis apparatus according to claim 6, wherein the case data includes case data in which a planned operation that is artificially and mechanically performed on the plant is performed.   When a deviation from the normal range of the monitoring index calculated from the process signal observed at the plant is detected as an abnormal sign, the start and end of every planned and artificially performed operation on the plant The change in the monitoring index that occurs, the analog monitoring index X that changes first by the operation, the other analog monitoring index {Yi} that may change under the influence, and the period in which the change appears are registered in advance. From the time when the change accompanying the start of the operation is detected to the time when the change accompanying the end of the fixed period or the end is detected, the analog monitoring index {Yi} has a normal change range with the relative change rate {Yi / X} as a new index. A plant monitoring diagnosis method characterized by detecting a deviation from the above as an abnormal sign. 7. The plant monitoring diagnosis apparatus according to claim 1, wherein the abnormal propagation path identification unit introduces a relationship held constant by feedback control in the influence propagation characteristics between monitoring indexes, and is transient due to a feedback effect of the control system. A plant monitoring diagnostic apparatus characterized in that a propagation path is considered to be established even when an influence does not appear apparently in an observation value of a monitoring index because a change is suppressed. In the plant monitoring diagnostic apparatus according to claim 1, 2, or 6 , the abnormal propagation path identification unit gives a time delay to the influence propagation characteristic between the monitoring indexes, and the monitoring processing unit has an abnormal sign in the monitoring indicator on the cause side. If it is detected but not detected in the monitoring indicator on the result side, the abnormal sign is re-established by limiting the time range for extracting the characteristics of the change in the monitoring indicator on the result side and the range of normal values based on this time delay. A plant monitoring diagnostic apparatus characterized by identifying an abnormal propagation path after evaluation. In the plant monitoring diagnostic apparatus according to claim 1, 2 or 6 , the abnormal propagation path identification unit gives a time delay to the effect propagation characteristics between the monitoring indices, and the identified propagation path constitutes a closed loop. The monitoring index in the closed loop in which the abnormal sign is first detected is the starting point of the closed loop, and when a different propagation path is apparently established between the two monitoring indices, the propagation time is delayed at the time of detecting the abnormal sign in the cause-side monitoring index A plant monitoring diagnostic apparatus characterized in that a true propagation path is estimated by comparing the time of adding to the abnormality sign detection time of the monitoring index on the result side. 7. The plant monitoring diagnosis apparatus according to claim 1, wherein the abnormality propagation path identification unit includes in the qualitative model the influence propagation characteristics that the operation of the protection sequence provided for the plant abnormality has on the monitoring index. Monitoring the response of the plant by comparing the abnormal sign of the monitoring index observed from the time when the sequence operation is detected with the sign of the subsequent monitoring index predicted based on the qualitative model. Diagnostic device. 11. The plant monitoring diagnosis apparatus according to claim 1, 2, 6 or 10, wherein in the output display of the diagnosis result by the abnormal propagation path identification unit, the gain is a positive propagation characteristic, the gain is a negative propagation characteristic, and a propagation characteristic showing a feedback effect. Draw a diagram connecting the names of process signals corresponding to monitoring indicators with the three types of directed segments, and process the results of classifying the observed change behavior of monitoring indicators into three types: increase, decrease, and no change. Displayed along with the signal name, highlighting the origin of the change and the name of the identified process signal, and changing the color of the directed line corresponding to the identified impact propagation path. A plant monitoring diagnostic device characterized by the above. 7. The plant monitoring diagnosis apparatus according to claim 1, wherein in the output display of the diagnosis result by the abnormal propagation path identification unit, the process signal observation point and the signal name are clearly shown in the plant system diagram, and the process signal Display the result of classifying the change behavior of the monitoring index calculated as increase, decrease, and no change together with the process signal name, and display the process signal corresponding to the monitoring index identified as the starting point of the change and its signal. Plant monitoring characterized by highlighting the equipment on the upstream side that has a direct impact and changing the color of the control circuit, piping, valves, pumps, etc. present in the identified impact propagation path Diagnostic device.

Images