JPWO2015011791A1 - Anomaly detection evaluation system - Google Patents

Anomaly detection evaluation system Download PDF

Info

Publication number
JPWO2015011791A1
JPWO2015011791A1 JP2015528053A JP2015528053A JPWO2015011791A1 JP WO2015011791 A1 JPWO2015011791 A1 JP WO2015011791A1 JP 2015528053 A JP2015528053 A JP 2015528053A JP 2015528053 A JP2015528053 A JP 2015528053A JP WO2015011791 A1 JPWO2015011791 A1 JP WO2015011791A1
Authority
JP
Japan
Prior art keywords
abnormality
sound
abnormal
degree
normal
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
JP2015528053A
Other languages
Japanese (ja)
Inventor
今朝明 峰村
今朝明 峰村
晋也 湯田
晋也 湯田
崇 佐伯
崇 佐伯
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Hitachi Ltd
Original Assignee
Hitachi Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Hitachi Ltd filed Critical Hitachi Ltd
Publication of JPWO2015011791A1 publication Critical patent/JPWO2015011791A1/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N29/00—Investigating or analysing materials by the use of ultrasonic, sonic or infrasonic waves; Visualisation of the interior of objects by transmitting ultrasonic or sonic waves through the object
    • G01N29/44—Processing the detected response signal, e.g. electronic circuits specially adapted therefor
    • G01N29/46—Processing the detected response signal, e.g. electronic circuits specially adapted therefor by spectral analysis, e.g. Fourier analysis or wavelet analysis
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01H—MEASUREMENT OF MECHANICAL VIBRATIONS OR ULTRASONIC, SONIC OR INFRASONIC WAVES
    • G01H3/00—Measuring characteristics of vibrations by using a detector in a fluid
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
    • G01M13/00—Testing of machine parts
    • G01M13/04—Bearings
    • G01M13/045—Acoustic or vibration analysis
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
    • G01M7/00—Vibration-testing of structures; Shock-testing of structures
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N29/00—Investigating or analysing materials by the use of ultrasonic, sonic or infrasonic waves; Visualisation of the interior of objects by transmitting ultrasonic or sonic waves through the object
    • G01N29/04—Analysing solids
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N29/00—Investigating or analysing materials by the use of ultrasonic, sonic or infrasonic waves; Visualisation of the interior of objects by transmitting ultrasonic or sonic waves through the object
    • G01N29/04—Analysing solids
    • G01N29/07—Analysing solids by measuring propagation velocity or propagation time of acoustic waves
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N29/00—Investigating or analysing materials by the use of ultrasonic, sonic or infrasonic waves; Visualisation of the interior of objects by transmitting ultrasonic or sonic waves through the object
    • G01N29/04—Analysing solids
    • G01N29/11—Analysing solids by measuring attenuation of acoustic waves
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N29/00—Investigating or analysing materials by the use of ultrasonic, sonic or infrasonic waves; Visualisation of the interior of objects by transmitting ultrasonic or sonic waves through the object
    • G01N29/44—Processing the detected response signal, e.g. electronic circuits specially adapted therefor
    • G01N29/4445—Classification of defects
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N2291/00—Indexing codes associated with group G01N29/00
    • G01N2291/02—Indexing codes associated with the analysed material
    • G01N2291/025—Change of phase or condition
    • G01N2291/0258—Structural degradation, e.g. fatigue of composites, ageing of oils

Abstract

診断対象機器の異常検知を行うシステムにおいて、診断対象機器の異常の程度と異常検出率との関係を評価することを可能とする。診断対象機器の異常検知を行うシステムにおいて、診断対象機器の異常の程度毎に模擬した異常音を用いて、異常の程度と異常検出率との対応関係を示す劣化曲線を算出する。In a system that detects an abnormality of a diagnosis target device, it is possible to evaluate the relationship between the degree of abnormality of the diagnosis target device and the abnormality detection rate. In a system for detecting an abnormality of a diagnosis target device, a deterioration curve indicating a correspondence relationship between the degree of abnormality and an abnormality detection rate is calculated using an abnormal sound simulated for each degree of abnormality of the diagnosis target device.

Description

本発明は、診断対象機器の音に基づいて正常か異常かを判定する正常異常判定部を備えた異常検知評価システムに関する。   The present invention relates to an abnormality detection / evaluation system including a normal / abnormality determination unit that determines normality / abnormality based on sound of a diagnosis target device.

本技術分野の背景技術として、特開２００３−２１５５５号公報（特許文献１）がある。特許文献１には、診断対象機械設備からの音に基づいて異常の有無を監視する異常監視装置に関し、模擬異音を用いて異常監視装置が異常と判定できるかを調べることにより、異常監視装置自体が正常であるかを判定する技術が記載されている。   As a background art of this technical field, there is JP-A-2003-21555 (Patent Document 1). Patent Document 1 relates to an abnormality monitoring device that monitors presence / absence of abnormality based on sound from a diagnosis target machine facility, and checks whether the abnormality monitoring device can be determined to be abnormal by using simulated abnormal noise. A technique for determining whether the device itself is normal is described.

特開２００３−２１５５５号公報JP 2003-21555 A

しかしながら、特許文献１では、異常の程度まで考慮した異常監視装置の評価や、異常検出率の評価については行われておらず、診断対象機器の異常検知を行うシステムにおいて、診断対象機器の異常の程度と異常検出率との関係を評価することができない。   However, in Patent Document 1, the evaluation of the abnormality monitoring device considering the degree of abnormality and the evaluation of the abnormality detection rate are not performed, and in the system that detects abnormality of the diagnosis target device, the abnormality of the diagnosis target device is not evaluated. The relationship between the degree and the abnormality detection rate cannot be evaluated.

上記課題を解決するために、本発明では、診断対象機器の異常検知を行うシステムにおいて、診断対象機器の異常の程度毎に模擬した異常音を用いて、異常の程度と異常検出率との対応関係を示す劣化曲線を算出する。   In order to solve the above-mentioned problem, in the present invention, in a system for detecting an abnormality of a diagnosis target device, correspondence between the degree of abnormality and the abnormality detection rate using an abnormal sound simulated for each degree of abnormality of the diagnosis target device A deterioration curve indicating the relationship is calculated.

具体的には、本発明の異常検知評価システムでは、例えば、診断対象機器からの音に基づいて前記診断対象機器が正常か異常かを判定する正常異常判定部と、前記診断対象機器の異常の程度毎に異常音を模擬する異常模擬音合成部と、前記異常模擬音合成部で模擬した異常音を用いて前記正常異常判定部で正常か異常かを複数回判定した場合の異常検出率と前記模擬した異常音に対応する前記異常の程度との対応関係を示す劣化曲線を算出する劣化曲線算出部とを有することを特徴とする。   Specifically, in the abnormality detection evaluation system of the present invention, for example, a normal abnormality determination unit that determines whether the diagnosis target device is normal or abnormal based on sound from the diagnosis target device, and abnormality of the diagnosis target device An abnormality simulation sound synthesizer that simulates an abnormal sound for each degree, and an abnormality detection rate when the normal abnormality determination unit uses the abnormal sound simulated by the abnormality simulation sound synthesizer to determine whether it is normal or abnormal multiple times A deterioration curve calculating unit that calculates a deterioration curve indicating a correspondence relationship with the degree of abnormality corresponding to the simulated abnormal sound.

本発明によれば、診断対象機器の異常検知を行うシステムにおいて、診断対象機器の異常の程度毎に模擬した異常音を用いて、異常の程度と異常検出率との対応関係を示す劣化曲線を算出することで、診断対象機器の異常の程度と異常検出率との関係を評価することができる。   According to the present invention, in a system for detecting an abnormality of a diagnosis target device, a deterioration curve indicating a correspondence relationship between the degree of abnormality and the abnormality detection rate is obtained using an abnormal sound simulated for each degree of abnormality of the diagnosis target device. By calculating, it is possible to evaluate the relationship between the degree of abnormality of the diagnosis target device and the abnormality detection rate.

これにより、実際の診断対象機器に異常が発生する前に異常検知を行うシステムの検証が可能となり、検証コスト、ならびに検証時間をかけずに異常検知システムを開発することが可能となる。   This makes it possible to verify a system that detects an abnormality before an abnormality occurs in an actual diagnosis target device, and to develop an abnormality detection system without spending verification cost and verification time.

異常検知評価システムの構成図の例である。It is an example of the block diagram of an abnormality detection evaluation system. 劣化曲線の例である。It is an example of a deterioration curve. データベースの例である。It is an example of a database.

本発明の実施例を、図面を参照しながら説明する。尚、各図において、同一又は類似の構成要素には同じ符号を付し、説明を省略する。   Embodiments of the present invention will be described with reference to the drawings. In the drawings, the same or similar components are denoted by the same reference numerals and description thereof is omitted.

図１は、異常検知評価システムの構成図の例である。本実施例の異常検知評価システムは、診断対象機器の正常異常判定と、その評価のための劣化曲線の算出を行う装置の例である。本実施例の異常検知評価システム００１は、データ取り込み部００３、異常程度入力部００４、データベース（以下、ＤＢ）００５、異常模擬音合成部００６、正常異常判定部００７、劣化曲線算出部００８、出力部００９、異常程度推定部０１０で構成される。尚、異常程度推定部０１０の機能は不要であれば省略しても良い。   FIG. 1 is an example of a configuration diagram of an abnormality detection evaluation system. The abnormality detection evaluation system of the present embodiment is an example of an apparatus that performs normal / abnormal determination of a diagnosis target device and calculates a deterioration curve for the evaluation. The abnormality detection evaluation system 001 of the present embodiment includes a data capturing unit 003, an abnormality degree input unit 004, a database (hereinafter referred to as DB) 005, an abnormal simulated sound synthesis unit 006, a normal abnormality determination unit 007, a deterioration curve calculation unit 008, and an output. Part 009 and an abnormality degree estimation part 010. Note that the function of the abnormality degree estimation unit 010 may be omitted if unnecessary.

本実施例の異常検知評価システム００１は、通常の診断時には、診断対象機器００２からの音を例えばマイクなどの音響センサを用いた音声入力部０１１で集音し、直接あるいは予め録音しておいてから、データ取り込み部００３にてその音のデータを取り込み、その音のデータに基づいて正常異常判定部００７にて診断対象機器０００２が正常か異常かを判定する。判定結果は、出力部００９によって出力される。出力部００９を例えばディスプレイで構成し、判定結果を表示するようにしても良いし、出力部００９から判定結果の信号を出力し、異常検知評価システム００１の外部のディスプレイに対して出力部００９が判定結果を表示するようにしても良い。あるいは、ディスプレイに表示はせず、異常検知評価システム００１の外部の遠隔監視装置などに、出力部００９から判定結果の信号を出力するようにしても良い。   The abnormality detection evaluation system 001 according to the present embodiment collects sound from the diagnosis target device 002 with a voice input unit 011 using an acoustic sensor such as a microphone and records it directly or in advance during normal diagnosis. Then, the data acquisition unit 003 acquires the sound data, and the normal / abnormality determination unit 007 determines whether the diagnosis target device 0002 is normal or abnormal based on the sound data. The determination result is output by the output unit 009. The output unit 009 may be configured with a display, for example, so that the determination result is displayed, or a signal of the determination result is output from the output unit 009, and the output unit 009 is connected to a display external to the abnormality detection evaluation system 001. The determination result may be displayed. Alternatively, the determination result signal may be output from the output unit 009 to a remote monitoring device outside the abnormality detection evaluation system 001 without displaying on the display.

次に、診断対象機器の異常の程度毎の異常音の模擬について説明する。   Next, simulation of abnormal sound for each degree of abnormality of the diagnosis target device will be described.

異常の程度毎の異常音の模擬を行う場合には、データ取り込み部００３から入力された音のデータは、異常模擬音合成部００６に入力される。また、例えばキーボードなどの異常程度入力部００４から、異常の程度が入力され、異常模擬音合成部００６に送られる。ここで、異常の程度としては、例えば診断対象機器００２がベアリングで、ベアリングの軸ずれを診断する場合においては、軸ずれの大きさ（０．０１ｍｍ、０．０５ｍｍ、０．１０ｍｍ等）、もしくは、初期、中期、末期といった入力方法が考えられる。   When simulating abnormal sound for each degree of abnormality, sound data input from the data capturing unit 003 is input to the abnormal simulated sound synthesizing unit 006. Also, for example, the degree of abnormality is input from an abnormality degree input unit 004 such as a keyboard, and is sent to the abnormality simulation sound synthesis unit 006. Here, as the degree of abnormality, for example, when the diagnosis target device 002 is a bearing and diagnoses a bearing shaft misalignment, the size of the shaft misalignment (0.01 mm, 0.05 mm, 0.10 mm, etc.), or Input methods such as initial, mid-term, and final term are conceivable.

異常模擬音合成部００６は、データ取り込み部００３から入力された音のデータである正常な診断対象機器００２の稼動音と、異常の程度に対応した特徴データとから、模擬した異常音を生成する。具体的には、異常模擬音合成部００６は、ＤＢ００５を参照して、異常の程度に対応する異常音の特徴データを取り込み、正常な診断対象機器００２の稼動音を、異常の程度に対応する異常音の特徴データを反映した音となるように信号処理することで、異常音を模擬して生成する。   The abnormal simulated sound synthesizing unit 006 generates a simulated abnormal sound from the operating sound of the normal diagnosis target device 002 that is the sound data input from the data capturing unit 003 and the feature data corresponding to the degree of abnormality. . Specifically, the abnormal simulated sound synthesizer 006 refers to the DB 005, captures abnormal sound feature data corresponding to the degree of abnormality, and corresponds the operating sound of the normal diagnosis target device 002 to the degree of abnormality. An abnormal sound is simulated and generated by performing signal processing so that the sound reflects the characteristic data of the abnormal sound.

図３は、データベースの例である。ＤＢ００５には、異常の程度として、軸ずれの大きさ０．０１ｍｍ、０．０５ｍｍ、０．１０ｍｍの３つを用意しそれぞれ、初期・中期・末期と定義する。ベアリングの軸ずれの場合、診断対象機器００２からの音の周波数において、本来生じない周波数帯域にピークが生じる。そこで、ＤＢ００５には、異常の程度に対応した特徴データとして、正常周波数、異常周波数、振幅が定義されており、例えば、軸ずれ０．０１ｍｍの場合には正常周波数３０Ｈｚ、異常周波数３２Ｈｚ、振幅＋１ｄＢ、軸ずれ０．０５ｍｍの場合には正常周波数３０Ｈｚ、異常周波数３５Ｈｚ、振幅＋２ｄＢ、軸ずれ０．１０ｍｍの場合には正常周波数３０Ｈｚ、異常周波数４０Ｈｚ、振幅＋３ｄＢの特徴データが格納されている。尚、これはあくまでＤＢ００５の一例を示しただけであり、図３に示した例に限定されるものではない。   FIG. 3 is an example of a database. DB005 prepares three types of abnormalities of the axis deviations of 0.01 mm, 0.05 mm, and 0.10 mm, which are defined as initial, intermediate, and final stages, respectively. In the case of a bearing shaft misalignment, a peak occurs in a frequency band that does not originally occur in the frequency of sound from the diagnosis target device 002. Therefore, the normal frequency, abnormal frequency, and amplitude are defined in the DB005 as feature data corresponding to the degree of abnormality. For example, when the axis deviation is 0.01 mm, the normal frequency is 30 Hz, the abnormal frequency is 32 Hz, and the amplitude is +1 dB. When the axis deviation is 0.05 mm, characteristic data of normal frequency 30 Hz, abnormal frequency 35 Hz, amplitude +2 dB, and when axis deviation 0.10 mm, normal frequency 30 Hz, abnormal frequency 40 Hz, amplitude +3 dB is stored. Note that this is only an example of DB005, and is not limited to the example shown in FIG.

異常模擬音合成部００６は、例えば異常の程度が軸ずれ０．０１ｍｍの場合には、データ取り込み部００３から入力された音（正常音）に対して、正常周波数である３０Ｈｚの音を小さくする、あるいは、なくし、その代わりに本来生じない周波数帯域である異常周波数の３２Ｈｚに振幅＋１ｄＢのピーク周波数を加える。ピークの加え方としては、正弦波を足し合わせる、あるいは、解析信号による初期位相を考慮した形での足し合わせ等が考えられる。例えば、３２Ｈｚが本来生じない周波数帯域であるとする。その場合、３２Ｈｚの正弦波もしくは解析信号を作成し、正常音の時間波形に足し算を行う。解析信号については、ヒルベルト変換により算出可能である。尚、正常周波数の音を小さくする、あるいは、なくすことを行わず、単に異常周波数の音を加えるだけとしても良い。その場合、ＤＢ００５の特徴データにおいては正常周波数の項目は不要である。   For example, when the degree of abnormality is an axis deviation of 0.01 mm, the abnormal simulated sound synthesizer 006 reduces the sound of 30 Hz that is a normal frequency with respect to the sound (normal sound) input from the data capturing unit 003. Alternatively, the peak frequency with an amplitude of +1 dB is added to the abnormal frequency of 32 Hz, which is a frequency band that does not occur originally. As a method of adding the peak, it is conceivable to add sine waves or to add in a form that takes into account the initial phase based on the analysis signal. For example, suppose that 32 Hz is a frequency band that does not originally occur. In that case, a 32 Hz sine wave or analysis signal is created and added to the time waveform of the normal sound. The analytic signal can be calculated by Hilbert transform. It should be noted that it is possible to simply add an abnormal frequency sound without reducing or eliminating the normal frequency sound. In that case, the item of normal frequency is not necessary in the feature data of DB005.

異常模擬音合成部００６における異常模擬音合成の一例を説明する。異常模擬音合成部００６は、データ取り込み部００３から入力された正常データから、音の特徴（例えば、周波数、振幅）を抽出する。音の特徴抽出方法については、周波数・振幅については、フーリエ変換による方法が広く行われている。計算機上においては、デジタルフーリエ変換（ＤＦＴ）およびＤＦＴを高速に計算するアルゴリズムであるＦＦＴが広く用いられている。フーリエ変換を行う際には、時間軸での窓長の設定ならびに適切な窓掛けが必要となる。窓長の設定により、取り扱うことが可能となる周波数（周波数分解能）が変化するためである。診断対象機器００２がモータ、ギア、ベアリング等の回転機器である場合、１回転相当の時間を窓長として設定することが好ましい。また、窓については、矩形窓等が望ましい。音の特徴として、位相（時間変動）を用いてもよい。位相（時間変動）については、時間情報を含むため直接取り扱うことが難しい。そのため、位相（時間変動）を直接扱わず、時間変動を示す別の特徴へ置き換えることが好ましい。例えば、フーリエ変換を行ってピーク周波数を算出し、ピーク周波数の時間変動を捕らえる方法、あるいは、スペクトルの零点に着目する方法等が考えられる。どのような音の特徴を抽出するかは、正常異常判定部００７の診断アルゴリズムに応じて定めればよい。そして、音の特徴毎に、ＤＢ００５に登録された特徴データを参照して、正常音の特徴に異常音の特徴を合成し、合成された音の特徴を有する音のデータに変換することで、模擬した異常音が生成される。尚、異常の程度と異常音との関係としては、例えば、異常の程度毎に音の高低を示す周波数が変動する、もしくは、音の大きさが変わる、もしくは、周波数が時間ごとに変動するといった例があげられる。   An example of the abnormal simulated sound synthesis in the abnormal simulated sound synthesis unit 006 will be described. The abnormal simulated sound synthesizer 006 extracts sound characteristics (for example, frequency and amplitude) from normal data input from the data capture unit 003. As a sound feature extraction method, a method using Fourier transform is widely used for frequency and amplitude. On a computer, digital Fourier transform (DFT) and FFT which is an algorithm for calculating DFT at high speed are widely used. When performing the Fourier transform, it is necessary to set the window length on the time axis and appropriate windowing. This is because the frequency (frequency resolution) that can be handled changes depending on the setting of the window length. When the diagnosis target device 002 is a rotating device such as a motor, a gear, or a bearing, it is preferable to set a time corresponding to one rotation as the window length. As for the window, a rectangular window or the like is desirable. As a feature of sound, phase (time variation) may be used. The phase (time variation) is difficult to handle directly because it includes time information. Therefore, it is preferable not to directly handle the phase (time variation) but to replace it with another feature indicating the time variation. For example, a method of calculating a peak frequency by performing Fourier transform and capturing time fluctuation of the peak frequency, a method of paying attention to the zero point of the spectrum, or the like can be considered. What kind of sound feature should be extracted may be determined according to the diagnosis algorithm of the normal / abnormality determination unit 007. Then, by referring to the feature data registered in DB 005 for each sound feature, the feature of the abnormal sound is synthesized with the feature of the normal sound, and converted into the sound data having the synthesized sound feature, Simulated abnormal sound is generated. The relationship between the degree of abnormality and the abnormal sound is, for example, that the frequency indicating the level of the sound fluctuates for each degree of abnormality, the sound volume changes, or the frequency fluctuates with time. An example is given.

このように、正常な診断対象機器００２の稼動音と、異常の程度に対応した特徴データとから、模擬した異常音を生成することにより、正常な診断対象機器００２の稼動音を使わずに異常音を模擬する場合に比べて、正常異常判定部００７を評価する際に実際の診断対象機器の正常異常判定に近い評価を行うことが可能になる。また、異常の程度を考慮して模擬した異常音を生成することで、異常の程度まで考慮した正常異常判定部００７の評価が可能となる。   In this way, by generating a simulated abnormal sound from the operating sound of the normal diagnosis target device 002 and the characteristic data corresponding to the degree of abnormality, the abnormal sound can be detected without using the operating sound of the normal diagnosis target device 002. Compared to the case of simulating sound, when evaluating the normal / abnormality determination unit 007, it is possible to perform an evaluation closer to the normal / abnormal determination of the actual diagnosis target device. In addition, by generating an abnormal sound that simulates the degree of abnormality, the normal abnormality determination unit 007 can be evaluated in consideration of the degree of abnormality.

次に、異常の程度毎に模擬した異常音を用いた劣化曲線の算出について説明する。   Next, calculation of a deterioration curve using an abnormal sound simulated for each degree of abnormality will be described.

正常異常判定部００７は、異常模擬音合成部００６で模擬した異常音を用いて正常か異常かを判定する。この判定を複数回行った場合の異常検出率と、模擬した異常音に対応する異常の程度との対応関係を示すのが劣化曲線であり、劣化曲線算出部００８は、正常異常判定部００７と異常程度入力部００４とのデータに基づいてこの劣化曲線を算出する。   The normality / abnormality determination unit 007 determines whether the sound is normal or abnormal by using the abnormal sound simulated by the abnormality simulation sound synthesis unit 006. The deterioration curve indicates the correspondence between the abnormality detection rate when this determination is performed a plurality of times and the degree of abnormality corresponding to the simulated abnormal sound, and the deterioration curve calculation unit 008 is a normal abnormality determination unit 007. This deterioration curve is calculated based on data with the abnormality degree input unit 004.

具体的には、正常異常判定部００７は、正常データを予め学習しておき、異常模擬音合成部００６で模擬した異常音を用いて診断を行い、診断対象機器００２の正常異常判定を行う。尚、正常データを予め学習しておくので、模擬した異常音を用いて診断を行う場合にはデータ取り込み部００３からの音データは正常異常判定部００７に入力しなくてもよい。本実施例では、例えばクラスタリングによる方法で判定を行う。正常異常判定部００７は、模擬した異常音について、予め学習しておいて正常データからの統計的な離れ具合である統計距離を算出し、閾値と比較することにより正常か異常かを判定する。算出される統計距離は標準偏差であり、閾値としては例えば標準偏差３（σ）を用いる。標準偏差３（σ）は、統計的な観点で９９．９９％包含される指標である。そのため、正常音と診断データ（サンプル）とで比較を行い統計距離の差が３（σ）以上であれば、正常時に含まれていない値が出現したとみなすことができる。また、クラスタリング方法については、ｋ−ｍｅａｎｓ法等が考えられる。   Specifically, the normal / abnormality determination unit 007 learns normal data in advance, performs diagnosis using the abnormal sound simulated by the abnormal simulation sound synthesis unit 006, and performs normal / abnormal determination of the diagnosis target device 002. Since normal data is learned in advance, the sound data from the data capturing unit 003 may not be input to the normal / abnormality determination unit 007 when diagnosis is performed using simulated abnormal sound. In the present embodiment, the determination is performed by a method using clustering, for example. The normal / abnormality determination unit 007 learns about the simulated abnormal sound in advance, calculates a statistical distance that is a statistical distance from normal data, and compares it with a threshold value to determine whether it is normal or abnormal. The calculated statistical distance is a standard deviation, and for example, standard deviation 3 (σ) is used as a threshold value. The standard deviation 3 (σ) is an index included 99.99% from a statistical viewpoint. Therefore, a comparison is made between normal sound and diagnostic data (sample), and if the difference in statistical distance is 3 (σ) or more, it can be considered that a value not included in normal time has appeared. As a clustering method, a k-means method or the like can be considered.

また、異常検出率を得るために、正常異常判定部００７は、異なるサンプルを用いて複数回の判定を行う。例えば、ある異常の程度に対応する模擬した異常音として１００秒間のデータを用い１秒間を１サンプルとする、もしくは、回転機器においてある異常の程度に対応する模擬した異常音として１００回転分のデータを用い、１回転を１サンプルとするなどして、１００サンプルのデータを準備し、正常異常判定部００７は各サンプル毎に正常か異常かを判定することで、１００回の判定を行う。ある異常の程度に対応する模擬した異常音を用いてＡ回の判定を行い、異常と判定された回数がＢ回であれば、（Ｂ／Ａ）×１００（％）が当該異常の程度における異常検出率である。本実施例では、この異常検出率を、異常の程度毎に求める。尚、異常検出率は、正常異常判定部００７で算出してもよいが、これに限定されず、正常異常判定部００７は正常か異常かを判定して出力するだけとし、次の劣化曲線算出部００８で異常検出率を算出してもよい。   In addition, in order to obtain an abnormality detection rate, the normal / abnormality determination unit 007 performs a plurality of determinations using different samples. For example, 100 seconds of data is used as a simulated abnormal sound corresponding to a certain degree of abnormality, and one second is taken as one sample, or data for 100 rotations as a simulated abnormal sound corresponding to a certain degree of abnormality in a rotating device. 100 samples of data are prepared by making one rotation one sample, and the normality / abnormality determination unit 007 determines 100 times by determining whether each sample is normal or abnormal. If a simulated abnormal sound corresponding to a certain degree of abnormality is used to perform A determination and the number of times determined as abnormal is B, (B / A) × 100 (%) is the degree of the abnormality Anomaly detection rate. In this embodiment, this abnormality detection rate is obtained for each degree of abnormality. The abnormality detection rate may be calculated by the normal / abnormality determination unit 007. However, the present invention is not limited to this, and the normal / abnormality determination unit 007 only determines whether or not it is normal and outputs it, and calculates the next deterioration curve. The part 008 may calculate the abnormality detection rate.

劣化曲線算出部００８は、異常の程度と、その異常の程度における異常検出率とを対応付けることで、劣化曲線を算出する。   The deterioration curve calculation unit 008 calculates a deterioration curve by associating the degree of abnormality with the abnormality detection rate at the degree of abnormality.

図２は、劣化曲線の例である。図２では、劣化曲線２００として、横軸に異常の程度（ｍｍ）、縦軸に異常検出率（％）をとった例を示している。図２では、異常の程度として軸ずれ０．０１ｍｍの場合は異常検出率が５％、異常の程度として軸ずれ０．０５ｍｍの場合は異常検出率が２０％、異常の程度として軸ずれ０．１０ｍｍの場合は異常検出率が７０％の例を示している。劣化曲線２００は、異常の程度と異常検出率との数組の組み合わせで構成されているが、劣化曲線算出部００８は、実際に異常検出率を算出するのに用いた異常の程度だけを用いて劣化曲線２００を算出してよいし、数組の組み合わせから間を直線または曲線で補完して劣化曲線２００を算出してもよい。   FIG. 2 is an example of a deterioration curve. FIG. 2 shows an example of the deterioration curve 200 in which the horizontal axis represents the degree of abnormality (mm) and the vertical axis represents the abnormality detection rate (%). In FIG. 2, when the axis deviation is 0.01 mm as the degree of abnormality, the abnormality detection rate is 5%. When the axis deviation is 0.05 mm as the degree of abnormality, the abnormality detection rate is 20%. In the case of 10 mm, an example in which the abnormality detection rate is 70% is shown. The deterioration curve 200 is composed of several combinations of the degree of abnormality and the abnormality detection rate, but the deterioration curve calculation unit 008 uses only the degree of abnormality actually used to calculate the abnormality detection rate. The deterioration curve 200 may be calculated, or the deterioration curve 200 may be calculated by complementing a gap between several combinations with a straight line or a curve.

出力部００９は、算出された劣化曲線２００を出力する。例えば、出力部００９は、異常の程度毎の異常検出率をディスプレイなどで表示する。図２は、出力部００９で表示された劣化曲線２００の一例である。   The output unit 009 outputs the calculated deterioration curve 200. For example, the output unit 009 displays an abnormality detection rate for each degree of abnormality on a display or the like. FIG. 2 is an example of the deterioration curve 200 displayed on the output unit 009.

このように、異常の程度毎に模擬した異常音を用いて異常の程度と異常検出率との対応関係を示す劣化曲線を算出することにより、正常異常判定部７の正常か異常かの判定の精度や、異常の程度と異常検出率との関係を評価、検証することが可能となる。   In this way, by calculating the deterioration curve indicating the correspondence between the degree of abnormality and the abnormality detection rate using the abnormal sound simulated for each degree of abnormality, the normal abnormality determining unit 7 can determine whether the abnormality is normal or abnormal. It is possible to evaluate and verify the accuracy and the relationship between the degree of abnormality and the abnormality detection rate.

例えば、正常異常判定部７の判定方法や閾値を調整し、再び劣化曲線２００を算出して評価、検証することにより、正常異常判定部７が必要な性能を得られるように調整するといった利用方法も考えられる。   For example, a method of adjusting the normal / abnormality determination unit 7 so that the normal / abnormality determination unit 7 can obtain the required performance by adjusting the determination method and threshold value, calculating the degradation curve 200 again, and evaluating and verifying it. Is also possible.

劣化曲線２００の他の利用方法の例として、診断対象機器００２からの音のデータを用いて診断を行った場合の異常検出率から、劣化曲線２００を用いて異常の程度を逆算して推定することで異常の程度を把握し、診断対象機器の保全計画や稼動計画を立てるという利用方法も考えられる。   As an example of another method of using the deterioration curve 200, the degree of abnormality is estimated by back-calculating using the deterioration curve 200 from the abnormality detection rate when diagnosis is performed using sound data from the diagnosis target device 002. Therefore, it is possible to grasp the degree of abnormality and to make a maintenance plan and an operation plan for the diagnosis target device.

この場合、異常検知評価システム００１は、診断対象機器００２からの音を用いて正常異常判定部７で正常か異常かを複数回判定した場合の異常検出率と、劣化曲線２００とから、診断対象機器００２の異常の程度を推定する異常程度推定部０１０を備える。まず、これまでに説明した方法で予め劣化曲線２００を算出しておく。そして、実際に診断したい診断対象機器００２からの音（劣化曲線２００を算出したときの音とは異なる）をデータ取り込み部００３で取り込み、異常検出率を求めるために正常異常判定部７で複数回の判定を行う。そして、正常異常判定部７で異常検出率を求めてから異常程度推定部０１０に送る、あるいは、正常異常判定部７の複数回の判定結果を異常程度推定部０１０に送るって異常程度推定部０１０で異常検出率を求める。異常程度推定部０１０は、異常検出率と、劣化曲線算出部００８で予め算出された劣化曲線２００とに基づいて、劣化曲線２００上で異常検出率に対応する異常の程度を求めることで、今回診断を行った診断対象機器００２の異常の程度を推定し、結果を出力部００９に送る。出力部００９は、例えば結果をディスプレイに表示するなどして出力する。これにより、異常の程度を把握することが可能となり、診断対象機器の保全計画や稼動計画を立てることも可能となる。   In this case, the abnormality detection evaluation system 001 uses the sound from the diagnosis target device 002 to determine whether the normal abnormality determination unit 7 determines normal or abnormal plural times and the abnormality detection rate and the deterioration curve 200. An abnormality degree estimation unit 010 that estimates the degree of abnormality of the device 002 is provided. First, the deterioration curve 200 is calculated in advance by the method described so far. Then, the sound from the diagnosis target device 002 to be actually diagnosed (different from the sound when the deterioration curve 200 is calculated) is captured by the data capturing unit 003, and the normal / abnormal determination unit 7 performs a plurality of times to obtain the abnormality detection rate. Judgment is made. Then, after obtaining the abnormality detection rate in the normal / abnormality determination unit 7, it is sent to the abnormality level estimation unit 010, or a plurality of determination results of the normal / abnormality determination unit 7 are sent to the abnormality level estimation unit 010. At 010, the abnormality detection rate is obtained. The abnormality degree estimation unit 010 obtains the degree of abnormality corresponding to the abnormality detection rate on the deterioration curve 200 based on the abnormality detection rate and the deterioration curve 200 calculated in advance by the deterioration curve calculation unit 008. The degree of abnormality of the diagnosis target device 002 that performed the diagnosis is estimated, and the result is sent to the output unit 009. The output unit 009 outputs the result by displaying the result on a display, for example. As a result, the degree of abnormality can be grasped, and the maintenance plan and operation plan for the diagnosis target device can be made.

以上、本発明の実施例を説明してきたが、これまでの各実施例で説明した構成はあくまで一例であり、本発明は、技術思想を逸脱しない範囲内で適宜変更が可能である。   As mentioned above, although the Example of this invention has been described, the structure demonstrated by each Example so far is an example to the last, and this invention can be suitably changed within the range which does not deviate from a technical idea.

００１ 異常検知評価システム
００２ 診断対象機器
００３ データ取り込み部
００４ 異常程度入力部
００５ データベース（ＤＢ）
００６ 異常模擬音合成部
００７ 正常異常判定部
００８ 劣化曲線算出部
００９ 出力部
０１０ 異常程度推定部
０１１ 音声入力部
２００ 劣化曲線001 Anomaly detection evaluation system 002 Diagnosis target device 003 Data acquisition unit 004 Abnormality degree input unit 005 Database (DB)
006 Abnormal simulation sound synthesis unit 007 Normal abnormality determination unit 008 Deterioration curve calculation unit 009 Output unit 010 Abnormal degree estimation unit 011 Voice input unit 200 Deterioration curve

Claims (4)

診断対象機器からの音に基づいて前記診断対象機器が正常か異常かを判定する正常異常判定部と、
前記診断対象機器の異常の程度毎に異常音を模擬する異常模擬音合成部と、
前記異常模擬音合成部で模擬した異常音を用いて前記正常異常判定部で正常か異常かを複数回判定した場合の異常検出率と前記模擬した異常音に対応する前記異常の程度との対応関係を示す劣化曲線を算出する劣化曲線算出部とを有することを特徴とする異常検知評価システム。A normal / abnormality determination unit that determines whether the diagnosis target device is normal or abnormal based on sound from the diagnosis target device;
An abnormal simulation sound synthesizer that simulates an abnormal sound for each degree of abnormality of the diagnosis target device;
Correspondence between the abnormality detection rate and the degree of abnormality corresponding to the simulated abnormal sound when the normal abnormality determination unit determines whether it is normal or abnormal multiple times using the abnormal sound simulated by the abnormal simulated sound synthesis unit An abnormality detection evaluation system comprising: a deterioration curve calculation unit that calculates a deterioration curve indicating a relationship. 前記異常模擬音合成部は、正常な前記診断対象機器の稼動音と、前記異常の程度に対応した特徴データとから、前記模擬した異常音を生成することを特徴とする請求項１に記載の異常検知評価システム。   The said abnormal simulation sound synthesizer generates the simulated abnormal sound from normal operating sounds of the diagnosis target device and feature data corresponding to the degree of the abnormality. Anomaly detection evaluation system. 前記異常の程度毎の前記異常検出率を表示する出力部を有することを特徴とする請求項１または２に記載の異常検知評価システム。   The abnormality detection evaluation system according to claim 1, further comprising an output unit that displays the abnormality detection rate for each degree of abnormality. 前記診断対象機器からの音を用いて前記正常異常判定部で正常か異常かを複数回判定した場合の異常検出率と、前記劣化曲線とから、前記診断対象機器の異常の程度を推定する異常程度推定部を有することを特徴とする請求項１から３の何れかに記載の異常検知評価システム。   An abnormality that estimates the degree of abnormality of the diagnosis target device from the abnormality detection rate when the normality / abnormality determination unit determines whether it is normal or abnormal multiple times using sound from the diagnosis target device and the deterioration curve The abnormality detection evaluation system according to claim 1, further comprising a degree estimation unit.
JP2015528053A 2013-07-24 2013-07-24 Anomaly detection evaluation system Pending JPWO2015011791A1 (en)

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/JP2013/069975 WO2015011791A1 (en) 2013-07-24 2013-07-24 Abnormality detection evaluation system

Publications (1)

Publication Number Publication Date
JPWO2015011791A1 true JPWO2015011791A1 (en) 2017-03-02

Family

ID=52392872

Family Applications (1)

Application Number Title Priority Date Filing Date
JP2015528053A Pending JPWO2015011791A1 (en) 2013-07-24 2013-07-24 Anomaly detection evaluation system

Country Status (2)

Country Link
JP (1) JPWO2015011791A1 (en)
WO (1) WO2015011791A1 (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2019174617A (en) * 2018-03-28 2019-10-10 コニカミノルタ株式会社 Failure diagnosis apparatus, failure diagnosis method and control program of failure diagnosis apparatus
US11513000B2 (en) 2016-12-12 2022-11-29 Rorze Corporation Waveform analysis device and waveform analysis method

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US11480497B2 (en) * 2016-04-01 2022-10-25 Nippon Telegraph And Telephone Corporation Anomalous sound detection training apparatus, acoustic feature extraction apparatus, anomalous sound sampling apparatus, and methods and programs for the same
WO2017212645A1 (en) * 2016-06-10 2017-12-14 株式会社日立製作所 Bearing diagnosing device and bearing diagnosing method, and rotating machine and maintenance method for same
WO2020026829A1 (en) * 2018-07-31 2020-02-06 パナソニックＩｐマネジメント株式会社 Sound data processing method, sound data processing device, and program

Family Cites Families (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP3275770B2 (en) * 1997-04-16 2002-04-22 日本鋼管株式会社 Defect inspection method for resin-coated steel
JP2001242089A (en) * 2000-02-28 2001-09-07 Nkk Corp Surface flaw detecting method
JP2003021555A (en) * 2001-07-06 2003-01-24 Yamatake Sangyo Systems Co Ltd Abnormality monitoring equipment
JP2005321946A (en) * 2004-05-07 2005-11-17 Mitsubishi Electric Corp Monitoring device
JP4202400B1 (en) * 2007-07-27 2008-12-24 三菱重工業株式会社 Crack growth prediction method and program

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US11513000B2 (en) 2016-12-12 2022-11-29 Rorze Corporation Waveform analysis device and waveform analysis method
JP2019174617A (en) * 2018-03-28 2019-10-10 コニカミノルタ株式会社 Failure diagnosis apparatus, failure diagnosis method and control program of failure diagnosis apparatus

Also Published As

Publication number Publication date
WO2015011791A1 (en) 2015-01-29

Similar Documents

Publication Publication Date Title
US20240068864A1 (en) Systems and methods for monitoring of mechanical and electrical machines
JP5162114B2 (en) Method and apparatus for signal signature analysis for event detection in rotating machinery
US10316849B2 (en) Method and system for detection of faults in pump assembly via handheld communication device
Zhen et al. Fault diagnosis of motor drives using stator current signal analysis based on dynamic time warping
WO2015011791A1 (en) Abnormality detection evaluation system
RU2017130315A (en) METHOD, SYSTEM AND COMPUTER PROGRAM FOR THE STAGE OF TEACHING ACOUSTIC OR VIBRATION ANALYSIS OF THE MACHINE
JP6922708B2 (en) Anomaly detection computer program, anomaly detection device and anomaly detection method
JP2013221877A (en) Abnormality inspection method and abnormality inspection device
JP2019067197A (en) Method for detecting trouble sign
US20120130682A1 (en) Real-time detection system and the method thereof
JP2019027814A (en) Method and device for analyzing vibration acoustic analysis and method and device for estimating abnormal part of apparatus
KR20110034281A (en) Method and analysis system for diagnosising exclusive vibration in generator stator windings
CN108629077B (en) Fault diagnosis device, monitoring device, fault diagnosis method, and recording medium
CN107192550B (en) Order analysis system
Dietel et al. Fault detection in rotating machinery using spectral modeling
KR102566810B1 (en) Motion signal extraction system and method based on vibration signal
JP4303072B2 (en) Engine evaluation device
US11439320B2 (en) Biological-sound analysis device, biological-sound analysis method, program, and storage medium
Tossavainen Sound based fault detection system
JP7272631B2 (en) Sound or vibration determination method and information processing system for determination
JP4349888B2 (en) Engine evaluation device
JP2013002825A (en) Abnormality checking method and abnormality checking apparatus
TW202121160A (en) Sound-based equipment surveillance method
JP2015175769A (en) Insulation abnormality diagnostic device, insulation abnormality diagnostic method and program for insulation abnormality diagnosis
JP2023030903A (en) Environment monitoring system