JP2018189522A - Method for diagnosing sign of facility failure - Google Patents

Method for diagnosing sign of facility failure Download PDF

Info

Publication number
JP2018189522A
JP2018189522A JP2017092656A JP2017092656A JP2018189522A JP 2018189522 A JP2018189522 A JP 2018189522A JP 2017092656 A JP2017092656 A JP 2017092656A JP 2017092656 A JP2017092656 A JP 2017092656A JP 2018189522 A JP2018189522 A JP 2018189522A
Authority
JP
Japan
Prior art keywords
image data
fourier transform
pseudo
sign
sound
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
JP2017092656A
Other languages
Japanese (ja)
Inventor
武 荻野
Takeshi Ogino
武 荻野
正彦 伊東
Masahiko Ito
正彦 伊東
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.)
QP Corp
Original Assignee
QP Corp
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 QP Corp filed Critical QP Corp
Priority to JP2017092656A priority Critical patent/JP2018189522A/en
Publication of JP2018189522A publication Critical patent/JP2018189522A/en
Pending legal-status Critical Current

Links

Images

Landscapes

  • Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)

Abstract

PROBLEM TO BE SOLVED: To provide a method for diagnosing a sign that allows not only experienced persons but also inexperienced persons to analyze sound and/or vibrations and to notice a sign of a facility failure that only experienced persons have noticed before.SOLUTION: The method for diagnosing a sign of a facility failure by analyzing sound and/or vibrations and detecting the sign in advance includes: a Fourier transform process of obtaining a Fourier transform signal by conducting fast-Fourier transform on sound and/or vibrations; a pseudo-image data creation process of obtaining pseudo-image data in which the intensity of a Fourier transform signal is plotted in a two dimension of wavelength and time; and a pseudo-image diagnosis process of inputting pseudo-image data into a convolution neural network and obtaining an analysis value.SELECTED DRAWING: Figure 2

Description

本発明は、音声及び又は振動を用いて、設備故障の予兆診断する方法に関する。   The present invention relates to a method for predicting equipment failure using sound and / or vibration.

<背景技術の説明>
被監視装置から発生する機械音を周波数と音圧の時系列変化に変換し、時系列変化イメージを正常機械音の周波数と音圧の時系列変化イメージと比較して被監視装置の状態の良否を判別する判別装置が知られている(特許文献1)。特許文献1には、異常の判定に人工知能要素を含むこともできると記載されている(第9頁)。
<Description of background technology>
Convert mechanical sound generated from the monitored device into time-series changes in frequency and sound pressure, compare the time-series change image with the time-series change image of normal machine sound frequency and sound pressure, and the status of the monitored device is good or bad A discriminating device for discriminating the above is known (Patent Document 1). Patent Document 1 describes that an artificial intelligence element can be included in the determination of abnormality (page 9).

特開昭63−261117号公報JP 63-261117 A

<背景技術の課題>
そこで、本発明は、このような事情に鑑みてなされたものであり、その目的は、いままで経験者でしか予め気づくことが出来なかった設備故障の予兆診断を、音声及び又は振動を分析することで、未経験者であっても可能にする予兆診断方法を提供することにある。
<Background technology issues>
Therefore, the present invention has been made in view of such circumstances, and its purpose is to analyze the sound and / or vibration for predictive diagnosis of equipment failure that could only be previously noticed by experienced persons. Thus, it is to provide a predictive diagnosis method that enables even an inexperienced person.

<請求項1の内容>
このような目的を達成するため、本発明は、以下の構成によって把握される。
(1)本発明は、音声及び又は振動を分析して設備故障の予兆診断する予兆診断方法であって、前記音声及び又は振動を高速フーリエ変換してフーリエ変換信号を得るフーリエ変換プロセス、 前記フーリエ変換信号の強度を、波長×時間の二次元にプロットした、疑似画像データを得る疑似画像データ作成プロセス、前記疑似画像データを畳み込みニューラルネットワークに入力して分析値を得る疑似画像診断プロセスと、を有することを特徴とするものである。
<Content of Claim 1>
In order to achieve such an object, the present invention is grasped by the following configuration.
(1) The present invention is a predictive diagnosis method for analyzing a sound and / or vibration to predict an equipment failure, and a Fourier transform process for obtaining a Fourier transform signal by performing fast Fourier transform on the sound and / or vibration, A pseudo image data creation process for obtaining pseudo image data, in which the intensity of the converted signal is plotted in two dimensions of wavelength × time, and a pseudo image diagnosis process for obtaining the analysis value by inputting the pseudo image data into a convolution neural network. It is characterized by having.

本発明によれば、いままで経験者でしか予め気づくことが出来なかった設備故障の予兆診断を、音声及び又は振動を分析することで、未経験者であっても可能にする予兆診断方法を提供することができる。   According to the present invention, there is provided a predictive diagnosis method that enables an inexperienced person to perform predictive diagnosis of equipment failure that has been previously noticed only by an experienced person by analyzing voice and / or vibration. can do.

本発明の実施形態に係る予兆診断方法の処理工程示したフローチャートである。It is the flowchart which showed the process process of the predictive diagnosis method which concerns on embodiment of this invention. 本発明の実施形態に係る予兆診断方法で得られた疑似画像データを表したものである。It represents the pseudo image data obtained by the predictive diagnosis method according to the embodiment of the present invention. 本発明の実施形態の変形例に係る予兆診断方法で得られた自己符号化された疑似画像データを表したものである。It represents the self-encoded pseudo image data obtained by the predictive diagnosis method according to a modification of the embodiment of the present invention.

<実施形態の説明−0>
以下、添付図面を参照して、本発明を実施するための形態(以下、「実施形態」と称する)について詳細に説明する。実施形態の説明の全体を通して同じ要素には同じ符号を付している。
<Description of Embodiment-0>
DESCRIPTION OF EMBODIMENTS Hereinafter, embodiments for carrying out the present invention (hereinafter referred to as “embodiments”) will be described in detail with reference to the accompanying drawings. The same symbols are attached to the same elements throughout the description of the embodiments.

<実施形態の説明−1>
まず、フーリエ変換プロセス1において、装置周辺に配置された集音マイク11により、あるいは回転軸等に設置された振動計12により得られた音声及び又は振動を示す電気信号13は、高速フーリエ変換されて、フーリエ変換信号の強度14が記録装置15に保存される。
<Description of Embodiment-1>
First, in the Fourier transform process 1, an electrical signal 13 indicating sound and / or vibration obtained by a sound collecting microphone 11 arranged around the apparatus or by a vibrometer 12 installed on a rotating shaft or the like is subjected to fast Fourier transform. Thus, the intensity 14 of the Fourier transform signal is stored in the recording device 15.

<実施形態の説明−2>
疑似画像データ作成プロセス2において、記録装置15に保存されたフーリエ変換信号の強度14を直近数ミリ秒間分〜数分間分読み出し、波長×時間の二次元にプロットして疑似画像データ21が作成され、記録装置22に保存される。
<Description of Embodiment-2>
In the pseudo image data creation process 2, the intensity 14 of the Fourier transform signal stored in the recording device 15 is read for the last several milliseconds to several minutes and plotted in two dimensions of wavelength × time to create pseudo image data 21. And stored in the recording device 22.

<実施形態の説明−3>
疑似画像診断プロセス3において、記録装置22に保存された疑似画像データ21を、畳み込みニューラルネットワーク31に入力して分析値32を得て、記録装置33に保存される。
<Description of Embodiment-3>
In the pseudo image diagnosis process 3, the pseudo image data 21 stored in the recording device 22 is input to the convolutional neural network 31 to obtain an analysis value 32 and stored in the recording device 33.

<実施形態の説明−4>
警報プロセス4において、記録装置33に保存された分析値32を読み出し、閾値41を超えた場合に警報42が発報され、あるいは複数の閾値群43に該当した警戒レベル表示44が表示される。
<Description of Embodiment-4>
In the alarm process 4, the analysis value 32 stored in the recording device 33 is read, and when the threshold value 41 is exceeded, an alarm 42 is issued, or a warning level display 44 corresponding to a plurality of threshold groups 43 is displayed.

<疑似画像データ21−1>
疑似画像データ21は、波長×時間の二次元に、フーリエ変換信号の強度14がプロットされている。フーリエ変換信号の強度14が高いほど該当ピクセルのグレースケールが濃くなる、あるいは薄くなるようにプロットされている。
<Pseudo-image data 21-1>
In the pseudo image data 21, the intensity 14 of the Fourier transform signal is plotted in two dimensions of wavelength × time. Plotting is such that the higher the intensity 14 of the Fourier transform signal, the darker or lighter the gray scale of the corresponding pixel.

<疑似画像データ21−変形例>
音声及び又は振動の入力が3チャンネル以内であれば、疑似的にRGBに割り当てて疑似RGB画像データ21−2としてプロットすることが望ましい。またさらに入力が4チャンネル以上に増える場合は、画面を上下に二分割あるいは複数に分割することで、複数の入力に対応させた分割疑似RGB画像データ21−3を用いることも可能である。
<Pseudo Image Data 21—Modification>
If the audio and / or vibration input is within 3 channels, it is desirable to assign them to pseudo RGB and plot them as pseudo RGB image data 21-2. When the number of inputs further increases to four or more channels, the divided pseudo RGB image data 21-3 corresponding to a plurality of inputs can be used by dividing the screen into two or more parts vertically.

<畳み込みニューラルネットワーク31>
疑似画像診断する方法としては、近年発展の目覚ましいディープラーニングを用いることができる。ただし、センサーの出力値を直接演算する場合は、汎用性のある演算プラットフォームが少ない。ここで、特に疑似RGB画像データ21−2を用いた場合、特に画像処理用として用いられる畳み込みニューラルネットワーク31を用いることができ、好ましい。画像処理用としては、汎用性のある演算プラットフォームとして、Google社のTensorFlow(TM)を用いることができる。
<Convolutional neural network 31>
As a method for diagnosing a pseudo image, it is possible to use deep learning that has been developed in recent years. However, there are few general-purpose computing platforms when directly calculating the sensor output value. Here, in particular, when the pseudo RGB image data 21-2 is used, the convolutional neural network 31 used particularly for image processing can be used, which is preferable. For image processing, Google's TensorFlow (TM) can be used as a versatile computing platform.

<畳み込みニューラルネットワーク31−変形例>
疑似画像診断する方法としてはさらに、エッジ検出を行ったうえで自己符号化させることがより好ましい。設備の正常時の音声及び又は振動の特徴を畳み込みニューラルネットワーク31に学習させた学習モデルで自己符号化させた場合、設備が正常であった場合の自己符号化疑似画像データ51が得られる(図3)。この自己符号化疑似画像データ51と疑似画像データ21とを演算すると、図2のZで示される異常例のみがピックアップされ、異常を鋭敏に検知することが可能となる。
<Convolution Neural Network 31-Modification>
As a method of performing pseudo image diagnosis, it is more preferable to perform self-encoding after performing edge detection. When self-encoding is performed with a learning model in which the convolutional neural network 31 learns the voice and / or vibration characteristics when the equipment is normal, self-coded pseudo image data 51 when the equipment is normal is obtained (FIG. 3). When the self-encoded pseudo image data 51 and the pseudo image data 21 are calculated, only the abnormality example indicated by Z in FIG. 2 is picked up, and the abnormality can be detected sharply.

以上、実施形態を用いて本発明を説明したが、本発明の技術的範囲は上記実施形態に記載の範囲には限定されないことは言うまでもない。上記実施形態に、多様な変更又は改良を加えることが可能であることが当業者に明らかである。また、その様な変更又は改良を加えた形態も本発明の技術的範囲に含まれ得ることが、特許請求の範囲の記載から明らかである。   As mentioned above, although this invention was demonstrated using embodiment, it cannot be overemphasized that the technical scope of this invention is not limited to the range as described in the said embodiment. It will be apparent to those skilled in the art that various modifications or improvements can be added to the above embodiment. Further, it is apparent from the description of the scope of claims that embodiments with such changes or improvements can be included in the technical scope of the present invention.

1 フーリエ変換プロセス、 11 集音マイク、 12 振動計、 13 電気信号、 14 フーリエ変換信号の強度、 15 記録装置
2 疑似画像データ作成プロセス、 21 疑似画像データ、 22 記録装置
3 疑似画像診断プロセス、 31 畳み込みニューラルネットワーク、32 分析値、 33 記録装置
4 警報プロセス
Z 異常例
DESCRIPTION OF SYMBOLS 1 Fourier transform process, 11 Sound collecting microphone, 12 Vibrometer, 13 Electric signal, 14 Intensity of Fourier transform signal, 15 Recording device 2 Pseudo image data creation process, 21 Pseudo image data, 22 Recording device 3 Pseudo image diagnosis process, 31 Convolutional neural network, 32 analysis values, 33 Recording device 4 Alarm process Z Abnormal example

Claims (1)

音声及び又は振動を分析して設備故障の予兆診断する予兆診断方法であって、
前記音声及び又は振動を高速フーリエ変換してフーリエ変換信号を得るフーリエ変換プロセス、
前記フーリエ変換信号の強度を、波長×時間の二次元にプロットした、疑似画像データを得る疑似画像データ作成プロセス、
前記疑似画像データを畳み込みニューラルネットワークに入力して分析値を得る疑似画像診断プロセス
を有することを特徴とする予兆診断方法。
A predictive diagnosis method for predicting equipment failure by analyzing voice and / or vibration,
A Fourier transform process for obtaining a Fourier transform signal by fast Fourier transforming the sound and / or vibration;
Pseudo image data creation process for obtaining pseudo image data, in which the intensity of the Fourier transform signal is plotted in two dimensions of wavelength × time;
A predictive diagnosis method comprising: a pseudo image diagnosis process in which the pseudo image data is input to a convolutional neural network to obtain an analysis value.
JP2017092656A 2017-05-08 2017-05-08 Method for diagnosing sign of facility failure Pending JP2018189522A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
JP2017092656A JP2018189522A (en) 2017-05-08 2017-05-08 Method for diagnosing sign of facility failure

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
JP2017092656A JP2018189522A (en) 2017-05-08 2017-05-08 Method for diagnosing sign of facility failure

Publications (1)

Publication Number Publication Date
JP2018189522A true JP2018189522A (en) 2018-11-29

Family

ID=64478475

Family Applications (1)

Application Number Title Priority Date Filing Date
JP2017092656A Pending JP2018189522A (en) 2017-05-08 2017-05-08 Method for diagnosing sign of facility failure

Country Status (1)

Country Link
JP (1) JP2018189522A (en)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2020195536A1 (en) * 2019-03-28 2020-10-01 ポート・アンド・アンカー株式会社 Structural anomaly assessment method and anomaly assessment system
WO2021059909A1 (en) 2019-09-27 2021-04-01 オムロン株式会社 Data generation system, learning device, data generation device, data generation method, and data generation program
WO2021100482A1 (en) 2019-11-21 2021-05-27 オムロン株式会社 Model generation device, estimation device, model generation method, and model generation program
CN117237359A (en) * 2023-11-15 2023-12-15 天津市恒一机电科技有限公司 Conveyor belt tearing detection method and device, storage medium and electronic equipment

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2020195536A1 (en) * 2019-03-28 2020-10-01 ポート・アンド・アンカー株式会社 Structural anomaly assessment method and anomaly assessment system
JP2020165672A (en) * 2019-03-28 2020-10-08 ポート・アンド・アンカー株式会社 Abnormality discrimination method and abnormality discrimination system of structure
WO2021059909A1 (en) 2019-09-27 2021-04-01 オムロン株式会社 Data generation system, learning device, data generation device, data generation method, and data generation program
WO2021100482A1 (en) 2019-11-21 2021-05-27 オムロン株式会社 Model generation device, estimation device, model generation method, and model generation program
CN117237359A (en) * 2023-11-15 2023-12-15 天津市恒一机电科技有限公司 Conveyor belt tearing detection method and device, storage medium and electronic equipment
CN117237359B (en) * 2023-11-15 2024-02-20 天津市恒一机电科技有限公司 Conveyor belt tearing detection method and device, storage medium and electronic equipment

Similar Documents

Publication Publication Date Title
JP2018189522A (en) Method for diagnosing sign of facility failure
Safizadeh et al. Using multi-sensor data fusion for vibration fault diagnosis of rolling element bearings by accelerometer and load cell
Amar et al. Vibration spectrum imaging: A novel bearing fault classification approach
Senanayaka et al. Multiple classifiers and data fusion for robust diagnosis of gearbox mixed faults
JP5197853B2 (en) Monitoring device
KR102209401B1 (en) Apparatus and method for fault diagnosis of gearbox using cnn
JP6882397B2 (en) Abnormal noise detection system, equipment, method and program
KR102261899B1 (en) Apparatus and method for diagnosing fault of drone
JP2019067197A (en) Method for detecting trouble sign
KR20100094452A (en) A method for detection and automatic identification of damage to rolling bearings
KR20130065621A (en) Abnormality detection apparatus for periodic driving system, processing apparatus including periodic driving system, abnormality detection method for periodic driving system, and computer program
JP2018163178A (en) Vibration analyzer and mechanical component diagnostic system
JP7407382B2 (en) Sound data processing method, sound data processing device and program
Senanayaka et al. Towards online bearing fault detection using envelope analysis of vibration signal and decision tree classification algorithm
Attoui et al. Novel machinery monitoring strategy based on time–frequency domain similarity measurement with limited labeled data
JP2012018066A (en) Device for inspecting abnormality
CN113125135A (en) Fault diagnosis method for rotary machine, storage medium, and electronic device
CN113551765A (en) Sound spectrum analysis and diagnosis method for equipment fault
KR102240775B1 (en) Deep learning-based apparatus and method for determining breakdown of power transfer device using noise data
Senanayaka et al. Multiple fault diagnosis of electric powertrains under variable speeds using convolutional neural networks
JP4412306B2 (en) Abnormality determination method and abnormality determination device
KR20210006832A (en) Method and apparatus for machine fault diagnosis
Li et al. On-line fault detection in wind turbine transmission system using adaptive filter and robust statistical features
CN114639391A (en) Mechanical failure prompting method and device, electronic equipment and storage medium
JP2006285884A (en) Failure diagnosis method and control apparatus therewith