CN113962264A - Aero-engine rotor system fault diagnosis algorithm based on deep learning - Google Patents

Aero-engine rotor system fault diagnosis algorithm based on deep learning Download PDF

Info

Publication number
CN113962264A
CN113962264A CN202111237432.0A CN202111237432A CN113962264A CN 113962264 A CN113962264 A CN 113962264A CN 202111237432 A CN202111237432 A CN 202111237432A CN 113962264 A CN113962264 A CN 113962264A
Authority
CN
China
Prior art keywords
data
sample
fault diagnosis
fault
sampling
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
CN202111237432.0A
Other languages
Chinese (zh)
Inventor
杨蒲
耿慧琳
柳鹏
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.)
Nanjing University of Aeronautics and Astronautics
Original Assignee
Nanjing University of Aeronautics and Astronautics
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 Nanjing University of Aeronautics and Astronautics filed Critical Nanjing University of Aeronautics and Astronautics
Priority to CN202111237432.0A priority Critical patent/CN113962264A/en
Publication of CN113962264A publication Critical patent/CN113962264A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/08—Feature extraction
    • 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
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/048—Activation functions
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/12—Classification; Matching

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Artificial Intelligence (AREA)
  • General Engineering & Computer Science (AREA)
  • General Health & Medical Sciences (AREA)
  • Biomedical Technology (AREA)
  • Evolutionary Computation (AREA)
  • Computational Linguistics (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Biophysics (AREA)
  • Data Mining & Analysis (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Signal Processing (AREA)
  • Acoustics & Sound (AREA)
  • Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
  • Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)

Abstract

The invention discloses a fault diagnosis algorithm of an aircraft engine rotor system based on deep learning, which is designed by considering the problems that a vibration signal of an aircraft engine rotor system is weak and sample collection is difficult; then, establishing a novel depth residual shrinkage network for feature extraction, setting the first layer convolution layer as a wide convolution kernel, and using a progressive half-soft threshold function algorithm as a shrinkage layer of the network; and finally, outputting a fault classification result through the full connection layer and the Softmax classifier. The invention takes the original vibration signal data of the aeroengine bearing as input, automatically and accurately identifies the fault characteristics in high noise, and directly outputs the fault classification identification result.

Description

Aero-engine rotor system fault diagnosis algorithm based on deep learning
Technical Field
The invention relates to a rotor system of an aircraft engine, designs a fault diagnosis algorithm based on deep learning, and belongs to the technical field of fault diagnosis.
Background
The aircraft engine, as a core component of the aircraft, plays an important role in providing aircraft power for the aircraft, and has an important influence on the performance and reliability of the aircraft. Moreover, the aeroengine has severe working conditions and is easy to break down, and the rolling bearing is an important component of the rotor system of the aeroengine, and if the rolling bearing is damaged, property loss and even casualties can be caused. Therefore, the fault diagnosis research of the aeroengine bearing has important significance.
The fault diagnosis method mainly comprises the steps of physical model based, signal processing and intelligent diagnosis. The diagnostic method of the fault diagnosis method based on the physical model has good robustness, but the method has the premise that an accurate model is established aiming at a target system, so the method is not easy to realize in practical application; the diagnosis method based on signal processing does not need to establish a quantitative or qualitative mathematical model of the system, only needs to collect original data, and extracts fault characteristics from the original data through signal processing so as to diagnose the system fault; with the development of computational intelligence, deep learning becomes a hotspot in the research field of fault detection, and the diagnostic method based on the neural network preprocesses original data and inputs the data into a neural network model to finally and directly obtain a fault diagnosis result.
Because of the problems of complex vibration transmission path and the like of the aeroengine bearing vibration signal, the obtained original vibration signal has the characteristics of high noise and multiple interference information, and the problem that the fault characteristics are difficult to extract is caused. Therefore, how to extract effective fault features from raw data in the background of high noise interference is a major concern in the field of fault diagnosis.
Disclosure of Invention
The purpose of the invention is as follows: aiming at the research background and the problems of high noise and difficult feature extraction of the vibration signal of the aircraft engine on the basis of the prior art method, the fault diagnosis algorithm of the aircraft engine rotor system based on deep learning is provided.
The technical scheme is as follows: in order to achieve the purpose, the invention provides a novel fault diagnosis algorithm of an aircraft engine rotor system based on deep learning, which is characterized in that: the method comprises the following steps of repeatedly sampling original data by using a sliding window algorithm to realize data enhancement, constructing a fault sample set after normalization, establishing a depth residual shrinkage network to perform automatic feature extraction, setting a first-layer convolution layer as a wide convolution kernel, using a progressive half-soft threshold function algorithm as a shrinkage layer of the network to realize data noise reduction, and obtaining a fault identification classification result through a full connection layer and a Softmax classifier, wherein the method comprises the following specific steps:
step 1) acquiring acceleration vibration signals of an aeroengine rolling bearing under different faults through a vibration signal data acquisition system, and constructing an original vibration signal sample;
step 2) carrying out overlapped sampling on the original data sample by a sampling sliding window:
Figure BSA0000255751240000021
where n is the number of samples after the overlapping sampling, LrawIs the original data length, LsampleIs the length of a single sample, i.e. the window width, and P is the moving step length of the sliding window, i.e. the sampling interval;
the Data after overlapped sampling is Data ═ x1，y1)，…，(xi，yi)，…，(xn，yn)]TData is a Data set after segmentation, xiIs a single vibration signal sample data, each sample contains 1024 sampling points, yiIs a fault category label for the sample data;
step 3) carrying out normalization processing on the data to realize unification of data dimensions:
Figure BSA0000255751240000022
wherein x ismaxAnd xminRespectively, the minimum value and the maximum value, and x' is a normalized sample;
step 4) randomly dividing the samples obtained after normalization, wherein 70% of data is used as a training set, and 30% of data is used as a testing set;
step 5) setting a novel depth residual shrinkage network fault diagnosis model under a Pythrch, training the fault diagnosis model by using a training set, wherein the network uses an Adam optimization algorithm and consists of a first-layer wide convolution layer, a plurality of residual shrinkage modules, a full-link layer and a Softmax classifier, and finally outputting a fault classification result through the Softmax classifier, wherein the structural parameters of the novel depth residual shrinkage network fault diagnosis model are as follows:
Figure BSA0000255751240000023
Figure BSA0000255751240000031
step 6) inputting the training set into the trained deep residual shrinkage network model, and identifying the fault type of the test sample;
has the advantages that: the invention provides a fault diagnosis algorithm of an aircraft engine rotor system based on deep learning, which uses an improved deep residual shrinkage network to provide an end-to-end fault diagnosis algorithm of the aircraft engine rotor system based on deep learning, and has the following specific advantages:
(1) the invention introduces a sliding window algorithm to realize repeated interception and sampling of the original data, so that the data sample is expanded, the data enhancement is realized, and the problem of less original data samples is solved;
(2) the first layer convolution layer in the neural network model is set as a wide convolution kernel, and the wide convolution kernel can effectively extract short-time features in a sample and improve the feature extraction capability and noise immunity of the model;
(3) the invention provides a novel depth residual shrinkage network model, which uses a progressive half-soft threshold shrinkage function, fully retains the effective characteristics of signals on the basis of realizing the noise reduction of vibration signals and avoids the problem of signal distortion caused by the soft threshold function;
the method provided by the invention is used as a fault diagnosis algorithm of the aircraft engine rotor system based on deep learning, has a certain practical application value, is easy to realize, has high accuracy, and can be widely applied to bearing fault diagnosis of the aircraft engine rotor system.
Drawings
FIG. 1 is a flow chart of a diagnostic algorithm of the present invention;
FIG. 2 is a schematic diagram of a basic module of the novel deep residual error shrinkage network according to the present invention;
FIG. 3 is a schematic diagram of the overall structure of the novel deep residual error shrinkage network of the present invention;
FIG. 4 is a graph of training loss and accuracy for the novel deep residual shrinkage network of the present invention;
FIG. 5 is a confusion matrix of the novel deep residual shrinkage network of the present invention;
Detailed Description
The invention is further explained below with reference to the drawings.
FIG. 1 is a flow chart of fault diagnosis of an aircraft engine rotor system based on deep learning, comprising the following specific steps:
step 1) acquiring acceleration vibration signals of an aeroengine rolling bearing under different faults through a vibration signal data acquisition system, and constructing an original vibration signal sample;
step 2) carrying out overlapped sampling on the original data sample by a sampling sliding window:
Figure BSA0000255751240000041
where n is the number of samples after the overlapping sampling, LrawIs the original data length, LsampleIs the length of a single sample, i.e. the window width, and P is the moving step length of the sliding window, i.e. the sampling interval;
the Data after overlapped sampling is Data ═ x1，y1)，…，(xi，yi)，…，(xn，yn)]TData is a Data set after segmentation, xiIs a single vibration signal sample data, each sample contains 1024 sampling points, yiIs a sampleA fault category label for the data;
step 3) carrying out normalization processing on the data to realize unification of data dimensions:
Figure BSA0000255751240000042
wherein x ismaxAnd xminRespectively, the minimum value and the maximum value, and x' is a normalized sample;
step 4) randomly dividing the samples obtained after normalization, wherein 70% of data is used as a training set, and 30% of data is used as a testing set;
step 5) setting a novel deep residual shrinkage network fault diagnosis model under the Pythrch, and training the fault diagnosis model by using a training set;
the basic module diagram of the depth residual shrinking network is shown in FIG. 2;
the method comprises the following specific steps:
and 5.1) after the input sample passes through the first convolutional layer, the input sample enters a second convolutional layer through a ReLU activation function. The second convolutional layer constructs a substructure for obtaining a noise threshold;
step 5.2) in the substructure, firstly, carrying out absolute value taking and global average pooling on input to obtain a mean value parameter, then excavating the characteristics of the channels through two full-connection layers, and finally obtaining attention weight parameters through a Sigmoid activation function, wherein each attention weight parameter acts on a characteristic vector of a corresponding characteristic channel;
step 5.3) multiplying the attention weight parameter and the mean value parameter to obtain a noise threshold value, so that each characteristic channel has an independent noise threshold value;
step 5.4) finally, the obtained threshold is reused to carry out progressive semi-soft threshold processing on the sample data, and the result after the threshold processing is added with the residual error item of the cross-layer identity mapping to obtain the final module output;
the progressive half-soft threshold calculation formula is as follows:
Figure BSA0000255751240000051
where τ is the threshold, a' and asRespectively is sample data before and after the progressive half-soft threshold processing;
step 5.5) the overall structure schematic diagram of the deep residual shrinkage network is shown in fig. 3, the overall deep residual shrinkage network uses an Adam optimization algorithm and consists of a first wide convolution layer, two residual shrinkage modules, a full connection layer and a Softmax classifier, and finally, a fault classification result is output through the Softmax classifier, so that the fault feature extraction in a high-noise vibration signal is realized, and the structural parameters of the novel deep residual shrinkage network fault diagnosis model are as follows:
Figure BSA0000255751240000052
and 6) inputting the training set into the trained deep residual shrinkage network model, and identifying the fault diagnosis result of the test sample.
The aeroengine fault diagnosis algorithm has high fault identification accuracy, the CWRU bearing data of the bearing data center of the university of Western storage in America verifies the fault diagnosis algorithm to finally obtain the accuracy of 99.21%, other noise reduction processing is not required to be carried out on original data by the algorithm, end-to-end fault diagnosis is realized through characteristic self-extraction, and the final accuracy of the fault diagnosis algorithm in a test set and a training set can reach more than 99% as shown in a training curve of a fault diagnosis model in FIG. 4.
The precision of the fault diagnosis algorithm is measured by using a confusion matrix, the confusion matrix realizes the performance evaluation of the diagnosis model by calculating the correct classification and the wrong classification number of the model, fig. 5 is the confusion matrix of the fault diagnosis algorithm, as shown in fig. 5, the abscissa of the confusion matrix is the fault diagnosis result of the diagnosis model, and the ordinate is the actual fault category label, so that the fault diagnosis algorithm can effectively identify various faults of the rolling bearing, and the diagnosis effect is more accurate.
It can be seen that the progressive semi-soft threshold mechanism is fused into the residual network by the deep residual shrinkage network, and different thresholds are automatically set for each threshold module by means of the attention mechanism, so that the processing of original data is realized, the interference caused by noise information is avoided, and the extraction of fault features in the noise information can be realized.

Claims (1)

1. An aeroengine rotor system fault diagnosis algorithm based on deep learning comprises the following specific steps:
step 1) acquiring acceleration vibration signals of an aeroengine rolling bearing under different faults through a vibration signal data acquisition system, and constructing an original vibration signal sample;
step 2) carrying out overlapped sampling on the original data sample by a sampling sliding window:
Figure FSA0000255751230000011
where n is the number of samples after the overlapping sampling, LrawIs the original data length, LsampleIs the length of a single sample, i.e. the window width, and P is the moving step length of the sliding window, i.e. the sampling interval;
the Data after overlapped sampling is Data ═ x1，y1)，…，(xi，yi)，…，(xn，yn)]TData is a Data set after segmentation, xiIs a single vibration signal sample data, each sample contains 1024 sampling points, yiIs a fault category label for the sample data;
step 3) carrying out normalization processing on the data to realize unification of data dimensions:
Figure FSA0000255751230000012
wherein x ismaxAnd xminRespectively, the minimum value and the maximum value, and x' is a normalized sample;
step 4) randomly dividing the samples obtained after normalization, wherein 70% of data is used as a training set, and 30% of data is used as a testing set;
step 5) setting a novel depth residual shrinkage network fault diagnosis model under a Pythrch, training the fault diagnosis model by using a training set, wherein the network uses an Adam optimization algorithm and consists of a first-layer wide convolution layer, a plurality of residual shrinkage modules, a full-link layer and a Softmax classifier, and finally outputting a fault classification result through the Softmax classifier, wherein the structural parameters of the novel depth residual shrinkage network fault diagnosis model are as follows:
Figure FSA0000255751230000013
Figure FSA0000255751230000021
and 6) inputting the training set into the trained deep residual shrinkage network model, and identifying the fault type of the test sample.
CN202111237432.0A 2021-10-22 2021-10-22 Aero-engine rotor system fault diagnosis algorithm based on deep learning Pending CN113962264A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202111237432.0A CN113962264A (en) 2021-10-22 2021-10-22 Aero-engine rotor system fault diagnosis algorithm based on deep learning

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202111237432.0A CN113962264A (en) 2021-10-22 2021-10-22 Aero-engine rotor system fault diagnosis algorithm based on deep learning

Publications (1)

Publication Number Publication Date
CN113962264A true CN113962264A (en) 2022-01-21

Family

ID=79466266

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202111237432.0A Pending CN113962264A (en) 2021-10-22 2021-10-22 Aero-engine rotor system fault diagnosis algorithm based on deep learning

Country Status (1)

Country Link
CN (1) CN113962264A (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114662546A (en) * 2022-04-06 2022-06-24 西北工业大学 Aero-engine sensor fault diagnosis method and device based on optimization of badger
CN116538092A (en) * 2023-07-06 2023-08-04 中国科学院理化技术研究所 Compressor on-line monitoring and diagnosing method, device, equipment and storage medium

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114662546A (en) * 2022-04-06 2022-06-24 西北工业大学 Aero-engine sensor fault diagnosis method and device based on optimization of badger
CN116538092A (en) * 2023-07-06 2023-08-04 中国科学院理化技术研究所 Compressor on-line monitoring and diagnosing method, device, equipment and storage medium
CN116538092B (en) * 2023-07-06 2023-11-14 中国科学院理化技术研究所 Compressor on-line monitoring and diagnosing method, device, equipment and storage medium

Similar Documents

Publication Publication Date Title
CN110991295B (en) Self-adaptive fault diagnosis method based on one-dimensional convolutional neural network
CN111709448A (en) Mechanical fault diagnosis method based on migration relation network
CN112284736B (en) Convolutional neural network fault diagnosis method based on multi-channel attention module
CN112257530B (en) Rolling bearing fault diagnosis method based on blind signal separation and support vector machine
CN113962264A (en) Aero-engine rotor system fault diagnosis algorithm based on deep learning
CN111580151B (en) SSNet model-based earthquake event time-of-arrival identification method
CN112378660A (en) Intelligent fault diagnosis method for aero-engine bearing based on data driving
CN113887136B (en) Electric automobile motor bearing fault diagnosis method based on improved GAN and ResNet
CN112651167A (en) Semi-supervised rolling bearing fault diagnosis method based on graph neural network
CN109214349A (en) A kind of object detecting method based on semantic segmentation enhancing
CN112487890B (en) Bearing acoustic signal fault diagnosis method based on parallel sparse filtering
CN112070104A (en) Main transformer partial discharge identification method
CN114993669A (en) Multi-sensor information fusion transmission system fault diagnosis method and system
CN115905976B (en) Attention mechanism-based fault diagnosis method, system and equipment for Highway Bi-LSTM bearing
CN113569990A (en) Performance equipment fault diagnosis model construction method oriented to strong noise interference environment
CN115238785A (en) Rotary machine fault diagnosis method and system based on image fusion and integrated network
CN114169377A (en) G-MSCNN-based fault diagnosis method for rolling bearing in noisy environment
CN114186617B (en) Mechanical fault diagnosis method based on distributed deep learning
CN112528774A (en) Intelligent sorting system and method for unknown radar signals in complex electromagnetic environment
CN115876473A (en) Bearing fault diagnosis method based on PWVD and DenseNet
CN114357372A (en) Aircraft fault diagnosis model generation method based on multi-sensor data driving
CN105823634A (en) Bearing damage identification method based on time frequency relevance vector convolution Boltzmann machine
CN112600659A (en) Method and system for detecting side channel leakage of safety chip based on neural network
CN110555457A (en) Engine lubricating oil wear signal characteristic processing system and method
CN114781427A (en) Wind generating set rolling bearing fault diagnosis system based on information fusion technology

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination