CN114004144A - Method for predicting remaining service life of aircraft engine - Google Patents

Method for predicting remaining service life of aircraft engine Download PDF

Info

Publication number
CN114004144A
CN114004144A CN202111240696.1A CN202111240696A CN114004144A CN 114004144 A CN114004144 A CN 114004144A CN 202111240696 A CN202111240696 A CN 202111240696A CN 114004144 A CN114004144 A CN 114004144A
Authority
CN
China
Prior art keywords
engine
layer
network model
data
layers
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
CN202111240696.1A
Other languages
Chinese (zh)
Inventor
段秉环
戴小氐
郝玉锴
赵根学
孙志颖
牛玥瑶
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.)
Xian Aeronautics Computing Technique Research Institute of AVIC
Original Assignee
Xian Aeronautics Computing Technique Research Institute of AVIC
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 Xian Aeronautics Computing Technique Research Institute of AVIC filed Critical Xian Aeronautics Computing Technique Research Institute of AVIC
Priority to CN202111240696.1A priority Critical patent/CN114004144A/en
Publication of CN114004144A publication Critical patent/CN114004144A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
    • G06F18/2135—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on approximation criteria, e.g. principal component analysis
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • 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
    • G06N3/044—Recurrent networks, e.g. Hopfield networks
    • 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
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
    • G06F2119/04—Ageing analysis or optimisation against ageing
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/02—Preprocessing
    • G06F2218/04—Denoising
    • G06F2218/06—Denoising by applying a scale-space analysis, e.g. using wavelet analysis
    • 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
    • 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)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • General Physics & Mathematics (AREA)
  • Artificial Intelligence (AREA)
  • General Engineering & Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Software Systems (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Molecular Biology (AREA)
  • Health & Medical Sciences (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • General Health & Medical Sciences (AREA)
  • Evolutionary Biology (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Medical Informatics (AREA)
  • Computer Hardware Design (AREA)
  • Geometry (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention belongs to the field of airborne equipment fault prediction and health management, and particularly relates to a method for predicting the remaining service life of an aircraft engine. The method comprises the following steps: processing data of an original time sequence of the engine to obtain first training data and first test data; constructing a network model mainly comprising two long-short time memory network layers and two full-connection layers, wherein the network model comprises seven layers, the first layer and the third layer are the long-short time memory network layers, the fifth layer and the seventh layer are full-connection layers, and the second layer, the fourth layer and the sixth layer are discarding layers; inputting first training data into the network model to perform network training to obtain a trained network model; and inputting the first test data into the trained network model for testing to obtain the remaining service life value of the engine. The method solves the problems of overfitting and high-dimensional disasters in the aspect of time sequence prediction of the traditional deep learning methods such as BP neural network and support vector machine.

Description

Method for predicting remaining service life of aircraft engine
Technical Field
The invention belongs to the field of airborne equipment fault prediction and health management, and particularly relates to a method for predicting the remaining service life of an aircraft engine.
Background
The aircraft engine is the heart of the aircraft, and the performance of the aircraft engine is closely related to the safe navigation of the aircraft, so that the Remaining service Life (RUL) and the health condition of the engine are predicted, protective measures can be taken before a fault occurs, and the safety of a flight system can be guaranteed effectively; in addition, the system can help equipment support personnel to determine the best time for maintaining the engine, and the maintenance cost is reduced. Compared with the traditional methods based on Kalman filtering, similar approximation and the like, the deep learning has been more and more widely applied to solving the problem of engine fault prediction due to the advantages that the deep learning can extract more useful features from complex data and is more adept in learning and identification.
Data collected from an aircraft engine generally has the characteristics of high dimensionality, inaccurate reading, noise pollution, complex time dependence among multiple parts and the like, deep learning methods such as a BP neural network and a support vector machine have achieved certain achievements in predicting the residual service life of the engine, but the deep learning methods have the problems of overfitting, dimensionality disaster, difficulty in parameter selection and the like, and a method capable of capturing complex machine behaviors and time sequence correlation from data samples of the engine is urgently needed.
Disclosure of Invention
The purpose of the invention is as follows:
the method for predicting the remaining service life of the aircraft engine is provided, and the problems of overfitting and high-dimensional disasters in the time sequence prediction aspect of the traditional deep learning methods such as a BP neural network and a support vector machine are solved.
The technical scheme of the invention is as follows:
in a first aspect, a method for predicting the remaining service life of an aircraft engine is provided, which includes:
carrying out data preprocessing on an original time sequence of an engine to obtain first training data and first test data;
constructing a network model mainly comprising two long-short time memory network layers and two full-connection layers, wherein the network model comprises seven layers, the first layer and the third layer are the long-short time memory network layers, the fifth layer and the seventh layer are full-connection layers, and the second layer, the fourth layer and the sixth layer are discarding layers;
inputting first training data into the network model to perform network training to obtain a trained network model;
and inputting the first test data into the trained network model for testing to obtain the remaining service life value of the engine.
Further, the method for preprocessing the original time sequence of the engine to obtain the first training data and the first testing data specifically comprises the following steps:
extracting effective sample characteristics from training data in an original time sequence of an engine, determining the remaining service life of the engine corresponding to the sample characteristics, and extracting the effective sample characteristics from test data in the original time sequence of the engine, wherein the sample characteristics comprise: operating conditions and sensor values;
and respectively carrying out wavelet transform denoising and principal component analysis dimensionality reduction on the extracted sample features aiming at the training data and the test data to obtain first training data and first test data.
Further, determining the remaining service life of the engine corresponding to the sample characteristics specifically as follows:
the full life cycle value of the engine is subtracted by the number of cycles currently in operation.
Further, the method comprises the following steps of preprocessing the original time sequence of the engine to obtain first training data and first test data, wherein the method comprises the following steps:
and modifying the remaining service life value of the engine, which is higher than a certain threshold value, in the first training data into the threshold value.
Further, inputting the first training data into the network model for network training to obtain a trained network model, specifically comprising:
at each time step, the remaining useful life of the engine at that time is collectively predicted using a time series sequence contained within a time window having a length less than the minimum of the lengths of the time series sequences of the respective engines in the first test data.
Further, inputting the first training data into the network model for network training to obtain a trained network model, specifically comprising:
and gradually reducing the learning rate in a segmentation way in the training process by adopting a variable learning rate strategy to accelerate the convergence of the network model.
Further, the wavelet transformation method is a wavelet transformation fixed threshold denoising method.
Further, the wavelet basis function is db 3.
Has the advantages that:
the invention provides a method for predicting the remaining service life of an aircraft engine. The engine time sequence data has the characteristics of nonlinearity and high noise, the wavelet transformation can extract time-frequency characteristics, and the method is very suitable for processing non-stable and nonlinear signals and can achieve a good noise reduction effect; the normalization enables the data to be limited in a certain range, so that adverse effects caused by singular sample data are eliminated, and the convergence speed and the prediction accuracy of the network are improved. And secondly, constructing a network model combining a long-time memory network layer, a full connection layer and a discarding layer. The two lstm layers can effectively improve the feature extraction capability of the network, and the learning capability of the network can be further improved on the premise that the network parameters are slightly increased by adding the two fully-connected layers at last. And a dropout layer is accessed after the LSTM layer and the full connection layer, so that the learning of the network can be accelerated, and overfitting can be prevented. Furthermore, the prediction accuracy of the network is further improved by setting a time window mechanism and adopting a variable learning rate strategy in the training process.
The method has high prediction accuracy and feasibility, and has positive promotion effects on improving equipment guarantee capability, reducing maintenance cost and the like.
Drawings
FIG. 1 is a schematic flow chart of a method for predicting the remaining service life of an aircraft engine;
FIG. 2 is a schematic diagram of a constructed LSTM network model;
FIG. 3 is a histogram of the root mean square error RMSE of a test engine;
FIG. 4 is a comparison of the predicted RUL value and the true RUL value for a test engine.
Detailed Description
The Long Short-Term Memory (LSTM) network is a cyclic neural network for processing sequence data, and connects intermediate results calculated by the neural network at different moments, and the prediction result at each moment is not only dependent on the input of the current moment, but also related to the intermediate results output at all past moments, so that the problems of time sequence prediction and Long-Term dependence can be effectively processed.
The method comprises the steps of firstly, carrying out data preprocessing operations such as effective sample feature extraction, wavelet transform threshold value method denoising, Principal Component Analysis (PCA) dimensionality reduction and the like on an original time sequence of an engine so as to improve the quality and the credibility of original data; then establishing a network model combining an LSTM layer, a full connection layer and a dropout layer; training the network by strategies such as label cutting, time window setting, parameter adjustment and the like; and finally, predicting the fault to obtain a predicted value of the residual service life of the engine. The method can effectively solve the problems of overfitting, high-dimensional disasters and the like in the aspect of time sequence prediction of the traditional deep learning methods such as the BP neural network, the support vector machine and the like, has higher prediction accuracy and feasibility, and has positive promotion effects on improving equipment guarantee capability, reducing maintenance cost and the like.
The following detailed description of embodiments of the present invention is provided in connection with the accompanying drawings and examples. The following examples are intended to illustrate the invention but are not intended to limit the scope of the invention.
The data set of the present embodiment is derived from operational monitoring data for an aircraft engine, and comprises a time series sequence of 260 engine operations of the same model. Each engine was used normally at the start of the test, and the test was continued under a combination of 3 different operating conditions until the engine degraded (reached a failure threshold), and data from 21 sensors was monitored during the test. The format of each row of the time sequence is: the number of the engine, the number of the current cycle, the operating conditions 1 to 3, and the sensors 1 to 21. Csv contains 53759 pieces of data of full life cycle monitoring data of each engine from normal use to degradation and no operation; csv only retains partial data of normal operation before engine degradation, and 19482 data are total; csv gives a true RUL value for 150 engines, which can be used to evaluate algorithm performance.
The invention provides a method for predicting the remaining service life of an aircraft engine, which comprises the following steps as shown in figure 1:
step 1), carrying out pretreatment operations such as effective sample characteristic column extraction, wavelet transform threshold value method denoising, Principal Component Analysis (PCA) dimension reduction and the like on an original time sequence of the engine to obtain pretreated training data and test data.
Step 1.1), extracting effective sample characteristics from original time sequence train.csv and test.csv of the engine, wherein the effective sample characteristics comprise 3 operating conditions and 21 sensor values; for training data, it is also necessary to label the remaining useful life of the engine by the full life cycle value of the engine minus the number of cycles currently in operation.
Step 1.2), denoising the sample characteristics by using a wavelet transform fixed threshold denoising method, which comprises the following specific steps:
orthogonal wavelet transform is carried out on the acquired signals f (t), the correlation of f (t) is greatly removed, the effective part is mainly concentrated on a small amount of wavelet coefficients with relatively large amplitude, the wavelet coefficients on all scales are greatly reduced after noise is processed, then threshold processing is carried out on all the wavelet coefficients, and finally signal reconstruction is carried out, so that the purpose of suppressing the noise is achieved.
The denoising effect mainly depends on the selection of a threshold denoising method and a wavelet basis function, and the change of the characteristic value of a time sequence is observed to be large, in order to keep lossless characteristics as much as possible, the denoising strength is small, the value of a wavelet series N is small, and preferably not more than 4, so that the wavelet basis function selects 'db3', the parameters are set as soft thresholds, and the scale value is set as 'mln', namely, the adjustment is carried out according to the noise level estimation of wavelet decomposition of each layer.
Step 1.3), mapping the sample characteristics to a low-dimensional space by utilizing a PCA method, and reducing the data scale, wherein the specific method comprises the following steps:
step 1.3.1), normalizing the sample characteristics, and the specific process is as follows: firstly, Z-score standardization is carried out on the characteristics of the training data, so that the characteristics fall into a small specific interval, normalization parameters such as mean values, standard deviations and the like are stored, and then Z-score standardization is carried out on the characteristics of the test data by utilizing the stored parameters.
Step 1.3.2), calculating a dimensionality reduction matrix, and the specific process is as follows: firstly, calculating a covariance matrix of sample characteristics; and then, calculating the eigenvalue and the eigenvector of the covariance matrix by adopting a singular value decomposition algorithm to obtain a dimension reduction matrix.
Step 1.3.3), reducing the dimension, namely mapping the sample characteristics to a low-dimensional space through a dimension reducing matrix.
Step 2), as shown in fig. 2, constructing a network model of two layers of LSTM and two layers of full-connected layers, wherein the two layers of LSTM and the next full-connected layer are mainly used for feature extraction; a dropout layer (a discarding layer) is connected behind the three network layers respectively, and partial nodes of the layer are invalid according to the probability of the discarding rate p, so that overfitting of the network model is prevented, and convergence of the network model is accelerated; and the last full-connection layer is mainly used for changing the output vector dimension of the network, and the node number of the full-connection layer is the engine class number.
Wherein, the node numbers of the first layer and the third layer of LSTM can be respectively set as 200 and 100; the node numbers of the fifth layer and the seventh layer of the full connecting layer can be respectively set to be 50 and 260 (engine type number); the discard rate p is generally between 0.2 and 0.5, the discard rates of the second layer and the fourth layer may be set to 0.2, and the discard rate of the sixth layer may be set to be greater than 0.4, so as to accelerate the convergence of the network model.
The two LSTM layers can effectively improve the feature extraction capability of the network, but in order to prevent the problems of difficult convergence and gradient disappearance caused by excessive network parameters, the number of nodes in the second LSTM layer is reduced; the first full-connection layer is used for further improving the learning capacity of the network on the premise that network parameters are increased by a small amount, the second full-connection layer is used for classification, and the number of nodes is the number of engine categories. And a dropout layer is accessed after the LSTM layer and the full connection layer, so that the learning of the network can be accelerated, and overfitting can be prevented.
And 3) inputting the training data preprocessed in the step 1) into the network model constructed in the step 2) to train the network, so as to obtain a trained network model.
Step 3.1), the label value higher than a certain threshold value is cut to be equal to the set threshold value, so that the network can regard the example with the higher residual service life value as the same, the engine can learn more from the sequence data when the fault is about to occur, and the prediction accuracy is improved.
The selection of the threshold value generally depends on the priority recognition of the engine, such as statistics that the remaining service life cycle of most engines is 0-200; if the threshold is set to 150, the priority of the engine having a remaining useful life of less than 150 is considered to be higher, and the accuracy of prediction of the remaining useful life of these engines is also required to be higher.
Step 3.2), a time window mechanism is adopted, namely, in each time step, the remaining service life of the engine at the moment is jointly predicted by using the time sequence contained in the time window, the length of the time window is not suitable to be overlarge and is generally set to be less than 30, and the minimum value of the length of the time sequence of each engine in the test data is preferably smaller; and adopting a variable learning rate strategy such as piewise and the like, and gradually reducing the learning rate in a segmentation way in the training process so as to accelerate the convergence of the network model.
And 4) inputting the test data preprocessed in the step 1) into the trained network model in the step 3), and outputting a predicted value of the residual service life of the engine.
The root mean square error RMSE is an evaluation index widely applied in a time sequence model, and is defined as shown in formula (1):
Figure BDA0003319123810000051
where m denotes the number of samples, yiThe true value of the ith sample is represented,
Figure BDA0003319123810000052
representing the predicted value of the ith sample. In the present embodiment, yiRepresents the true value of the RUL,
Figure BDA0003319123810000053
representing the predicted RUL value.
Fig. 3 is a histogram of the root mean square error RMSE of the engine under test in the present embodiment, in which the abscissa is the difference between the predicted value of RUL and the true value of RUL, and the ordinate is the frequency of occurrence, and the higher the frequency of occurrence in the region where the error value is smaller represents the more accurate the prediction result. FIG. 4 is a comparison graph of the predicted RUL value and the true RUL value of a part of the tested engines in the present embodiment, wherein the abscissa is the serial number of the engine, the ordinate is the RUL value, the real star points represent the true values, and the real dots represent the predicted values. The two prediction result graphs can show that the RUL prediction value and the RUL true value of the tested engine are relatively close, and the method can accurately predict the residual service life of the engine.

Claims (8)

1. A method for predicting the remaining service life of an aircraft engine is characterized by comprising the following steps:
carrying out data preprocessing on an original time sequence of an engine to obtain first training data and first test data;
constructing a network model of two long-short time memory network layers and two full-connection layers, wherein the network model comprises seven layers, the first layer and the third layer are the long-short time memory network layers, the fifth layer and the seventh layer are full-connection layers, and the second layer, the fourth layer and the sixth layer are discarding layers;
inputting first training data into the network model to perform network training to obtain a trained network model;
and inputting the first test data into the trained network model for testing to obtain the remaining service life value of the engine.
2. The method of claim 1, wherein preprocessing the raw time series of the engine to obtain first training data and first test data comprises:
extracting effective sample characteristics from training data in an original time sequence of an engine, determining the remaining service life of the engine corresponding to the sample characteristics, and extracting the effective sample characteristics from test data in the original time sequence of the engine, wherein the sample characteristics comprise: operating conditions and sensor values;
and respectively carrying out wavelet transform denoising and principal component analysis dimensionality reduction on the extracted sample features aiming at the training data and the test data to obtain first training data and first test data.
3. The method according to claim 2, characterized in that the remaining useful life of the engine corresponding to the sample characteristics is determined, in particular:
the full life cycle value of the engine is subtracted by the number of cycles currently in operation.
4. The method of claim 2, wherein the pre-processing of the data for the original time series of engines results in first training data and first test data, and wherein the method further comprises:
and modifying the remaining service life value of the engine, which is higher than a certain threshold value, in the first training data into the threshold value.
5. The method according to claim 1, wherein inputting first training data into the network model for network training to obtain a trained network model specifically comprises:
at each time step, the remaining useful life of the engine at that time is collectively predicted using a time series sequence contained within a time window having a length less than the minimum of the lengths of the time series sequences of the respective engines in the first test data.
6. The method according to claim 1, wherein inputting first training data into the network model for network training to obtain a trained network model specifically comprises:
and gradually reducing the learning rate in a segmentation way in the training process by adopting a variable learning rate strategy to accelerate the convergence of the network model.
7. The method of claim 2, wherein the wavelet transform method is a wavelet transform fixed threshold denoising method.
8. The method of claim 7, wherein the wavelet basis function is db 3.
CN202111240696.1A 2021-10-25 2021-10-25 Method for predicting remaining service life of aircraft engine Pending CN114004144A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202111240696.1A CN114004144A (en) 2021-10-25 2021-10-25 Method for predicting remaining service life of aircraft engine

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202111240696.1A CN114004144A (en) 2021-10-25 2021-10-25 Method for predicting remaining service life of aircraft engine

Publications (1)

Publication Number Publication Date
CN114004144A true CN114004144A (en) 2022-02-01

Family

ID=79923738

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202111240696.1A Pending CN114004144A (en) 2021-10-25 2021-10-25 Method for predicting remaining service life of aircraft engine

Country Status (1)

Country Link
CN (1) CN114004144A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117192063A (en) * 2023-11-06 2023-12-08 山东大学 Water quality prediction method and system based on coupled Kalman filtering data assimilation

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117192063A (en) * 2023-11-06 2023-12-08 山东大学 Water quality prediction method and system based on coupled Kalman filtering data assimilation
CN117192063B (en) * 2023-11-06 2024-03-15 山东大学 Water quality prediction method and system based on coupled Kalman filtering data assimilation

Similar Documents

Publication Publication Date Title
CN109492193B (en) Abnormal network data generation and prediction method based on deep machine learning model
CN111830408A (en) Motor fault diagnosis system and method based on edge calculation and deep learning
CN111562108A (en) Rolling bearing intelligent fault diagnosis method based on CNN and FCMC
CN111861272B (en) Multi-source data-based complex electromechanical system abnormal state detection method
CN112785091A (en) Method for performing fault prediction and health management on oil field electric submersible pump
CN114167838B (en) Multi-scale health assessment and fault prediction method for servo system
CN115017826B (en) Method for predicting residual service life of equipment
CN115409131B (en) Production line abnormity detection method based on SPC process control system
CN113449473B (en) Wind power bearing degradation trend prediction method, system and device based on DFCAE-BiLSTM
CN115791174B (en) Rolling bearing abnormality diagnosis method, system, electronic equipment and storage medium
CN111504647A (en) AR-MSET-based performance degradation evaluation method for rolling bearing
CN116150901A (en) Rolling bearing residual life prediction method based on attention-enhanced time-frequency converter
CN115858794B (en) Abnormal log data identification method for network operation safety monitoring
CN115290326A (en) Rolling bearing fault intelligent diagnosis method
CN114004144A (en) Method for predicting remaining service life of aircraft engine
CN114741945B (en) On-line fault diagnosis method for aero-engine
CN114492534A (en) Construction method and application of cross-size motor bearing fault diagnosis model
CN114330430A (en) Elevator fault judgment method and system based on big data characteristic analysis
CN117592543A (en) Aeroengine rolling bearing fault diagnosis method based on self-supervision learning
CN116861343A (en) Bearing fault diagnosis method
CN113051809A (en) Virtual health factor construction method based on improved restricted Boltzmann machine
CN112834194B (en) Fault intelligent detection method based on soft target measurement under fault-free sample
CN116361728B (en) Civil aircraft system level abnormal precursor identification method based on real-time flight data
CN115580445B (en) Unknown attack intrusion detection method, unknown attack intrusion detection device and computer readable storage medium
Toledo et al. Acoustic emission based fault diagnosis via a novel deep convolutional neural network method

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