CN116340751A - Neural network-based aeroengine sensor fault diagnosis method and system - Google Patents

Neural network-based aeroengine sensor fault diagnosis method and system Download PDF

Info

Publication number
CN116340751A
CN116340751A CN202310155116.1A CN202310155116A CN116340751A CN 116340751 A CN116340751 A CN 116340751A CN 202310155116 A CN202310155116 A CN 202310155116A CN 116340751 A CN116340751 A CN 116340751A
Authority
CN
China
Prior art keywords
parameters
characteristic
feature
parameter
matching degree
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
CN202310155116.1A
Other languages
Chinese (zh)
Inventor
陈映雪
冯冠翔
缑林峰
陈华涛
迟浩然
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.)
Northwestern Polytechnical University
Original Assignee
Northwestern Polytechnical University
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 Northwestern Polytechnical University filed Critical Northwestern Polytechnical University
Priority to CN202310155116.1A priority Critical patent/CN116340751A/en
Publication of CN116340751A publication Critical patent/CN116340751A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
    • G01M15/00—Testing of engines
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/004—Artificial life, i.e. computing arrangements simulating life
    • G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
    • 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
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
    • Y02T10/00—Road transport of goods or passengers
    • Y02T10/10—Internal combustion engine [ICE] based vehicles
    • Y02T10/40—Engine management systems

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Computational Linguistics (AREA)
  • General Health & Medical Sciences (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Health & Medical Sciences (AREA)
  • Testing Or Calibration Of Command Recording Devices (AREA)

Abstract

The invention discloses a neural network-based aero-engine sensor fault diagnosis method and system, which comprises the steps of firstly obtaining a sample data set of an aero-engine sensor, and extracting characteristic parameters in the sample data set to obtain a characteristic set; determining the optimal characteristic parameter with the highest matching degree, and updating the balance parameter and the self-adaptive parameter in the HPO algorithm according to the optimal characteristic parameter and the corresponding matching degree; updating the feature set according to the updated balance parameters and the adaptive parameters and the average value of all the feature parameters in the feature set; determining the matching degree of each characteristic parameter in the updated characteristic set, and updating the optimal characteristic parameter according to the matching degree of each characteristic parameter; and predicting the fault type of the aeroengine sensor according to the optimal characteristic parameters. The method combines the feature extraction and selection technology, and performs feature selection by utilizing a hunter-prey optimization algorithm so as to realize multi-objective optimization for reducing information redundancy and improving the accuracy and efficiency of fault diagnosis and prediction of the fault air engine.

Description

Neural network-based aeroengine sensor fault diagnosis method and system
Technical Field
The invention relates to the field of fault detection of aeroengines, in particular to a neural network-based aeroengine sensor fault diagnosis method and system.
Background
The aeroengine is a complex multi-component coupling system, and in order to ensure the safe operation of the aeroengine and realize the health management of the aeroengine, the operation condition of the aeroengine needs to be monitored in real time so as to take timely and effective measures to cope with various conditions. Real-time monitoring of the operation of an aeroengine is mainly implemented by a large number of various sensors. The sensor itself is extremely prone to failure over time and under the influence of severe operating conditions, which presents difficulties in diagnosing the failure of the aircraft engine.
Currently, three methods are mainly used for fault diagnosis of an aeroengine, namely a method based on a physical model, a data driving method and a method combining the physical model and the data driving method. The method based on the physical model is the earliest developed method and has been widely applied at present. The method is used for diagnosing faults of the aero-engine, such as a Kalman filter, a parity space and the like, according to a physical model and related knowledge rules of the aero-engine. However, because the aeroengine has a dynamically changing operating environment and a complex coupling mechanism, high-precision modeling of the aeroengine is difficult to realize, and the performance of the fault diagnosis technology based on the physical model is greatly affected. Although the method based on the fusion of the physical model and the data driving has better results compared with the single model method, the matrix between the data and the physical rules in the fusion process is difficult to process. Therefore, this method is poorly applicable in terms of fault diagnosis.
Compared with the two methods, the method based on data driving has the unique advantages that the method does not need to deeply explore the principle of a physical model or support related prior knowledge of an aeroengine, and can carry out fault diagnosis only depending on the original data or the characteristics of the original data. The fault diagnosis method based on data driving mainly comprises machine learning, neural network, deep learning and other methods, and the methods are applied to fault diagnosis in different fields. The machine learning method has limited learning ability in the face of nonlinear signal characteristics of increasingly complex aeroengines, and is difficult to obtain ideal fault diagnosis results. In addition, most of the neural networks are simpler in structure and difficult to well represent when facing complex fault characteristics, so that the neural network-based method is poor in applicability in the field of aeroengine fault diagnosis and difficult to obtain ideal effects. Moreover, the diagnosis method without feature extraction often takes a long time, so that feature extraction and selection play an important role in neural network-based fault diagnosis.
Disclosure of Invention
Aiming at the defects of low efficiency and poor applicability of the traditional model-based and hybrid fault diagnosis technology of the aero-engine, the invention provides a neural network-based aero-engine sensor fault diagnosis method and system, and a hunter-prey optimization (HPO) algorithm is utilized for feature selection so as to realize multi-objective optimization for reducing information redundancy and improving fault diagnosis accuracy and efficiency.
The invention is realized by the following technical scheme:
an aeroengine sensor fault diagnosis method based on a neural network comprises the following steps:
step 1, acquiring a sample data set of an aeroengine sensor, and extracting characteristic parameters in the sample data set to obtain a characteristic set;
step 2, calculating the matching degree of each group of characteristic parameters in the characteristic set, and determining the optimal characteristic parameter with the highest matching degree;
step 3, updating balance parameters and self-adaptive parameters in the HPO algorithm according to the optimal characteristic parameters and the corresponding matching degree;
step 4, updating the feature set obtained in the step 2 according to the updated balance parameters and the updated self-adaptive parameters and the average value of all the feature parameters in the feature set;
step 5, determining the matching degree of each group of characteristic parameters in the characteristic set updated in the step 4, and updating the positions of hunters and prey in the characteristic set in the step 2 according to each characteristic parameter to obtain an updated characteristic set;
and 6, repeating the steps 3-5 until the set iteration condition is reached, outputting the optimal characteristic parameters, and predicting the fault type of the aero-engine sensor according to the optimal characteristic parameters.
Preferably, in step 1, operation data of an aeroengine sensor is obtained, and the operation data is normalized to obtain a sample data set, and the normalization method is as follows:
Figure BDA0004092101280000031
wherein y is m Representing the operation data of the sensor, m represents the actual fault class number, y m，max ，y m，min Maximum and minimum values of the operation data corresponding to each fault type,
Figure BDA0004092101280000032
for normalized sample data, n represents the actual failure category number.
Preferably, in step 1, sample data in the sample data set are grouped, feature parameters of each group of sample data are extracted, and a binary method is adopted to select the feature parameters, so as to obtain N groups of feature parameters and a feature set formed by the N groups of feature parameters.
Preferably, the method for extracting the characteristic parameters comprises the following steps:
x i ＝lb i +rand×(ub i -lb i )；rand∈[0，1]
Figure BDA0004092101280000033
wherein x is i Is indicative of characteristic parameter lb i And ub i Representing the lower and upper bounds of the characteristic parameter, respectively.
Preferably, the method for calculating the matching degree in the step 2 is as follows:
Figure BDA0004092101280000034
wherein Fitness is provided i Represents the matching degree of the characteristic parameters of the i group, err i To classify errors, w 1 As main weight, w 2 As a secondary weight, d i And the number of the features contained in the selected feature parameters is represented, and D is the dimension.
Preferably, the updating method of the balance parameter in the step 3 is as follows:
Figure BDA0004092101280000035
wherein T is max The maximum iteration number is t is the current iteration number;
the updating method of the self-adaptive parameters comprises the following steps:
Figure BDA0004092101280000036
Figure BDA0004092101280000037
wherein, the liquid crystal display device comprises a liquid crystal display device,
Figure BDA0004092101280000038
and->
Figure BDA0004092101280000039
Represents a random vector, each value of the vector is between 0 and 1, R 2 Is a random number between 0 and 1.
Preferably, in step 4, the feature set is updated in combination with a decrementing mechanism, so as to increase the convergence speed of the HPO algorithm.
Preferably, the updating method of the feature set is as follows:
Figure BDA0004092101280000041
wherein beta is a constant, P pos(j) Representing the position of the best feature parameter in the parameter set matrix, R 4 Represents a random number within 0 to 1, D euc(i) Represents Euclidean distance, R 5 Is a random number.
Preferably, in step 6, the optimal characteristic parameters are input into the LSTM neural network to predict the sensor fault type of the aeroengine.
A system for an aeroengine sensor fault diagnosis method based on a neural network comprises,
the characteristic parameter selection module is used for acquiring a sample data set of the aeroengine sensor, extracting characteristic parameters in the sample data set and obtaining a characteristic set;
the feature matching module is used for calculating the matching degree of each feature parameter in the feature set and determining the optimal feature parameter with the highest matching degree;
the algorithm updating module is used for updating the balance parameter and the self-adaptive parameter in the HPO algorithm according to the optimal characteristic parameter and the corresponding matching degree;
the feature updating module is used for updating the feature set according to the updated balance parameters, the self-adaptive parameters and the average value of all the feature parameters in the feature set;
the optimal feature optimization module is used for determining the matching degree of each feature parameter in the updated feature set and updating the optimal feature parameters according to the matching degree of each feature parameter;
and the prediction module is used for outputting the optimal characteristic parameters according to the set iteration conditions and predicting the fault type of the aeroengine sensor according to the optimal characteristic parameters.
A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that the processor implements the steps of the neural network based aircraft engine sensor fault diagnosis method when the computer program is executed.
Compared with the prior art, the invention has the following beneficial technical effects:
according to the neural network-based aeroengine sensor fault diagnosis method provided by the invention, based on original fault data, data multi-domain feature extraction is carried out, then a hunter-prey optimization (HPO) algorithm is adopted to carry out optimization selection on the extracted multi-domain features, a multi-domain feature set which can fully embody the fault features and does not contain excessive redundant information is determined, the fault diagnosis accuracy and efficiency are improved, the HPO algorithm can well balance the exploration and development stages of the algorithm, and better performance is achieved in terms of convergence speed and local optimization avoidance compared with other methods; finally, based on the selected multi-domain feature set, a long-short-term memory neural network (LSTM) which is one of the deep learning algorithms is adopted for fault diagnosis, and through verification, the diagnosis method is more reliable and has strong generalization capability, has good fault type identification capability, can be suitable for fault diagnosis of the aero-engine, and provides more reliable guarantee for safe operation of the aero-engine.
Drawings
FIG. 1 is a flow chart of a method of diagnosing an aircraft engine sensor fault in accordance with the present invention;
FIG. 2 is a diagram of an LSTM based confusion matrix in accordance with the present invention;
FIG. 3 is a diagram of an SVM-based confusion matrix of the present invention;
fig. 4 is a confusion matrix diagram of BP of the present invention.
Detailed Description
The invention will now be described in further detail with reference to the accompanying drawings, which illustrate but do not limit the invention.
Referring to fig. 1-4, the neural network-based aero-engine sensor fault diagnosis method comprises the following steps:
step 1, acquiring operation data of an aeroengine sensor, and normalizing the operation data to obtain a sample data set.
Specifically, the operation data of the aeroengine sensor is normalized by using a formula 1 to reduce adverse effects caused by fluctuation of transportation data, the operation data after normalization is used as a sample data set, the sample data set is divided into two groups, 80% of sample data is used as a training sample set, and 20% of sample data is used as a test data set.
The method for normalizing the operation data is as follows:
Figure BDA0004092101280000061
wherein y is m Representing the operation data of the sensor, m represents the actual fault class number, y m，max ，y m，min Maximum and minimum values of the operation data corresponding to each fault type,
Figure BDA0004092101280000062
for operating data y of the sensor m Normalized data, n represents the actual number of fault classes and +.>
Figure BDA0004092101280000063
And 2, extracting characteristic parameters of each sample data in the sample data set, and converting the characteristic parameters by adopting a binary method to obtain a characteristic set formed by N groups of characteristic parameters.
Specifically, the characteristic parameters of the sample data are selected according to the formula (2 a), and as the characteristic parameters can only be selected or not selected, the selection process has binary characteristics, and then binary conversion is performed through the formula (2 b), so that a characteristic set formed by N groups of characteristic parameters is obtained.
x i ＝lb i +rand×(ub i -lb i )；rand∈[0，1] (2a)
Figure BDA0004092101280000064
Wherein x is i Is indicative of characteristic parameter lb i And ub i Representing the lower and upper bounds of the feature parameters, respectively, the process generates N sets of feature parameters, the dimension of each set of feature parameters is D, and the whole feature parameter set is represented by a matrix. 0 indicates that the corresponding feature parameter is not selected, and 1 indicates that the corresponding feature parameter is selected.
Step 3, calculating the matching degree of each group of characteristic parameters in the characteristic set, and selecting the optimal characteristic parameter T with the highest matching degree pos The matching degree calculating method comprises the following steps:
Figure BDA0004092101280000065
wherein Fitness is provided i Represents the matching degree of the characteristic parameters of the i group, err i Representing the diagnostic error rate of fault diagnosis using the set of features, the primary weight w is therefore 1 Set to 0.99, the secondary weight w 2 Set to 0.01, d i Representing the number of features contained in the selected feature parameters, fitness in FIG. 1 pos Indicating the current best match.
And 4, updating balance parameters in the HPO algorithm according to the optimal characteristic parameters and the corresponding matching degree obtained in the step 3, wherein the updating method comprises the following steps:
Figure BDA0004092101280000071
wherein T is max Is the maximum number of iterations, and t is the current number of iterations.
The self-adaptive parameters are updated according to the updated balance parameters, and the updating method comprises the following steps:
Figure BDA0004092101280000072
Figure BDA0004092101280000073
wherein, the liquid crystal display device comprises a liquid crystal display device,
Figure BDA0004092101280000074
and->
Figure BDA0004092101280000078
Represents a random vector, each value of the vector is between 0 and 1, R 2 Is a random number between 0 and 1.
And 5, updating the positions of hunters and prey in the feature set in the step 2 according to the updated balance parameters, the updated self-adaptive parameters and the average value of the feature parameters and combining a decremental mechanism to obtain an updated feature set.
Figure BDA0004092101280000075
Where β is a constant with a value of 0.1, the HPO algorithm includes 10% hunter agent and 90% hunter agent, when the random number R 5 When the random number R is smaller than 0.1, the hunter agent position is updated, and when the random number R is the random number 5 When the hunting agent position is greater than 0.1, performing hunting agent position updating; p (P) pos(j) Representing the position of the best feature parameter in the parameter set matrix, R 4 Represents a random number within 0 to 1, D euc(i) Representing the euclidean distance.
α index Is a characteristic parameter index of a vector satisfying a condition (p= =0), and an average value of all characteristic parameters in the characteristic set is found as follows:
Figure BDA0004092101280000076
and adding a decrementing mechanism into the HPO algorithm to influence the position of the optimal characteristic parameter in the parameter set matrix, so as to improve the convergence rate of the HPO algorithm.
K best ＝round(C×N) (5d)
Step 6, calculating the matching degree of each characteristic parameter in the characteristic set updated in the step 5, comparing with the optimal matching degree, and updating the optimal characteristic parameter T if the matching degree is better than the optimal matching degree pos 。
Figure BDA0004092101280000077
Step 7, repeating the steps 4-6 until the maximum number of times of iteration is reached, ending the iteration and outputting the optimal characteristic parameter T pos 。
Step 8, the optimal characteristic parameters T pos The method comprises the steps of inputting the sensor fault type of an aeroengine in an LSTM neural network in an MATLAB tool box to predict, and evaluating a prediction result, wherein an average accuracy ACC is mainly adopted as an evaluation index mean The calculation method is as follows
Figure BDA0004092101280000081
Wherein Q represents the running times of the program, M represents the number of data of the verification set of each running program, y a To predict fault categories.
Example 1
The SVM support vector machine, the BP neural network and the LSTM are selected for comparison, and the three classification models are trained and verified based on the feature subset selected by the HPO optimization algorithm, and the SVM support vector machine and the BP network are also selected by adopting a trial calculation method. The average classification accuracy of the three models for the different faults is given in table 1. The behavior of the different classifiers for each type of fault is clearly demonstrated using a confusion matrix diagram, and figures 2-4 are confusion matrices for three model validation sets.
Table 1 precision based on different models
Figure BDA0004092101280000082
The confusion matrix of fig. 2-4 is obtained under the condition of 20 runs based on the selection characteristics, the verification set has 700 data each time, each type of faults is 100, and 2000 types of faults are obtained under 20 runs, so as to eliminate the uncertainty of the network in the training and verification process.
From the three pictures and table 1, it can be seen that the average classification accuracy of the three classification models is very high, the LSTM model reaches 97.2%, the BP neural network is 96.1%, and the SVM accuracy is 94.3%. For the classification condition of single faults, the classification accuracy of the three models on pulse faults, normal operation and periodic faults is 100%. It is worth mentioning that LSTM has achieved 100% for other types of faults besides the deviation faults, while SVM and BP have more or less erroneous recognition of these faults.
The error recognition rate of the three models on deviation faults is worst, and when the deviation value is smaller, the models are poor, the LSTM with the best performance has the fault recognition rate of 80.3 percent, but the error recognition rate is still improved greatly compared with 61.0 percent and 76.0 percent of SVM and BP. The result shows that the LSTM has the capability of identifying fault information, is greatly improved compared with SVM and BP, and can better perform aeroengine fault diagnosis.
The results show that: the fault diagnosis method based on the LSTM and the optimized feature subset is reliable, strong in generalization capability, good in fault type identification capability, applicable to fault diagnosis of the aero-engine, and capable of providing more reliable guarantee for safe operation of the aero-engine.
The neural network-based aero-engine sensor fault diagnosis method provided by the invention is used for diagnosing faults by adopting a long-term and short-term memory neural network aiming at the defects of low efficiency and poor applicability of the traditional model-based and hybrid fault diagnosis technology of an aero-engine. Because LSTM processes high-dimensional data under the condition of non-characteristic selection and requires a large amount of time, the invention combines characteristic extraction and selection technology, utilizes hunter-prey optimization (HPO) algorithm to perform characteristic selection, and the HPO algorithm can well balance the exploration and development stages of the algorithm, has better performance than other methods in the aspects of convergence speed and avoiding local optimization, so as to realize multi-objective optimization of reducing information redundancy and improving the accuracy and efficiency of fault diagnosis and prediction of a fault air engine.
The invention also provides an aeroengine sensor fault diagnosis system based on the neural network, which comprises,
the characteristic parameter selection module is used for acquiring a sample data set of the aeroengine sensor, extracting characteristic parameters in the sample data set and obtaining a characteristic set;
the feature matching module is used for calculating the matching degree of each feature parameter in the feature set and determining the optimal feature parameter with the highest matching degree;
the algorithm updating module is used for updating the balance parameter and the self-adaptive parameter in the HPO algorithm according to the optimal characteristic parameter and the corresponding matching degree;
the feature updating module is used for updating the feature set according to the updated balance parameters, the self-adaptive parameters and the average value of all the feature parameters in the feature set;
the optimal feature optimization module is used for determining the matching degree of each feature parameter in the updated feature set and updating the optimal feature parameters according to the matching degree of each feature parameter;
and the prediction module is used for outputting the optimal characteristic parameters according to the set iteration conditions and predicting the fault type of the aeroengine sensor according to the optimal characteristic parameters.
The division of the modules in the embodiments of the present invention is schematically only one logic function division, and there may be another division manner in actual implementation, and in addition, each functional module in each embodiment of the present invention may be integrated in one processor, or may exist separately and physically, or two or more modules may be integrated in one module. The integrated modules may be implemented in hardware or in software functional modules.
In yet another embodiment of the present invention, a computer device is provided that includes a processor and a memory for storing a computer program including program instructions, the processor for executing the program instructions stored by the computer storage medium. The processor may be a central processing unit (Central Processing Unit, CPU), but may also be other general purpose processors, digital signal processors (Digital Signal Processor, DSP), application specific integrated circuits (Application Specific Integrated Circuit, ASIC), off-the-shelf Programmable gate arrays (FPGAs) or other Programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc., which are the computational core and control core of the terminal adapted to implement one or more instructions, in particular adapted to load and execute one or more instructions within a computer storage medium to implement the corresponding method flow or corresponding functions; the processor provided by the embodiment of the invention can be used for the operation of the neural network-based aeroengine sensor fault diagnosis method.
In yet another embodiment of the present invention, a storage medium, specifically a computer readable storage medium (Memory), is a Memory device in a computer device, for storing a program and data. It is understood that the computer readable storage medium herein may include both built-in storage media in a computer device and extended storage media supported by the computer device. The computer-readable storage medium provides a storage space storing an operating system of the terminal. Also stored in the memory space are one or more instructions, which may be one or more computer programs (including program code), adapted to be loaded and executed by the processor. The computer readable storage medium herein may be a high-speed RAM memory or a non-volatile memory (non-volatile memory), such as at least one magnetic disk memory. One or more instructions stored in a computer-readable storage medium may be loaded and executed by a processor to implement the respective steps of the above-described embodiments with respect to a neural network-based aircraft engine sensor fault diagnosis method.
The above is only for illustrating the technical idea of the present invention, and the protection scope of the present invention is not limited by this, and any modification made on the basis of the technical scheme according to the technical idea of the present invention falls within the protection scope of the claims of the present invention.

Claims (10)

1. The aeroengine sensor fault diagnosis method based on the neural network is characterized by comprising the following steps of:
step 1, acquiring a sample data set of an aeroengine sensor, and extracting characteristic parameters in the sample data set to obtain a characteristic set;
step 2, calculating the matching degree of each group of characteristic parameters in the characteristic set, and determining the optimal characteristic parameter with the highest matching degree;
step 3, updating balance parameters and self-adaptive parameters in the HPO algorithm according to the optimal characteristic parameters and the corresponding matching degree;
step 4, updating the feature set obtained in the step 2 according to the updated balance parameters and the updated self-adaptive parameters and the average value of all the feature parameters in the feature set;
step 5, determining the matching degree of each group of characteristic parameters in the characteristic set updated in the step 4, and updating the positions of hunters and prey in the characteristic set in the step 2 according to each characteristic parameter to obtain an updated characteristic set;
and 6, repeating the steps 3-5 until the set iteration condition is reached, outputting the optimal characteristic parameters, and predicting the fault type of the aero-engine sensor according to the optimal characteristic parameters.
2. The neural network-based aeroengine sensor fault diagnosis method according to claim 1, wherein in step 1, operation data of the aeroengine sensor is obtained, the operation data is normalized to obtain a sample data set, and the normalization method is as follows:
Figure FDA0004092101270000011
wherein y is m Representing the operation data of the sensor, m represents the actual fault class number, y m,max ，y m,min Maximum and minimum values of the operation data corresponding to each fault type,
Figure FDA0004092101270000012
for normalized sample data, n represents the actual failure category number.
3. The neural network-based aeroengine sensor fault diagnosis method according to claim 1, wherein in step 1, sample data in the sample data set are grouped, characteristic parameters of each group of sample data are extracted, and the characteristic parameters are selected by adopting a binary method to obtain N groups of characteristic parameters and form a characteristic set.
4. The neural network-based aeroengine sensor fault diagnosis method according to claim 3, wherein the characteristic parameter extraction method is as follows:
x i ＝lb i +rand×(ub i -lb i )；rand∈[0,1]
Figure FDA0004092101270000021
wherein x is i Is indicative of characteristic parameter lb i And ub i Representing the lower and upper bounds of the characteristic parameter, respectively.
5. The neural network-based aeroengine sensor fault diagnosis method according to claim 1, wherein the matching degree calculation method in step 2 is as follows:
Figure FDA0004092101270000022
wherein Fitness is provided i Represents the matching degree of the characteristic parameters of the i group, err i To classify errors, w 1 As main weight, w 2 As a secondary weight, d i And the number of the features contained in the selected feature parameters is represented, and D is the dimension.
6. The neural network-based aeroengine sensor fault diagnosis method according to claim 1, wherein the method for updating the balance parameter in step 3 is as follows:
Figure FDA0004092101270000023
wherein T is max The maximum iteration number is t is the current iteration number;
the updating method of the self-adaptive parameters comprises the following steps:
Figure FDA0004092101270000024
α index ＝(P＝＝0)
Figure FDA0004092101270000025
wherein, the liquid crystal display device comprises a liquid crystal display device,
Figure FDA0004092101270000026
and->
Figure FDA0004092101270000027
Represents a random vector, each value of the vector is between 0 and 1, R 2 Is a random number between 0 and 1.
7. The neural network-based aircraft engine sensor fault diagnosis method according to claim 1, wherein the feature set is updated in step 4 in combination with a decrementing mechanism for improving the convergence speed of the HPO algorithm.
8. The neural network-based aeroengine sensor fault diagnosis method according to claim 6, wherein the feature set updating method comprises the following steps:
Figure FDA0004092101270000031
wherein beta is a constant, P pos(j) Representing the position of the best feature parameter in the parameter set matrix, R 4 Represents a random number within 0 to 1, D euc(i) Represents Euclidean distance, R 5 Is a random number.
9. The neural network-based aircraft engine sensor fault diagnosis method according to claim 1, wherein in step 6, the optimal characteristic parameter is input into the LSTM neural network to predict the sensor fault type of the aircraft engine.
10. A system for performing a neural network-based aircraft engine sensor fault diagnosis method according to any one of claims 1-9, characterized in that,
the characteristic parameter selection module is used for acquiring a sample data set of the aeroengine sensor, extracting characteristic parameters in the sample data set and obtaining a characteristic set;
the feature matching module is used for calculating the matching degree of each feature parameter in the feature set and determining the optimal feature parameter with the highest matching degree;
the algorithm updating module is used for updating the balance parameter and the self-adaptive parameter in the HPO algorithm according to the optimal characteristic parameter and the corresponding matching degree;
the feature updating module is used for updating the feature set according to the updated balance parameters, the self-adaptive parameters and the average value of all the feature parameters in the feature set;
the optimal feature optimization module is used for determining the matching degree of each feature parameter in the updated feature set and updating the optimal feature parameters according to the matching degree of each feature parameter;
and the prediction module is used for outputting the optimal characteristic parameters according to the set iteration conditions and predicting the fault type of the aeroengine sensor according to the optimal characteristic parameters.
CN202310155116.1A 2023-02-22 2023-02-22 Neural network-based aeroengine sensor fault diagnosis method and system Pending CN116340751A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202310155116.1A CN116340751A (en) 2023-02-22 2023-02-22 Neural network-based aeroengine sensor fault diagnosis method and system

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202310155116.1A CN116340751A (en) 2023-02-22 2023-02-22 Neural network-based aeroengine sensor fault diagnosis method and system

Publications (1)

Publication Number Publication Date
CN116340751A true CN116340751A (en) 2023-06-27

Family

ID=86883102

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202310155116.1A Pending CN116340751A (en) 2023-02-22 2023-02-22 Neural network-based aeroengine sensor fault diagnosis method and system

Country Status (1)

Country Link
CN (1) CN116340751A (en)

Similar Documents

Publication Publication Date Title
US10713597B2 (en) Systems and methods for preparing data for use by machine learning algorithms
JP2008536221A (en) Control system and method
Lanubile et al. Comparing models for identifying fault-prone software components.
JP2008536219A (en) Diagnosis and prediction method and system
CN110442516B (en) Information processing method, apparatus, and computer-readable storage medium
US8738271B2 (en) Asymmetric wavelet kernel in support vector learning
Singh et al. Software defect prediction tool based on neural network
CN113923104B (en) Network fault diagnosis method, equipment and storage medium based on wavelet neural network
CN116881832B (en) Construction method and device of fault diagnosis model of rotary mechanical equipment
WO2022077907A1 (en) Adversarial attack detection method, system and device, and computer-readable storage medium
CN116668083A (en) Network traffic anomaly detection method and system
CN114609994A (en) Fault diagnosis method and device based on multi-granularity regularization rebalance incremental learning
Xiu et al. Variational disentanglement for rare event modeling
Song et al. Software defect prediction based on elman neural network and cuckoo search algorithm
Ding et al. A zero-shot soft sensor modeling approach using adversarial learning for robustness against sensor fault
CN113541985B (en) Internet of things fault diagnosis method, model training method and related devices
Byun et al. Black-box testing of deep neural networks
Hond et al. Verifying artificial neural network classifier performance using dataset dissimilarity measures
CN116340751A (en) Neural network-based aeroengine sensor fault diagnosis method and system
CN116910526A (en) Model training method, device, communication equipment and readable storage medium
US20220269991A1 (en) Evaluating reliability of artificial intelligence
CN114637620A (en) Database system abnormity classification prediction method based on SVM algorithm
CN112906751A (en) Method for identifying abnormal value through unsupervised learning
Lu et al. Robust least-squares support vector machine using probabilistic inference
Boateng Unsupervised Ensemble Methods for Anomaly Detection in PLC-based Process Control

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