[go: up one dir, main page]

CN120197005A - A method for fault diagnosis of aeroengine gas path system based on data and mechanism fusion - Google Patents

A method for fault diagnosis of aeroengine gas path system based on data and mechanism fusion Download PDF

Info

Publication number
CN120197005A
CN120197005A CN202510632525.5A CN202510632525A CN120197005A CN 120197005 A CN120197005 A CN 120197005A CN 202510632525 A CN202510632525 A CN 202510632525A CN 120197005 A CN120197005 A CN 120197005A
Authority
CN
China
Prior art keywords
neural network
fault
data
circuit system
sequence
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
CN202510632525.5A
Other languages
Chinese (zh)
Inventor
年夫强
刘太秋
杨光红
李霄剑
董久祥
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.)
AECC Shenyang Engine Research Institute
Original Assignee
AECC Shenyang Engine Research Institute
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 AECC Shenyang Engine Research Institute filed Critical AECC Shenyang Engine Research Institute
Priority to CN202510632525.5A priority Critical patent/CN120197005A/en
Publication of CN120197005A publication Critical patent/CN120197005A/en
Pending legal-status Critical Current

Links

Classifications

    • GPHYSICS
    • G06COMPUTING OR CALCULATING; COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00Pattern recognition
    • G06F18/20Analysing
    • G06F18/24Classification techniques
    • GPHYSICS
    • G06COMPUTING OR CALCULATING; COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00Computing arrangements based on biological models
    • G06N3/004Artificial life, i.e. computing arrangements simulating life
    • G06N3/006Artificial 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]
    • GPHYSICS
    • G06COMPUTING OR CALCULATING; COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00Computing arrangements based on biological models
    • G06N3/02Neural networks
    • G06N3/08Learning methods
    • G06N3/084Backpropagation, e.g. using gradient descent

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Health & Medical Sciences (AREA)
  • Software Systems (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Biophysics (AREA)
  • Biomedical Technology (AREA)
  • Mathematical Physics (AREA)
  • Computational Linguistics (AREA)
  • Health & Medical Sciences (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Biology (AREA)
  • Testing And Monitoring For Control Systems (AREA)

Abstract

The application belongs to the technical field of fault diagnosis of an aeroengine gas circuit system, and particularly relates to a fault diagnosis method of the aeroengine gas circuit system with data and mechanism integrated, which comprises a fault detection method, wherein the fault detection method comprises the following steps of constructing a BP neural network, and initializing the weight and bias of the BP neural network; the method comprises the steps of obtaining a BP neural network output estimation model by training the BP neural network based on input and measurement output data of a gas circuit system, optimizing weight and bias of the BP neural network by a particle swarm optimization PSO, obtaining a BP neural network output estimation model by training the BP neural network based on the input and measurement output data of the gas circuit system, obtaining a gas circuit system output estimation sequence by calculating the BP neural network output estimation model in real time, calculating a residual sequence and a residual estimation function thereof, and detecting faults of the gas circuit system.

Description

Data and mechanism fusion type aeroengine gas circuit system fault diagnosis method
Technical Field
The application belongs to the technical field of aero-engine gas circuit system fault diagnosis, and particularly relates to a data and mechanism fusion aero-engine gas circuit system fault diagnosis method.
Background
Aircraft engines, which are one of the most important components of aircraft, are complex aerodynamic thermodynamic systems integrating aircraft, electric, gas, fluid and other technologies, and generally operate for a long time in demanding environments, with extremely high temperatures, pressures, rotational speeds, vibrations and loads. Thus, as the operating time increases, the reliability of the aero-engine gradually decreases and the occurrence of faults is unavoidable.
In aeroengine systems, faults are generally classified into gas circuit system faults, bearing system faults and control system faults, wherein the gas circuit system faults are generally caused by factors such as component aging, mechanical friction increase and the like, and are most difficult to accurately detect and position. The method can accurately and effectively diagnose the fault of the gas circuit system, take corresponding remedy and maintenance measures, and has important significance for improving the stability and safety of the aero-engine and prolonging the service life of the high aero-engine.
At present, for the fault diagnosis of an aeroengine gas circuit system, a model-based diagnosis method and a data-driven diagnosis method are mainly available.
The diagnosis method based on the model is designed by using a Kalman filter, such as a linear Kalman filter, an extended Kalman filter, a unscented Kalman filter design and the like, but due to the complexity of the air circuit system structure of the aeroengine, the related parameters of the model are difficult to measure, so that the diagnosis method based on the model cannot be well applied, and even false alarm or missing alarm can be generated.
The data-driven diagnosis method utilizes historical data of the aeroengine to carry out fault diagnosis, and mostly adopts a subspace identification-based method, wherein the method needs multiple steps of identification represented by a kernel function, comprises strict equation operation, the order of a system and other prior knowledge, and has the following remarkable defects:
1) The neural network is directly introduced to carry out training and verification, theoretical analysis and proof are lacked, and the interpretation is poor;
2) Certain limitations are imposed on the fault diagnosis of a linear system with unknown parameters, including the requirement that the system order is known and the matrix is full of lines and ranks;
3) By adopting the neural network method, the network optimization is insufficient, the algorithm execution period is long, the operation efficiency is low, the requirement on the computer performance is high, and the method is not suitable for rapid processing analysis and fault detection of massive aeroengine operation data.
The present application has been made in view of the above-described technical drawbacks.
Disclosure of Invention
The application aims to provide a fault diagnosis method for an air circuit system of an aeroengine, which integrates data and mechanisms, is a fault diagnosis method without strict requirements on the air circuit system, and can achieve the purposes of fault detection and fault isolation only through input and output data of the air circuit system of the aeroengine under the condition of not depending on parameters of an air flow system, so as to overcome or alleviate the technical defects of at least one aspect of the prior art.
The technical scheme of the application is as follows:
a method for diagnosing faults of an aeroengine gas circuit system by fusing data and mechanisms comprises a fault detection method, wherein the fault detection method comprises the following steps:
Step one, constructing a BP neural network, and initializing the weight and bias of the BP neural network;
Step two, based on input and measurement output data of the gas circuit system, optimizing the weight and bias of the BP neural network by using a particle swarm algorithm PSO;
Training the BP neural network based on input and measurement output data of the gas circuit system to obtain an output estimation model of the BP neural network;
And fourthly, calculating an output estimation sequence of the gas circuit system in real time by using the BP neural network output estimation model, calculating a residual sequence and a residual estimation function thereof, and detecting faults of the gas circuit system.
Optionally, in the method for diagnosing the fault of the air circuit system of the aeroengine by fusing the data and the mechanism, the step two of the fault detection method specifically comprises the following steps:
S21, initializing a particle swarm in a particle swarm algorithm PSO for the weight and the bias of the BP neural network;
S22, calculating an adaptability function value of a particle swarm algorithm PSO to the BP neural network based on input, measurement and output data of the gas circuit system;
s23, calculating the optimal position of single particle in particle swarm algorithm PSO And the optimal position of particle swarm;
S24, judging whether the maximum iteration times are reached or not, or judging whether the fitness function value is smaller than a set error threshold value or not:
if so, the optimal position of the single particle is obtained And the optimal position of particle swarmObtaining an optimized value of the weight and the bias of the BP neural network;
if not, then based on the best position of the individual particles And the optimal position of particle swarmThe particle position and velocity are updated and the process returns to S22.
Optionally, in the method for diagnosing the fault of the air path system of the aeroengine by fusing the data and the mechanism, in the step four of the fault detection method, an output estimation sequence of the air path system is calculated in real time by using a BP neural network output estimation model, and specifically the method comprises the following steps:
;
wherein, the
The output estimation sequence of the gas circuit system is selfLength from moment to moment isIs a column vector of a stack of (a) columns;
is an input sequence of the gas circuit system, is self-contained Length from moment to moment isIs a column vector of a stack of (a) columns;
is an input sequence of the gas circuit system, is self-contained Length from moment to moment isIs a column vector of a stack of (a) columns;
is a health parameter sequence of the gas circuit system, is self-contained Length from moment to moment isIs a column vector of a stack of (a) columns;
is a health parameter sequence of the gas circuit system, is self-contained Length from moment to moment isIs a column vector of a stack of (a) columns;
the measuring output sequence of the gas circuit system is self Length from moment to moment isIs a column vector of a stack of (a) columns;
is the super parameter of BP neural network;
is a structural function of the BP neural network.
Optionally, in the method for diagnosing the fault of the air circuit system of the aeroengine by fusing the data and the mechanism, in the step four of the fault detection method, a residual sequence is calculated, and the method specifically comprises the following steps:
;
wherein, the
Is a residual sequence;
is a measuring output sequence of the gas circuit system, is self-contained Length from moment to moment isIs a column vector of a stack of (a) columns.
Optionally, in the method for diagnosing a fault of an air circuit system of an aeroengine by fusing data and mechanisms, in the step four of the fault detection method, a residual error evaluation function is calculated, which specifically includes:
;
wherein, the
Evaluating a function for the residual error;
Is that Is not limited to the above-described embodiments.
Optionally, in the method for diagnosing a fault of an air path system of an aeroengine by integrating the data and the mechanism, in the step four of the fault detection method, the fault detection is performed on the air path system, specifically:
residual error evaluation function And residual error thresholdComparing, judging whether the gas circuit system has faults, if soJudging that the gas circuit system has not failed, ifAnd judging that the gas circuit system fails.
Optionally, in the method for diagnosing the fault of the air circuit system of the aeroengine by fusing the data and the mechanism, in the step four of the fault detection method, a residual error threshold value is setThe specific calculation is as follows:
;
wherein, the
A level of significance acceptable for a gas circuit system fault;
Is the degree of freedom of chi-square distribution.
Optionally, the method for diagnosing the fault of the air circuit system of the aeroengine by fusing the data and the mechanism further comprises a fault isolation method, and the fault isolation method comprises the following steps:
Step one, constructing a fault database of a gas circuit system, wherein each fault data comprises a process input and a measurement output;
Step two, determining fault types corresponding to the fault data, and setting corresponding labels for the fault data;
Step three, fault data are selected as input, corresponding labels are used as output, and the fault isolation neural network is trained to obtain a fault isolation neural network model;
And fourthly, calculating fault data of the gas circuit system in real time by using a fault isolation neural network model, outputting a corresponding label, and further judging and obtaining a corresponding fault type.
Optionally, in the method for diagnosing the fault of the air circuit system of the aeroengine by fusing the data and the mechanism, in the step three of the fault isolation method, the fault isolation neural network adopts a BP neural network.
The application has at least the following beneficial technical effects:
The method has the advantages that a model for estimating and outputting is constructed by using a BP neural network, fault detection is carried out, feasibility and interpretability are realized, a residual sequence and a residual evaluation function thereof are calculated, fault judgment is carried out by comparing the residual sequence with a residual threshold, fault detection can be realized only by inputting and outputting data, the parameter of the gas circuit system is not depended, the method has higher engineering application value, a model for fault isolation is constructed by using the neural network model on the basis of fault diagnosis, and fault types are identified with high accuracy.
Drawings
Fig. 1 is a schematic diagram of a BP neural network provided by an embodiment of the present application;
FIG. 2 is a schematic diagram of fault detection provided by an embodiment of the present application;
fig. 3 is a schematic diagram of an aeroengine gas circuit system fault diagnosis method with data and mechanism fusion provided by the embodiment of the application.
For the purpose of better illustrating the embodiments, the drawings are certain drawings that are omitted, enlarged or reduced in size, and are not to be construed as limiting the application.
Detailed Description
In order to make the technical solution of the present application and its advantages more clear, the technical solution of the present application will be further and completely described in detail with reference to the accompanying drawings, it being understood that the specific embodiments described herein are only some of the embodiments of the present application, which are for explanation of the present application and not for limitation of the present application. It should be noted that, for convenience of description, only a portion related to the present application is shown in the drawings, and other related portions may refer to a general design.
Furthermore, unless defined otherwise, technical or scientific terms used in the description of the application should be given the ordinary meaning as understood by one of ordinary skill in the art to which the application pertains. As used in this description of the application, the word "comprising" means that the element preceding the word covers the elements listed after the word and equivalents thereof without excluding other associated elements.
According to the embodiment of the application, under a data driving framework, a fault diagnosis method of an aeroengine gas circuit system with data and mechanism fused is designed, aiming at the limitation of a subspace identification method, a residual error generator is designed and a residual error evaluation function is established by introducing a BP neural network, under the condition that the system parameters are not depended, fault detection and fault isolation are realized by only inputting and outputting data through the operation of an aeroengine, and meanwhile, the related parameters of the network are optimized by introducing a particle swarm algorithm PSO aiming at the problems of the traditional BP neural network, so that the performance and fitting precision of the BP neural network are improved.
The discrete time invariant equation of the aeroengine gas circuit system is established as follows:
............(1.1)
wherein, the Is the firstThe state of the gas circuit system at the moment,Is the firstThe process input of the time gas circuit system,Is the firstHealth parameters of the air path system at the moment,Is the firstThe measurement output of the gas circuit system at the moment,Is the firstProcess noise of the air circuit system at the moment,Is the firstMeasuring noise of the moment air circuit system, andAre uncorrelated zero-mean white noise signals.
For a given operating point, equation (1.1) is linearized, which can be expressed in the form:
............(1.2)
wherein, the Is a gas circuit system parameter with proper dimension.
In an aero-engine system, gas circuit system component efficiency and flow are critical health parameters, which are non-measurable parameters, and the gas circuit system matrix is unknown through a regression linearization process, in particular, if the gas circuit system is affected by a health parameter failure, equation (1.2) can be changed into the following form:
............(1.3)
wherein, the Is the fault vector of the gas circuit system to be diagnosed.
Faults in aircraft engine systems can lead to abrupt changes in health parameters, for example, if a fan in an aircraft engine system fails, the health parameters such as fan efficiency and fan airflow can also change accordingly.
Because the air circuit system of the aeroengine is complex in structure, the model parameters of the air circuit system are difficult to accurately measure, fault detection and fault isolation are realized through input and output data under the condition of not depending on the parameters of the air circuit system, and an equation (1.3) can be written into the form of the following transfer function:
......(1.4)
wherein, the To expand the observability matrix, the specific form is as follows:
............(1.5)
For the block Toeplitz matrix, the specific form is as follows:
............(1.6)
for the health parameter-output transfer matrix, For a process noise-output transfer matrix, both have the same structure, andIn addition,The noise-output transfer matrix is measured as an identity matrix.
Is a measuring output sequence of the gas circuit system, is self-containedLength from moment to moment isIs used to determine the column vector of the stack, the definition is as follows:
............(1.7)
Gas circuit system input sequence Sequence of health parametersSequence of faultsProcess noise sequenceMeasuring noiseAnd (3) withThe same structure.
Both aeroengine performance decay and component anomaly faults can cause changes in health parameters, the health parameters caused by performance decay are relatively slow to change, and generally treated as no fault, the occurrence of component anomaly faults is often abrupt, and the component anomaly faults can cause changes in health parameters related to faults and cause actual output to deviate from normal output. For related fault diagnosis, the following can be assumed:
Assume 1 that under normal conditions, the health parameters of an aeroengine are unchanged throughout the operating cycle;
Assume 2 that in each fault condition, only the health parameters associated with the fault will change, and the remaining health parameters remain unchanged, with a default value of 1.
Currently, in the fault diagnosis method based on subspace identification, for a linear system with unknown parameters, equation (1.3) can be written into the form of equation state estimation, specifically as follows:
......(2.1)
wherein, the Is the firstThe estimated gas circuit system state at the moment,As covariance matrixIs used for the zero-mean signal of (c),Is the kalman gain.
Based on equation (2.1), equation (1.4) can be rewritten as follows:
......(2.2)
Wherein the method comprises the steps of Is a block Toeplitz matrix, andHas the same structure as that of the conventional one,For stacked column vectors, like output sequences
The corresponding observer of equation (1.3) can be established as follows:
......(2.3)
wherein, the ,,
To eliminate what is generally unknown in practiceThe following form of equation can be obtained by equation (2.3):
......(2.4)
due to Is located within a unit circle for a large,0, Then, stateCan be further expressed in the form:
......(2.5)
in the case of no fault, substituting equation (2.5) into equation (2.2) yields:
......(2.6)
In equation (2.6), all the parameters of the gas path system are unknown, and the fault diagnosis method based on subspace identification is realized by identifying Developed by left-hand null space ofRepresenting and incorporating a data matrix of past input and output data. However, this method, when applied, has a system orderMust be known and a matrixMust be a constraint in terms of rank of line full, once the orderUnknown or matrixNot row full rank will not be usable.
Aiming at the limitation of the subspace identification-based diagnosis method, the design can be carried out by utilizing the strong fitting regression capability and learning capability of the neural network and combining with periodic aeroengine operation data.
In general, the learning process of the BP neural network includes two stages of signal forward propagation and error feed forward propagation, which include an input layer, a hidden layer, and an output layer, and as shown in fig. 1, the mathematical relationship between the three layers can be described as follows:
............(3.1)
wherein, the Respectively input of the input layer, value of the hidden layer and output of the output layer,The activation functions of the hidden layer and the output layer are respectively; The weights and the biases of the hidden layer and the output layer are respectively.
Super parameter of BP neural networkIs defined asThe building of the BP neural network overall model is specifically as follows:
............(3.2)
wherein, the Respectively the input and the output of the BP neural network,Is a structural function of the network.
By equation (2.6), the BP neural network of equation (1.2) can be expressed as follows:
...(3.3)
wherein, the
The output estimation sequence of the gas circuit system is selfLength from moment to moment isIs a column vector of a stack of (a) columns;
is an input sequence of the gas circuit system, is self-contained Length from moment to moment isIs a column vector of a stack of (a) columns;
is an input sequence of the gas circuit system, is self-contained Length from moment to moment isIs a column vector of a stack of (a) columns;
is a health parameter sequence of the gas circuit system, is self-contained Length from moment to moment isIs a column vector of a stack of (a) columns;
is a health parameter sequence of the gas circuit system, is self-contained Length from moment to moment isIs a column vector of a stack of (a) columns;
the measuring output sequence of the gas circuit system is self Length from moment to moment isIs a column vector of a stack of (a) columns.
Based on equation (3.3), the loss function of BP neural networkThe method comprises the following steps:
............(3.4)
Super parameter The following can be calculated:
............(3.5)
Residual signal of BP neural network The calculation is as follows:
............(3.6)
the BP neural network expression equation of the gas circuit system is constructed, and the relation between the BP neural network model and the state space model is established, so that the feasibility of the related fault diagnosis method can be proved in theory. In addition, the BP neural network can be trained and learned to reduce residual signals by using available data of the aeroengine health management period
While BP neural network can be used to learn the input-output relationship of the gas circuit system, it has some limitations such as slow convergence speed and long training time. On the other hand, the selection of the initial weight and bias of the BP neural network has great influence on the network performance, for example, the improper selection of the initial weight and bias of the network can influence the training time and convergence and even lead to local optimization, and aiming at the problem, a particle swarm algorithm PSO can be introduced to optimize the initial weight and bias, so that the performance and fitting precision of the BP neural network are improved.
Particle swarm Algorithm PSO is a global stochastic algorithm that searches for the best solution by tracking the best position of individual particlesAnd the optimal position of particle swarmTo obtain an optimization, dynamically updating the position of the particles in each iteration of the search process.
Assume thatThe individual particles are formed asParticle swarm, existence population, in a dimensional search spaceThen (1)The position of the individual particles in the search space isAt a speed ofFirst, theIndividual extremum of individual particles isThe global extremum of the population is
During each search iteration, the best position of a single particle is passedOptimum position of particle swarmThe speed and position of the particles are updated, the update being specifically as follows:
............(4.1)
Wherein, superscript +1 Represents the number of iterations and,In the form of an inertial weight,Is a non-negative acceleration factor and,Is a random number between 0, 1.
The particle swarm algorithm PSO does not directly optimize the objective function, but optimizes through the fitness function corresponding to the objective function. Therefore, to optimize the initial weights and bias of the BP neural network, the fitness function of the particle swarm algorithm PSOThe method can be constructed as follows:
;
wherein, the Is the first gas circuit systemThe number of measurement outputs is calculated,Is the BP neural networkThe estimates are output.
Based on the above, the fault diagnosis method of the air circuit system of the aeroengine with the data and the mechanism fused is designed, and the fault detection method is shown in fig. 2, and is specifically referred to as follows.
Step one, constructing a BP neural network.
Setting the number of neurons and the network layer number of the BP neural network, learning rate and activation function, and initializing the weight and bias of the BP neural network.
And step two, optimizing the weight and the bias of the BP neural network by using a particle swarm algorithm PSO based on the input and measurement output data of the gas circuit system.
S21, initializing particle swarm in a particle swarm algorithm PSO, including particle position and speed, and including particle number, for weight and bias of BP neural networkSearch space dimensionInertial weightNon-negative acceleration factorRandom numberEtc.
S22, calculating the fitness function value of the particle swarm algorithm PSO on the BP neural network based on the input and measurement output data of the gas circuit system.
S23, calculating the optimal position of single particle in particle swarm algorithm PSOAnd the optimal position of particle swarm
The fitness function value of each particle can be calculated and compared with the fitness of the historical position to obtain the optimal individual positionAnd comparing the fitness of all particle positions in the current iteration to obtain an optimal global position
S24, judging whether the maximum iteration times are reached or not, or judging whether the fitness function value is smaller than a set error threshold value, namely judging whether constraint conditions are met or not:
if so, the optimal position of the single particle is obtained And the optimal position of particle swarmObtaining an optimized value of the weight and the bias of the BP neural network;
if not, then based on the best position of the individual particles And the optimal position of particle swarmThe particle position and velocity are updated, specifically, referring to equation (4.1), and the process returns to S22.
Training the BP neural network based on input and measurement output data of the gas circuit system to obtain an BP neural network output estimation model, wherein the BP neural network output estimation model specifically represents a reference equation (3.3), and the reference equation is a super-parameterCan take optimal super parameters*。
The PSO algorithm is adopted to optimize the initial value of the weight and the bias of the BP neural network particle swarm algorithm PSO to the BP neural network, so that the BP neural network has the optimized initial value of the weight and the bias, the performance of the BP neural network can be improved, the BP neural network can have better capability to learn the input-output relation of the equation (3.3), and the BP neural network output estimation model can be obtained efficiently.
The BP neural network output estimation model is obtained by training by using normal gas circuit system input and output data, if the BP neural network training is accurate, once faults occur, deviation occurs between the output and the actual output of the BP neural network output estimation model, and therefore, the faults can be detected by designing residual errors.
And fourthly, calculating an output estimation sequence of the gas circuit system in real time by using the BP neural network output estimation model, calculating a residual sequence and a residual estimation function thereof, and detecting faults of the gas circuit system.
The residual sequence is calculated as follows:
............(5.1)
wherein, the Is a residual sequence.
Calculating a residual evaluation function, specifically as follows:
............(5.2)
wherein, the
Evaluating a function for the residual error;
Is that Is not limited to the above-described embodiments.
Assuming that the process noise and the measurement noise are gaussian-like, then one can useThe test is performed to perform a residual evaluation,, wherein,The degree of freedom of chi-square distribution is determined by the shape of the distribution, and usually a natural number is used.
Residual error evaluation functionAnd residual error thresholdComparing, judging whether the gas circuit system has faults, if soJudging that the gas circuit system has not failed, ifAnd judging that the gas circuit system fails.
For residual thresholdThe specific design and calculation are as follows:
;
wherein, the The specific numerical value of the acceptable significance level of the gas circuit system fault can be selected and determined by a person skilled in the art according to specific actual practice when the technical scheme disclosed by the application is applied.
In the method for diagnosing the fault of the air circuit system of the aeroengine by fusing the disclosed data and the mechanism, a fault detection method based on a PSO-BP neural network is designed in a data-driven framework, the BP neural network is constructed and trained by using normal data, in order to overcome the defects of low convergence speed, long training time and the like of the BP neural network, the initial weight and bias of the BP neural network are optimized by using a particle swarm algorithm PSO, an output estimation model of the BP neural network is obtained by efficient training, an output estimation sequence is calculated by using the output estimation model of the BP neural network, and a residual error sequence and a residual error estimation function thereof are calculated, so that the fault of the air circuit system is detected.
After detecting the fault of the gas circuit system, the fault type needs to be further identified, so that fault isolation is realized, the fault isolation method has strong classification capability and can be used for isolating faults, and based on the fault isolation method, the fault isolation method is designed, and specific reference is made to the following steps:
Step one, constructing a fault database of the gas circuit system, wherein each fault data comprises a process input and a measurement output.
Historical fault data of the gas circuit system can be collected as much as possible, and an overfill fault database is built so that the candidate fault data can be selected with a larger degree of freedom, and the number of fault types contained in the candidate fault data is far greater than the required number.
Step two, determining fault types corresponding to the fault data, and setting corresponding labels for the fault data.
And thirdly, selecting fault data as input, and training the fault isolation neural network by taking a corresponding label as output to obtain a fault isolation neural network model.
The fault isolation neural network can specifically adopt a BP neural network.
And fourthly, calculating fault data of the gas circuit system in real time by using a fault isolation neural network model, outputting a corresponding label, further judging and obtaining a corresponding fault type, and realizing fault isolation.
After fault detection, inputting data corresponding to the fault of the gas circuit system into a fault isolation neural network model for calculation, if the fault type is overlapped with one fault type corresponding to the fault of the gas circuit system in a fault database, the label obtained by calculating the fault isolation neural network model is obviously close to 1 with the label corresponding to the fault type, and is close to 0 with other labels, and a fault table can be constructed as follows:
In the method for diagnosing the fault of the air path system of the aeroengine, which is formed by integrating the data and the mechanism, two neural network operations are executed in the designed fault detection and isolation method, firstly, in the fault detection stage, one neural network is trained by utilizing normal input and output data of the air path system to obtain an output estimated neural network model, then the fault detection is carried out by utilizing residual errors, which can be regarded as utilizing the fitting regression capability of the neural network, and in the fault isolation stage, the other neural network is trained by utilizing the fault data and the labels of the air path system to obtain a fault isolated neural network model, and the fault identification is carried out, which can be regarded as utilizing the classification capability of the neural network, as shown in fig. 3.
In the method for diagnosing the fault of the air circuit system of the aeroengine, which is formed by integrating the data and the mechanism, the model which is formed by constructing the estimation output by using the BP neural network is designed, the fault detection is feasible and interpretable, the fault judgment is performed by calculating the residual sequence and the residual evaluation function thereof and comparing the residual threshold value, the fault detection can be realized only by inputting and outputting the data, the parameter of the air circuit system is not depended on, the method has higher engineering application value, the model which is formed by constructing the fault isolation model by using the neural network model on the basis of fault diagnosis, and the fault type is identified with high accuracy.
In addition, in the aeroengine gas circuit system fault diagnosis method with the data and mechanism integrated disclosed by the embodiment, aiming at the problems of low convergence speed, long training time, high calculation cost and the like of the BP neural network, the particle swarm algorithm PSO is introduced to optimize the network related parameters, the performance and fitting precision of the BP neural network are improved, and the construction of the BP neural network estimation output model can be rapidly completed.
Having thus described the technical aspects of the present application with reference to the preferred embodiments shown in the drawings, it should be understood by those skilled in the art that the scope of the present application is not limited to the specific embodiments, and those skilled in the art may make equivalent changes or substitutions to the related technical features without departing from the principle of the present application, and those changes or substitutions will fall within the scope of the present application.

Claims (8)

1.一种数据与机理融合的航空发动机气路系统故障诊断方法,其特征在于,包括故障检测方法,该故障检测方法包括:1. A method for diagnosing faults in an aero-engine gas path system by integrating data and mechanism, characterized in that it comprises a fault detection method, the fault detection method comprising: 步骤一、构建BP神经网络,初始化BP神经网络的权重、偏置;Step 1: Construct a BP neural network and initialize the weights and biases of the BP neural network; 步骤二、基于气路系统的输入、测量输出数据,以粒子群算法PSO对BP神经网络的权重、偏置进行优化;Step 2: Based on the input and measured output data of the gas path system, the weight and bias of the BP neural network are optimized using the particle swarm algorithm PSO; 步骤三、基于气路系统的输入、测量输出数据,对BP神经网络进行训练,得出BP神经网络输出估计模型;Step 3: Based on the input and measured output data of the gas path system, the BP neural network is trained to obtain a BP neural network output estimation model; 步骤四、以BP神经网络输出估计模型实时计算得出气路系统的输出估计序列,计算残差序列及其残差评估函数,对气路系统进行故障检测;Step 4: Use the BP neural network output estimation model to calculate the output estimation sequence of the gas path system in real time, calculate the residual sequence and its residual evaluation function, and perform fault detection on the gas path system; 故障检测方法的步骤二具体为:Step 2 of the fault detection method is specifically as follows: S21、对于BP神经网络的权重、偏置,初始化粒子群算法PSO中粒子群;S21, for the weight and bias of the BP neural network, initialize the particle swarm in the particle swarm algorithm PSO; S22、基于气路系统的输入、测量输出数据,计算粒子群算法PSO对于BP神经网络的适应度函数值;S22, based on the input and measured output data of the gas path system, calculating the fitness function value of the particle swarm algorithm PSO for the BP neural network; S23、计算粒子群算法PSO中单个粒子的最佳位置和粒子群的最佳位置S23. Calculate the optimal position of a single particle in the particle swarm algorithm PSO and the optimal position of the particle swarm ; S24、判断是否到达最大迭代次数,或者适应度函数值是否小于设定误差阈值:S24, determine whether the maximum number of iterations has been reached, or whether the fitness function value is less than the set error threshold: 若是,则得到单个粒子的最佳位置和粒子群的最佳位置,得到BP神经网络的权重、偏置的优化值;If so, the optimal position of a single particle is obtained and the optimal position of the particle swarm , get the optimized values of weight and bias of BP neural network; 若否,则基于单个粒子的最佳位置和粒子群的最佳位置,更新粒子位置、速度,返回S22。If not, then based on the best position of a single particle and the optimal position of the particle swarm , update the particle position and velocity, and return to S22. 2.根据权利要求1所述的数据与机理融合的航空发动机气路系统故障诊断方法,其特征在于,2. The method for diagnosing faults in an aircraft engine gas path system by integrating data and mechanism according to claim 1, characterized in that: 故障检测方法的步骤四中,以BP神经网络输出估计模型实时计算得出气路系统的输出估计序列,具体为:In step 4 of the fault detection method, the output estimation sequence of the gas path system is calculated in real time using the BP neural network output estimation model, specifically: ; 其中,in, 为气路系统的输出估计序列,为自时刻起长度为的堆叠的列向量; is the output estimation sequence of the gas system, The length from time A stacked column vector of ; 为气路系统的输入序列,为自时刻起长度为的堆叠的列向量; is the input sequence of the gas system, The length from time A stacked column vector of ; 为气路系统的输入序列,为自时刻起长度为的堆叠的列向量; is the input sequence of the gas system, The length from time A stacked column vector of ; 为气路系统的健康参数序列,为自时刻起长度为的堆叠的列向量; is the health parameter sequence of the gas system, The length from time A stacked column vector of ; 为气路系统的健康参数序列,为自时刻起长度为的堆叠的列向量; is the health parameter sequence of the gas system, The length from time A stacked column vector of ; 为气路系统的测量输出序列,为自时刻起长度为的堆叠的列向量; is the measurement output sequence of the gas system, The length from time A stacked column vector of ; 为BP神经网络的超参数; is the hyperparameter of BP neural network; 为BP神经网络的结构函数。 is the structural function of the BP neural network. 3.根据权利要求2所述的数据与机理融合的航空发动机气路系统故障诊断方法,其特征在于,3. The method for diagnosing faults in an aircraft engine gas path system by integrating data and mechanism according to claim 2, characterized in that: 故障检测方法的步骤四中,计算残差序列,具体为:In step 4 of the fault detection method, the residual sequence is calculated, specifically: ; 其中,in, 为残差序列; is the residual sequence; 为气路系统的测量输出序列,是自时刻起长度为的堆叠的列向量。 It is the measurement output sequence of the gas system. The length from time A stacked column vector of . 4.根据权利要求3所述的数据与机理融合的航空发动机气路系统故障诊断方法,其特征在于,4. The method for diagnosing faults in an aircraft engine gas path system by integrating data and mechanism according to claim 3 is characterized in that: 故障检测方法的步骤四中,计算残差评估函数,具体为:In step 4 of the fault detection method, the residual evaluation function is calculated, specifically: ; 其中,in, 为残差评估函数; is the residual evaluation function; 的期望。 for expectations. 5.根据权利要求4所述的数据与机理融合的航空发动机气路系统故障诊断方法,其特征在于,5. The method for diagnosing faults in an aircraft engine gas path system by integrating data and mechanism according to claim 4, characterized in that: 故障检测方法的步骤四中,对气路系统进行故障检测,具体为:In step 4 of the fault detection method, the gas circuit system is subjected to fault detection, specifically: 将残差评估函数与残差阈值进行对比,判断气路系统是否发生故障,若,则判断气路系统未发生故障,若,则判断气路系统发生故障。The residual evaluation function With residual threshold Compare and judge whether the gas system fails. , it is judged that there is no fault in the gas system. , it is judged that the gas system is faulty. 6.根据权利要求5所述的数据与机理融合的航空发动机气路系统故障诊断方法,其特征在于,6. The method for diagnosing faults in an aircraft engine gas path system by integrating data and mechanism according to claim 5, characterized in that: 故障检测方法的步骤四中,残差阈值具体计算为:In step 4 of the fault detection method, the residual threshold The specific calculation is: ; 其中,in, 为可气路系统故障可接受的显著性水平; The acceptable significance level of gas path system failure; 为卡方分布的自由度。 is the degrees of freedom of the chi-square distribution. 7.根据权利要求6所述的数据与机理融合的航空发动机气路系统故障诊断方法,其特征在于,还包括故障隔离方法,该故障隔离方法包括:7. The method for diagnosing faults in an aircraft engine gas path system by integrating data and mechanism according to claim 6, characterized in that it also includes a fault isolation method, the fault isolation method comprising: 步骤一、构建气路系统的故障数据库,其中每个故障数据包括一个过程输入、一个测量输出;Step 1: construct a fault database of the gas circuit system, where each fault data includes a process input and a measurement output; 步骤二、确定各个故障数据对应的故障类型,为各个故障数据设置相应的标签;Step 2: determine the fault type corresponding to each fault data, and set a corresponding label for each fault data; 步骤三、选取故障数据作为输入,以相应的标签作为输出,对故障隔离神经网络进行训练,得到故障隔离神经网络模型;Step 3: Select fault data as input, take corresponding labels as output, train the fault isolation neural network, and obtain a fault isolation neural network model; 步骤四、以故障隔离神经网络模型对气路系统的故障数据进行实时计算,输出相应的标签,进而判断得出相应的故障类型。Step 4: Use the fault isolation neural network model to perform real-time calculations on the fault data of the gas circuit system, output corresponding labels, and then determine the corresponding fault type. 8.根据权利要求7所述的数据与机理融合的航空发动机气路系统故障诊断方法,其特征在于,8. The method for diagnosing faults in an aircraft engine gas path system by integrating data and mechanism according to claim 7, characterized in that: 故障隔离方法的步骤三中,故障隔离神经网络采用BP神经网络。In step three of the fault isolation method, the fault isolation neural network adopts a BP neural network.
CN202510632525.5A 2025-05-16 2025-05-16 A method for fault diagnosis of aeroengine gas path system based on data and mechanism fusion Pending CN120197005A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202510632525.5A CN120197005A (en) 2025-05-16 2025-05-16 A method for fault diagnosis of aeroengine gas path system based on data and mechanism fusion

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202510632525.5A CN120197005A (en) 2025-05-16 2025-05-16 A method for fault diagnosis of aeroengine gas path system based on data and mechanism fusion

Publications (1)

Publication Number Publication Date
CN120197005A true CN120197005A (en) 2025-06-24

Family

ID=96065801

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202510632525.5A Pending CN120197005A (en) 2025-05-16 2025-05-16 A method for fault diagnosis of aeroengine gas path system based on data and mechanism fusion

Country Status (1)

Country Link
CN (1) CN120197005A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN120974282A (en) * 2025-10-20 2025-11-18 山东大学 A Fault Diagnosis Method and System for Wind Turbines Based on Mechanism Data Fusion

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108462191A (en) * 2018-04-08 2018-08-28 福州大学 One kind being based on O3The low-frequency oscillation of electric power system discrimination method of KID algorithms
CN109710983A (en) * 2018-12-04 2019-05-03 北京大学 A Cylinder Stratification Fault Diagnosis Method Based on Key Performance Indicators in Diesel Engines
CN110017857A (en) * 2019-04-04 2019-07-16 广东工业大学 Nonlinear transducer method for diagnosing faults based on adaptive learning and neural network
CN112749789A (en) * 2020-12-30 2021-05-04 西北工业大学 Aero-engine multiple fault diagnosis device based on self-association neural network
CN116956719A (en) * 2023-07-18 2023-10-27 中国航发沈阳发动机研究所 Aeroengine fault mode modeling and detecting method based on IWOA (IWOA optimization) BP (back propagation)

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108462191A (en) * 2018-04-08 2018-08-28 福州大学 One kind being based on O3The low-frequency oscillation of electric power system discrimination method of KID algorithms
CN109710983A (en) * 2018-12-04 2019-05-03 北京大学 A Cylinder Stratification Fault Diagnosis Method Based on Key Performance Indicators in Diesel Engines
CN110017857A (en) * 2019-04-04 2019-07-16 广东工业大学 Nonlinear transducer method for diagnosing faults based on adaptive learning and neural network
CN112749789A (en) * 2020-12-30 2021-05-04 西北工业大学 Aero-engine multiple fault diagnosis device based on self-association neural network
CN116956719A (en) * 2023-07-18 2023-10-27 中国航发沈阳发动机研究所 Aeroengine fault mode modeling and detecting method based on IWOA (IWOA optimization) BP (back propagation)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN120974282A (en) * 2025-10-20 2025-11-18 山东大学 A Fault Diagnosis Method and System for Wind Turbines Based on Mechanism Data Fusion

Similar Documents

Publication Publication Date Title
Hanachi et al. Performance-based gas turbine health monitoring, diagnostics, and prognostics: A survey
Chen et al. A risk-averse remaining useful life estimation for predictive maintenance
CN109857094B (en) Aero-engine fault diagnosis method based on two-stage Kalman filter algorithm
US6898554B2 (en) Fault detection in a physical system
CN114943281B (en) Intelligent decision-making method and system for heat pipe cooling reactor
Ntalampiras Fault identification in distributed sensor networks based on universal probabilistic modeling
CN115081484B (en) Aero-engine sensor fault diagnosis method based on CRJ-OSELM algorithm
CN120197005A (en) A method for fault diagnosis of aeroengine gas path system based on data and mechanism fusion
CN117470274A (en) An online fault detection and repair method and system for a multi-navigation sensor system
CN114818116B (en) Aircraft engine failure mode identification and life prediction method based on joint learning
Zhong et al. An improved correlation-based anomaly detection approach for condition monitoring data of industrial equipment
US20240241806A1 (en) Estimation of a diagnosis of a physical system by predicting fault indicators by machine learning
Zhang et al. Research on sensor fault tolerance technology in nuclear power plant control system
CN111076744B (en) Satellite sensor fault detection positioning method based on self-coding observer
CN112801267A (en) Multiple fault diagnosis device for aircraft engine with dynamic threshold value
Shah et al. Multi-Auto Associative Neural Network based sensor validation and estimation for aero-engine
Rabenoro et al. A methodology for the diagnostic of aircraft engine based on indicators aggregation
CN114112374B (en) Fault diagnosis method of weapon equipment system based on orthogonal sub-block projection
José et al. Improvements in failure detection of DAMADICS control valve using neural networks
Li et al. Predicting remaining useful life of industrial equipment based on multivariable monitoring data analysis
Loboda et al. Neural networks for gas turbine fault identification: multilayer perceptron or radial basis network?
Nozari et al. Model-based fault detection and isolation using neural networks: An industrial gas turbine case study
Tao et al. Incipient fault online estimation based on Kullback-Leibler divergence and fast moving window PCA
Arogeti et al. Mode tracking and FDI of hybrid systems
Andreev et al. Simulation-based system reliability analysis of electrohydraulic actuator with dual modular redundancy

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