CN117421556A - Trusted aviation equipment fault diagnosis method for solving ambiguity and uncertainty - Google Patents
Trusted aviation equipment fault diagnosis method for solving ambiguity and uncertainty Download PDFInfo
- Publication number
- CN117421556A CN117421556A CN202311395944.9A CN202311395944A CN117421556A CN 117421556 A CN117421556 A CN 117421556A CN 202311395944 A CN202311395944 A CN 202311395944A CN 117421556 A CN117421556 A CN 117421556A
- Authority
- CN
- China
- Prior art keywords
- fault diagnosis
- model
- representing
- rule
- fault
- 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
Links
- 238000003745 diagnosis Methods 0.000 title claims abstract description 74
- 238000000034 method Methods 0.000 title claims abstract description 47
- 238000006243 chemical reaction Methods 0.000 claims abstract description 22
- 238000005457 optimization Methods 0.000 claims abstract description 19
- 238000001514 detection method Methods 0.000 claims description 16
- 230000008569 process Effects 0.000 claims description 16
- 238000011156 evaluation Methods 0.000 claims description 12
- 239000013598 vector Substances 0.000 claims description 12
- 239000011159 matrix material Substances 0.000 claims description 11
- 230000004913 activation Effects 0.000 claims description 10
- 238000005315 distribution function Methods 0.000 claims description 4
- 238000005070 sampling Methods 0.000 claims description 4
- 238000012549 training Methods 0.000 claims description 4
- 230000007704 transition Effects 0.000 claims description 4
- 238000010606 normalization Methods 0.000 claims description 3
- 238000004880 explosion Methods 0.000 abstract description 3
- 238000010586 diagram Methods 0.000 description 5
- 230000006870 function Effects 0.000 description 4
- 238000010276 construction Methods 0.000 description 3
- 230000003044 adaptive effect Effects 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 239000000446 fuel Substances 0.000 description 2
- 230000004927 fusion Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 238000012546 transfer Methods 0.000 description 2
- 230000009466 transformation Effects 0.000 description 2
- 238000012795 verification Methods 0.000 description 2
- 230000006399 behavior Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000002349 favourable effect Effects 0.000 description 1
- 238000012804 iterative process Methods 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 230000008439 repair process Effects 0.000 description 1
- 230000009897 systematic effect Effects 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/12—Computing arrangements based on biological models using genetic models
- G06N3/126—Evolutionary algorithms, e.g. genetic algorithms or genetic programming
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N5/00—Computing arrangements using knowledge-based models
- G06N5/01—Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N5/00—Computing arrangements using knowledge-based models
- G06N5/02—Knowledge representation; Symbolic representation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N5/00—Computing arrangements using knowledge-based models
- G06N5/04—Inference or reasoning models
- G06N5/045—Explanation of inference; Explainable artificial intelligence [XAI]; Interpretable artificial intelligence
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N7/00—Computing arrangements based on specific mathematical models
- G06N7/01—Probabilistic graphical models, e.g. probabilistic networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N7/00—Computing arrangements based on specific mathematical models
- G06N7/02—Computing arrangements based on specific mathematical models using fuzzy logic
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Software Systems (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Mathematical Physics (AREA)
- Computing Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Pure & Applied Mathematics (AREA)
- Biomedical Technology (AREA)
- Algebra (AREA)
- Computational Mathematics (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Fuzzy Systems (AREA)
- Automation & Control Theory (AREA)
- Probability & Statistics with Applications (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physiology (AREA)
- Genetics & Genomics (AREA)
- Medical Informatics (AREA)
- Test And Diagnosis Of Digital Computers (AREA)
Abstract
The invention discloses a reliable aviation equipment fault diagnosis method for solving ambiguity and uncertainty, which comprises the following steps: step one, constructing an aviation equipment fault relation: analyzing various indexes of the aviation equipment and constructing a reasonable fault relation; step two, establishing an equivalence relation: establishing an equivalence relation between fuzzy fault tree analysis and an interval confidence rule base; firstly, relevant attributes of aviation equipment are researched, and logic relations among fault events are constructed; secondly, an IBRB model constructed after FFTA assistance avoids the explosion problem of a certain combination rule; thirdly, the IP-CMA-ES optimization algorithm supports the interpretability of the model, so that the fault diagnosis result of the aviation equipment is credible. The Bayesian network is used as a conversion bridge between the FFTA and the IBRB, fault diagnosis under the condition of ambiguity and uncertainty is realized, and meanwhile, the diagnosis result is ensured to be credible by using an optimization algorithm with an interpretability constraint.
Description
Technical Field
The invention relates to the technical field of aviation, in particular to a trusted aviation equipment fault diagnosis method for solving ambiguity and uncertainty.
Background
The method has practical significance for fault diagnosis of the aviation equipment, can ensure safe and reliable operation of the aircraft, and can timely discover and repair the problems possibly influencing the operation of the aircraft by detecting, analyzing and diagnosing faults.
At present, the fault diagnosis of aviation equipment is researched and can be divided into three types, namely a black box model, a white box model and a gray box model, the black box model judges whether faults exist by observing the output of the equipment under different input conditions, long et al put forward a multi-layer neural network model aiming at an aircraft fuel system, a proper reasoning strategy is designed by establishing fault trees (Fault Tree Analysis, FTA), the guiding significance of the method in the aircraft fuel fault diagnosis is verified, the white box model is a method for carrying out the fault diagnosis based on the internal structure and the working principle of the equipment, the internal structure and the parameters of the method can be completely understood and interpreted, the Yang et al apply the FTA to the reliability analysis of an arrow-mounted recorder, and have important significance for mastering the experience of the fault diagnosis of an arrow-mounted micro-electromechanical system, however, the white box model has the problems of higher data requirements, the requirement of professional knowledge in the field and the like, and the gray box model is a fault diagnosis method between the black box model and the white box model, and the external observation data of the equipment is utilized, and partial internal information is also used for carrying out the fault diagnosis.
However, the conventional fault diagnosis method has the following disadvantages:
as a typical gray box model, a confidence Rule Base (BRB) can construct a Rule Base through expert knowledge, and ER is taken as an inference engine to combine multiple independent evidences to effectively diagnose faults, and a specific BRB method can be used for fault diagnosis of aviation equipment, but firstly, the influence of uncertainty of expert knowledge and ambiguity of input data on modeling is not considered, which results in that a complete Rule Base cannot be constructed, so that the intuitiveness of the obtained model is not strong, which makes it difficult to evaluate the rationality of the BRB model in an initial stage, in other words, whether the models can accurately capture the characteristics of the faults of the aviation equipment or not can not be determined, and secondly, the methods do not consider the problem of combined Rule explosion possibly occurring under the complex condition of excessive attributes. This means that when the number of combinations of attributes is large, the rule base may become very large and difficult to manage, which results in a resulting model that is not intuitive, making it difficult to evaluate the rationality of the BRB model in an initial stage, and finally, at the time of model optimization, the interpretability may be destroyed, which means that the result may not fit the reality.
Disclosure of Invention
The purpose of the invention is that: taking the influence of ambiguity and uncertainty on model construction into consideration, modeling the aviation equipment by using FFTA to assist IBRB, and providing interpretability support for the model by adopting an IP-CMA-ES algorithm. And the comprehensive evaluation is provided for the running condition of the aviation equipment, and the reliability of the aviation equipment is improved.
In order to achieve the above purpose, the present invention provides the following technical solutions: a trusted aviation equipment fault diagnosis method for solving ambiguity and uncertainty comprises the following steps:
step one, constructing an aviation equipment fault relation: analyzing various indexes of the aviation equipment and constructing a reasonable fault relation;
step two, establishing an equivalence relation: establishing an equivalence relation between fuzzy fault tree analysis and an interval confidence rule base;
step three, a fault diagnosis model is established: establishing an aviation equipment fault diagnosis model based on FFTA and IBRB;
step four, model optimization: optimizing the model based on a projection covariance matrix adaptive evolution strategy algorithm with interpretable constraints;
step five, fault diagnosis: FFTA and IBRB based avionics fault diagnosis.
As a preferable technical scheme of the invention, the relation establishment in the second step is to construct a fuzzy fault tree specifically for the acquired fault relation, and the Bayesian network is used as a conversion space to assist expert knowledge in establishing an equivalent relation with the IBRB.
As a preferable technical scheme of the invention, the fault diagnosis model establishment in the third step is concretely to successfully establish a reasonable conversion rule, and the preliminary modeling is carried out on the fault diagnosis model of the aviation equipment.
As a preferred technical scheme of the invention, the model optimization in the fourth step is specifically to optimize parameters of an initial model, and to simultaneously support the interpretability and high accuracy of the model in the optimization process, three model parameters of confidence level, rule weight and rule reliability are optimized by adopting an IP-CMA-ES algorithm.
As a preferable technical scheme of the invention, in the step five, the fault diagnosis is specifically carried out by processing the ER opposite confidence level, the evidence weight and the evidence reliability, so as to obtain the fault diagnosis result of the aviation equipment, and the diagnosis result is displayed through final confidence distribution.
As a preferred embodiment of the present invention, the index in the step one includes, but is not limited to, voltage, current, and friction torque.
As a preferred technical solution of the present invention, in the second step, attention should be paid to the equivalent transformation proof of the IBRB and the bayesian network, and when the events are independent of each other, the confidence level of the IBRB and the conditional probability of the bayesian network are equivalent, and their relationships are expressed as follows: h ab expressed in evidence as the a-th hypothesis ζ a B-th detection result e in the established case b Probability of expected occurrence, i.e.)>γ ab Representing the detection result e b Pointing hypothesis ζ a Confidence of (1), and->Psi represents a set of propositions for fault diagnosis, h ψ,b Representing the b-th detection result e under the proposition psi b Is a generalized likelihood of (1), and>γ ψ,b representing the detection result e b Confidence pointing to proposition ψ, and +.>
As a preferred embodiment of the present invention, in the second step, when the FFTA performs conversion under different logic gates and the bayesian network, the conversion method needs to be distinguished, and first, the following logic gate is determined as follows, p (T) =logic g (p (B) 1 ),...,p(B n ) Where p (·) represents the conditional probability corresponding to FFTA, p (T) represents the probability of occurrence of a top event in FFTA, logicG (·) represents the logic gate in FFTA, p (B) 1 ),...,p(B n ) Representing the failure probability of the ith elementary event in the FFTA, the conditional probability of a node in the bayesian network is expressed in terms of the transition between the and gate and the or gate, respectively, u in the above i A reference value indicating the i-th basic event in FFTA, and G indicating the failure diagnosis result.
As the inventionIn the second step, the reliability of the activation rule of the IBRB under different logic gates should be differentiated, and the reliability of the activation rule under the and gate or gate is as follows: where α is the activation rule reliability, related to the confidence factor χ and the index reliability δ, and M is the number of indexes.
In the third step, as a preferred technical scheme of the present invention, the construction process of the confidence rule is as follows: firstly, analyzing a fault state of aviation equipment, and acquiring an index system of fault diagnosis through FFTA; secondly, constructing a corresponding IRBR model, and performing ER fusion by taking the rule weight, the rule reliability and the confidence coefficient as the evidence weight, the evidence reliability and the confidence coefficient; finally, the final confidence distribution of fault diagnosis is obtained.
In the third step, as a preferred technical solution of the present invention, the reasoning process of the IBRB is: the Θ is marked as an identification frame and is formed by N evaluation grades { G ] 1 ,G 2 ,...,G N Composition, beta Θ,i Indicating global unknowing, normalizing the data, and then obtaining the data e i The confidence profile is described as follows,evidence weight reuse i ∈[0,1]Expressed by evidence reliability i ∈[0,1]Representation, will ew i And ed i The new confidence profile is generated after the mixing weights as follows:
wherein m is n,i Representing the quality of the basic probability that it is high,indicating the remaining basic probability mass without any evaluation level assigned,/->Representing a globally unknown basic probability mass, representing a single evidence imperfection, f dw,i Represents the normalization coefficient, satisfiesJoint support beta for L independent evidence n,e(L) Is represented as follows: in the above formula, beta is represented n,e(k) (k=1, 2,., L) after fusing the first n pieces of evidence, rating G was evaluated n And satisfies m n,e(1) =m n,1 ,m β(Θ),e(1) ,m β(Θ),1 From the above information, the belief distribution and expected utility value of the following output, e (L) = { (G) n ,β n,e(L) ),n=1,...,N,(Θ,β Θ,e(L) },/>Where y represents the desired utility value, i.e. the predicted output, v (G n ) Expressed in the evaluation level G n Utility values below.
As a preferable technical scheme of the invention, in the fourth step, the interpretability and the high precision of the IBRB are ensured in the optimization process, and the MSE (-) is assumed to be the deviation degree of the utility value and the actual value output by the IBRB model, O T Representing the number of training samples, y representing the desired output value, y # Representing the result of the actual output of the model,describing a parameter set to be optimized, wherein the parameter set comprises confidence coefficient, rule reliability and rule weight, and the parameter set is as follows: omega 0 =Ω 0 {β 1,1 ,...,β 1,1 ,ζ 1 ,...,ζ L ,λ 1 ,...,λ L Obtaining parameters of each generation by sampling, < -> The ith solution representing the d+1th generation, phi represents the evolution step of the d generation, Ω d Mean value of d-th generation search distribution, C d The covariance matrix of the d generation is represented, N (-) represents the normal distribution function of the parameter, g represents the number of offspring, and then the explanatory constraint condition is added, namely, the explanatory property of the IBRB model is not destroyed, the rationality of model output is guaranteed, the constraint condition added to belief distribution is as follows>In the above formula, I k Representing an interpretable constraint on belief distribution under a first k rule, beta up 、ζ up 、λ up Respectively represent confidence, rule reliability, upper bound of rule weight, beta low 、ζ low 、θ low The lower limit values of the confidence degree, the rule reliability and the rule weight are respectively expressed, no specific standard exists for constraint conditions of belief distribution, the constraint conditions are usually determined by experts according to experience and reality conditions, in order to meet equality constraint, the equality constraint is converted into constraint in a hyperplane through projection operation,
wherein sigma e =1..n represents the number of variables in the equality constraint, η=1..n+1 represents +.>The number of the equation constraints in (V) represents the parameter vector, then the optimal subgroup is filtered out in the population, the average value of the next generation is updated,/->W in the above i Refers to the firstWeight coefficient of i solutions, τ represents the size of offspring population, +.>Representing the solution in the g solution, finally updating the covariance matrix and the evolution steps and the step length thereof according to the strategy, and recording the evolution steps under the d+1th generation as +.>a 1 And a 2 Record as learning rate, < >>Representing the ith solution vector in the g solution vectors at the d-th generation, the number of offspring is recorded as theta d The step size in the d generation is marked as phi d The expression pattern is as follows:
compared with the prior art, the invention has the beneficial effects that:
firstly, researching related attributes of aviation equipment and constructing a logic relationship between fault events; secondly, an IBRB model constructed after FFTA assistance avoids the explosion problem of a certain combination rule; the invention uses Bayesian network as the conversion bridge between FFTA and IBRB to realize the fault diagnosis under the condition of ambiguity and uncertainty, and simultaneously, uses the optimization algorithm with the interpretation constraint to ensure the credibility of the diagnosis result.
Drawings
FIG. 1 is a flow chart of the present invention;
FIG. 2 is a schematic diagram of the transitions of FFTA and Bayesian networks in accordance with the present invention;
FIG. 3 is a schematic representation of the conversion of IBRB and Bayesian networks of the present invention;
FIG. 4 is a schematic diagram of the conversion of FFTA and IBRB of the present invention;
FIG. 5 is a diagram of a model reasoning process of the present invention;
FIG. 6 is a graph of the optimization process of the IP-CMA-ES algorithm of the model of the present invention;
FIG. 7 is a schematic diagram of a friction torque fault sub-tree and IBRB of the present invention;
FIG. 8 is a plot of a first set of experimental expected outputs versus actual outputs of the present invention;
FIG. 9 is a graph of events of a friction torque subtree of the present invention;
FIG. 10 is a corresponding reference value diagram of the flywheel system fault condition of the present invention;
FIG. 11 is a graph of optimized model parameters of the present invention;
FIG. 12 is a graph of accuracy of the model verification results of the present invention.
Detailed Description
The technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the embodiments of the present invention, and it is apparent that the described embodiments are only some embodiments of the present invention, not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Referring to fig. 1-12, the present invention provides a method for diagnosing a fault of a trusted aviation device for resolving ambiguity and uncertainty, comprising the steps of:
step one, constructing an aviation equipment fault relation: analyzing various indexes of the aviation equipment and constructing a reasonable fault relation;
step two, establishing an equivalence relation: establishing an equivalence relation between fuzzy fault tree analysis and an interval confidence rule base;
step three, a fault diagnosis model is established: establishing an aviation equipment fault diagnosis model based on FFTA and IBRB;
step four, model optimization: optimizing the model based on a projection covariance matrix adaptive evolution strategy algorithm with interpretable constraints;
step five, fault diagnosis: FFTA and IBRB based avionics fault diagnosis.
And step two, establishing a relation, namely establishing a fuzzy fault tree by using the acquired fault relation, and establishing an equivalent relation with the IBRB by using a Bayesian network as a conversion space and assisting expert knowledge.
And thirdly, establishing a fault diagnosis model, namely successfully constructing a reasonable conversion rule, and carrying out preliminary modeling on the fault diagnosis model of the aviation equipment.
And step four, model optimization is specifically to optimize parameters of an initial model, and to simultaneously support the interpretability and high accuracy of the model in the optimization process, three model parameters of confidence level, rule weight and rule reliability are optimized by adopting an IP-CMA-ES algorithm.
And step five, performing fault diagnosis, namely processing the confidence level, the evidence weight and the evidence reliability through ER to obtain a fault diagnosis result of the aviation equipment, and displaying the diagnosis result through final confidence distribution.
The first step index includes, but is not limited to, voltage, current, and friction torque.
In the second step, note that the equivalent transformation between the IBRB and the bayesian network proves that when the events are independent of each other, the confidence level of the IBRB and the conditional probability of the bayesian network are equivalent, and their relationships are expressed as follows:h ab expressed in evidence as the a-th hypothesis ζ a B-th detection result e in the established case b Probability of expected occurrence, i.e.)>γ ab Representing the detection result e b Pointing hypothesis ζ a Confidence of (1), and->Psi represents a set of propositions for fault diagnosis, h ψ,b Representing the b-th detection result e under the proposition psi b Is a generalized likelihood of (1), and>γ ψ,b representing the detection result e b Confidence pointing to proposition ψ, and +.>
In the second step, when the FFTA performs conversion under different logic gates and the bayesian network, the conversion method needs to be distinguished, firstly, the following logic gate is determined to be described as follows, p (T) =logic g (p (B) 1 ),...,p(B n ) Where p (·) represents the conditional probability corresponding to FFTA, p (T) represents the probability of occurrence of a top event in FFTA, logicG (·) represents the logic gate in FFTA, p (B) 1 ),...,p(B n ) Representing the failure probability of the ith elementary event in the FFTA, the conditional probability of the node in the bayesian network is expressed as the transition between the and gate and the or gate, respectively,u in the above i A reference value indicating the i-th basic event in FFTA, and G indicating the failure diagnosis result.
In the second step, the reliability of the activation rule of the IBRB under different logic gates should be differentiated, and the reliability of the activation rule under the and gate and the or gate are as follows: where α is the activation rule reliability, related to the confidence factor χ and the index reliability δ, and M is the number of indexes.
In the third step, the construction process of the confidence rule is as follows: firstly, analyzing a fault state of aviation equipment, and acquiring an index system of fault diagnosis through FFTA; secondly, constructing a corresponding IRBR model, and performing ER fusion by taking the rule weight, the rule reliability and the confidence coefficient as the evidence weight, the evidence reliability and the confidence coefficient; finally, the final confidence distribution of fault diagnosis is obtained.
In the third step, the reasoning process of the IBRB is as follows: the Θ is marked as an identification frame and is formed by N evaluation grades { G ] 1 ,G 2 ,...,G N Composition, beta Θ,i Indicating global unknowing, normalizing the data, and then obtaining the data e i The confidence profile is described as follows,evidence weight reuse i ∈[0,1]Expressed by evidence reliability i ∈[0,1]Representation, will ew i And ed i The new confidence profile is generated after the mixing weights as follows: f dw,i =1/(1+ew 1 -ed i );m n,i =ew i β n,i the method comprises the steps of carrying out a first treatment on the surface of the Wherein m is n,i Representing basic probability mass, ++>Indicating the remaining basic probability mass without any evaluation level assigned,/->Representing a globally unknown basic probability mass, representing a single evidence imperfection, f dw,i Represents the normalization coefficient, satisfiesJoint support beta for L independent evidence n,e(L) Is represented as follows: in the above formula, beta is represented n,e(k) (k=1, 2,., L) after fusing the first n pieces of evidence, rating G was evaluated n And satisfies m n,e(1) =m n,1 ,m β(Θ),e(1) ,m β(Θ),1 From the above information, the belief distribution and expected utility value of the following output, e (L) = { (G) n ,β n,e(L) ),n=1,...,N,(Θ,β Θ,e(L) },/>Where y represents the desired utility value, i.e. the predicted output, v (G n ) Expressed in the evaluation level G n Utility values below.
In the fourth step, the interpretability and high precision of the IBRB are ensured in the optimization process, and the MSE (-) is assumed to be the deviation degree between the utility value and the actual value output by the IBRB model, O T Representing the number of training samples, y representing the desired output value, y # Representing the result of the actual output of the model,
describing a parameter set to be optimized, wherein the parameter set comprises confidence coefficient, rule reliability and rule weight, and the parameter set is as follows: omega 0 =Ω 0 {β 1,1 ,...,β 1,1 ,ζ 1 ,...,ζ L ,λ 1 ,...,λ L Obtaining parameters of each generation by sampling, < -> The ith solution representing the d+1th generation, phi represents the evolution step of the d generation, Ω d Mean value of d-th generation search distribution, C d The covariance matrix of the d generation is represented, N (-) represents the normal distribution function of the parameter, g represents the number of offspring, and then the explanatory constraint condition is added, namely, the explanatory property of the IBRB model is not destroyed, the rationality of model output is guaranteed, the constraint condition added to belief distribution is as follows>In the above formula, I k Representing an interpretable constraint on belief distribution under a first k rule, beta up 、ζ up 、λ up Respectively represent confidence, rule reliability, upper bound of rule weight, beta low 、ζ low 、θ low The lower limit values of the confidence degree, the rule reliability and the rule weight are respectively expressed, no specific standard exists for constraint conditions of belief distribution, the constraint conditions are usually determined by experts according to experience and reality conditions, in order to meet equality constraint, the equality constraint is converted into constraint in a hyperplane through projection operation,
wherein sigma e =1..n represents the number of variables in the equality constraint, η=1..n+1 represents +.>The number of the equation constraints in (V) represents the parameter vector, then the optimal subgroup is filtered out in the population, the average value of the next generation is updated,/->W in the above i Refers to the weight coefficient of the ith solution, τ represents the size of the offspring population, +.>Representing the solution in the g solution, finally updating the covariance matrix and the evolution steps and the step length thereof according to the strategy, and recording the evolution steps under the d+1th generation as +.>a 1 And a 2 Record as learning rate, < >>Representing the ith solution vector in the g solution vectors at the d-th generation, the number of offspring is recorded as theta d The step size in the d generation is marked as phi d The expression mode is as follows:
in the invention, for the fault diagnosis of the aviation equipment, a complete logic relation needs to be constructed to judge the root cause of the faultThe reason is that when constructing the logic relationship between events, the actual situation should be met, and the principles of systematics, scientificity and the like are followed, therefore, a fuzzy fault tree is constructed in a bottom-up mode, the basic events are gradually constructed, the connection between fault combinations of aviation equipment is described through graphic symbols of the fault tree, the reason causing the fault of the top-level event is gradually analyzed, the connection between the event and the fault is clarified, when the initial data can be relied on, the occurrence probability of the top-level event is calculated through the failure probability of the bottom event, a reasonable fuzzy fault tree is successfully constructed, the reason causing the fault to occur is quickly traced, the expert knowledge is assisted to construct a complete rule base, the method is favorable for solving uncertainty of expert knowledge and uncertainty of input information, FFTAs are embedded into IBRBs through a Bayesian network, conversion modes of the FFTAs under different logic gates are different, conversion of the logic gates is analyzed in detail to construct reasonable equivalence relations, bottom events, logic gates and top events in the FFTAs correspond to father nodes, conditional probability distribution and child nodes of the Bayesian network respectively, the bottom events, the logic gates and the top events correspond to input, confidence rules and output of the IBRBs, when the different logic gates in the FFTAs are converted with the Bayesian network, corresponding conversion methods are different, p (·) is assumed to represent conditional probabilities in the FFTAs, logicG (·) represents the logic gates in the FFTAs, and p (B) 1 ),...,p(B n ) Representing the basic event in FFTA, the description of the logic gate is as follows, p (T) =logicg (p (B) 1 ),...,p(B n )),u i The reference value representing the ith basic event in the FFTA, G represents the fault diagnosis result, and the conditional probability of the node in the bayesian network and the conversion formula between the and gate and the or gate are expressed as follows: the Bayesian rule is performed in a symmetric process of ER paradigm, wherein each evidence is described in the same confidence distribution format, and when events are independent of each other, the confidence level and conditional probability are equivalent if all are used to generate a generalized likelihood h ψ,b Is independent of the test of (a) and (b),bayesian reasoning is equivalent to ER reasoning, assuming h ab Expressed in evidence as the a-th hypothesis ζ a B-th detection result e in the established case b Probability of expected occurrence, i.eγ ab Representing the detection result e b Pointing hypothesis ζ a Confidence of (1), and->Psi represents a set of propositions for fault diagnosis, h ψ,b Representing the b-th detection result e under the proposition psi b And (2) generalized likelihood of (2) andγ ψ,b representing the detection result e b Confidence of pointing to proposition psi, andthe relationship is demonstrated as follows:in the IBRB, the input and output confidence levels are obtained through a series of confidence rules, the function components of the input and output confidence levels are marked as BD (-), the function components of the confidence rules of the IBRB are marked as BF (-), and the input and output confidence levels are marked as BF (input) 1 ),...,bf(input n ),BD(output)=BF(bf(input 1 ),...,bf(input n ) Obtaining a computational conversion process from FFTA to Bayesian network to IBRB and from Convert F Representing the transfer function of FFTA, convertet B Representing the transfer function of IBRB, θ is the parameter set of IBRB, ρ represents the parameter set of FFTA, logicG (= Convert) F (BF(·),ρ),BF(·)=Convert B (LogicG (·), θ), let u be 1 ,u 2 ,...,u M Represents the ith precondition attribute as input, M refers to the number of precondition attributes, [ M ] 1 ,n 1 ],[m 2 ,n 2 ],...,[m M ,n M ]Reference interval representing reference point, G 1 ,G 2 ,...,G N Indicating the result of fault diagnosis, (G) 1 ,β 1,k ),(G 2 ,β 2,k ),...,(G N ,β N,k ) Is their corresponding confidence, beta 1,k ,β 2,k ,...,β N,k Represents the confidence coefficient, lambda corresponding to the fault diagnosis result k Represents rule weights ζ k Representing rule reliability, alpha 1 ,α 2 ,...,α M Indicating the reliability of the index, r 1 ,r 2 ,...,r f The expression can explain constraint, the AND logic gate is expressed by V, the OR logic gate is expressed by V, and the confidence rule for obtaining the IBRB is described as follows: />The calculation method of the activation rule reliability alpha in the IBRB under different logic gates is different, and the calculation process is related to the confidence factor χ and the index reliability delta: /> Based on this, the equivalent conversion between FFTA and IBRB is introduced. The IBRB model expresses the probability of an event through the fuzzy triangle number of FFTA, the fuzzy triangle number is divided into three groups, one group expresses an upper bound value, one group expresses a lower bound value, and the other group expresses a reference value with representative meaning, which corresponds to a reference interval in the IBRB, the IBRB reasoning uses ER rules to fuse evidences, the ER rules solve the problem of evidence reliability, the reasoning of the IBRB model is perfected, Θ is marked as an identification frame, and N assessment grades { G } 1 ,G 2 ,...,G N Composition, beta Θ,i Indicating global unknowing, normalizing the data, and then obtaining the data e i The confidence profile is described as follows: />Evidence weight reuse i ∈[0,1]Expressed by evidence reliability i ∈[0,1]Representation, will ew i And ed i The new confidence profile is generated after the mixing weights as follows:
f dw,i =1/(1+ew 1 -ed i ),m n,i =ew i β n,i (1) Wherein m is n,i Representing basic probability mass, ++>Indicating the remaining basic probability mass without any evaluation level assigned,/->Representing a globally unknown basic probability mass, representing a single evidence imperfection, f dw,i Represents a normalized coefficient, satisfies->Joint support beta for L independent evidence n,e(L) Is represented as follows:
in the above formula, beta is represented n,e(k) (k=1, 2,., L) after fusing the first n pieces of evidence, rating G was evaluated n And satisfies m n,e(1) =m n,1 ,m β(Θ),e(1) ,m β(Θ),1 From the above information, the belief distribution and expected utility value of the following output, e (L) = { (G) n ,β n,e(L) ),n=1,...,N,(Θ,β Θ,e(L) },Where y represents the desired utility value, i.e. the predicted output, v (G n ) Expressed in the evaluation level G n At the lower utility value, the IBRB model itself has interpretability, however, the optimization algorithm usually converges to high accuracy to ensure the accuracy of the result, which can destroy the interpretability of the model, aeronautical facilitiesIf a fault occurs, a great impact is generated on the environment where the fault occurs, so that the IP-CMA-ES algorithm is necessary to keep the model interpretability, MSE (DEG) is the deviation degree of the utility value and the actual value output by the IBRB model, O T Representing the number of training samples, y representing the desired output value, y # The result representing the actual output of the model: /> Describing a parameter set to be optimized, wherein the parameter set comprises confidence coefficient, rule reliability and rule weight, and the parameter set is as follows: omega 0 =Ω 0 {β 1,1 ,...,β 1,1 ,ζ 1 ,...,ζ L ,λ 1 ,...,λ L Obtaining parameters of each generation by sampling, < -> The ith solution representing the d+1th generation, phi represents the evolution step of the d generation, Ω d Mean value of d-th generation search distribution, C d The covariance matrix of the d generation is represented, N (), the normal distribution function of the parameters is represented, g represents the number of offspring, and then the explanatory constraint condition is added, namely, the explanatory property of the IBRB model is not destroyed, the rationality of model output is guaranteed, and the constraint condition added to belief distribution is as follows:in the above formula, I k Representing an interpretable constraint on belief distribution under a first k rule, beta up 、ζ up 、λ up Respectively represent confidence, rule reliability, upper bound of rule weight, beta low 、ζ low 、λ low The confidence, rule reliability and lower limit value of rule weight are respectively expressed, no specific standard is used for constraint conditions of belief distribution, and expert usually determines according to experience and reality conditionTo satisfy the equality constraint, the equality constraint is converted into a constraint in the hyperplane by a projection operation, +.>Wherein sigma e =1..n represents the number of variables in the equality constraint, η=1..n+1 represents +.>The number of the equation constraints in (V) represents the parameter vector, then the optimal subgroup is filtered out in the population, the average value of the next generation is updated,/->W in the above i Refers to the weight coefficient of the ith solution, τ represents the size of the offspring population, +.>The solution in the g solution is represented, and finally, the covariance matrix is updated according to the strategy, and the evolution step and the step length are recorded as +.>a 1 And a 2 Record as learning rate, < >>Representing the ith solution vector in the g solution vectors at the d-th generation, the number of offspring is recorded as theta d The step size in the d generation is marked as phi d The expression pattern is as follows: />The flywheel system is used as an important component in a space attitude controller and is used for adjusting and controlling the attitude, stability and accurate positioning of a spacecraft, experimental verification is carried out by using a flywheel data set, fault diagnosis indexes comprise current, voltage, shaft temperature, friction moment and rotating speed, a friction moment fault subtree part is used as an experimental object, the experimental result of fault diagnosis is divided into three states, namely mild fault, medium fault and serious fault, and the fault diagnosis is carried out by using the friction moment fault subtree partThe parameter optimization of the model is an iterative process, the performance of the model is required to be continuously fed back, then the model is adjusted and optimized according to the feedback, the data of the flywheel system is well adapted, the running state and the performance of the flywheel system are accurately predicted, namely, the uncertainty of expert knowledge and the ambiguity of input data are solved by the model, the model with good fitting effect provides important information including parameters of aviation equipment and prediction of potential faults and problems, and helps us to better know the behaviors and trends of the aviation system, provide valuable references for maintaining and optimizing the aviation system, and meanwhile, help predict the future running state and performance, so that we can take measures in time to prevent faults and losses.
Although the present invention has been described with reference to the foregoing embodiments, it will be apparent to those skilled in the art that modifications may be made to the embodiments described, or equivalents may be substituted for elements thereof, and any modifications, equivalents, improvements and changes may be made without departing from the spirit and principles of the present invention.
Claims (10)
1. A trusted aviation equipment fault diagnosis method for solving ambiguity and uncertainty is characterized by comprising the following steps:
step one, constructing an aviation equipment fault relation: analyzing various indexes of the aviation equipment and constructing a reasonable fault relation;
step two, establishing an equivalence relation: establishing an equivalence relation between fuzzy fault tree analysis and an interval confidence rule base;
step three, a fault diagnosis model is established: establishing an aviation equipment fault diagnosis model based on FFTA and IBRB;
step four, model optimization: optimizing the model based on a projection covariance matrix self-adaptive evolution strategy algorithm with interpretable constraint;
step five, fault diagnosis: FFTA and IBRB based avionics fault diagnosis.
2. A trusted avionics fault diagnosis method to resolve ambiguity and uncertainty as claimed in claim 1, wherein: and in the second step, the relation establishment is specifically to construct a fuzzy fault tree for the acquired fault relation, a Bayesian network is used as a conversion space, and the expert knowledge is assisted to establish an equivalent relation with the IBRB.
3. A trusted avionics fault diagnosis method to resolve ambiguity and uncertainty as claimed in claim 1, wherein: and in the third step, the fault diagnosis model is established, namely a reasonable conversion rule is successfully established, and preliminary modeling is carried out on the fault diagnosis model of the aviation equipment.
4. A trusted avionics fault diagnosis method to resolve ambiguity and uncertainty as claimed in claim 1, wherein: the model optimization in the fourth step is specifically to optimize parameters of an initial model, and to simultaneously support the interpretability and high accuracy of the model in the optimization process, three model parameters of the opposite belief, the rule weight and the rule reliability are optimized by adopting an IP-CMA-ES.
5. A trusted avionics fault diagnosis method to resolve ambiguity and uncertainty as claimed in claim 1, wherein: in the fifth step, the fault diagnosis is specifically that the reliability, the evidence weight and the evidence reliability are processed through evidence reasoning (Evidential Reasoning, ER), so that the fault diagnosis result of the aviation equipment is obtained, and the diagnosis result is displayed through final confidence distribution.
6. A trusted avionics fault diagnosis method to resolve ambiguity and uncertainty as claimed in claim 2, wherein: in the second step, it is noted that the IBRB and the bayesian network are equivalent in terms of the confidence level of the IBRB and the conditional probability of the bayesian network when the events are independent of each other, and their relationships are expressed as follows:h ab expressed in evidence as the a-th hypothesis ζ a B-th detection result e in the established case b Probability of expected occurrence, i.e.)>γ ab Representing the detection result e b Pointing hypothesis ζ a Confidence of (1), and->Psi represents a set of propositions for fault diagnosis, h ψ,b Representing the b-th detection result e under the proposition psi b Is a generalized likelihood of (1), and>γ ψ,b representing the detection result e b Confidence pointing to proposition ψ, and +.>
7. A trusted avionics fault diagnosis method to resolve ambiguity and uncertainty as claimed in claim 2, wherein: in the second step, when the FFTA performs conversion under different logic gates and the bayesian network, the conversion method needs to be distinguished, firstly, the logic gate is determined as follows, p (T) =logic g (p (B) 1 ),...,p(B n ) Where p (·) represents the conditional probability corresponding to FFTA, p (T) represents the probability of occurrence of a top event in FFTA, logicG (·) represents the logic gate in FFTA, p (B) 1 ),...,p(B n ) Representing the failure probability of the ith elementary event in the FFTA, the conditional probability of a node in the bayesian network is expressed in terms of the transition between the and gate and the or gate, respectively,u in the above i A reference value indicating the i-th basic event in FFTA, and G indicating the failure diagnosis result.
8. A trusted avionics fault diagnosis method to resolve ambiguity and uncertainty as claimed in claim 2, wherein: in the second step, the reliability of the activation rule of the IBRB under different logic gates should be differentiated, and the reliability of the activation rule under the and gate or gate is as follows:where α is the activation rule reliability, related to the confidence factor χ and the index reliability δ, and M is the number of indexes.
9. A trusted avionics fault diagnosis method to resolve ambiguity and uncertainty as claimed in claim 3, wherein: in the third step, the reasoning process of the IBRB is as follows: the Θ is marked as an identification frame and is formed by N evaluation grades { G ] 1 ,G 2 ,...,G N Composition, beta Θ,i Indicating global unknowing, normalizing the data, and then obtaining the data e i The confidence profile is described as follows,evidence weight reuse i ∈[0,1]Expressed by evidence reliability i ∈[0,1]Representation, will ew i And ed i The new confidence profile is generated after the mixing weights as follows: f dw,i =1/(1+ew 1 -ed i );m n,i =ew i β n,i the method comprises the steps of carrying out a first treatment on the surface of the Wherein m is n,i Representing basic probability mass, ++>Indicating the remaining basic probability mass without any evaluation level assigned,/->Representing a globally unknown basic probability mass, representing a single evidence imperfection, f dw,i Represents the normalization coefficient, satisfiesJoint support beta for L independent evidence n,e(L) Is represented as follows: in the above formula, beta is represented n,e(k) (k=1, 2,., L) after fusing the first n pieces of evidence, rating G was evaluated n And satisfies m n,e(1) =m n,1 ,m β(Θ),e(1) ,m β(Θ),1 From the above information, the belief distribution and expected utility value of the following output, e (L) = { (G) n ,β n,e(L) ),n=1,...,N,(Θ,β Θ,e(L) },/>Where y represents the desired utility value, i.e. the predicted output, v (G n ) Expressed in the evaluation level G n Utility values below.
10. A trusted avionics fault diagnosis method to resolve ambiguity and uncertainty as claimed in claim 4, wherein: in the fourth step, the interpretability and high precision of the IBRB are ensured in the optimization process, and the MSE (-) is assumed to be the deviation degree between the utility value and the actual value output by the IBRB model, O T Representing the number of training samples, y representing the desired output value, y # Representing the result of the actual output of the model,describing the parameter set to be optimized, including confidence, rule reliability, rule weight, e.g.The following is shown: omega 0 =Ω 0 {β 1,1 ,...,β 1,1 ,ζ 1 ,...,ζ L ,λ 1 ,...,λ L Obtaining parameters of each generation by sampling, < -> The ith solution representing the d+1th generation, phi represents the evolution step of the d generation, Ω d Mean value of d-th generation search distribution, C d The covariance matrix of the d generation is represented, N (-) represents the normal distribution function of the parameter, g represents the number of offspring, and then the explanatory constraint condition is added, namely, the explanatory property of the IBRB model is not destroyed, the rationality of model output is guaranteed, the constraint condition added to belief distribution is as follows>In the above formula, I k Representing an interpretable constraint on belief distribution under a first k rule, beta up 、ζ up 、λ up Respectively represent confidence, rule reliability, upper bound of rule weight, beta low 、ζ low 、λ low The lower limit values of the confidence level, the rule reliability and the rule weight are respectively expressed, no specific standard exists on constraint conditions of belief distribution, the constraint conditions are usually determined by experts according to experience and reality conditions, and in order to meet equality constraint, the equality constraint needs to be converted into constraint in a hyperplane through projection operation>Wherein sigma e =1..n represents the number of variables in the equality constraint, η+1..n+1 represents +.>The number of the equation constraints in (V) represents the parameter vector, then the optimal subgroup is filtered out in the population, the average value of the next generation is updated,/->W in the above i Refers to the weight coefficient of the ith solution, τ represents the size of the offspring population, +.>Representing the solution in the g solution, finally updating the covariance matrix and the evolution steps and the step length thereof according to the strategy, and recording the evolution steps under the d+1th generation as +.>a 1 And a 2 Record as learning rate, < >>Representing the ith solution vector in the g solution vectors at the d-th generation, the number of offspring is recorded as theta d The step size in the d generation is marked as phi d The expression pattern is as follows: />
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202311395944.9A CN117421556A (en) | 2023-10-25 | 2023-10-25 | Trusted aviation equipment fault diagnosis method for solving ambiguity and uncertainty |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202311395944.9A CN117421556A (en) | 2023-10-25 | 2023-10-25 | Trusted aviation equipment fault diagnosis method for solving ambiguity and uncertainty |
Publications (1)
Publication Number | Publication Date |
---|---|
CN117421556A true CN117421556A (en) | 2024-01-19 |
Family
ID=89532087
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202311395944.9A Pending CN117421556A (en) | 2023-10-25 | 2023-10-25 | Trusted aviation equipment fault diagnosis method for solving ambiguity and uncertainty |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN117421556A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117629637A (en) * | 2024-01-24 | 2024-03-01 | 哈尔滨师范大学 | Aeroengine bearing fault diagnosis method and diagnosis system |
-
2023
- 2023-10-25 CN CN202311395944.9A patent/CN117421556A/en active Pending
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117629637A (en) * | 2024-01-24 | 2024-03-01 | 哈尔滨师范大学 | Aeroengine bearing fault diagnosis method and diagnosis system |
CN117629637B (en) * | 2024-01-24 | 2024-04-30 | 哈尔滨师范大学 | Aeroengine bearing fault diagnosis method and diagnosis system |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Khelif et al. | RUL prediction based on a new similarity-instance based approach | |
CN110286333B (en) | Fault diagnosis method for lithium power battery system | |
CN111339712A (en) | Method for predicting residual life of proton exchange membrane fuel cell | |
CN117421556A (en) | Trusted aviation equipment fault diagnosis method for solving ambiguity and uncertainty | |
CN110705812A (en) | Industrial fault analysis system based on fuzzy neural network | |
CN112100403A (en) | Knowledge graph inconsistency reasoning method based on neural network | |
CN112906764B (en) | Communication safety equipment intelligent diagnosis method and system based on improved BP neural network | |
CN116448419A (en) | Zero sample bearing fault diagnosis method based on depth model high-dimensional parameter multi-target efficient optimization | |
CN115600150A (en) | Multi-mode gearbox fault diagnosis method based on deep migration learning | |
CN112947385B (en) | Aircraft fault diagnosis method and system based on improved Transformer model | |
CN115791174B (en) | Rolling bearing abnormality diagnosis method, system, electronic equipment and storage medium | |
Agarwal et al. | Technical Basis for Advanced Artificial Intelligence and Machine Learning Adoption in Nuclear Power Plants | |
Pérez-Pérez et al. | Robust fault diagnosis of wind turbines based on MANFIS and zonotopic observers | |
Liu et al. | Residual useful life prognosis of equipment based on modified hidden semi-Markov model with a co-evolutional optimization method | |
Farahani et al. | A novel method for designing transferable soft sensors and its application | |
CN116596107A (en) | Method for predicting residual life of equipment in multiple failure modes based on small sample learning | |
Liu et al. | A new fault diagnosis model of flywheel system based on belief rule base and fuzzy fault tree analysis | |
Zhang et al. | A flexible monitoring framework via dynamic-multilayer graph convolution network | |
Liu et al. | Causality-based PCA Methods for Condition Modeling of Mechatronic Systems | |
Yin et al. | Interpretable belief rule base for safety state assessment with reverse causal inference | |
CN113792776A (en) | Interpretation method of deep learning model in network security anomaly detection | |
Li et al. | A New Fault Diagnosis Method Based on Belief Rule Base With Attribute Reliability Considering Multi-Fault Features | |
Ramachandran et al. | Case-Based Reasoning for System Anomaly Detection and Management | |
Li et al. | On Testing and Evaluation of Artificial Intelligence Models | |
CN117499199B (en) | VAE-based information enhanced decoupling network fault diagnosis method and system |
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 |