Civil aircraft undercarriage system fault parameter importance analysis method

Abstract

The invention discloses a civil aircraft undercarriage system fault parameter importance analysis method, and belongs to the field of fault diagnos.The method comprises the steps that fault parameter correlation and fault parameter redundancy of fault parameters under different working conditions are calculated based on the mutual information theory, and the characteristics of the fault parameter correlation and the fault parameter redundancy are integrated for representing the parameter importance degree; in addition, a plurality of groups of fault parameter importance degree sequences are calculated through a plurality of feature importance degree calculation methods; and performing weighted calculation on the multiple groups of fault parameter importance degree sequences by combining the fault parameter importance degree weights and the feature importance degree calculation method weights to obtain the fusion importance degree of each fault parameter, and obtaining the fusion importance ranking of each fault parameter according to the importance degree ranking of the fault parameters so as to realize importance evaluation of the fault parameters. According to the method, from the perspective of improving fault interpretability, an importance measurement and integrated fusion method is constructed to evaluate the importance of the civil aircraft undercarriage system fault parameters, and an analysis thought for monitoring the parameter priority degree is provided for civil aircraft health management system design.

Description

Technical Field

[0001] The present invention relates to the technical field of fault diagnosis, and particularly to a method for analyzing the importance of fault parameters of a civil aircraft landing gear system. Background Art

[0002] As a non-linear system with complex structural cross-links, the civil aircraft landing gear system is likely to fail during long-term operation on multiple routes. Due to the high reliability and safety design requirements of civil aircraft, it is difficult to obtain a large amount of complete labeled fault data of civil aircraft systems, which poses challenges to the actual fault analysis of civil aircraft systems. In order to further ensure the flight safety of civil aircraft, it is necessary to analyze the fault principle and characteristics from the perspective of fault mechanism, select key fault parameters for monitoring, and optimize the Prognostics Health Management (PHM) system of civil aircraft.

[0003] Although there are numerous existing airborne parameters and a large amount of flight monitoring data for civil airliners, the parameters related to the fault mechanism are few, and there is currently no unified standard for how to select necessary airborne health monitoring parameters. Therefore, in order to achieve the necessary selection of fault parameter monitoring, it is necessary to further analyze the importance of fault parameters. The theory of feature importance stems from a core concept in machine learning and statistics, which characterizes the contribution of input features to the model output. This theory is also commonly used to assist feature selection methods to obtain optimal features. Common feature importance methods evaluate importance based on the relationship between features and target variables or model outputs, pay less attention to the interaction between features, and do not consider the importance evaluation scenario where multiple features jointly represent the same parameter. In addition, traditional importance evaluation methods tend to provide a basis for feature selection and are more suitable for high-dimensional feature datasets, but there are problems of high computational cost, loss of time series information, and large estimation errors for high-dimensional time series sample datasets with few parameters.

[0004] In summary, how to further analyze the key fault parameters of the civil aircraft landing gear system that need to be monitored to cooperate with the improvement of the civil aircraft PHM system is an urgent problem to be solved by those skilled in the art at present. Summary of the Invention

[0005] The present invention provides a method for analyzing the importance of fault parameters of a civil aircraft landing gear system, introducing a new measure of fault parameter importance to characterize the importance degree of fault parameters, and hierarchically fusing the results of multiple evaluation models based on the ranking variable weight method and the rule weight fine-tuning method to solve the problem that traditional feature importance methods cannot be directly applied to the importance analysis of key fault parameters.

[0006] An embodiment of the present invention provides a method for analyzing the importance of fault parameters of a civil aircraft landing gear system, including the following steps:

[0007] S1. Collect the sensing signals of multiple fault parameters of the civil aircraft landing gear system, and establish a multi-condition time series data set of multiple fault parameters;

[0008] S2. For the multi-condition time series data set of the multiple fault parameters, calculate the fault parameter correlation and fault parameter redundancy of the multiple fault parameters under different conditions based on the mutual information theory, and integrate the fault parameter correlation and fault parameter redundancy to characterize the importance of the fault parameters;

[0009] S3. Objectively evaluate and analyze the importance of multiple fault parameters from the statistical perspective and the classification perspective through multiple feature importance calculation methods, and obtain multiple groups of fault parameter importance sequences;

[0010] S4. Use the sorting variable weight method to calculate the importance weights of each fault parameter, and use the rule-based weight fine-tuning method to adjust the weights of multiple feature importance calculation methods;

[0011] S5. Combine the importance weights of each fault parameter and the weights of multiple feature importance calculation methods to perform weighted calculation on the importance of each fault parameter in the multiple groups of fault parameter importance, obtain the fusion importance of each fault parameter, obtain the fusion importance ranking list of the fault parameters through sorting, and obtain the importance ranking of each fault parameter according to the fusion importance ranking list, so as to realize the importance evaluation of the fault parameters.

[0012] Optionally, in an embodiment of the present invention, in step S1, at least three fault parameter sensing signals are collected by using sensors or simulation models, and each fault parameter sensing signal includes time series data under normal conditions and at least one fault condition.

[0013] Optionally, in an embodiment of the present invention, in step S2, the fault parameter correlation includes fault identifiability, fault distribution correlation, and hierarchical fault correlation, and the fault parameter redundancy includes parameter global redundancy and hierarchical fault redundancy. Step S2 specifically includes the following steps:

[0014] S201. The mutual information I(X; Y) between variables X = {x1, x2,..., x n} and Y = {y1, y2,…, y m} is expressed as:

[0015]

[0016] S202. Calculate the fault identifiability I rel1 based on the mutual information, which is expressed as the mutual information between the multi-parameter time series data matrix X nor under normal conditions and the multi-parameter time series data matrix X fault under fault conditions. The calculation formula is:

[0017]

[0018] Among them, X nor ={p1, p2, …, p N}} represents the time-series data matrix under normal working conditions, and p i represents the time series of the i-th fault parameter, and there are N fault parameters in total; represents the time-series data matrix under the k-th fault condition, represents the time series of the i-th fault parameter under the k-th fault condition, and there are K fault conditions in total;

[0019] S203. Calculate the fault distribution correlation I based on mutual information rel2 , which is expressed as the mutual information between the multi-parameter time-series data matrix P and the fault label matrix C under all working conditions. The calculation formula is:

[0020]

[0021] Among them, P t represents the t-th row sample data of the multi-parameter time-series data matrix P, and C t represents the t-th label of the fault label matrix C;

[0022] S204. Calculate the hierarchical fault correlation I based on mutual information rel3 , if there are faults with the same fault mode but different fault degrees in the fault conditions, they are considered hierarchical faults; assuming that there are H fault modes in a certain type of hierarchical fault, then the multi-parameter time-series data matrix under the h-th fault mode of this type of hierarchical fault is expressed as The hierarchical fault correlation I rel3 The calculation formula is:

[0023]

[0024] S205. Calculate the parameter global redundancy I based on mutual information redu , and the calculation formula is:

[0025]

[0026] Among them, P i,k represents the time series of the i-th fault parameter under the k-th working condition, and P j,k represents the time series of the j-th fault parameter under the k-th working condition;

[0027] S206. The hierarchical fault redundancy refers to considering the superimposed redundancy effect of the existence of hierarchical faults. Assuming that L hf is the set of hierarchical fault numbers, then the hierarchical fault redundancy I hf is expressed as:

[0028]

[0029] S207, the importance I of fault parameters integrating the characterization of fault parameter correlation and fault parameter redundancy FIRM is as follows:

[0030] I FIRM = γ1I rel1 + γ2I rel2 + γ3I rel3 - λI hf

[0031] where γ i and λ are weight factors.

[0032] Optionally, in an embodiment of the present invention, in step S3, at least one model-independent method and one model analysis method are included in multiple feature importance calculation methods.

[0033] Optionally, in an embodiment of the present invention, in step S4, the sorting variable weight method refers to adjusting the weights of different fault parameters based on the ranking of the importance of fault parameters obtained by the feature importance calculation method, and the rule-based weight fine-tuning method refers to assigning the weights of different feature importance calculation methods according to objective rules. Step S4 specifically includes:

[0034] Calculating the importance weights of each fault parameter by using the sorting variable weight method: It is assumed that there are Q feature importance calculation methods to calculate the importance of N fault parameters. The ranking of the jth fault parameter of the ith feature importance calculation method is rank(i, j), and the weight w i,j of the jth fault parameter of the ith feature importance calculation method is as follows:

[0035]

[0036] Adjusting the weights of multiple feature importance calculation methods by using the rule-based weight fine-tuning method: Combining objective evaluation rules to fine-tune the weights of different feature importance calculation methods. The adjustment formula is:

[0037]

[0038] where ω i represents the weight of the ith feature importance calculation method, r represents the total number of objective evaluation rules, β j represents the influence coefficient of the jth objective evaluation rule, and α j represents the number of parameters that satisfy the jth objective evaluation rule.

[0039] Optionally, in an embodiment of the present invention, in step S5, the fusion importance calculation formula for all fault parameters is:

[0040]

[0041] Among them, I CMPI = {χ1, χ2, …, χ N} represents the importance of each fault parameter calculated by a variety of feature importance calculation methods, and χ j represents the importance value of the jth fault parameter, q represents the total number of feature importance calculation methods, and ρ i,j represents the importance value of the jth fault parameter calculated by the ith feature importance calculation method;

[0042] For the importance of each fault parameter in I CMPI , sort them from large to small according to the size of the importance value of the fault parameter to obtain the importance ranking list of the fault parameter, which is expressed as follows:

[0043] σ(I CMPI ) = {χ σ(1) , χ σ(1) , …, χ σ(N)}, χ σ(1) ≥ χ σ(1) ≥ … ≥ χ σ(N)

[0044] Among them, σ(·) represents the sorting function;

[0045] According to the importance ranking list, obtain the importance ranking rank CMPI which is expressed as follows:

[0046] rank CMPI (χ σ(1) ) ≤ rank CMPI (χ σ(2) ) ≤ … ≤ rank CMPI (χ σ(N) )

[0047] Among them, the ranking value retains the mapping relationship with the original fault parameter order to obtain the importance ranking of each fault parameter.

[0048] The method for analyzing the importance of fault parameters of the civil aircraft landing gear system in the embodiments of the present invention has the following beneficial effects:

[0049] 1) The method of the present invention considers the fault identifiability, fault distribution correlation, hierarchical fault correlation, parameter global redundancy, and hierarchical fault redundancy of the high-dimensional time series data set, constructs a new parameter importance measurement strategy, retains the time series, classifiability, and fault differences of the original data as much as possible, and enhances the interpretability of the parameter importance evaluation model.

[0050] 2) The method of the present invention proposes a multi-model parameter importance fusion strategy based on sorting variable weights and rule weight fine-tuning, combines objective rules to assist in judging the reliability of single-model evaluation, fine-tunes the integrated weights of parameters and the integrated weights of models, and improves the reliability of model importance analysis.

[0051] 3) The method of the present invention introduces statistical metrics and model classification performance metrics to form a hybrid evaluation model suitable for parameter importance evaluation, considers the statistical characteristics and classification characteristics of parameters to comprehensively evaluate the importance of civil aircraft landing gear fault parameters, and provides an analysis idea for monitoring parameter priorities in the design of civil aircraft PHM systems.

[0052] Additional aspects and advantages of the present invention will be given in part in the following description, will become apparent in part from the following description, or will be learned through the practice of the present invention. Brief Description of the Drawings

[0053] The above-mentioned and / or additional aspects and advantages of the present invention will become apparent and easy to understand from the following description of the embodiments in conjunction with the drawings, wherein:

[0054] Figure 1 is a framework diagram of a method for analyzing the importance of fault parameters of a civil aircraft landing gear system according to an embodiment of the present invention;

[0055] Figure 2 is a schematic diagram of the execution process of a method for analyzing the importance of fault parameters of a civil aircraft landing gear system according to an embodiment of the present invention;

[0056] Figure 3 is a comparison diagram of the importance evaluation results of the evaluation model involved in the embodiment of the present invention;

[0057] Figure 4 is a diagram of the importance evaluation result of the fault parameter output by the method of the present invention. Detailed Embodiment

[0058] The embodiments of the present invention will be described in detail below. The examples of the embodiments are shown in the drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below by referring to the drawings are exemplary and are intended to explain the present invention and should not be construed as limiting the present invention.

[0059] As Figure 1 shown, the method for analyzing the importance of fault parameters of the civil aircraft landing gear system includes the following steps:

[0060] S1, collect the sensing signals of multiple fault parameters of the civil aircraft landing gear system, and establish a multi-condition time series data set of multiple fault parameters.

[0061] Collect at least three fault parameter sensing signals using sensors or simulation models, and establish a multi-condition time series data set for multiple fault parameters. The sensors are arranged at important measurement points related to the fault state of the landing gear system, or important measurable parameters related to the fault state in the landing gear simulation model, ensuring that the time series data of the fault parameters associated with the health state of the landing gear system can be obtained. To construct a multi-condition time series data set of fault parameters, at least three important fault parameters should be selected. For the same fault parameter, the time series data including normal conditions and at least one fault condition should be collected.

[0062] S2. For the multi-condition time series data set of multiple fault parameters, calculate the fault parameter correlation and fault parameter redundancy of multiple fault parameters under different conditions based on the mutual information theory, and integrate the fault parameter correlation and fault parameter redundancy to characterize the importance of fault parameters. The fault parameter correlation includes fault identifiability, fault distribution correlation, and hierarchical fault correlation, and the fault parameter redundancy includes parameter global redundancy and hierarchical fault redundancy.

[0063] The present invention proposes a parameter importance measurement strategy (Fault Identifiability and Redundancy Mitigation, FIRM) considering fault identifiability and redundancy reduction. The calculation steps of this strategy are as follows:

[0064] S201. The parameter importance measurement strategy calculates the fault parameter correlation and fault parameter redundancy respectively based on the mutual information theory. The mutual information I(X; Y) of variables X = {x1, x2,..., x n} and Y = {y1, y2,..., y m} is expressed as follows:

[0065]

[0066] S202. Calculate the fault identifiability I rel1 based on the mutual information, which is expressed as the mutual information between the multi-parameter time series data matrix X nor under normal conditions and the multi-parameter time series data matrix X fault under fault conditions. The calculation formula is:

[0067]

[0068] Among them, X nor = {p1, p2,..., p N} represents the time series data matrix under normal conditions, and p i represents the time series of the i-th fault parameter, and there are N fault parameters in total; represents the time series data matrix under the k-th fault condition, Denote the time series of the \(i\)-th fault parameter under the \(k\)-th fault condition, and there are \(K\) fault conditions in total;

[0069] S203. Calculate the fault distribution correlation \(I\) based on mutual information rel2 , which is expressed as the mutual information between the multi-parameter time series data matrix \(P\) and the fault label matrix \(C\) under all conditions, and the calculation formula is:

[0070]

[0071] where, \(P\) t denotes the \(t\)-th row sample data of the multi-parameter time series data matrix \(P\), and \(C\) t denotes the \(t\)-th label of the fault label matrix \(C\);

[0072] S204. Calculate the hierarchical fault correlation \(I\) based on mutual information rel3 , if there are faults with the same fault mode but different fault degrees in the fault conditions, they are considered hierarchical faults; assume that there are \(H\) fault modes in a certain type of hierarchical fault, then the multi-parameter time series data matrix under the \(h\)-th fault mode of this type of hierarchical fault is expressed as The hierarchical fault correlation \(I\) rel3 The calculation formula is:

[0073]

[0074] S205. Calculate the global redundancy \(I\) of parameters based on mutual information redu , the calculation formula is:

[0075]

[0076] where, \(P\) i,k denotes the time series data of the \(i\)-th fault parameter under the \(k\)-th condition, and \(P\) j,k denotes the time series data of the \(j\)-th fault parameter under the \(k\)-th condition;

[0077] S206. The hierarchical fault redundancy refers to considering the superimposed redundancy effect of the existence of hierarchical faults. Assume that \(L\) hf is the set of hierarchical fault numbers, then the hierarchical fault redundancy \(I\) hf is expressed as:

[0078]

[0079] S207. The importance \(I\) of fault parameters integrating the characterization of fault parameter correlation and fault parameter redundancy FIRM is:

[0080] \(I\) FIRM =\(\gamma_1I\) rel1 +\(\gamma_2I\) rel2 +\(\gamma_3I\) rel3-λI hf

[0081] where γ i and λ are weight factors.

[0082] S3. Objectively evaluate and analyze the importance of multiple fault parameters from both statistical and classification perspectives through multiple feature importance calculation methods, and obtain multiple groups of fault parameter importance sequences.

[0083] At least one model-independent method and one model analysis method are included in the multiple feature importance calculation methods to ensure that the importance of parameters can be objectively evaluated and analyzed from both statistical and classification perspectives.

[0084] The model-independent method refers to measuring the importance of data based on the characteristics of the data itself without relying on a specific model structure, and can be applied to any machine learning model, such as mutual information measurement, correlation measurement, variance analysis, etc.; the model analysis method depends on the model structure, and the evaluation result of feature importance depends on a specific machine learning model. Common methods include feature permutation method, importance analysis method based on tree model, feature importance analysis method based on regularization, SHAP method, etc.

[0085] S4. Calculate the importance weights of each fault parameter using the ranking variable weight method, and adjust the weights of multiple feature importance calculation methods using the rule-based weight fine-tuning method.

[0086] As Figure 2 shown, the ranking variable weight method refers to adjusting the weights of different fault parameters based on the ranking of the importance of fault parameters obtained by the feature importance calculation method, and the rule-based weight fine-tuning method refers to giving the weights of different feature importance calculation methods according to objective rules. The specific calculation steps of step S4 are as follows:

[0087] S401. Calculate the importance weights of each fault parameter using the ranking variable weight method: Assume that there are Q feature importance calculation methods to calculate the importance of N fault parameters. The ranking of the jth fault parameter calculated by the ith feature importance calculation method is rank(i, j), and the weight w i,j of the jth fault parameter calculated by the ith feature importance calculation method is:

[0088]

[0089] S402. Adjust the weights of multiple feature importance calculation methods using the rule-based weight fine-tuning method: Fine-tune the weights of different feature importance calculation methods in combination with objective evaluation rules. The adjustment formula is:

[0090]

[0091] where, ω i represents the weight of the i-th characteristic importance calculation method, r represents the total number of objective evaluation rules, β j represents the influence coefficient of the j-th objective evaluation rule, α j represents the number of parameters that satisfy the j-th objective evaluation rule.

[0092] The rules should conform to objective facts and have significant reference value for fault analysis; assuming the parameter importance ranking calculated by a certain importance evaluation method, such as Rule 1, if the system input-output parameters are ranked in the top 20% in the evaluation result of this method; such as Rule 2, if the constant parameters of the system are ranked in the bottom 20% in the evaluation conclusion of this method; when at least one objective rule is set, the formula described in S402 is used to fine-tune the method weight.

[0093] S5. Combine the importance weights of each fault parameter and the weights of multiple characteristic importance calculation methods to perform weighted calculation on the importance of each fault parameter in multiple groups of fault parameter importance, obtain the fusion importance of each fault parameter, obtain the fusion importance ranking list of fault parameters through sorting, and obtain the importance ranking of each fault parameter according to the fusion importance ranking list, so as to realize the importance evaluation of fault parameters.

[0094] In S5, apply the importance weights of the fault parameters and the weights of the characteristic importance calculation methods in the integration steps S2 and S3 to comprehensively integrate the importance of the fault parameters, and give the importance ranking of the fault parameters according to the comprehensive importance of the fault parameters. The specific steps are as follows:

[0095] S501. The fusion importance calculation formula for all fault parameters is:

[0096]

[0097] where, I CMPI ={χ1, χ2, …, χ N} represents the importance of each fault parameter calculated by multiple characteristic importance calculation methods, χ j represents the importance value of the j-th fault parameter, q represents the total number of characteristic importance calculation methods, ρ i,j represents the importance value of the j-th fault parameter of the i-th characteristic importance calculation method;

[0098] S502. For the importance of each fault parameter in I CMPI , sort them in descending order according to the size of the fault parameter importance value, and obtain the importance ranking list of fault parameters, which is expressed as follows:

[0099] σ(I CMPI )={χ σ(1) , χσ(1) , …, χ σ(N)},χ σ(1) ≥ χ σ(1) ≥ … ≥ χ σ(N)

[0100] Among them, σ(·) represents the sorting function;

[0101] S503, sort the list according to the importance, and obtain the importance ranking rank of the fault parameters CMPI It is expressed as follows:

[0102] rank CMPI (χ σ(1) ) ≤ rank CMPI (χ σ(2) ) ≤... ≤ rank CMPI (χ σ(N) )

[0103] Among them, the ranking value retains the mapping relationship with the original order of the fault parameters, and the importance ranking of each fault parameter is obtained.

[0104] The method of the present invention will be described in detail below through a specific embodiment.

[0105] In the simulation model of the landing gear system of a certain type of aircraft, select the front wheel turning process of the landing gear system. The total simulation time is 85 s, and the sampling frequency is 10 kHz. Obtain the fault signal set of the landing gear system under 23 working conditions, and the fault modes are shown in Table 1. From the perspective of fault mechanism, select 10 analysis parameters that can reflect the system operation state. The 10 fault parameters are shown in Table 2. To facilitate data processing, stack the time series data of different working conditions of the same parameter vertically and stack the data of different parameters horizontally to form a two-dimensional sample data set.

[0106] Table 1 Landing gear system fault mode table

[0107]

[0108]

[0109] Table 2 Aircraft landing gear system fault parameter table

[0110] Serial number Parameter name Parameter abbreviation 1 External pressure of the turning actuator Force 2 Control current of the turning control valve i 3 Front wheel turning angle jd 4 Pressure in the rod chamber of the turning actuator PA 5 Pressure in the non-rod chamber of the turning actuator PB 6 Output pressure of the hydraulic pump PS 7 Flow rate in the rod chamber of the turning actuator QA 8 Flow rate in the non-rod chamber of the turning actuator QB 9 Moving speed of the rod of the turning actuator Vp 10 Displacement of the spool of the turning control valve Xv

[0111] In the embodiment of the present invention, a total of 13 feature importance calculation methods including the method of the present invention were selected for operation, as shown in Table 3. Among them, Method 1 refers to a feature importance measurement strategy in S2 that considers fault identifiability and redundancy reduction, and Method 13 refers to a method for analyzing the importance of landing gear system parameters combining composite metrics (Composite Metric for Parameter Importance, CMPI) of the present invention. All methods use default parameter settings, and the parameter settings involved in the CMPI method are shown in Table 4.

[0112] Table 3 Feature importance calculation methods involved in the invention embodiment

[0113]

[0114] Table 4 CMPI method parameter setting table

[0115] Parameter name <![CDATA[γ1]]> <![CDATA[γ2]]> <![CDATA[γ3]]> λ <![CDATA[β1]]> <![CDATA[β2]]> Value 1 1 0.2 1 0.01 0.05

[0116] Table 5 Reference table for the priority of landing gear system fault parameters

[0117] Serial number Parameter type Parameter Priority 1 Parameters directly affecting flight safety PS, PA, PB Level 1 (highest) 2 Parameters directly reflecting control effect jd, Xv, i Level 2 3 Auxiliary parameters reflecting the working state of the system QA, QB, Vp Level 3 4 Parameters indirectly reflecting the system state Force Level 4 (relatively low)

[0118] The experimental results of the embodiment of the present invention are as Figure 3 shown. According to the simulation structure and fault mode of the landing gear system, it can be known that the control current i of the turning control valve is the input parameter, the front wheel turning angle jd is the output parameter, and the output pressure PS of the hydraulic pump simulates the hydraulic system to supply energy to the landing gear system. Table 5 gives a reference table for the priority of relevant fault parameters in the actual civil aircraft landing gear system. This table is jointly obtained by combining knowledge such as maintenance manuals, expert experience, and model principles. Based on the priority information of the fault parameters in this table, qualitative analysis can be carried out on Figure 3 the importance results. It should be particularly noted that the simulation model in this embodiment is based on the premise assumption of stable energy supply of the hydraulic system. Except for the low supply pressure fault mode, the PS parameter maintains a constant working value in the non-low supply pressure fault mode, that is, the output pressure PS of the hydraulic pump in the simulation model is used as a constant parameter. Therefore, the PS parameter is not considered in this embodiment for evaluation.

[0119] From Figure 3It can be seen that the FIRM metric proposed by the method of the present invention ranks the importance of 10 fault parameters from a statistical perspective, can identify parameter i and parameter jd as important parameters for system fault analysis, basically reflects the importance trend of system parameters, and can find that the importance of parameters PA and PB in terms of statistical characteristics is not very obvious. The MIFS and mRMR methods are also importance calculated based on statistical metrics. The results show that the statistical perspective pays more attention to the system structure characteristics, and the relative importance of the reflected parameters is basically the same. Only some intermediate importance variables, such as QA, QB, Force, etc., will change with the characteristics of different methods. This phenomenon also indicates that the use of a fusion method can more objectively analyze the importance of parameters. The SHAP, feature permutation method, and feature elimination method are all model-based analysis methods, which pay more attention to the contribution of parameters to the classification performance during the calculation process. It can be seen from the results that this type of method is significantly different from the method based on the statistical perspective, and can observe the importance of parameters PA and PB related to system safety; from the overall results of the above methods, it can be seen that the model analysis method tends to select parameters that are easily recognized as features by the model as more important parameters.

[0120] The experimental results of the CMPI method are as Figure 4 shown. It can be seen that the method of the present invention fully combines the characteristics of statistical metrics and classification performance, can identify the output parameter jd as an important monitoring parameter, retains the relatively unimportant ranking of the Force parameter, and comprehensively considers the correlation and redundancy between other parameters, which also conforms to the priority ranking in Table 5 to a certain extent, and gives a relatively objective comprehensive ranking. In summary, the method of the present invention uses a composite metric to integrate the results of multiple methods, can combine the characteristics of statistical metrics and classification metrics, and obtain a more objective importance evaluation result. By analyzing the importance of the fault parameters of the landing gear system, it can give reference suggestions on the monitoring necessity of the parameters, guide relevant practitioners to further improve the civil aircraft PHM system, and ensure the safety of the civil aircraft electromechanical system.

[0121] The method for analyzing the importance of fault parameters of a civil aircraft landing gear system according to the embodiments of the present invention calculates the fault parameter correlation and fault parameter redundancy of fault parameters under different working conditions based on the mutual information theory, and integrates the characteristics of the two to characterize the parameter importance. In addition, multiple groups of fault parameter importance sequences are calculated by multiple characteristic importance calculation methods; the sorting variable weight method is used to calculate the importance weights of each fault parameter, and the weights of multiple characteristic importance calculation methods are adjusted by the rule-based weight fine-tuning method; the weighted calculation of multiple groups of fault parameter importance sequences is carried out by combining the fault parameter importance weights and the characteristic importance calculation method weights to obtain the fusion importance of each fault parameter, and the fusion importance ranking of each fault parameter is obtained according to the importance ranking of the fault parameters, so as to realize the importance evaluation of the fault parameters. Starting from the perspective of improving fault interpretability, the method of the present invention constructs an importance measure and an integration and fusion method to evaluate the importance of fault parameters of a civil aircraft landing gear system, providing an analysis idea for the priority degree of monitoring parameters in the design of a civil aircraft health management system.

[0122] In the description of this specification, the descriptions referring to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples", etc. mean that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described may be combined in any one or N embodiments or examples in a suitable manner. In addition, without contradiction, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of different embodiments or examples.

[0123] In addition, the terms "first" and "second" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include at least one of the features. In the description of the present invention, the meaning of "N" is at least two, such as two, three, etc., unless otherwise specifically defined.

[0124] Any process or method description in the flowchart or described in other ways herein may be understood to represent a module, segment, or portion of code including one or N executable instructions for implementing a customized logic function or process, and the scope of the preferred embodiments of the present invention includes additional implementations, where the functions may be executed in a substantially simultaneous manner or in the reverse order according to the functions involved, rather than in the order shown or discussed, which should be understood by those skilled in the art to which the embodiments of the present invention belong.

Claims

1. A method for analyzing the importance of fault parameters of a civil aircraft landing gear system, characterized in that, It includes the following steps: S1. Collect the sensing signals of multiple fault parameters of the civil aircraft landing gear system, and establish a multi-condition time series data set of multiple fault parameters; S2. For the multi-condition time series data set of the multiple fault parameters, calculate the fault parameter correlation and fault parameter redundancy of the multiple fault parameters under different conditions based on the mutual information theory, and integrate the fault parameter correlation and fault parameter redundancy to characterize the importance of the fault parameters; S3. Objectively evaluate and analyze the importance of the multiple fault parameters from the statistical perspective and the classification perspective respectively through a variety of feature importance calculation methods, and obtain multiple groups of fault parameter importance sequences; S4. Use the sorting variable weight method to calculate the importance weights of each fault parameter, and use the rule-based weight fine-tuning method to adjust the weights of a variety of feature importance calculation methods; S5. Combine the importance weights of each fault parameter and the weights of a variety of feature importance calculation methods to perform weighted calculation on the importance of each fault parameter in the multiple groups of fault parameter importance, obtain the fusion importance of each fault parameter, obtain the fusion importance ranking list of the fault parameters through sorting, and obtain the importance ranking of each fault parameter according to the fusion importance ranking list, so as to realize the importance evaluation of the fault parameters.

2. The method according to claim 1, characterized in that, In step S1, use sensors or simulation models to collect at least three fault parameter sensing signals, and each fault parameter sensing signal includes the time series data of normal conditions and at least one fault condition.

3. The method according to claim 1, wherein In step S2, the fault parameter correlation includes fault identifiability, fault distribution correlation, and hierarchical fault correlation, and the fault parameter redundancy includes parameter global redundancy and hierarchical fault redundancy. Step S2 specifically includes the following steps: S201, the mutual information I(X; Y) of variables X = {x1, x2,..., x n} and Y = {y1, y2,..., y m} is expressed as: S202, calculate the fault identifiability I based on mutual information rel1 , which is represented as the multi-parameter time series data matrix X under normal working conditions nor and the multi-parameter time series data matrix X under fault conditions fault . The calculation formula is as follows: Among them, X nor = {p1, p2,..., p N} represents the time series data matrix under normal working conditions, and p i represents the time series of the i-th fault parameter, and there are N fault parameters in total; represents the time series data matrix under the k-th fault condition, represents the time series of the i-th fault parameter under the k-th fault condition, and there are K fault conditions in total; S203, calculate the correlation I of fault distribution based on mutual information rel2 , which is expressed as the mutual information between the multi-parameter time series data matrix P and the fault label matrix C under all working conditions. The calculation formula is as follows: Among them, P t represents the t-th row sample data of the multi-parameter time series data matrix P, and C t represents the t-th label of the fault label matrix C; S204, calculate the hierarchical fault correlation I based on mutual information rel3 , if there are faults with the same fault mode but different fault degrees in the fault conditions, they are considered hierarchical faults; assuming that there are H fault modes in a certain type of hierarchical fault, the multi-parameter time series data matrix under the h-th fault mode of this type of hierarchical fault is expressed as Hierarchical fault correlation I rel3 The calculation formula is as follows: S205, calculate the global redundancy I of the parameter based on mutual information redu , and the calculation formula is as follows: Among them, P i,k represents the time-series data of the i-th fault parameter under the k-th working condition, and P j,k represents the time-series data of the j-th fault parameter under the k-th working condition; S206. The hierarchical fault redundancy refers to considering the superimposed redundancy effect of the existence of hierarchical faults, and setting L hf as the hierarchical fault serial number set. Then the hierarchical fault redundancy I hf is expressed as: S207, the importance I of fault parameters that integrates the characterization of the correlation and redundancy of fault parameters FIRM is as follows: I FIRM = γ1I rel1 + γ2I rel2 + γ3I rel3 - λI hf Among them, γ i and λ are weighting factors.

4. The method according to claim 1, characterized in that, In step S3, the variety of feature importance calculation methods include at least one model-independent method and one model analysis method.

5. The method according to claim 1, characterized in that In step S4, the sorting variable weight method refers to adjusting the weights of different fault parameters based on the importance ranking of the fault parameters obtained by the feature importance calculation method, and the rule-based weight fine-tuning method refers to giving the weights of different feature importance calculation methods according to objective rules. Step S4 specifically includes: The importance weights of each fault parameter are calculated using the sorting variable weight method: It is assumed that there are Q feature importance calculation methods to calculate the importance of N fault parameters. The rank of the j-th fault parameter in the i-th feature importance calculation method is rank(i, j), and the weight w i,j of the j-th fault parameter in the i-th feature importance calculation method is: Use the rule-based weight fine-tuning method to adjust the weights of a variety of feature importance calculation methods: fine-tune the weights of different feature importance calculation methods in combination with objective evaluation rules, and the adjustment formula is: Among them, ω i represents the weight of the i-th feature importance calculation method, r represents the total number of objective evaluation rules, and β j represents the influence coefficient of the j-th objective evaluation rule, and α j represents the number of parameters that satisfy the j-th objective evaluation rule.

6. The method according to claim 1, wherein In step S5, the fusion importance calculation formula of all fault parameters is: Among them, I CMPI ={χ1, χ2,..., χ N} represents the importance degrees of various fault parameters calculated by multiple feature importance calculation methods, χ j represents the importance degree value of the j-th fault parameter, q represents the total number of feature importance calculation methods, ρ i,j represents the importance degree value of the j-th fault parameter calculated by the i-th feature importance calculation method; For I CMPI For each fault parameter importance in it, sort them from large to small according to the magnitude of the fault parameter importance value to obtain the importance ranking list of fault parameters, which is shown as follows: σ(I CMPI ) = {χ σ(1) , χ σ(1) ,..., χ σ(N)}, χ σ(1) ≥ χ σ(1) ≥... ≥ χ σ(N) Among them, σ(·) represents the sorting function; Sort the list according to importance to obtain the importance ranking rank of the fault parameters CMPI It is shown as follows: rank CMPI (χ σ(1) ) ≤ rank CMPI (χ σ(2) ) ≤... ≤ rank CMPI (χ σ(N) ) Among them, the ranking value retains the mapping relationship with the original fault parameter order to obtain the importance ranking of each fault parameter.

Images