Inertial measurement unit health state determination method and system based on interpretable statistical characteristic evidence reasoning
Through the reasoning method based on the explanatory statistical characteristics evidence, the problems of transparency and interpretability in the health status assessment of laser stroller groups are solved, and the transparency and credibility of the evaluation process and results are achieved, providing clear health Status evaluation results.
Patent Information
- Application Number
- CN202411487443.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-24
- Publication Date
- 2024-12-06
AI Technical Summary
The prior art is difficult to achieve transparency of the evaluation process and results in the assessment of the health status of laser strap-inert groups, especially in multivariate test data uncertainty and complex systems, feature extraction and information based on black box and gray box methods. Fusion is difficult to provide explainable results.
Using an inference method based on explanatory statistical characteristics, the confidence distribution of multiple sets of samples is obtained, the state space matrix is constructed, principal component analysis and normalization is performed, and multivariate information fusion is combined with evidence reasoning to obtain the confidence distribution of the evaluation results. This determines the health status of the inertia group.
The transparency of the evaluation process and results is achieved, and clear physical significance and explainable health status assessment results are provided, which enhances the credibility of the evaluation results.
Smart Images

Figure CN119089391A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of health status management of laser strapdown inertial navigation systems, and in particular to a method and system for determining the health status of inertial navigation systems based on interpretable statistical feature evidence reasoning. Background Art
[0002] The health status assessment of the laser strapdown inertial group (abbreviated as "inertial group") is a process of evaluating its health status by comprehensively utilizing multivariate test data. Timely understanding of the health status of the aircraft is a key link in achieving health management and provides a theoretical basis for subsequent maintenance decisions.
[0003] In practice, the health status assessment of inertial groups faces the challenge of test data uncertainty, so it is necessary to fuse multivariate information to enhance the accuracy of the assessment results. Health status assessment methods based on white boxes, such as decision trees and fault trees, provide more accurate and effective results based on the accurate establishment of the physical model of the system. However, with the increase in the complexity of the system, it is difficult to grasp a clear and clear mechanism. Health status assessment methods based on black boxes, such as support vector machines and artificial neural networks, describe the input-output relationship of the model through data, have strong data mining and feature extraction capabilities, and can obtain highly accurate assessment results through training. However, the reasoning mechanism within the model is difficult to understand, and it is also difficult to analyze the assessment results. Health status assessment methods based on gray boxes combine data and mechanism information to complement each other. For example, the evidence reasoning method, as a multivariate information fusion method, converts variable data into a unified confidence structure and then fuses it to achieve the integration of uncertainty information. However, in the process of applying this method, some harsh assumptions need to be met: the fused evidence is independent. The feature extraction method based on principal components can construct independent fusion evidence, but the extracted features are linear combinations of variables, with unclear physical meanings, and are difficult to correspond to clear variables during the reasoning process. In addition, the health status assessment of aircraft pays more attention to interpretability, including the transparency of the model and the comprehensibility of the assessment results, requiring the assessment process to be transparent so that the credibility of the assessment results is guaranteed. Summary of the invention
[0004] The purpose of the present invention is to provide a method and system for determining the health status of a group based on interpretable statistical feature evidence reasoning, which can achieve transparency of the evaluation process and evaluation results.
[0005] To achieve the above object, the present invention provides the following solution.
[0006] An embodiment of the present invention provides a method for determining a health status of a group of people based on interpretable statistical feature evidence reasoning, the method comprising:
[0007] Obtain multiple groups of samples for determining the health status of the laser strapdown inertial system, and calculate the confidence distribution of each variable in each group of samples under different indicators;
[0008] According to the confidence distribution of each variable in each group of samples under different indicators, the state space matrix of each evaluation level under different indicators is constructed;
[0009] According to the state space matrix of each evaluation level under different indicators, the statistics of the state space of each evaluation level under different indicators are calculated and normalized to obtain the confidence distribution of different indicators;
[0010] According to the confidence distribution of different indicators, the evidence reasoning method is used to fuse multivariate information to obtain the confidence distribution of the evaluation results;
[0011] The health status of the laser strapdown inertial system is determined based on the confidence distribution of the evaluation results.
[0012] Optionally, the confidence distribution of the mth sample of the jth variable under the lth indicator is:
[0013]
[0014] When R j,n,l ≤x mj <R j,n+1,l hour:
[0015]
[0016]
[0017]
[0018] When x mj ≤R j,1,l hour:
[0019]
[0020]
[0021] in,
[0022] When R j,N,l ≤x mj hour:
[0023]
[0024]
[0025] Wherein, n=1, 2…, N-1, s1 is an integer greater than or equal to 1 and less than or equal to N, and s1≠n, n+1, s2 is an integer greater than or equal to 1 and less than or equal to N, s3 is an integer greater than or equal to 1 and less than N, and N is the number of evaluation levels;
[0026] x mj,l is the confidence distribution of the mth sample of the jth variable under the lth indicator, Know The mth sample of the jth variable under the lth indicator belongs to the evaluation level h1, evaluation level h2 and evaluation level h N confidence level;
[0027] R j,1,l , R j,n,l , R j,n+1,l , R j,N,l is the evaluation level h1 and evaluation level h of the jth variable under the lth indicator. n , Assessment level h n+1 and evaluation level h N The reference value, x mj is the mth sample of the jth variable, and The mth sample of the jth variable under the lth indicator belongs to the evaluation level h1 and the evaluation level h n , Assessment level h n+1 and evaluation level h N The confidence level, The mth sample of the jth variable under the lth indicator belongs to the evaluation level h s1 , Assessment level h s2 and evaluation level h s3 confidence level.
[0028] Optionally, according to the confidence distribution of each variable in each group of samples under different indicators, the state space matrix of each evaluation level under different indicators is constructed as follows:
[0029]
[0030] Among them, H n,l h is the evaluation level under the lth indicator n The state space matrix, Know The first, mth and Mth samples of the first variable under the lth indicator belong to the evaluation level h n The confidence level, Know The first, mth and Mth samples of the second variable under the lth indicator belong to the evaluation level h n The confidence level, Know The first, mth and Mth samples of the Jth variable under the lth indicator belong to the evaluation level h n , N is the number of evaluation levels, J is the number of variables, and M is the number of samples.
[0031] Optionally, the statistics of the state space of each evaluation level under different indicators are calculated and normalized to obtain the confidence distribution of different indicators, including:
[0032] The principal component analysis is performed on the state space matrix of each evaluation level under different indicators respectively to obtain the principal component matrix of each evaluation level under different indicators;
[0033] According to the principal component matrix of each evaluation level under different indicators, the statistics of the state space of each evaluation level under different indicators are calculated using the following formula;
[0034]
[0035] in, h is the evaluation level under the lth indicator n The statistics of the state space are h is the evaluation level under the lth indicator n The principal component matrix of for The covariance matrix of , the superscript T indicates transpose;
[0036] The statistics of the state space of each evaluation level under different indicators are normalized by using the following formula to obtain the normalized statistics of the state space of each evaluation level under different indicators;
[0037]
[0038] in, Indicates the evaluation level h under the lth indicator n A normalized statistic in the state space of Indicates the evaluation level h under the lth indicator n The control limits of the state space are is the first intermediate quantity, and Respectively The variance and mean of , α is the confidence level, The confidence level is α and the degrees of freedom are The chi-square distribution of is the second intermediate quantity,
[0039] According to the first normalized statistics of the state space of each evaluation level under different indicators, the first normalized statistics of the state space of each evaluation level under different indicators are respectively normalized twice to obtain the second normalized statistics of the state space of each evaluation level under different indicators; the second normalized statistics of the state space of all evaluation levels under different indicators constitute the confidence distribution of the different indicators.
[0040] Optionally, based on the confidence distribution of different indicators, an evidential reasoning method is used to perform multivariate information fusion to obtain the confidence distribution of the evaluation results, specifically including:
[0041] The confidence distribution of different indicators is weighted by using a relationship matrix to obtain the weighted confidence distribution of different indicators; the relationship matrix is a matrix that represents whether different indicators affect each other;
[0042] According to the weighted confidence distribution of different indicators, the following formula is used to determine the distribution probability of different indicators;
[0043]
[0044] in, The mth group of samples under the lth indicator belongs to the evaluation level h n The probability of distribution, The mth group of samples under the lth indicator belongs to the evaluation level h n The weighted confidence of , Θ is the set of evaluation levels, is an empty set;
[0045] According to the distribution probability of different indicators, the method based on evidence reasoning is iterated L-1 times to fuse L indicators, and the fusion results of L indicators are obtained to support the evaluation of the probability quality of each evaluation level; where L is the number of indicators;
[0046] According to the probability mass of each evaluation level supported by the fusion results of L indicators, the confidence of each evaluation level supported by the fusion results of L indicators is calculated to obtain the confidence distribution of the evaluation results.
[0047] Optionally, the health status of the laser strapdown inertial system is determined based on the confidence distribution of the evaluation result, including:
[0048] According to the Euclidean distances of the confidence distribution of the evaluation results and the standard state modes of each evaluation level in the standard state mode matrix, the evaluation level corresponding to the standard state mode with the smallest Euclidean distance to the confidence distribution of the evaluation results is determined as the health state of the laser strapdown inertial system.
[0049] The present invention provides a computer system, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above-mentioned method for determining the health status of a habit group based on interpretable statistical feature evidence reasoning.
[0050] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects:
[0051] The embodiment of the present invention provides a method and system for determining the health status of an inertial system based on interpretable statistical feature evidence reasoning, the method comprising: obtaining multiple groups of samples for determining the health status of a laser strapdown inertial system, and calculating the confidence distribution of each variable in each group of samples under different indicators; constructing a state space matrix of each evaluation level under different indicators according to the confidence distribution of each variable in each group of samples under different indicators; calculating the statistics of the state space of each evaluation level under different indicators according to the state space matrix of each evaluation level under different indicators, and normalizing to obtain the confidence distribution of different indicators; using the evidence reasoning method to perform multivariate information fusion according to the confidence distribution of different indicators to obtain the confidence distribution of the evaluation result; determining the health status of the laser strapdown inertial system according to the confidence distribution of the evaluation result. The present invention realizes the unified expression of information by calculating the confidence distribution of each variable in each group of samples under different indicators, constructs the statistics of the state space of each evaluation level under different indicators based on interpretable statistical theory, uses the interpretable evidence reasoning method to perform multivariate information fusion, determines the state evaluation result, and realizes the transparency of the evaluation process and the evaluation result. BRIEF DESCRIPTION OF THE DRAWINGS
[0052] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0053] Figure 1 A flowchart of a method for determining a health status of a habit group based on interpretable statistical feature evidence reasoning provided by an embodiment of the present invention.
[0054] Figure 2 A schematic diagram of a hierarchical group health status assessment indicator system provided in an embodiment of the present invention.
[0055] Figure 3 A health status change curve diagram of a single variable (accelerometer X-axis first-order calibration coefficient) provided in an embodiment of the present invention.
[0056] Figure 4A schematic diagram of comprehensive indicators of a gyroscope provided in an embodiment of the present invention.
[0057] Figure 5 A schematic diagram of the confidence distribution of the health status assessment results of the habit group provided in an embodiment of the present invention.
[0058] Figure 6 A schematic diagram of the correspondence between samples of the four indicators provided in an embodiment of the present invention and health patterns.
[0059] Figure 7 An internal structure diagram of a computer system provided in accordance with an embodiment of the present invention. DETAILED DESCRIPTION
[0060] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0061] The purpose of the present invention is to provide a method and system for determining the health status of a group based on interpretable statistical feature evidence reasoning, which can achieve transparency of the evaluation process and evaluation results.
[0062] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.
[0063] Example 1
[0064] Embodiment 1 of the present invention proposes a method for assessing the health status of a group based on interpretable statistical feature evidence reasoning, aiming to establish an interpretable health status assessment model, so that a model with clear parameter meanings and transparent reasoning process can be established in the forward direction, and the cause of the change in health status can be traced back based on the assessment results in the reverse direction. In the establishment of the model, not only test data is used, but also expert knowledge or mechanism information is incorporated. The main contents are: the data conversion process based on the reference value realizes the unified expression of information, the feature extraction based on the principal component analysis method realizes the compression of high-dimensional variables and the establishment of evidence, and the framework based on evidence reasoning realizes the fusion of irrelevant evidence. Figure 1 As shown, the method for determining the health status of a group based on interpretable statistical feature evidence reasoning in this embodiment includes the following steps 101 to 105.
[0065] Step 101, obtaining multiple groups of samples for determining the health status of the laser strapdown inertial system, and calculating the confidence distribution of each variable in each group of samples under different indicators.
[0066] In step 101 of the present invention, by establishing a corresponding relationship between the original variables and the health status assessment level, the problem of maintaining the physical meaning of the reference value and the assumption of evidence independence in the process of evidence establishment is solved.
[0067] In the embodiment of the present invention, the confidence distribution of each variable in each group of samples under different indicators is calculated by converting the variable data based on the reference value.
[0068] In this step, the data of each variable is converted based on the variable reference value to correspond to the final evaluation level. The reference value is determined based on the value range of the variable and has a clear and definite physical meaning. Assume that there are J variables and M groups of samples. The sample of the jth variable is: j ={x m,j |m=1,2,…,M},j=1,2,…,J,where m represents the mth sample. Under any indicator, the conversion relationship between data-variable reference value-assessment level is established as formula (1).
[0069]
[0070] Among them, f(x j ,R j ,h) represents the conversion relationship between data-variable reference value-assessment level, R j = {R j,k |k=1,2,…,K} represents the reference value of the jth variable, h={h n |n=1,2,…,N} represents the evaluation level of health status, β n,k (n=1,2,…,N;k=1,2,…,K) represents the variable x j Based on the reference value R j,k After conversion, relative to the evaluation level h n The confidence level indicates the degree to which the variable value belongs to this level, R j,k represents the kth reference value of the jth variable, h n represents the nth evaluation level, K represents the number of reference values, and N represents the number of evaluation levels. Experts can establish data conversion relationships based on different forms of variables, such as triangular membership function, Gaussian membership function, etc.
[0071] The data conversion process based on the reference value is shown in the following formula.
[0072]
[0073] Among them, R j,k+1 represents the k+1th reference value of the jth variable, β n+1,k Represents variable x j Based on the reference value R j,kAfter conversion, relative to the evaluation level h n+1 The confidence level, β s,k Represents variable x j Based on the reference value R j,k After conversion, relative to the evaluation level h s The confidence level, h n+1 、h s They represent the n+1th and sth evaluation levels respectively.
[0074] According to the aforementioned conversion relationship, the sample of the j-th variable is expressed in the form of a confidence distribution relative to the evaluation level as shown in formula (2).
[0075] x j ={(h1,β 1,k ),(h2,β 2,k ),…,(h N ,β N,k )} (2)
[0076] The data of other variables are also converted into confidence distribution. After data conversion, a one-to-one correspondence between the data of the variables and the health status assessment level is established. And the health status change of the variable can be preliminarily judged from the confidence distribution of a single variable.
[0077] Specifically, the confidence distribution of the mth sample of the jth variable under the lth indicator is:
[0078]
[0079] When R j,n,l ≤x mj <R j,n+1,l hour:
[0080]
[0081]
[0082]
[0083] When x mj ≤R j,1,l hour:
[0084]
[0085]
[0086] When R j,N,l ≤x mj hour:
[0087]
[0088]
[0089] Wherein, n=1, 2…, N-1, s1 is an integer greater than or equal to 1 and less than or equal to N, and s1≠n, n+1, s2 is an integer greater than 1 and less than or equal to N, s3 is an integer greater than or equal to 1 and less than N, and N is the number of evaluation levels; x mj,l is the confidence distribution of the mth sample of the jth variable under the lth indicator, and The mth sample of the jth variable under the lth indicator belongs to the evaluation level h1, evaluation level h2 and evaluation level h N Confidence level; R j,1,l , R j,n,l , R j,n+1,l , R j,N,l is the evaluation level h1 and evaluation level h of the jth variable under the lth indicator. n , Assessment level h n+1 and evaluation level h N The reference value, x mj is the mth sample of the jth variable, and The mth sample of the jth variable under the lth indicator belongs to the evaluation level h1 and the evaluation level h n , Assessment level h n+1 and evaluation level h N The confidence level, The mth sample of the jth variable under the lth indicator belongs to the evaluation level h s1 , Assessment level h s2 and evaluation level h s3 confidence level.
[0090] Step 102, constructing a state space matrix of each evaluation level under different indicators according to the confidence distribution of each variable in each group of samples under different indicators.
[0091] Because there are many variables in the system and they are correlated, feature transformation is needed to eliminate the correlation. In the state space composed of health status assessment levels, a feature transformation method based on principal component statistics-control limits is proposed to realize the confidence distribution form of converting correlated variables into uncorrelated features.
[0092] The confidence levels of all data confidence distributions of each variable at the same evaluation level are combined into a state space H = {H n |n=1,2,…,N}, as shown in formula (3).
[0093]
[0094] Among them, H n For the evaluation level h n The state space of Indicates that the mth sample of the jth variable belongs to the evaluation level h n The confidence level, j = 1, 2, ..., J, m = 1, 2, ..., M. The state space H corresponds to the health assessment level h, that is,
[0095] For example, construct the evaluation level h under the lth indicator n The state space matrix of is shown below.
[0096]
[0097] Among them, H n,l h is the evaluation level under the lth indicator n The state space matrix, and The first, mth and Mth samples of the first variable under the lth indicator belong to the evaluation level h n The confidence level, and The first, mth and Mth samples of the second variable under the lth indicator belong to the evaluation level h n The confidence level, and The first, mth and Mth samples of the Jth variable under the lth indicator belong to the evaluation level h n , N is the number of evaluation levels, J is the number of variables, and M is the number of samples.
[0098] Step 103, according to the state space matrix of each evaluation level under different indicators, calculate the statistics of the state space of each evaluation level under different indicators, and normalize them to obtain the confidence distribution of different indicators, which specifically includes the following steps.
[0099] Step A1, perform principal component analysis on the state space matrix of each evaluation level under different indicators respectively, and obtain the principal component matrix of each evaluation level under different indicators.
[0100] Perform principal component analysis on the variables in each state space to achieve dimensionality reduction and obtain comprehensive indicators. n The state space H n For example, find H n The covariance matrix of As well as the eigenvalue λ and eigenvector V n ; According to the eigenvalue contribution rate greater than 85%, the unrelated projection matrix Q is obtained n and the principal component matrix As shown in formula (4) to formula (6).
[0101]
[0102] The superscript T indicates transposition.
[0103] Step A2, according to the principal component matrix of each evaluation level under different indicators, calculate the statistics of the state space of each evaluation level under different indicators.
[0104] Calculate the state space H n Statistics As shown in formula (7), the state space H is realized n A synthesis of correlated variables.
[0105]
[0106] in, express The covariance matrix of .
[0107] For example, the evaluation level h under the lth indicator n The calculation formula of the state space statistics is:
[0108] in, h is the evaluation level under the lth indicator n The statistics of the state space are h is the evaluation level under the lth indicator n The principal component matrix of for The covariance matrix of , the superscript T indicates the transpose.
[0109] Step A3, normalizing the statistics of the state space of each evaluation level under different indicators to obtain the normalized statistics of the state space of each evaluation level under different indicators.
[0110] According to the control limit of the state space statistics, the index of the space is normalized to reflect the degree of index change in each state space, as shown in formula (8).
[0111]
[0112] in, Indicates the evaluation level h n A normalized statistic in the state space of Indicates the evaluation level h n The state space H n The control limits, is the first intermediate function, and Respectively represent statistics The variance and mean of The confidence level is α and the degrees of freedom are The chi-square distribution of , α is the confidence level, is the second intermediate function,
[0113] For example, the evaluation level h under the lth indicator n The calculation formula of the normalized statistics of the state space is: in, Indicates the evaluation level h under the lth indicator n A normalized statistic in the state space of Indicates the evaluation level h under the lth indicator n The control limits of the state space are is the first intermediate quantity, and Respectively The variance and mean of , α is the confidence level, The confidence level is α and the degrees of freedom are The chi-square distribution of is the second intermediate quantity,
[0114] Step A4, according to the first normalized statistics of the state space of each evaluation level under different indicators, perform second normalization on the first normalized statistics of the state space of each evaluation level under different indicators to obtain the second normalized statistics of the state space of each evaluation level under different indicators; the second normalized statistics of the state space of all evaluation levels under different indicators constitute the confidence distribution of the different indicators.
[0115] The indicators of N state spaces are normalized to form the confidence distribution of the comprehensive indicator, as shown in formula (9).
[0116]
[0117] Among them, F H is the confidence distribution of the comprehensive indicator, β n For the evaluation level h n The confidence level, and They are evaluation levels h1, h2, h N A normalized statistic in the state space of .
[0118] The extractable principal component features can be interpreted as follows: 1) extracting features in the state space corresponding to the evaluation level, realizing the integration of variables and reflecting the changes of indicators in the state space; 2) the comprehensive indicators composed of the extracted features are expressed in the form of confidence distribution, which still corresponds to the evaluation level one by one. Similarly, other relevant variables are also expressed in the form of confidence distribution.
[0119] Step 104, based on the confidence distribution of different indicators, adopt the evidential reasoning method to perform multivariate information fusion to obtain the confidence distribution of the evaluation result, which specifically includes the following steps.
[0120] Step B1, using a relationship matrix to weight the confidence distribution of different indicators to obtain weighted confidence distributions of different indicators; the relationship matrix is a matrix that characterizes whether different indicators affect each other.
[0121] Each comprehensive indicator in the form of confidence distribution is used as evidence to be integrated, and the weight is used as a parameter to measure the importance of evidence. It is determined by the relationship matrix of mechanism information, which can ensure the interpretability of the parameters.
[0122] According to the aircraft mechanism information, the relationship matrix I between indicators is established L×L , reflecting the mutual influence relationship between indicators as shown in formula (10).
[0123]
[0124] Index F p Can directly affect indicator F q When the relationship matrix I L×L Element i pq =1, otherwise it is recorded as 0. The mutual influence between indicators can reflect the hierarchical relationship of the system, and determine the weight of each piece of evidence in the fusion process. The weighted representation of evidence based on the relationship matrix of mechanism information is shown in Equation (11) and Equation (12).
[0125]
[0126] Among them, e1, e2, e L are the weighted representations of the first, second, and Lth indicators, respectively, β L×N×M represents the confidence matrix, represents the confidence matrix after weighted representation, Indicates that the mth sample under the lth indicator belongs to the evaluation level h n confidence level.
[0127] Step B2, determining the allocation probability of different indicators according to the weighted confidence distribution of different indicators.
[0128] The aircraft health status assessment mainly reflects the idea of hierarchical fusion. The fusion process is: the relevant variables within each indicator are first fused, then the indicators are fused, and finally a comprehensive evaluation is achieved. Through the operations of the first two steps, the premise assumptions of the application of the evidence reasoning method are strictly met, so fusion is performed.
[0129] The basic probability distribution of evidence m n,l It is generated by the weighted confidence distribution as shown in formula (13).
[0130]
[0131] in, Represents the weighted confidence, indicating that the support assessment is h n The degree of grade.
[0132] For example, the mth group of samples under the lth indicator belongs to the evaluation level h n The probability of allocation is: in, The mth group of samples under the lth indicator belongs to the evaluation level h n The probability of distribution, The mth group of samples under the lth indicator belongs to the evaluation level h n The weighted confidence of , Θ is the set of evaluation levels, Is an empty set.
[0133] Step B3, according to the allocation probability of different indicators, the L indicators are iterated L-1 times based on the evidential reasoning method to obtain the probability quality of each evaluation level supported by the fusion results of the L indicators; wherein L is the number of indicators.
[0134] The method based on evidence reasoning iterates L-1 times to fuse L pieces of evidence (each indicator is regarded as one piece of evidence during fusion), as shown in formula (14).
[0135]
[0136] Among them, m n,e(l) Indicates that the synthesis result of the first l pieces of evidence supports the assessment level h n The probability mass of h is the evaluation level of the first l pieces of evidence n The level of support, h is the evaluation level of the first l pieces of evidence D The support level, m B,e(l-1) h represents the support assessment level of the synthesis result of the first l-1 pieces of evidence B The probability mass, m C,l To indicate the level of evidence support assessment for Article 1 CThe probability mass of D, C and B are all subsets of the evaluation level set Θ, and B∩C=n represents the evaluation level h supported by two pieces of evidence. B and h C The intersection of n .
[0137] Step B4, according to the probability mass of the L indicator fusion results supporting the evaluation of each evaluation level, calculate the confidence of the L indicator fusion results supporting the evaluation of each evaluation level, and obtain the confidence distribution of the evaluation result.
[0138] The formula for calculating the confidence of each evaluation level supported by the fusion results of L indicators is shown in formula (15).
[0139]
[0140] Among them, P n,e(L) Indicates that the integration of L pieces of evidence supports the assessment level h n The confidence level, h is the evaluation level of the first L pieces of evidence n The level of support, h is the evaluation level of the first L pieces of evidence D level of support.
[0141] The confidence distribution A(x) of the evaluation result determined after the final fusion of multiple indicators is shown in the following formula:
[0142] A(x)={(h n ,P n,e(L) ),n=1,2,…,N}.
[0143] Step 105: determining the health status of the LSRMU according to the confidence distribution of the evaluation result.
[0144] According to the mechanism information and prior knowledge, the relationship between the health status of each indicator and the health status of the system is established, and the standard state mode matrix D = {D d |d=1,2,…,D}, including the state modes under different evaluation levels. It is used to judge the different state modes reflected by each indicator under the health status evaluation level according to the fusion results, realize the backtracking of the evaluation results, and reflect the interpretability. It also realizes the monitoring of the health status of the subsystem.
[0145]
[0146] Among them, d dl Elements in Indicates that it belongs to h n Assessment level, Indicates that it does not belong to any evaluation level, where d=1, 2,…, D, l=1, 2,…, L, n=1, 2,…, N.
[0147] The degree of closeness between the health status of each indicator and the standard status model is measured based on the Euclidean distance, as shown in formula (18).
[0148]
[0149] Among them, β m It represents the confidence that the mth test data of all L indicators belongs to each evaluation level. d Represents the dth mode in the standard state mode matrix D. β m and D d There are L×N elements in each.
[0150] The final health status assessment result relative to the judgment rule of the health status model is shown in formula (19).
[0151] μ D (β m )=min(dist(β m ,D d )) (19)
[0152] In the formula, μ D (β m ) indicates that the health status of each indicator belongs to μ D The state mode of the representative.
[0153] In order to verify the effectiveness of the proposed method, the laser inertial group is taken as an example to demonstrate the health status assessment process.
[0154] 1. Establishment of evaluation index system.
[0155] According to the test process, it is necessary to monitor the voltage, temperature, accelerometer and gyroscope. Establish a hierarchical inertial group health status assessment indicator system such as Figure 2 As shown in Figure 1, the health status assessment levels of the habit group are determined as follows: healthy (h1), sub-healthy (h2), and faulty (h3).
[0156] 2. Variable data conversion based on reference values.
[0157] In the present invention, the inertial group data is divided into two categories: state quantity and calibration quantity. The inertial group data has different units and orders of magnitude. In order to express the data uniformly and correspond to the evaluation level, the data is converted and expressed in the form of confidence distribution, that is, the conversion relationship of the data is given according to the mechanism information.
[0158] 2.1. State quantity, including variables in voltage monitoring and temperature monitoring. During the test, the state quantity records the maximum and minimum values within a period of time, and judges its health status based on the relationship between the extreme value and the judgment threshold. The data of the state quantity is recorded as I within the rated range, and O when it exceeds the range. The conversion relationship of the state quantity is shown in Table 1.
[0159] Table 1 Conversion relationship of inertia group state quantity
[0160] Health status I O <![CDATA[Health (h1)]]> 1 0 <![CDATA[Sub-health (h2)]]> 0 0 <![CDATA[Fault (h3)]]> 0 1
[0161] 2.2. Calibration, including 10 calibration coefficients of gyroscope and 10 calibration coefficients of accelerometer. Gyroscope calibration coefficients require c g ≤C g (C g =2.7σ g ), where c g Indicates the gyroscope calibration coefficient, C g represents the gyroscope calibration coefficient threshold, σ g is the standard deviation of the gyroscope calibration coefficient, and the accelerometer calibration coefficient requires c a ≤C a (C a =2.7σ a ), where c a is the accelerometer calibration coefficient, C a is the accelerometer calibration coefficient threshold, σ a is the standard deviation of the accelerometer calibration coefficient, and the conversion relationship of the calibration coefficient is shown in Table 2.
[0162] Table 2 Conversion relationship of inertial group calibration quantity
[0163] Health status c≤C(C=2.7σ) c>C(C=2.7σ) <![CDATA[Health (h1)]]> 1-c / C 0 <![CDATA[Sub-health (h2)]]> c / C 0 <![CDATA[Fault (h3)]]> 0 1
[0164] In Table 2, c represents the calibration coefficient, C represents the calibration coefficient threshold, and σ represents the calibration coefficient standard deviation.
[0165] After the data conversion matrix, a one-to-one correspondence between the original variables of the inertial group and the health status assessment level of the inertial group is achieved. And from the confidence distribution of a single variable, we can preliminarily judge the change in health status it reflects. Taking the first-order calibration coefficient of the accelerometer X-axis as an example, the health status change curve after data conversion is as follows: Figure 3 As shown in Figure 2, the first 60 samples reflect a healthy state, and then they change to a sub-healthy state.
[0166] 3. Independent evidence acquisition based on feature extraction.
[0167] Gyroscopes and accelerometers are multi-meter redundant configurations, and their calibration coefficients are correlated, so a step-by-step fusion method is used to obtain the indicators of the health status of gyroscopes and accelerometers. The process of obtaining comprehensive indicators is demonstrated using the gyroscope as an example. Similarly, the determination of the comprehensive indicators of the accelerometer can be achieved.
[0168] Firstly, the feature conversion method based on principal component is applied to extract the features of the five variables of the zero-order term and the five variables of the first-order term of the gyroscope calibration coefficient, respectively, to form the indicators of the zero-order term and the first-order term.
[0169] Secondly, according to the mechanism information, the two indicators of zero-order term and first-order term are independent of each other and do not affect each other. The relationship matrix I1 of the two indicators is established as shown in formula (20).
[0170]
[0171] Finally, the method based on evidence reasoning fuses the zero-order and first-order indicators of the gyroscope to determine the comprehensive indicator of the gyroscope, which is expressed in the form of confidence distribution, such as Figure 4 As shown. Figure 4 It can be seen that the comprehensive indicators of the gyroscope reflect that the first 30 groups of samples are in a healthy state with a high confidence level. As the test time increases, the healthy state slowly decreases. The last 10 groups of samples reflect that the gyroscope is in a sub-healthy state.
[0172] 4. Multi-index information fusion based on evidence reasoning method.
[0173] According to the inertial control unit health assessment index system, the health status of the inertial control unit is mainly affected by four indicators. Through the second and third steps, the comprehensive indicators of the gyroscope and accelerometer can be obtained. Because voltage monitoring and temperature monitoring are both state quantities represented by 0 or 1, they can be directly integrated and judged to be completely healthy. After obtaining the four comprehensive indicators of voltage, temperature, gyroscope, and accelerometer, these four comprehensive indicators are integrated based on the method of evidence reasoning to obtain the evaluation result of the health status of the inertial control unit.
[0174] Based on the mechanism information, the relationship matrix I2 of these four indicators is established as shown in formula (21).
[0175]
[0176] From the relationship matrix, we can see that voltage monitoring has a higher priority and will directly affect the other three indicators, while temperature monitoring is second and directly affects the gyroscope and accelerometer indicators. After integrating the comprehensive indicators of these four parts, the evaluation results are as follows: Figure 5 As shown. Figure 5It can be seen that the habitual group was in a healthy state during the test period. After the 80th sample, the health status declined and the sub-health status increased, and it was in a sub-healthy state with a relatively high confidence.
[0177] 5. Review of evaluation results.
[0178] Based on the mechanism information and expert knowledge, a standard state mode matrix is established to determine which specific monitoring indicator changes caused the change in the health status of the inertial group. The standard state mode is shown in Table 3.
[0179] Table 3 Standard status mode
[0180]
[0181]
[0182] Among them, D1 represents healthy mode, D2~D4 represent sub-healthy mode, D5~D 15 Represents the fault mode. Written in matrix form as shown in formula (22).
[0183]
[0184] The samples of the four indicators are matched to the standard state pattern, such as Figure 6 As shown. Figure 6 It can be seen that, except for a few samples that were unstable and misjudged, the first 50 samples were basically in a healthy mode. After 50 samples, the inertial group entered a sub-healthy state, matching the sub-healthy mode represented by D3. It was judged that this was caused by the sub-healthy state of the test data of a few gyroscopes. After more than 60 samples, the sub-healthy mode represented by D4 was matched. It can be seen that the test data of both the gyroscope and the accelerometer entered a sub-healthy state. The tracing of the evaluation results has achieved the Figure 5 Explanation of the evaluation results.
[0185] Example 2
[0186] Embodiment 2 of the present invention provides a computer system, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the method in Embodiment 1.
[0187] The internal structure diagram of the computer device can be shown as Figure 7As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. Among them, the processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store pending transactions. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, the method in Example 1 is implemented.
[0188] It should be noted that the object information (including but not limited to object device information, object personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in the present invention are all information and data authorized by the object or fully authorized by all parties.
[0189] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided by the present invention can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided by the present invention may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited thereto. The processor involved in each embodiment provided by the present invention may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but are not limited thereto.
[0190] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.
[0191] The principles and implementation methods of the present invention are described in this article using specific examples. The description of the above embodiments is only used to help understand the method and core idea of the present invention. At the same time, for those skilled in the art, according to the idea of the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting the present invention.
Claims
1. A method for determining the health status of a group of people based on interpretable statistical feature evidence reasoning, characterized in that: The method comprises: Obtain multiple groups of samples for determining the health status of the laser strapdown inertial system, and calculate the confidence distribution of each variable in each group of samples under different indicators; According to the confidence distribution of each variable in each group of samples under different indicators, the state space matrix of each evaluation level under different indicators is constructed; According to the state space matrix of each evaluation level under different indicators, the statistics of the state space of each evaluation level under different indicators are calculated and normalized to obtain the confidence distribution of different indicators; According to the confidence distribution of different indicators, the evidence reasoning method is used to fuse multivariate information to obtain the confidence distribution of the evaluation results; The health status of the laser strapdown inertial system is determined based on the confidence distribution of the evaluation results.
2. The method for determining the health status of a group based on interpretable statistical feature evidence reasoning according to claim 1, characterized in that: The confidence distribution of the mth sample of the jth variable under the lth indicator is: When R j,n,l ≤x mj <R j,n+1,l hour: When x mj ≤R j,1,l hour: in, When R j,N,l ≤x mj hour: Wherein, n=1, 2…, N-1, s1 is an integer greater than or equal to 1 and less than or equal to N, and s1≠n, n+1, s2 is an integer greater than 1 and less than or equal to N, s3 is an integer greater than or equal to 1 and less than N, and N is the number of evaluation levels; x mj,l is the confidence distribution of the mth sample of the jth variable under the lth indicator, and The mth sample of the jth variable under the lth indicator belongs to the evaluation level h1, evaluation level h2 and evaluation level h N confidence level; R j,1,l , R j,n,l , R j,n+1,l , R j,N,l is the evaluation level h1 and evaluation level h of the jth variable under the lth indicator. n , Assessment level h n+1 and evaluation level h N The reference value, x mj is the mth sample of the jth variable, and The mth sample of the jth variable under the lth indicator belongs to the evaluation level h1 and the evaluation level h n , Assessment level h n+1 and evaluation level h N The confidence level, The mth sample of the jth variable under the lth indicator belongs to the evaluation level h s1 , Assessment level h s2 and evaluation level h s3 confidence level.
3. The method for determining the health status of a group based on interpretable statistical feature evidence reasoning according to claim 1, characterized in that: According to the confidence distribution of each variable in each group of samples under different indicators, the state space matrix of each evaluation level under different indicators is constructed as follows: Among them, H n,l h is the evaluation level under the lth indicator n The state space matrix, and The first, mth and Mth samples of the first variable under the lth indicator belong to the evaluation level h n The confidence level, and The first, mth and Mth samples of the second variable under the lth indicator belong to the evaluation level h n The confidence level, and The first, mth and Mth samples of the Jth variable under the lth indicator belong to the evaluation level h n , N is the number of evaluation levels, J is the number of variables, and M is the number of samples.
4. The method for determining the health status of a group based on interpretable statistical feature evidence reasoning according to claim 1, characterized in that: The statistics of the state space of each evaluation level under different indicators are calculated and normalized to obtain the confidence distribution of different indicators, including: The principal component analysis is performed on the state space matrix of each evaluation level under different indicators respectively to obtain the principal component matrix of each evaluation level under different indicators; According to the principal component matrix of each evaluation level under different indicators, the statistics of the state space of each evaluation level under different indicators are calculated using the following formula; in, h is the evaluation level under the lth indicator n The statistics of the state space are h is the evaluation level under the lth indicator n The principal component matrix of for The covariance matrix of , the superscript T indicates transpose; The statistics of the state space of each evaluation level under different indicators are normalized by using the following formula to obtain the normalized statistics of the state space of each evaluation level under different indicators; in, Indicates the evaluation level h under the lth indicator n A normalized statistic in the state space of Indicates the evaluation level h under the lth indicator n The control limits of the state space are is the first intermediate quantity, and Respectively The variance and mean of , α is the confidence level, The confidence level is α and the degrees of freedom are The chi-square distribution of is the second intermediate quantity, According to the first normalized statistics of the state space of each evaluation level under different indicators, the first normalized statistics of the state space of each evaluation level under different indicators are respectively normalized twice to obtain the second normalized statistics of the state space of each evaluation level under different indicators; the second normalized statistics of the state space of all evaluation levels under different indicators constitute the confidence distribution of the different indicators.
5. The method for determining the health status of a group based on interpretable statistical feature evidence reasoning according to claim 1, characterized in that: According to the confidence distribution of different indicators, the evidence reasoning method is used to fuse multivariate information to obtain the confidence distribution of the evaluation results, including: The confidence distribution of different indicators is weighted by using a relationship matrix to obtain the weighted confidence distribution of different indicators; the relationship matrix is a matrix that represents whether different indicators affect each other; According to the weighted confidence distribution of different indicators, the following formula is used to determine the distribution probability of different indicators; in, The mth group of samples under the lth indicator belongs to the evaluation level h n The probability of distribution, The mth group of samples under the lth indicator belongs to the evaluation level h n The weighted confidence of , Θ is the set of evaluation levels, is an empty set; According to the distribution probability of different indicators, the method based on evidence reasoning is iterated L-1 times to fuse L indicators, and the fusion results of L indicators are obtained to support the evaluation of the probability quality of each evaluation level; where L is the number of indicators; According to the probability mass of each evaluation level supported by the fusion results of L indicators, the confidence of each evaluation level supported by the fusion results of L indicators is calculated to obtain the confidence distribution of the evaluation results.
6. The method for determining the health status of a group based on interpretable statistical feature evidence reasoning according to claim 1, characterized in that: According to the confidence distribution of the evaluation results, the health status of the laser strapdown inertial system is determined, including: According to the Euclidean distances of the confidence distribution of the evaluation results and the standard state modes of each evaluation level in the standard state mode matrix, the evaluation level corresponding to the standard state mode with the smallest Euclidean distance to the confidence distribution of the evaluation results is determined as the health state of the laser strapdown inertial system.
7. A computer system comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the method for determining the health status of a habit group based on interpretable statistical feature evidence reasoning as described in any one of claims 1 to 6.
