Equipment health assessment method and device based on multi-dimensional data fusion and dynamic weight

Abstract

The embodiment of the invention discloses an equipment health assessment method and device based on multi-dimensional data fusion and dynamic weight, and relates to an intelligent monitoring and health assessment technology of an equipment complex system in the field of equipment health management. The method comprises the following steps: describing the change condition of health parameters along with the increase of equipment storage life from the angles of the regression trend of the health parameters along with time, the incidence relation among different health parameters and the probability distribution of health parameter data, and accurately reflecting the difference between health parameter monitoring data and baseline data. A dynamic weight updating method in the whole storage process is provided, and the weights of different health parameters in the process of calculating the health indexes are scientifically calculated. Therefore, the problem that health parameter monitoring data are scarce due to the fact that the equipment cannot be frequently powered on for testing in the actual storage and use process is effectively solved, and the assessment accuracy of the health state of the equipment is improved under the condition that the monitoring data are limited.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of equipment health management, and particularly relates to an equipment health evaluation method and device based on multi-dimensional data fusion and dynamic weight. BACKGROUND

[0002] The performance degradation and reliability guarantee of the complex system of modern equipment directly affect the use efficiency and life cycle cost of the equipment in the long-term storage and use process. In order to ensure the reliable operation of the system, prolong the service life and realize the accurate maintenance, the fault prediction and health management (Prognostics and Health Management, PHM) technology has become the core means to improve the reliability, safety and reduce the operation and maintenance cost of the system. It should be noted that health evaluation is a core concept in PHM, and its essence is the depth analysis and decision based on state monitoring data - monitoring is the basis of "data acquisition and state perception" (such as obtaining vibration, voltage, resistance and other parameters), and evaluation is the key of "state quantification and trend judgment" (such as calculating health index, judging degradation stage and predicting residual life through multi-dimensional data fusion). Health state evaluation, as a core link of PHM technology, identifies the system degradation trend through real-time monitoring of multi-source performance parameters, provides decision basis for condition-based maintenance (CBM) system, and supports the task priority sorting and active support optimization of equipment. In the face of the challenge of multi-level coupled failure of complex systems, a high-reliability health evaluation method is of great significance to ensure the availability of equipment and efficient use of resources. The health evaluation method based on machine learning has become an important research direction in the field of PHM through data-driven modeling and adaptive feature extraction. Although such methods can quantitatively analyze the health status of equipment based on performance parameters and output individual or batch health grades, they still face many significant bottlenecks in practical application.

[0003] Therefore, in further research and development, the health evaluation method based on multi-dimensional data fusion has gradually attracted attention. This method integrates multi-source data throughout the life cycle of the equipment, builds a cross-level and cross-dimensional performance degradation analysis framework, realizes multi-feature collaborative mining and dynamic fusion. Compared with single parameter or model-driven evaluation methods, multi-dimensional data fusion technology can more comprehensively capture the coupled failure mechanism of complex systems, enhance the integrity of health characterization, and improve the characterization ability of the model to nonlinear degradation patterns through multi-source feature correlation analysis and adaptive weight allocation, providing technical support for accurate health quantification and active maintenance decision of complex systems. Although the multi-dimensional data fusion method has shown theoretical advantages, existing health evaluation methods have significant shortcomings in feature parameter completeness, feature fusion mechanism and dynamic weight optimization, which restrict the refinement level of complex system health management.

[0004] The main cause of the current main constraint is that the complex equipment system contains electronic, mechanical, kinetic energy and other subsystems, and there is functional coupling between the subsystems, which belongs to a "multi-domain nonlinear coupled system". However, a more optimized equipment health evaluation scheme often requires more abundant monitoring data, but the more complex the system, the less health parameter monitoring data can be obtained during actual storage and use due to the inability to frequently power on test, resulting in a lack of data samples, which makes it impossible to test each batch of equipment, and thus the difference between the health parameter monitoring data and the baseline data cannot be accurately reflected, making it difficult to improve the accuracy of equipment health state evaluation. SUMMARY

[0005] Embodiments of the present application provide an equipment health evaluation method and device based on multi-dimensional data fusion and dynamic weight, which can improve the accuracy of equipment health state evaluation under limited monitoring data.

[0006] To achieve the above-mentioned purpose, the embodiments of the present application adopt the following technical solutions: In a first aspect, an equipment health evaluation method based on multi-dimensional data fusion and dynamic weight is designed, comprising: S1, establishing an equipment health baseline model for the health characteristic parameters of the equipment; S2, dynamically adjusting the health indicators output by the equipment health baseline model based on weight, and then obtaining comprehensive health indicators of the equipment; S3, determining the health state of the equipment according to the health indicator failure determination condition.

[0007] In this embodiment, S1 includes: collecting key performance monitoring parameters of the equipment, and extracting health characteristic parameters representing the degradation trend of the equipment. The types of health characteristic parameters include: system characteristic voltage, system characteristic current, resistance and system power of the equipment. In actual application, the influence of equipment function and structure and its failure mode needs to be studied, including electrical systems, mechanical control systems, kinetic control systems, etc., to locate the key performance monitoring parameters. Analyze the equipment task profile and the existing state monitoring parameter system, study the data acquisition and storage method, and extract health characteristic parameters representing the degradation trend of the equipment from three aspects of electronic products, mechanical control products and kinetic control products, including equipment system characteristic voltage, characteristic current, resistance, power, etc. According to the health characteristic parameters and the nominal data and storage initial state data of the target equipment when it leaves the factory, an equipment health baseline model is established. The parameter types in the equipment health baseline model include Mahalanobis distance, mean absolute error and probability density distribution coincidence percentage.

[0008] The Mahalanobis distance index is used to represent the overall deviation between the health characteristic parameters and the standard matrix of the health baseline model, and the value of the Mahalanobis distance index is negatively correlated with the health state of the equipment; the Mahalanobis distance index of the i-th health characteristic parameter is represented as , represents the standard matrix of the health baseline model corresponding to the i-th parameter, represents the i-th sample data, and S represents the covariance matrix; the sample set includes p test samples, and each test sample contains m health characteristic parameters; is recorded as a matrix with a size of p x m; The mean absolute error index is used to measure the deviation between the actual value of the health characteristic parameter and the predicted value of the health baseline model, and the value of the mean absolute error index is negatively correlated with the health state of the equipment; the mean absolute error index of the i-th health characteristic parameter is represented as , represents the true value, represents the predicted value.

[0009] The probability density distribution coincidence percentage index is used to evaluate the similarity of the health characteristic parameters and the standard sample distribution shape, and the value of the probability density distribution coincidence percentage index is positively correlated with the health state of the equipment. The probability density distribution coincidence percentage index of the i-th health characteristic parameter is represented as , is the probability density value of the health characteristic parameter in the k-th interval, is the probability density value of the standard sample in the k-th interval, is the interval width of the probability density distribution, and n is the total number of intervals of the probability density distribution.

[0010] In this embodiment, S2 includes: adaptively updating the weights of the health indicators by using the CRITIC method and the entropy weight method; and generating a comprehensive health indicator according to the updated weights and the health indicators. For example: the time and space characteristics of the health evaluation parameter data sample can be expressed by three health indicators: mean absolute error, Mahalanobis distance, and probability density coincidence percentage. The weight calculation method is studied to dynamically adjust the health indicator fusion weights of different categories of parameters. The standard deviation and the correlation coefficient of the CRITIC method are used to reflect the variability and conflictivity of the data, respectively. The information entropy of the entropy weight method is used to reflect the degree of disorder of the data. The health indicators of each category of parameters and the weight calculation results are combined to construct an equipment comprehensive health indicator, which provides an index basis for equipment state evaluation and sorting.

[0011] The adaptive updating of the weights of the health indicators by using the CRITIC method and the entropy weight method includes: calculating the initial weights , , and respectively represent the weight of the i-th parameter based on the baseline data of the first year when calculating the health index by CRITIC method and entropy weight method respectively; wherein the health baseline model is calculated based on the nominal data and the initial storage state data when the equipment is shipped, and the baseline data is the general term of the nominal data and the initial storage state data. That is, the health baseline model is calculated based on the baseline data. The nominal data and the initial storage state data when the equipment is shipped focus on the production and application of the equipment, while the baseline data is a term in the method and model. After the parameter data is expanded, iterative calculation is performed again, and the parameter weight after the k-th iteration is: .

[0012] For example: assuming that there are p samples to be tested and q evaluation indexes, the original index data matrix is formed: , wherein represents the value of the j-th evaluation index of the i-th sample.

[0013] The process of calculating the objective weight by CRITIC weight method includes: dimensionless processing: in order to eliminate the influence of different dimensions on the evaluation results, CRITIC weight method generally uses positive or negative processing to perform dimensionless processing on each index.

[0014] If the greater the value of the used index, the better (positive index): , x max and x min represent the upper limit value and the lower limit value; if the smaller the value of the used index, the better (negative index, the relationship with the aforementioned positive index is or; essentially, the positive or negative of the index is or, so the index still uses to represent), represents the i-th evaluation index: j .

[0015] The index variability is expressed in the form of standard deviation: , represents the standard deviation of the j-th index, which is used to represent the difference and fluctuation of the internal value of each index. The greater the standard deviation, the greater the difference in the value of the index, the more information it can reflect, and the stronger the evaluation strength of the index itself, so the index should be allocated more weight, represents the average value of the j-th evaluation index, j represents the dimensionless processing result of the j-th evaluation index of the i-th sample.

[0016] The index conflict is represented by the correlation coefficient: ​​, This represents the correlation coefficient between evaluation indicators i and j. The correlation coefficient is used to represent the correlation between indicators. The stronger the correlation with other indicators, the less conflict there is between that indicator and the other indicators, the more identical information it reflects, and the more repetitive the evaluation content it conveys. This weakens the evaluation strength of that indicator to some extent, and the weight assigned to that indicator should be reduced.

[0017] Information content , The larger the value, the greater the role of the j-th evaluation indicator in the entire evaluation indicator system, and therefore the more weight should be assigned to it.

[0018] Objective weight: The objective weight of the j-th indicator is: .

[0019] The process of determining the weights of each indicator using the entropy weight method includes: Data standardization: Given m samples to be evaluated and n evaluation indicators, form a standardized original data matrix. , ,in Let be the evaluation value of the i-th sample data under the j-th indicator. Calculate the weight (also known as the prior probability) of the indicator value of the i-th sample under the j-th indicator. Therefore, a weighting matrix of the data can be established. .

[0020] Calculate the information entropy of each indicator: Calculate the entropy weight of the j-th indicator. , P ij This represents the contribution (probability value) of the information entropy of the j-th indicator for the i-th sample, a constant. Information utility value .

[0021] The information utility value of a certain indicator depends on the information entropy of that indicator (the j-th indicator). The difference between 1 and 0 directly affects the weight. The greater the information utility value, the greater its importance to the evaluation, and the greater its weight.

[0022] Determine the weight of each indicator: Estimate the weight of each indicator using the entropy weight method. Essentially, this involves calculating the weight using the value coefficient of the indicator's information. The higher the value coefficient, the greater its importance to the evaluation (or the greater the weight, the greater its contribution to the evaluation result). Calculate the entropy weight of the i-th indicator. .

[0023] Determine the comprehensive weight of the indicators Assume the evaluator determines the weight of the indicators' importance based on their own objectives and requirements. Combining the entropy weight of the indicator The comprehensive weight of index j can be obtained When the values of each alternative project on index j are exactly the same, the entropy of the index reaches the maximum value 1, and the entropy weight is zero. This shows that the index fails to provide useful information to the decision maker, i.e. all alternative projects are indistinguishable to the decision maker under the index, and the index can be considered to be removed. Therefore, the entropy weight itself is not an importance coefficient of the index, but indicates the degree of differentiation of the evaluation object under the index.

[0024] Further, the comprehensive health index of the equipment is obtained by: H i using the dynamically updated weights w i performing weighted fusion to obtain a preliminary comprehensive health index H param : H i defined as the average of the m sub-health indexes corresponding to the parameter, i.e. using the equal weight method to fuse: wherein, H i,j represents the jth sub-health index of the ith health characteristic parameter, m is the number of sub-health indexes, and n represents the number of health characteristic parameters; the initial model of the comprehensive health index of the equipment is: H including Mahalanobis distance index, mean absolute error index and probability density distribution overlap percentage index, so m=3.

[0025] The complete expansion of the comprehensive health index is obtained by substituting the specific calculation formula of the weight and the sub-health index, i.e. the final obtained comprehensive health index of the equipment H is equal to .

[0026] ​​S3 comprises: screening the equipment with the best health status according to the obtained equipment comprehensive health index; obtaining the health index failure determination threshold and delimiting the failure determination baseline after normalizing the comprehensive health index calculation result of the equipment with the best health status; evaluating the health status of all the equipment to be analyzed by using the failure determination baseline, and then sorting the equipment according to the health status according to the actual service years of the equipment. After calculating the comprehensive health index of the equipment, the equipment with the best health status is selected, the comprehensive health index calculation result thereof is normalized, the health index failure determination threshold is determined, and the failure determination baseline is delimited. On this basis, the health status of all the equipment is evaluated by using the failure determination baseline; the health evaluation ability of the method is evaluated by using the accuracy and other indicators. Finally, the equipment is sorted according to the health status according to the actual service years of the equipment, thereby providing a reference for equipment selection.

[0027] In a second aspect, an equipment health evaluation device based on multi-dimensional data fusion and dynamic weight is designed, comprising: A model maintenance module is configured to establish an equipment health baseline model for health characteristic parameters of the equipment. An analysis module is configured to perform dynamic adjustment on the health index output by the equipment health baseline model based on weight, and then obtain an equipment comprehensive health index. A monitoring module is configured to determine the health status of the equipment according to a health index failure determination condition.

[0028] The model maintenance module is specifically configured to collect key performance monitoring parameters of the equipment, and extract health characteristic parameters representing the degradation trend of the equipment. The types of the health characteristic parameters include system characteristic voltage, system characteristic current, resistance and system power of the equipment. An equipment health baseline model is established according to the extracted health characteristic parameters and the nominal data and the storage initial state data of the target equipment at the time of leaving the factory. The equipment health baseline model is used to analyze the health index, and the health index includes Mahalanobis distance index, mean absolute error index and probability density distribution coincidence percentage index.

[0029] In the embodiment, the average absolute error, Mahalanobis distance and probability density coincidence percentage are used to describe the change of health parameters with the increase of equipment storage time from the regression trend of health parameters over time, the correlation between different health parameters and the probability distribution of health parameter data, and accurately reflect the difference between health parameter monitoring data and baseline data. The CRITIC method is used to reflect the variability and conflict between the data at the beginning of equipment storage, and the entropy weight method is used to reflect the degree of disorder of the data at the middle and late stages of equipment storage. The advantages of the two weight calculation methods are combined to propose a storage whole process weight dynamic updating method, which scientifically calculates the weight of different health parameters in calculating the health index. Thus, the problem of lack of health parameter monitoring data due to the inability to frequently test the equipment during actual storage and use, and the real bottleneck of small data samples caused by the inability to test batch equipment one by one, is effectively solved, and a comprehensive health index is designed to accurately evaluate the health status of the equipment. Thus, the accuracy of the evaluation of the health status of the equipment is improved under the condition of limited monitoring data. BRIEF DESCRIPTION OF DRAWINGS

[0030] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative effort.

[0031] Figure 1 The method flowchart provided by the embodiments of the present application is provided. Figure 2 The comparison chart of the average absolute error of equipment 1 (normal) and equipment 11 (degradation failure) in the specific example provided by the embodiments of the present application with the change of storage time is shown in the following figure. Figure 3 The comparison chart of the Mahalanobis distance of equipment 1 and equipment 11 in the specific example provided by the embodiments of the present application with the change of storage time is shown in the following figure. Figure 4 The comparison chart of the probability density coincidence percentage of equipment 1 and equipment 11 in the specific example provided by the embodiments of the present application with the change of storage time is shown in the following figure. Figure 5 The schematic diagram of the weight of three types of health parameters of equipment 11 in the specific example provided by the embodiments of the present application with the change of storage time is shown in the following figure. Figure 6 The schematic diagram of the average absolute error of the equipment in the specific example provided by the embodiments of the present application with the change of storage time is shown in the following figure. Figure 7 The schematic diagram of the Mahalanobis distance of the equipment in the specific example provided by the embodiments of the present application with the change of storage time is shown in the following figure. Figure 8 Fig. 4 is a schematic diagram of the probability density overlap percentage of the equipment in the specific example provided by the embodiment of the present application as a function of the storage period; Figure 9 Fig. 5 is a schematic diagram of the comprehensive health index of the failed equipment in the specific example provided by the embodiment of the present application as a function of the storage period; Figure 10 Fig. 6 is a schematic diagram of the normalized result of the comprehensive health index of the failed equipment in the specific example provided by the embodiment of the present application as a function of the storage period. DETAILED DESCRIPTION

[0032] To enable persons skilled in the art to better understand the technical solutions of the present application, the present application will be further described in detail below in combination with the drawings and specific embodiments. In the following, the embodiments of the present application will be described in detail, and the examples of the embodiments are shown in the drawings, in which the same or similar reference numerals represent the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the drawings are exemplary and are only used to explain the present application, and cannot be interpreted as a limitation on the present application. Those skilled in the art can understand that, unless specifically stated, the singular forms "a", "an" and "the" used herein can also include the plural forms. It should be further understood that the phrase "comprising" used in the specification of the present application means that the features, integers, steps, operations, elements and / or components exist, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or groups thereof. It should be understood that when we say that an element is "connected" or "coupled" to another element, it can be directly connected or coupled to the other element, or there can be intermediate elements. In addition, "connected" or "coupled" used herein can include wireless connection or coupling. The phrase "and / or" used herein includes any one of the associated listed items and all combinations of the associated listed items. Those skilled in the art can understand that, unless otherwise defined, all terms (including technical terms and scientific terms) used herein have the same meaning as that generally understood by those skilled in the art in the field to which the present application belongs. It should also be understood that terms such as those defined in a general dictionary should be understood to have meanings consistent with those in the context of the prior art, and should not be interpreted with idealized or overly formal meanings unless defined as such.

[0033] The design idea of the embodiment scheme is to construct a health evaluation system covering the whole life cycle of the system through multi-source parameter collaborative screening, nonlinear feature fusion modeling, dimension processing and dynamic weight optimization driven by physical-data combination, so as to provide general technical support for equipment task planning and maintenance support, promote the transformation of health state evaluation from qualitative analysis to quantitative evaluation, and finally realize the comprehensive goal of improving equipment use efficiency, optimizing operation and maintenance cost and guaranteeing whole life cycle reliability. The equipment complex system health evaluation method based on multi-dimensional data aims to construct a system health evaluation model and realize sorting based on health state. The technical route is as shown in Figure 1 According to the execution logic, it can be divided into three links: complex system state monitoring parameter system and extraction analysis, health index construction based on regression trend and probability density distribution, and state evaluation and sorting based on health index. Specifically: I. Complex system state monitoring parameter system and extraction analysis 1.1. Research on equipment system function and structure and its failure mode influence, and positioning of key performance monitoring parameters. The key performance monitoring parameters are the parameters directly reflecting the health state of the subsystems for the three types of subsystems of electronic products, mechanical control products and kinetic control products of the equipment, which are determined through function-failure analysis. It should be noted that positioning the key performance parameters and key monitoring parameters refers to the structured method based on system engineering analysis, and the core means is "function-structure-failure" three-in-one analysis, and the specific steps include "function and structure analysis", "failure mode influence analysis (FMEA)" and "key parameter screening". The key performance parameters, key monitoring parameters and the like mentioned in this embodiment can be combined and referred to as "key performance monitoring parameters", which are the original physical quantities directly measured or monitored by sensors, instruments and the like during the operation of the equipment for the performance of key equipment and systems, and are the data source for health evaluation; the equipment health baseline model refers to the systematized model for establishing the "normal health state reference benchmark" by using the algorithm in this paper based on the nominal data and storage initial state data of the target equipment at the time of leaving the factory, which is used to quantitatively describe the characteristic mode or health index range of the equipment in the normal operating state, and provide objective numerical reference for judging whether the equipment has degradation, abnormality or failure.

[0034] 1.2. Analyze the equipment task profile and the existing state monitoring parameter system. The existing equipment state monitoring parameter system mainly includes electronic products, mechanical control products and kinetic control products. The key health evaluation parameters of the electronic products of the equipment are determined as: system power bus voltage and communication unit working current, the key monitoring health parameters of the mechanical control products are inertial measurement unit acceleration and actuator displacement, and the key monitoring health parameters of the kinetic control products are power device ignition loop resistance, control system power supply loop resistance and auxiliary system power supply loop resistance.

[0035] 1.3, Collecting equipment storage initial state data, building equipment health baseline model. According to the selected key monitoring health parameters in 1.2, collect the nominal data and storage initial state data of the equipment when it leaves the factory, and jointly build the equipment health baseline model. Among them, the manufacturer will provide the nominal data of each subsystem and key components and equipment when designing the equipment, and the storage initial state data refers to the test data when the equipment is put into the warehouse for storage for a short time (the user will generally detect the equipment regularly), both of which are easy to obtain.

[0036] II. Health index construction based on regression trend and probability density distribution: capture the dynamic change rule of parameter characteristics and reveal the time-varying characteristics of health state. The Mahalanobis distance, mean absolute error, and probability density distribution overlap percentage are used to calculate the equipment degradation, and the equipment health index is jointly measured.

[0037] 2.1, Calculate 3 types of health indexes for electronic products, kinetic control products and mechanical control products respectively: mean absolute error, Mahalanobis distance and probability density overlap percentage. Randomly select several equipment that have failed during the storage period, and calculate the average absolute error, Mahalanobis distance and probability density overlap percentage of each type of health parameter for electronic products, kinetic control products and mechanical control products. For each selected equipment, a total of 9 health index calculation results are obtained.

[0038] 2.2, Calculate the weight of electronic products, kinetic control products and mechanical control products when each type of health index is fused, including the initial weight and the self-adaptive dynamic update value changing with the storage period. Under different environmental and operating conditions, the importance of equipment health characteristic parameters is different. Therefore, dynamic weight adjustment is crucial to accurately describe the change of component operating state. Combined with CRITIC method and entropy weight method, the self-adaptive updating algorithm of health index weight is studied to cope with the dynamic change of multiple health index weights during equipment storage period and reflect the relative importance of different indexes. It should be noted that the skilled person in the art will adopt different calls according to his own work habits, such as: "key (performance) monitoring parameter" refers to the original physical quantity directly measured or monitored by sensors, instruments, etc. during the operation of the equipment, which is the data source for health assessment; "health characteristic parameter" refers to the derived feature sensitive to health status extracted from the key monitoring parameter, which is the information carrier for health assessment; "health index" is a comprehensive evaluation quantity obtained by fusing health characteristic parameters, which is the decision basis for health assessment; there is a progressive relationship among the three, the key monitoring parameter is essentially data, the health characteristic parameter is essentially information, and the health index is essentially a kind of knowledge that can assist maintenance and decision-making after being processed by algorithm model.

[0039] The CRITIC method relies on the data itself, fully utilizes the operating environment and state information contained therein, and determines the weight according to the variability and conflict of the indicators. The variability is measured by the standard deviation, and the conflict is measured by the correlation between indicators. The correlation between indicators is reflected. The entropy weight method is based on the concept of information entropy, which measures the uncertainty or disorder degree in the system; the greater the information quantity, the higher the mutual relationship between parameters, and the entropy weight method determines the weight according to the data dispersion degree of each parameter. Combining the two methods can alleviate their respective limitations, so that the weight distribution is more scientific. The specific steps of the multi-index weight self-adaptive updating process based on the CRITIC method and the entropy weight method are as follows: Initial weight calculation: after completing the baseline data collection, the CRITIC method and the entropy weight method are used to calculate the weight of each health characteristic parameter, and the average value is taken as the initial weight , , and respectively represent the weight of the ith parameter calculated by the CRITIC method and the entropy weight method using the baseline data of the first year.

[0040] Extension of parameter data: in the health assessment stage, the actual monitoring data of each health characteristic parameter is recorded each time, and is added to the parameter data set in order for weight calculation.

[0041] Dynamic weight updating: after the extension of parameter data, the CRITIC method and the entropy weight method are used to recalculate the weight, and the average value is taken as the dynamically adjusted weight. The parameter weight calculation after the first iteration: .

[0042] Iteration cycle: repeat the extension of parameter data and dynamic weight updating, continuously incorporate new parameter running data into the weight calculation data set, and ensure the continuous updating of parameter data set and the self-adaptive updating of weight.

[0043] The CRITIC method and the entropy weight method are used to dynamically update the weight of the health characteristic parameter. Through the dynamic integration of new running data, the iterative updating of the weight is realized, so as to realize the self-adaptive updating of the weight, and accurately reflect the changes of the equipment health status. This self-adaptive weight updating mechanism can ensure that the health assessment model always reflects the representation ability of each parameter to the equipment health status, so as to provide more comprehensive and reliable evaluation.

[0044] 2.3, calculate the 3 types of health indicators of the equipment. Multiply the calculation results of the 9 health indicators of each equipment by the corresponding weight, and for each selected equipment, obtain the calculation results of the 3 types of health indicators.

[0045] III. State assessment and ranking method based on health indicators: Two health indicators (Mahalanobis distance and mean absolute error) of the regression trend model and a single health indicator (percentage of overlap) of the probability density distribution model are calculated. The mean of the weights of the three health indicators in the equipment health assessment is combined to obtain the comprehensive health indicator of the equipment H The state of the equipment is assessed and ranked by the comprehensive indicator.

[0046] 3.1. Calculate the comprehensive health indicator of the equipment. According to the time of failure of each equipment, the comprehensive health indicator of the equipment at the starting time of failure is calculated. The health status of the equipment and the comprehensive health indicator value are negatively correlated, so the smaller the comprehensive health indicator, the better the health status of the equipment. Therefore, the minimum value of the comprehensive health indicator of all the equipment used for training at the starting time of failure is taken.

[0047] 3.2. Determine the failure judgment threshold of the health indicator. The failure judgment threshold of the health indicator is determined by comparing the health status of several pieces of equipment participating in the test. The first data point of the equipment with the smallest comprehensive health indicator is taken as the starting point, and the value of the equipment with the smallest comprehensive health indicator at the starting time of failure is taken as the endpoint for data normalization. On this basis, through engineering practice, the failure judgment margin is defined as 5%, that is, the failure judgment threshold of the health indicator is 0.95, and a unified equipment failure judgment baseline is drawn.

[0048] 3.3, Equipment health state evaluation. The comprehensive health index of the equipment used for testing (including all health states of the equipment reaching the specified storage life) is calculated, and the data samples (including health samples and failure samples) contained in all equipment are evaluated according to the health index failure determination threshold. The "health index failure determination threshold" refers to a specific quantitative value (0.95 in this paper), which is the critical value for determining whether the equipment health state is "failure". The essence is to normalize the health index data of the equipment participating in the test, and combine the failure determination standard set by engineering practice. The "equipment failure determination baseline" refers to the unified evaluation baseline established based on the failure determination threshold, which is the common reference framework for all equipment participating in the evaluation. The essence is to map the health index data of different equipment and different use environments to a unified evaluation scale through normalization from "starting point" to "end point" and threshold setting with "5% margin", forming a baseline of "0 (complete failure) ~ 1 (complete health)", and 0.95 is the "failure critical point" on this line. In this scheme, the threshold is determined by the health index of a few equipment combined with the engineering practice, and the health index failure determination threshold 0.95 is the specific critical value in the baseline, which is used for failure judgment of a single equipment; the equipment failure determination baseline is a unified evaluation scale, which is used to eliminate differences and realize horizontal comparison of multiple equipment, and the emphasis of the two is different. Simply put, the baseline is a "ruler", and the threshold is a "scale line" on the ruler - through the ruler (baseline) to unify the measurement standard, and through the scale line (threshold) to clearly define the failure limit.

[0049] After that, the health evaluation ability of the equipment failure determination baseline is evaluated by four indexes: accuracy, precision, recall, and F1 score. For example: the equipment failure determination baseline refers to the 0-1 interval obtained by "taking the first data point of the equipment with the smallest comprehensive health index as the starting point, and taking the value of the equipment with the smallest comprehensive health index at the failure starting time as the end point for data normalization" in "determination of health index failure determination threshold". In order to realize the health evaluation of equipment, the formulas of the four indexes of accuracy, precision, recall, and F1 score are not attached. According to the content in the title row of Table 6, the above four indexes are calculated based on "actual normal determination normal sample number", "actual normal determination failure sample number", "actual failure determination normal sample number", and "actual failure determination failure sample number". Among them, the smaller the number of "actual normal determination failure sample number" and "actual failure determination normal sample number", the better; the smaller the two values, the larger the value of the four indexes, that is, the more accurate the health evaluation, and the evaluation result is more in line with the actual situation.

[0050] Assume that TP (True Positive) is a true example, that is, the number of samples correctly predicted as positive; TN (True Negative) is a true negative example, that is, the number of samples correctly predicted as negative; FP (False Positive) is a false positive example, that is, the number of negative samples incorrectly predicted as positive; and FN (False Negative) is a false negative example, that is, the number of positive samples incorrectly predicted as negative. In the health assessment process in this paper, the positive class refers to the normal state of equipment, and the negative class refers to the failure of equipment. True example TP refers to the actual normal state of equipment and the health assessment determines that it is normal; true negative TN refers to the actual failure of equipment and the health assessment determines that it fails; false positive FP refers to the actual failure of equipment but the health assessment determines that it is normal; and false negative FN refers to the actual normal state of equipment but the health assessment determines that it fails.

[0051] Accuracy (Accuracy) refers to the proportion of correctly classified samples to the total number of samples, which reflects the classification ability of the classifier to the entire sample set. The formula of accuracy is: .

[0052] Precision (Precision) refers to the proportion of true positive samples in the samples classified as positive, which measures the accuracy of the classifier in predicting positive samples, and is also called "precision". The formula of precision is: .

[0053] Recall (Recall) is also called "recall rate" and refers to the proportion of samples correctly predicted as positive in all true positive samples, which reflects the recall ability of the classifier to positive samples, and is also called "recall rate". The formula of recall is: .

[0054] F1 score (F1-score) is the harmonic mean of precision and recall, which considers precision and recall comprehensively and can more comprehensively evaluate the performance of the classifier. The formula of F1 score is: .

[0055] The health assessment ability is evaluated by the four evaluation indexes of accuracy, precision, recall and F1 score.

[0056] 3.4 Equipment health state sorting. According to the actual use requirements of equipment, four time limits of 5 years, 10 years, 15 years and 20 years for health state sorting are defined. In the four time limits, all equipment is sorted according to the comprehensive health index value, and the best equipment is selected to provide reference for equipment use In this embodiment, to solve the problem of lack of health parameter monitoring data due to infrequent power-on testing during actual storage and use, and the problem of lack of data samples due to the inability to test each batch of equipment, a comprehensive health index is designed to accurately evaluate the health status of equipment. Using the mean absolute error, Mahalanobis distance, and probability density overlap percentage, the changes in health parameters with the increase of equipment storage time are described from the perspectives of the regression trend of health parameters over time, the correlation between different health parameters, and the probability distribution of health parameter data, accurately reflecting the differences between health parameter monitoring data and baseline data. The CRITIC method is used to reflect the variability and conflict between the initial data of equipment storage, and the entropy weight method is used to reflect the degree of disorder of the data in the middle and later stages of equipment storage. By combining the advantages of the two weight calculation methods, a dynamic weight updating method for the whole storage process is proposed to scientifically calculate the weight of different health parameters in calculating the health index.

[0057] Specific experimental cases are provided below for analysis: To verify the effectiveness of the health evaluation method for complex equipment systems based on multi-dimensional data fusion and dynamic weights proposed in this paper, simulated monitoring data of a batch of equipment are used to evaluate the health status of equipment in normal state, degradation failure, sudden failure, and degradation-sudden failure, and obtain the health status ranking, providing a reliable basis for equipment maintenance decision-making. In addition, performance comparison experiments, including ablation experiments and controlled variable experiments, are conducted to verify the superiority and robustness of the method. The computer hardware configuration used in the experiment is as follows: CPU is AMD Ryzen 7 5800X, running memory size is 32GB, and hard disk capacity is 1TB. The health evaluation method proposed in this paper is implemented using Python programming language based on Visual Studio Code code compilation software, and the Python version is 3.9.19.

[0058] Application scenario: By fusing multi-dimensional health evaluation parameter data, longitudinal analysis of equipment development, production, testing, and use life cycle data can be conducted to capture early performance degradation characteristics and slow cumulative degradation patterns that cannot be identified by traditional single-parameter threshold health evaluation methods, and to effectively identify equipment that has failed although it has passed the test. In the equipment that has passed the test, the health index is quantified as an accurate numerical value and a health status level (such as healthy, sub-healthy, medium, and warning) is established. Based on the quantification results, the health status evaluation and ranking are conducted to determine the equipment with the best health status at the current time, providing a scientific basis for equipment selection.

[0059] Parameter selection and data update: This method has the characteristics of universality, and does not depend on specific health assessment parameters. Different monitoring parameters can be selected to build a health evaluation model according to the actual needs of users. The health baseline model is built by using the health parameter nominal data and early detection data of the equipment at the time of delivery. The early health of the equipment is evaluated by comprehensive health indicators. In Table 1, the nominal data obtained at the time of delivery of the equipment is taken as the first detection, and in Table 2, sample No. 1 is taken as the health baseline sample of the equipment. The current health status of the equipment is quantified by calculating the comprehensive health indicators through continuous construction of health evaluation samples. As the use time and storage period of the equipment increase, the health evaluation model can be dynamically updated, so as to build more sensitive and generalizable health evaluation indicators and achieve more accurate health status evaluation.

[0060] Table 1 Equipment health parameter monitoring data

[0061] Table 2 Equipment health evaluation samples and comprehensive health indicators

[0062] Data description: The analysis object of this experimental case is a batch of equipment, a total of 28 pieces, with a normal storage period of 20 years. Among them, in order to simulate the actual situation, the batch of equipment is divided into 10 normal equipment (No. 1-10), 9 equipment with degradation failure (No. 11-19), 7 equipment with sudden failure (No. 20-26), and 2 equipment with degradation-sudden failure (No. 27-28). The health parameters used in the case are 7 key parameters, including system power bus voltage, communication unit working current, power device ignition loop resistance, control system power supply loop resistance, auxiliary system power supply loop resistance, inertial measurement unit acceleration and actuator displacement. In addition, according to the test cycle of the actual equipment, the monitoring data of the 7 key health parameters is obtained every half year. Therefore, all the data simulated in this paper is shown in the appendix, and the health parameter monitoring data of a normal equipment is shown in Table 3.

[0063] Table 3 Health parameter monitoring data of a normal equipment

[0064] The product nominal data of the equipment is taken as the standard data, which is defined as the basis for establishing the health evaluation model. When the health parameter monitoring data of the component drifts beyond the specified value, it is determined that the equipment has degradation failure. When the health parameter monitoring data of the component suddenly changes to 0 or exceeds 1 times the standard data, it is determined that the equipment has sudden failure. If the equipment has both degradation failure and sudden failure, it is determined that the equipment has degradation-sudden failure. Therefore, combined with this data, the degradation failure and sudden failure threshold values of the batch of equipment are shown in Table 4.

[0065] Table 4 Degradation failure and sudden failure threshold of key parameters of equipment system

[0066] Health parameter system and health baseline model construction: define the nominal data of 7 health parameters of equipment products as the standard data at the 0th year, the data format is 7x1, combine with the data obtained by the first test after the equipment enters the storage period, and jointly construct the health baseline model of the equipment product. The data format of the health baseline model is 7x2. Taking the health parameter monitoring data of this batch of equipment as an example, the sample data of the health baseline model is shown in Table 5.

[0067] Table 5 Sample data of health baseline model of this batch of equipment

[0068] According to the data format of the health baseline model, data samples are sequentially constructed (sample 1 is composed of data of the 0th year and the 0.5th year, sample 2 is composed of data of the 0.5th year and the 1st year, and so on...), then the number of samples of each equipment is 40, and 28 pieces of equipment construct 1120 data samples. Among them, the number of health data samples is 794, and the number of failure data samples is 326.

[0069] According to the demand of realizing equipment health evaluation under small sample condition in engineering practice, all data samples are divided into training data and test data according to small sample ratio, 120 data samples of 3 failure equipments (equipment 11, equipment 12 and equipment 13 in appendix) are selected as training samples, and 1000 samples of the remaining 25 equipments are selected as test samples. Among the 120 training samples, there are 63 health data samples and 57 failure data samples.

[0070] Health index construction: the health index for evaluating the health status of equipment system is constructed according to the following steps: (1) Calculate the average absolute error, Mahalanobis distance and probability density coincidence percentage of 3 types of health indexes for electronic product parameters, kinetic control product parameters and mechanical control product parameters respectively.

[0071] The average absolute error, Mahalanobis distance and probability density coincidence percentage of the selected electronic product parameters, kinetic control product parameters and mechanical control product parameters of the 3 failure equipments are calculated respectively. The change of the average absolute error, Mahalanobis distance and probability density coincidence percentage calculation results of equipment 11 with storage life and normal equipment 1 are compared, which shows the effectiveness of the construction of 3 types of health indexes for measuring each health parameter of the equipment. The comparison results are shown in Figure 2 , Figure 3 and Figure 4 respectively.

[0072] (2) Calculate the weight of the electronic product parameters, kinetic control product parameters and mechanical control product parameters in the fusion of each health index, including the initial weight and the adaptive dynamic update value with the change of storage time.

[0073] The weight of the selected electronic product parameters, kinetic control product parameters and mechanical control product parameters 3 health parameters of the failed equipment 11 in the fusion of health indicators changes with the storage time as shown in Figure 5 Figure 5 It can be seen from that the dynamic changes of the weights calculated by CRITIC method and entropy weight method have different characteristics. Specifically: 1) The weight change and update of CRITIC method mainly occur in the early stage of storage time, and the weight gradually tends to be stable with the passage of storage time. This is because CRITIC method mainly considers the contrast intensity and conflict between indexes when calculating the weight. In the early stage, the sample data is less, and the contrast intensity and conflict between indexes are more obvious, resulting in a large change in weight in the early stage. With the increase of storage time, the data gradually becomes rich, but CRITIC method has determined the general trend of weight according to the characteristics of initial data in the early stage, so the change in the later stage is relatively small. 2) Entropy weight method gives similar weights to each health parameter in the calculation of the initial stage, and the change and update of the weight mainly occur in the middle and late stages. This is because entropy weight method is based on information entropy to determine the weight. At the beginning, because the data is less, the calculation result of information entropy is close, so each health parameter is given similar weight. With the increase of storage time, the amount of data gradually increases, and the difference of information entropy of different indexes gradually appears, resulting in a more obvious change of weight in the middle and late stages. 3) The comprehensive weight method can combine the advantages of the two to dynamically adjust the parameter weight at different time periods of equipment storage, so as to more comprehensively and objectively reflect the objective law of the change of each health evaluation parameter of the equipment in the whole storage period.

[0074] (3) Calculate the three types of health indexes of the equipment.

[0075] Figure 6 Figure 7 Figure 8 The calculation results of the three types of health indexes of the selected three failed equipment are shown in

[0076] From Figure 6 ​​​It can be seen that the mean absolute error of equipment 11, 12, and 13 all show an upward trend as the storage period increases. When the storage period is 0 years, all health parameters are the nominal data at the time of manufacture, so the mean absolute error is the smallest. As the storage period increases, the health parameters gradually degrade, so the gap between the data sample and the baseline sample gradually increases.

[0077] Depend on Figure 7 It can be seen that the Mahalanobis distances of equipment 11, 12, and 13 also show an increasing trend with the increase of storage years. The Mahalanobis distance is the smallest when the storage years are 0. As time goes by, the health parameters gradually deteriorate, and the similarity between the data samples and the baseline samples decreases. Furthermore, unlike the calculation results of the mean absolute error index—the error of equipment 13 is greater than that of equipment 12, which is greater than that of equipment 11—the calculation results of the Mahalanobis distances of the three pieces of equipment show that the similarity of the data samples of equipment 11 is less than that of equipment 12, which is less than that of equipment 13.

[0078] Depend on Figure 8 It can be seen that as the storage years increase, the probability density overlap percentage of equipment 11, 12, and 13 shows a downward trend (the indicators in the figure have been normalized, so they show an upward trend). In the early stage of storage, the probability density overlap percentage is relatively high. As time goes by, the degree of overlap of the relevant probability distributions gradually decreases, that is, the difference in distribution between the data sample and the baseline sample gradually increases.

[0079] The equipment health status can then be evaluated and ranked according to the following steps: 1. Calculate the comprehensive health index of the equipment: Calculate the comprehensive health index of the three selected failed pieces of equipment, and observe how it changes with the storage years. Figure 9 As shown. By Figure 9It can be seen that the power supply loop resistance of the control system and the working current of the communication unit of equipment 11 begin to degrade and fail at the 11th year and the 13.5th year, respectively, and the comprehensive health index at the time of failure is 0.294, and the comprehensive health index at the 20th year of storage is 0.368. The power supply loop resistance of equipment 12 begins to degrade and fail at the 11.5th year, and the comprehensive health index at the time of failure is 0.301. The equipment and equipment 11 have different amounts of degraded parameters, but the overall parameter drift amount is similar to that of equipment 11, so the comprehensive health index at the 20th year of storage is similar to that of equipment 11, which is 0.369. The system power bus voltage and the control system power supply loop resistance of equipment 13 begin to degrade and fail at the 10.5th year and the 11th year, respectively, and the comprehensive health index at the time of failure is 0.298. Compared with equipment 11 and 12, the overall parameter drift amount of this equipment is larger, so the comprehensive health index at the 20th year of storage is the largest, which is 0.382. The above analysis shows that the comprehensive health index can accurately track and objectively reflect the changes of the equipment health evaluation parameters.

[0080] 2. Determine the failure determination threshold of the health index: Determine the failure determination threshold of the health index by comparing the health states of the three equipment. It is known that the health state of the equipment is negatively correlated with the comprehensive health index value, so among the three equipment, the health state of equipment 11 is the best. Take the comprehensive health index 0 of equipment 11 as the starting point, and take the comprehensive health index 0.294 of equipment 11 at the starting year of failure (the 11th year) as the endpoint to normalize the data, and the normalized result of the change of the comprehensive health index of the three failed equipment with the storage time is shown in the figure. On this basis, take 5% as the margin, and the failure determination threshold of the health index is 0.95.

[0081] 3. Health state evaluation: Calculate the comprehensive health index of 25 pieces of equipment used for testing (including 1-10 pieces of equipment without failure, 14-19 pieces of equipment with degradation failure, 20-26 pieces of equipment with sudden failure, and 27 and 28 pieces of equipment with degradation-sudden failure), and evaluate 1000 samples contained in the 25 equipment according to the failure determination threshold of the health index.

[0082] The results of the comprehensive health index calculation and sample determination of all 25 test equipment at the arrival of 20 years of storage life are shown in Table 6. The accuracy, precision, recall and F1 score are used as evaluation indexes to evaluate the ability of the proposed method for equipment product health state evaluation. Different colors in Table 6 are used to distinguish the state of the equipment, from top to bottom, they are normal, degradation failure, sudden failure and degradation-sudden failure. As can be seen from Table 6, among the 1000 samples of the 25 failed equipment, the actual normal samples and the samples determined as normal by threshold are 686, the actual normal samples and the samples determined as failure by threshold are 45, the actual failure samples and the samples determined as normal by threshold are 10, and the actual failure samples and the samples determined as failure by threshold are 278. The calculation of the determination accuracy, precision, recall and F1 score of the health index determination threshold is 94.5%, 98.6%, 93.8% and 96.1% respectively.

[0083] Table 6 Comprehensive health index calculation and sample determination results of 25 test equipment

[0084] Health state ranking: according to the calculation results of the comprehensive health index of equipment, the health state of all 28 equipment is ranked, which provides a reasonable basis for the scientific development of equipment selection. The order of the health state of the 28 equipment is shown in Table 7.

[0085] Table 7 Health state ranking results of 28 equipment

[0086] In the 28 pieces of equipment participating in the health assessment in Table 7, the top 5 equipment in the order from low to high according to the comprehensive health index at the 5th year of the storage period are equipment 24, equipment 3, equipment 10, equipment 7 and equipment 6, wherein the health state of the XX equipment 24 is the best, and the comprehensive health index is 0.3583. At the 10th year of the storage period, the top 5 equipment in the order from low to high according to the comprehensive health index are equipment 24, equipment 10, equipment 3, equipment 7 and equipment 6, wherein the health state of the XX equipment 24 is the best, and the comprehensive health index is 0.5304. At the 15th year of the storage period, the top 5 equipment in the order from low to high according to the comprehensive health index are equipment 3, equipment 10, equipment 4, equipment 1 and equipment 5, wherein the health state of the XX equipment 3 is the best, and the comprehensive health index is 0.8030. When the storage period reaches the specified period of 20 years, the comprehensive health index of equipment 8 is the lowest, so it is determined that the health state is the best. At this time, the equipment ranked 1-10 corresponds to 10 pieces of equipment in normal state without failure, and the equipment ranked 11-28 corresponds to the equipment with failure, which shows that the proposed health assessment method can correctly distinguish the equipment in different states, and the health index calculation result has reference value for the actual use of the equipment.

[0087] In summary, the case study of 28 pieces of equipment in different health states shows that the proposed health assessment method determines the failure determination threshold by calculating the comprehensive health index of 3 pieces of failed equipment and evaluates the health state of the remaining 25 pieces of equipment. The accuracy, precision, recall rate and F1 score reach 94.5%, 98.6%, 93.8% and 96.1% respectively, verifying the accuracy of the health assessment ability of the method under small sample.

[0088] Each of the embodiments in the specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other. Each embodiment focuses on the difference from other embodiments. In particular, the device embodiment is described simply because it is basically similar to the method embodiment, and the relevant parts can be referred to the part of the method embodiment. The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any changes or replacements within the technical range disclosed by the present application can be easily thought by those skilled in the art, which should be covered in the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A method for equipment health assessment based on multi-dimensional data fusion and dynamic weights, characterized in that, The method comprises the following steps: S1, establishing an equipment health baseline model for health characteristic parameters of equipment; S2, performing dynamic adjustment on health indexes output by the equipment health baseline model based on weights, and then obtaining an equipment comprehensive health index; S3, determining a health state of the equipment according to a health index failure determination condition.

2. The method of claim 1, wherein, S1 comprises the following steps: Collecting key performance monitoring parameters of the equipment, and extracting health characteristic parameters representing a degradation trend of the equipment, wherein the types of the health characteristic parameters include system characteristic voltage, system characteristic current, resistance and system power of the equipment; Establishing an equipment health baseline model according to the extracted health characteristic parameters and nominal data and storage initial state data of the target equipment when the target equipment is shipped, wherein the equipment health baseline model is used to analyze health indexes, and the health indexes include Mahalanobis distance indexes, mean absolute error indexes and probability density distribution coincidence percentage indexes.

3. The method of claim 2, wherein, The Mahalanobis distance indexes are used to represent an overall deviation degree between the health characteristic parameters and a standard matrix of the health baseline model, and the values of the Mahalanobis distance indexes are negatively correlated with the health state of the equipment; The mean absolute error indexes are used to measure a deviation between actual values of the health characteristic parameters and predicted values of the health baseline model, and the values of the mean absolute error indexes are negatively correlated with the health state of the equipment; The probability density distribution coincidence percentage indexes are used to evaluate the similarity between the health characteristic parameters and a standard sample distribution shape, and the values of the probability density distribution coincidence percentage indexes are positively correlated with the health state of the equipment.

4. The method according to claim 2 or 3, characterized in that, The Mahalanobis distance index of the i-th health characteristic parameter is represented as: , , represents the standard matrix of the health baseline model corresponding to the i-th parameter, represents the i-th sample data, and S represents the covariance matrix; the mean absolute error index of the i-th health characteristic parameter is represented as , represents the true value, represents the predicted value; The probability density distribution coincidence percentage index of the first health characteristic parameter is expressed as wherein, is the probability density value of the health characteristic parameter in the kth interval, is the probability density value of the standard sample in the kth interval, is the interval width of the probability density distribution, and n is the total number of intervals of the probability density distribution.

5. The method of claim 4, wherein S2 The method comprises the following steps: Adaptively updating the weights of the health indexes by using the CRITIC method and the entropy weight method; Generating a comprehensive health index according to the updated weights and the health indexes.

6. The method of claim 5, wherein, The self-adaptive updating of the weight of the health index by the CRITIC method and the entropy weight method comprises: calculating the initial weight of the health index , , and respectively represent the weight of the i-th parameter when the health index is calculated by the CRITIC method and the entropy weight method respectively based on the baseline data of the first year. After the parameter data is expanded, the parameter weights after the kth iteration are calculated again. , and are the weights calculated by CRITIC method and entropy weight method for the ith health feature parameter at the kth iteration, respectively.

7. The method of claim 6, wherein, The process of generating the comprehensive health index comprises: weighting and fusing the parameter health indexes of the i th health characteristic parameter to obtain a preliminary comprehensive health index , using the dynamically updated adaptive weights to perform weighting and fusion, to obtain a preliminary comprehensive health index : , , wherein represents the j th sub-health index of the i th health characteristic parameter, m is the number of sub-health indexes, and n represents the number of health characteristic parameters; and the finally obtained equipment comprehensive health index H is equal to 。 8. The method of claim 1, wherein S3 The method comprises the following steps: Screening the equipment with the best health state according to the obtained equipment comprehensive health index; Determining a health index failure determination threshold and a failure determination baseline after normalizing the comprehensive health index of the equipment with the best health state; Evaluating the health state of the equipment to be analyzed by using the failure determination baseline, and then sorting the equipment to be analyzed according to the health state according to the actual service years of the equipment.

9. An apparatus health assessment device based on multi-dimensional data fusion and dynamic weights, characterized in that, The method comprises the following steps: A model maintenance module is configured to establish an equipment health baseline model for health characteristic parameters of equipment; An analysis module is configured to perform dynamic adjustment on health indexes output by the equipment health baseline model based on weights, and then obtain an equipment comprehensive health index; A monitoring module is configured to determine a health state of the equipment according to a health index failure determination condition.

10. The apparatus of claim 9, wherein, The model maintenance module is specifically configured to collect key performance monitoring parameters of the equipment, and extract health characteristic parameters representing a degradation trend of the equipment, wherein the types of the health characteristic parameters include system characteristic voltage, system characteristic current, resistance and system power of the equipment; and establish an equipment health baseline model according to the extracted health characteristic parameters and nominal data and storage initial state data of the target equipment when the target equipment is shipped, wherein the equipment health baseline model is used to analyze health indexes, and the health indexes include Mahalanobis distance indexes, mean absolute error indexes and probability density distribution coincidence percentage indexes.

Images