System-level health state evaluation method and device driven by multi-component state fusion

Abstract

The application relates to a kind of multi-component state fusion driven system-level health state evaluation method and equipment, the method is used to carry out health state evaluation to the equipment system containing multi-component, comprising the following steps: collecting real-time data set, the real-time data set includes preselected characteristic parameter for representing equipment state change;Real-time data set is preprocessed by component, and the sample to be evaluated of each component is obtained;Based on the normal sample set obtained in advance, the deviation of the sample to be evaluated of each component is calculated, the health degree of the corresponding component is obtained, and the health degree sequence of each component in a period of evaluation time is obtained;The health degree sequence of multiple components is fused based on entropy weight method, and the current final equipment system comprehensive health degree is obtained.Compared with the prior art, the application has the advantages of accuracy, reliability, quantification and the like.

Description

Technical Field

[0001] This invention relates to the field of equipment health status management, and in particular to a system-level health status assessment method and device driven by multi-component status fusion. Background Technology

[0002] In the industrial sector, both mechanical and electronic systems require high reliability. Health status assessment is the foundation and prerequisite for condition-based equipment maintenance, and the key to assessment is establishing a suitable model of the equipment system. Based on the assessment level, it can be divided into component-level health status assessment and system-level health status assessment.

[0003] (I) Component-level health status assessment

[0004] Health status assessment algorithms can be categorized into three types based on their driving mechanism: model-based, expert experience-based, and data-based. Currently, data-based health status assessment methods are more widely used. Generally speaking, health status assessment analysis methods can be divided into qualitative analysis and quantitative analysis.

[0005] Qualitative analysis typically categorizes equipment health status into different levels, such as normal, degraded, or failed states. Commonly used algorithms include DBN (Database-Based Nomenclature), grey relational analysis, Analytic Hierarchy Process (AHP), Data Structures (DS) evidence theory, and fuzzy logic. For example, Ma et al. divided the equipment's run-to-failure cycle into five time intervals: 0-20%, 20%-40%, 40%-60%, 60%-80%, and 80%-100%, representing five levels of equipment status from high to low. They then trained a DBN model to establish a health status assessment classification model for qualitative analysis of equipment health status. However, qualitative analysis groups similar states into the same set, often providing only a rough evaluation of equipment status. When multiple devices are classified into the same status level, qualitative analysis methods cannot determine which device is in better or worse health, making it difficult for relevant personnel to prioritize equipment maintenance operations. Therefore, further quantitative description of equipment health status is needed, i.e., quantitative analysis.

[0006] Quantitative analysis quantifies equipment health status by establishing continuous-valued health indicators (such as health score) that describe equipment performance degradation. Common methods include Hidden Markov Models (HMMs) and Support Vector Data Descriptions (SVDDs). For example, Jiang et al. applied HMMs to bearing performance degradation assessment, using the probability that the current bearing condition belongs to a normal state as a performance indicator to achieve quantitative analysis and assessment of the bearing's health status. Compared to qualitative analysis, quantitative analysis describes the equipment's health level with continuous values, resulting in more explicit condition evaluation results.

[0007] (II) System-level health status assessment

[0008] Equipment systems typically consist of multiple components that interact and influence each other. When one component experiences performance degradation or failure, it often affects the health and even lifespan of other components in the system. Condition monitoring research should extend from the component level to the system level, exploring the knowledge of overall system performance degradation. Simultaneously, system-level health status assessments should reflect the performance change trends at the component level. Therefore, this invention, based on the performance assessment of a single component, considers the interactions between components during equipment degradation, assessing the overall health status at the system level. Modern systems often possess topological complexity, making system-level health status analysis a challenging task. A feasible approach is to define the health status of the component with the worst health or most critical function as the overall system health status. For example, Zhao et al. pointed out that system-level reliability can be decomposed into component-level reliability. Therefore, the key components of the entire system can be identified first, and the system-level reliability can be determined by assessing the reliability of these key components. The degradation process of key components can be viewed as a continuous-time Markov chain, and component reliability can be assessed based on service life and degradation status.

[0009] However, most current research focuses on component-level health status assessment, while research on multi-component equipment systems is relatively limited, making it difficult to accurately assess equipment health status and lacking quantitative status assessment methods. Summary of the Invention

[0010] The purpose of this invention is to overcome the shortcomings of the prior art by providing an accurate, reliable, and quantitative system-level health status assessment method and device driven by multi-component state fusion.

[0011] The objective of this invention can be achieved through the following technical solutions:

[0012] A system-level health status assessment method driven by multi-component state fusion, which is used to assess the health status of a device system containing multiple components, includes the following steps:

[0013] Collect a real-time dataset containing pre-selected feature parameters to characterize changes in device state;

[0014] The real-time dataset is preprocessed for each component to obtain the evaluation sample for each component.

[0015] Based on the pre-acquired normal sample set, the deviation of the sample to be evaluated for each component is calculated to obtain the health status of the corresponding component, and then the health status sequence of each component within a certain evaluation period is obtained.

[0016] The final overall health status of the equipment system is obtained by fusing the health status sequences of multiple components using the entropy weight method.

[0017] Furthermore, the preprocessing includes normalization and sample construction. The sample construction involves splicing together different feature parameter data of the same component within the same time period to form a sample of the component to be evaluated.

[0018] Furthermore, the normalization process involves restricting the values ​​of each feature parameter to [0, 1].

[0019] Furthermore, the normal sample set is extracted from historical data based on the SVDD algorithm.

[0020] Furthermore, the formula for calculating the deviation of the sample to be evaluated is as follows:

[0021]

[0022] Where D(x) t,i ) represents the sample x to be evaluated t,i The degree of deviation This represents the normal sample set of the currently evaluated component, where n1 represents the number of samples within the normal sample set.

[0023] Furthermore, the formula for calculating the health level is:

[0024]

[0025] Where h(x) t,i ) represents the sample x to be evaluated t,i Health status, D min The minimum average distance between samples within the normal sample set is used as the relative health score.

[0026] Furthermore, the calculation process for the overall health of the equipment system includes:

[0027] Calculate the information entropy Ent of the health sequence of each component. j :

[0028]

[0029] in, Indicates the current time point, r j,i G represents the proportion of the i-th health status in the health status sequence of the j-th component to the total health status sequence of that component, where G is the number of components.

[0030] Calculate the information entropy redundancy d of the health sequence of the j-th component. j :

[0031] d j =1-Ent j

[0032] Calculate the weight ω of the health status of each component. j :

[0033]

[0034] The overall health of the equipment system is obtained by weighted summation of the current health status of each component. The calculation formula is as follows:

[0035]

[0036] in, For the j-th component in the th... Health status at each point in time.

[0037] Furthermore, the proportion r of the i-th health value in the health value sequence of the j-th component to the total health value sequence of that component is... j,i The calculation formula is:

[0038]

[0039] Among them, h j,i Let represent the health status of the j-th component at the i-th time point, which is the i-th health status in the health status sequence of the j-th component.

[0040] Furthermore, the selection of the feature parameters used to characterize changes in device state specifically involves:

[0041] Furthermore, multiple sets of full lifecycle data are selected from the historical system-level equipment status dataset, and parameters that can characterize the equipment degradation state and can be continuously monitored and recorded are selected as the feature parameters.

[0042] The present invention also provides an electronic device including one or more processors, a memory, and one or more programs stored in the memory, said one or more programs including instructions for executing the system-level health status assessment method driven by multi-component state fusion as described above.

[0043] Compared with the prior art, the present invention has the following beneficial effects:

[0044] 1. As a current research hotspot in the field of equipment health management, equipment health assessment can detect internal defects as early as possible, avoiding serious accidents or significant losses caused by malfunctions. Compared with the "coarse" nature of the qualitative analysis methods commonly used in equipment health status assessment, this invention uses a quantitative analysis method to quantify the real-time health status of the equipment, more accurately describe the current equipment health status, and provide relatively clear guidance on equipment maintenance priorities.

[0045] 2. Current industrial equipment typically consists of multiple components. Before equipment failure points, a comprehensive, real-time assessment of the overall system health status is required. While component-level health status assessment is extensive and in-depth, research on system-level health status assessment for multi-component equipment is very limited. Existing research remains at the level of methods with high subjective judgment requirements, failing to provide a truly objective and reasonable health status assessment service for the entire system. To address this issue, this invention, based on the quantitative assessment of the health status of each component in the equipment system, comprehensively considers the health status of all components using the entropy weight method, thereby obtaining the real-time health status of the system. This provides an effective solution for comprehensive system-level equipment health status assessment. Attached Figure Description

[0046] Figure 1 This is a flowchart of the method of the present invention;

[0047] Figure 2 This is a schematic diagram of sample division in an embodiment of the present invention;

[0048] Figure 3 This is a schematic diagram of the bearing system test bench used in the embodiments of the present invention;

[0049] Figure 4 This is a schematic diagram of the health status of each bearing component (experimental group 1) in an embodiment of the present invention;

[0050] Figure 5 This is a schematic diagram of the experimental results (experimental group 1) of the system-level health status assessment driven by multi-component state fusion in an embodiment of the present invention;

[0051] Figure 6 This is a schematic diagram of the health status of each bearing component in an embodiment of the present invention (experimental group 2);

[0052] Figure 7 This is a schematic diagram of the experimental results (experimental group 2) of system-level health status assessment driven by multi-component state fusion in an embodiment of the present invention. Detailed Implementation

[0053] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. These embodiments are based on the technical solution of the present invention and provide detailed implementation methods and specific operating procedures. However, the scope of protection of the present invention is not limited to the following embodiments.

[0054] This invention uses a data-driven health status assessment algorithm, approaching the issue from both quantitative and system-level perspectives. It selects key characteristic indicators to quantitatively describe the real-time health status of the equipment system. For example... Figure 1 As shown, this invention provides a system-level health status assessment method driven by multi-component state fusion, the method comprising the following steps:

[0055] S1. Select characteristic parameters from the device system that can characterize its state changes and collect the dataset.

[0056] First, several sets of relatively complete run-to-failure lifecycle data are selected from the system-level equipment status dataset. From these, feature parameters that can characterize the equipment degradation state and can be continuously monitored and recorded are chosen as the equipment degradation variable dataset. Different types of industrial equipment can select their own monitored parameters, mainly including vibration, acceleration, torque, speed, load, temperature, and current.

[0057] S2. Preprocess the dataset.

[0058] The data sets of each component in the equipment system typically consist of sensor signals for various important indicators of the component. Data preprocessing includes the following steps:

[0059] (1) Normalization processing. The signal source of the component is normalized to limit the signal value to [0, 1].

[0060] (2) Construct samples. For example... Figure 2 As shown, all sensor signals of the component are divided into segments of the same length, and then segments of the same time period are horizontally spliced ​​together to form a new sample.

[0061] (3) Sample set partitioning. The sample set of each component is divided into a training set and a test set according to a certain ratio.

[0062] S3. Extraction of normal sample sets based on SVDD.

[0063] The SVDD algorithm with a Gaussian kernel function is used to solve the problem of normal sample set extraction. Its objective function and constraints can be expressed by the following formula.

[0064]

[0065] Where R is the radius of the hypersphere, and C s It is a penalty parameter that balances the volume of the hypersphere and the error rate. It is a relaxation vector, ξ i (i = 1, 2, ..., N) s ) is sample x 1, The corresponding relaxation factor, o, is the center of the hypersphere. k g (x 1, The Gaussian kernel function can be transformed into two high-dimensional vectors through Taylor expansion, k g (x 1,1 The expression for ) is as follows:

[0066]

[0067] N represents the state sample at time point i. s Indicates the number of state samples. The existence of ξ is intended to give the SVDD model a certain degree of fault tolerance. When all samples are located inside the hypersphere, then...

[0068] The original problem is transformed into a dual problem using the Lagrange multiplier method, and the Lagrange function is shown below:

[0069]

[0070] in and All are Lagrange multiplier vectors. By minimizing L with respect to the partial derivatives of L with respect to R, o, and ξ, we obtain the center of the hypersphere as shown below:

[0071]

[0072] Due to Both have α i ≥0 and β i =α i -C s If ≥0, the original problem is transformed into a dual problem with the objective function and constraints shown below.

[0073]

[0074] To solve this dual problem, the Lagrange coefficients must satisfy 0 < α. i <C sThe samples are called support vectors, and the set containing all support vectors is called the support vector set, denoted by SV. The radius of the hypersphere is calculated by arbitrarily selecting a support vector, as shown in the following formula.

[0075]

[0076] Where, x v Let x be any support vector. v ∈SV. For the test set sample x t The formula for the distance to the center of the ball is shown below. If D(x) t ) <R indicates x t If it is a normal sample, then D(x) t )>R indicates x t It is an abnormal sample.

[0077]

[0078] After filtering using the SVDD algorithm, a normal sample set of components is obtained. Where n1 represents the number of normal samples. right All

[0079] S4. Real-time health calculation.

[0080] After extracting the normal sample set based on the SVDD algorithm, it is necessary to find the normal sample with the highest similarity by calculating the distance between the test sample and the normal sample set, and then convert the distance into a health score through a negative transformation to evaluate the health score of the test sample.

[0081] Let the test sample set corresponding to component 1 in the equipment system be in This indicates the number of test samples. The traditional calculations for deviation and health are shown below.

[0082]

[0083]

[0084] in, D(x) represents the normal sample set of component 1; t,i ) and h(x t,i ) represents the i-th test sample x t,i Deviation and health, i = 1, ..., N tUsing the above formula as a negative transformation function requires obtaining the minimum and maximum values ​​to normalize the health score to [0, 1]. However, in actual operation, it is difficult to know the minimum and maximum values ​​in advance when calculating the real-time health score. Therefore, this invention proposes an improved method for calculating deviation and health score, which uses the average distance between the test set sample and all normal samples as the deviation of the sample. The calculation formula is shown below.

[0085]

[0086] After obtaining the deviation of the test samples, the health score is calculated as follows: First, calculate the minimum average distance between normal samples, and use this as the minimum value D for health score calculation. min ,as follows.

[0087]

[0088] With D min The formula for calculating the relative health of the test set samples, used as a benchmark, is as follows:

[0089]

[0090] Where, when D(x) t,i )>D min Directly use h(x) t,i )=(1+D min ) / (1+D(x t,i To calculate health status, we can obtain h(x). t,i When D(x) < 1; t,i )≤D min , there is (1+D min ) / (1+D(x t,i If ))≥1, it means that the distance between this sample and all normal samples is very small, that is, it has a very high health score. Therefore, the health score h(x) is set to... t,i Set it to 1.

[0091] The improved formula does not require prior knowledge of the minimum and maximum health values ​​of the test set samples; it only needs to calculate the baseline D on the set of normal samples extracted from the training set. min This is sufficient. Therefore, it can be applied to real-time equipment health assessment, where the final equipment deviation vector and health vector are respectively... and

[0092] S5. Repeat the above steps to calculate the health status of each component.

[0093] Assuming a device system contains G components, based on the steps described above, data preprocessing, normal sample set extraction based on the SVDD algorithm, and real-time health calculation are performed on each component in the device system to obtain the health sequence H1, H2, ..., H for each component. G ,in, Let h represent the health sequence of the j-th component. j,i This represents the health status of the j-th component at time i.

[0094] S6. The final overall health of the equipment system is obtained by fusing the health status of multiple components based on the entropy weight method.

[0095] From a statistical perspective, the greater the variation in a component's health status sequence, the greater its dispersion, and the smaller its corresponding information entropy value. This means that the component's health status sequence contains a greater amount of information, thus warranting a higher weight and a greater impact on the overall health status of the system. Assume a device system contains G components, with sample sets for each component as follows: All contain N s m-dimensional samples, Current time The specific steps are as follows:

[0096] (1) Calculate the proportion r of the i-th health status of the j-th component to the health status sequence of that component. ij As shown below.

[0097]

[0098] (2) Calculate the information entropy Ent of the health sequence of the j-th component. j As shown below.

[0099]

[0100] (3) Calculate the information entropy redundancy (d) of the health sequence of the j-th component. j As shown below.

[0101] d j =1-Ent j j = 1, 2, ..., G

[0102] (4) Calculate the weight ω of the health status of each component. j As shown below.

[0103]

[0104] (5) Calculate the current overall health of the system. Based on the current health of each component (the j-th component in the j-th state),... Health samples at each time point are We perform a weighted summation to obtain the overall health score of the system, as shown below.

[0105]

[0106] Finally, the system health sequence is obtained.

[0107] Since it's impossible to know the health sequence of all components in advance for the real-time health of a computing device system, we need to consider the health assessment problem based on the incomplete lifecycle operating status data of the equipment. To address this, the solution of this invention is to start from a certain time step, and for each health sample that appears, apply the entropy weight method to calculate the health weight of each component in the existing health sequence. Then, perform a weighted calculation on the latest health sample to obtain the overall system-level equipment health.

[0108] If the above methods are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0109] The following will illustrate this with specific examples:

[0110] To verify the effectiveness and superiority of the multi-component state fusion-driven system-level health status assessment method proposed in this invention, a bearing dataset provided by the Intelligent Maintenance System (IMS) Center at the University of Cincinnati, also known as the IMS dataset, was used. The IMS dataset contains three sub-datasets, namely dataset_1, dataset_2, and dataset_3. Each sub-dataset describes an experiment from the start of operation to failure, and the basic information is shown in Table 1.

[0111] Table 1. Dataset Introduction

[0112]

[0113]

[0114] The bearing test bench used to collect this dataset is as follows: Figure 3 As shown, four test bearings are mounted on a shaft driven by an AC motor and coupled via a friction belt. Since the bearings are interconnected through the shaft, they influence each other during operation, essentially forming a multi-component equipment system. The bearings operate at a constant speed of 2000 rpm, with a radial load of 6000 lbs, added to the shaft and bearings via a spring mechanism. All bearings are forced lubricated, with an oil circulation system regulating the flow and temperature of the lubricant. Accelerometers are used to collect vibration signals during bearing operation; in dataset_1, each bearing is equipped with two accelerometers (x-axis and y-axis), while in datasets_2 and_3, each bearing is equipped with one accelerometer. All failures occurred after exceeding the bearing's design life by more than 100 million revolutions. Datasets_2 and_3, with the same number of sensor signals per component, can serve as training and test sets for each other. For experimental groups where the training and test sets belong to the same sample set, the ratio of training to test sets is 3:7. For experimental groups where the training and test sets are different sample sets, sampling is performed based on the number of sample sets. The final experimental group settings are shown in Table 2. In the SVDD model, the Gaussian kernel function parameter γ = 1.

[0115] Table 2. Experimental group setup for health status assessment

[0116] experimental group training set test set Number of training sets Number of test sets 1 dataset_1 dataset_1 2584 6030 2 dataset_3 dataset_2 2529 3936

[0117] Based on the above steps, the health status assessment results of each bearing in experimental group 1 are as follows: Figure 4 As shown, both the training and test sets are derived from dataset_1. The test set samples are sorted chronologically, and the horizontal axis "time" essentially refers to the sequence number of the test set samples. Around time 500, the health of all four bearings decreased to some extent. Afterward, bearings 1 and 2 maintained relatively good condition; bearing 3 began to decline around time 5500 and then exhibited a failure state; bearing 4 had the worst health among the four bearings, with its health curve showing pulses during operation. It began to decline around time 3500, and around time 4750, the slope of the health curve further increased, significantly deviating from a healthy state, ultimately exhibiting a failure state. These experimental results are consistent with the descriptions in Table 1 of bearings 3 and 4 exhibiting inner and outer ring damage, respectively, at the end of the IMS dataset run-to-failure experiment.

[0118] The system-level health status assessment results driven by multi-component state fusion in Experiment Group 1 are as follows: Figure 5As shown, to highlight the effectiveness of real-time health calculation based on the entropy weight method, we also compare the trend changes of the non-real-time health calculation results based on the entropy weight method, the health calculation results based on the mean method, and the first principal component extracted from the health sequence of multiple components using principal component analysis (PCA). Non-real-time calculation refers to obtaining the health sequence of all components throughout their entire lifecycle before performing the entropy weight method calculation. For ease of explanation, these four methods will be referred to as "real-time entropy weight method," "entropy weight method," "mean method," and "principal component 1," respectively.

[0119] Figure 5 It can be seen that during the time period from 0 to 3500, the health curves of the "real-time entropy weight method," "entropy weight method," and "mean method" basically overlap. After 3500 time, that is, when bearing 4 shows obvious failure, the health begins to decline, with the "real-time entropy weight method" and "entropy weight method" declining at a faster rate than the "mean method." In general, the "real-time entropy weight method" can track the health changes of the equipment system more accurately throughout its entire life cycle, and its health curve is quite similar to the health curve of bearing 4 in dataset_1.

[0120] Based on the above steps, the health status assessment results of each bearing in experimental group 2 are as follows: Figure 6 As shown, the training set comes from dataset_3, and the test sets all come from dataset_2. Figure 6 As can be seen, the health status of bearing 1 in experimental group 2 began to decline at 2100 and oscillated significantly at 2800, indicating a clear failure in bearing 1. At this point, the health status curves of other bearings showed a dip and then gradually declined, meaning that the operating status of other bearings was affected by bearing 1, but had not yet reached the point of complete failure. Overall, the results of the component-level health status assessment are consistent with the description in dataset_2 that "bearing 1 suffered outer ring damage".

[0121] The system-level health status assessment results driven by multi-component state fusion in Experiment Group 2 are as follows: Figure 7 As shown, similarly, Figure 7 In the data, during the time period from 0 to 2000, the "entropy weight method" yielded the highest health value, exceeding 0.90, followed by the "mean method," with the "real-time entropy weight method" being the lowest, with a numerical difference within 0.2. When time reached 2100, i.e., when bearing 1 exhibited a failure, the overall system health began to decline. Subsequently, the "real-time entropy weight method" was able to track the "entropy weight method" curve well, and it was noted that this health change trend most closely matched the health status change of the failed bearing 1 in dataset_2.

[0122] Figure 5 and Figure 7In the analysis, "Principal Component 1" shows a negative correlation with health status and a positive correlation with deviation, exhibiting an overall trend similar to that of the "Entropy Weight Method" and the "Real-time Entropy Weight Method." According to the principles of PCA, it maximizes the sample variance by transforming the orthogonal basis. When the angle between the direction of equipment health degradation (i.e., the direction of deviation change) and the direction of the orthogonal basis corresponding to "Principal Component 1" is acute, "Principal Component 1" will gradually increase as the equipment operates.

[0123] In summary, this embodiment demonstrates that the calculation results of the "real-time entropy weight method" and the "entropy weight method" have very small errors, and can effectively assess the overall health of the current equipment system based on the current and historical component health sequences. The health curve obtained by the "mean method" is the flattest because it treats the importance of each component as equal; however, the curve's trend still indicates that some components in the equipment system have failed. The multi-component state fusion-driven system-level health status assessment method proposed in this invention can effectively capture the early failure characteristics of each component in the current system, accurately evaluate the overall health status of the current system, and provide a valid basis for subsequent equipment health status management.

[0124] The preferred embodiments of the present invention have been described in detail above. It should be understood that those skilled in the art can make numerous modifications and variations based on the concept of the present invention without creative effort. Therefore, all technical solutions that can be obtained by those skilled in the art based on the concept of the present invention through logical analysis, reasoning, or limited experimentation on the basis of existing technology should be within the scope of protection defined by the claims.

Claims

1. A system-level health status assessment method driven by multi-component state fusion, characterized in that, This method is used to assess the health status of a multi-component equipment system, and includes the following steps: Collect a real-time dataset containing pre-selected feature parameters to characterize changes in device state; The real-time dataset is preprocessed for each component to obtain the evaluation sample for each component. Based on the pre-acquired normal sample set, the deviation of the sample to be evaluated for each component is calculated to obtain the health status of the corresponding component, and then the health status sequence of each component within a certain evaluation period is obtained. Based on the entropy weight method, the health sequence of multiple components is fused to obtain the final comprehensive health of the equipment system. The formula for calculating the deviation of the sample to be evaluated is: in, Indicates the sample to be evaluated The degree of deviation This represents the normal sample set of the currently evaluated component. This represents the number of samples in the normal sample set; The formula for calculating the health status is: in, Indicates the sample to be evaluated health The minimum average distance between samples within the normal sample set is used as the relative health score.

2. The system-level health status assessment method driven by multi-component state fusion according to claim 1, characterized in that, The preprocessing includes normalization and sample construction. The sample construction involves splicing together different feature parameter data of the same component within the same time period to form a sample to be evaluated for that component.

3. The system-level health status assessment method driven by multi-component state fusion according to claim 2, characterized in that, The normalization process involves restricting the values ​​of each feature parameter to [0, 1].

4. The system-level health status assessment method driven by multi-component state fusion according to claim 1, characterized in that, The normal sample set is extracted from historical data based on the SVDD algorithm.

5. The system-level health status assessment method driven by multi-component state fusion according to claim 1, characterized in that, The calculation process for the overall health of the equipment system includes: Calculate the information entropy of the health sequence of each component. : in, Indicates the current time point, Indicates the first In the health sequence of component, the first... The proportion of each health level in the component's health level sequence. This refers to the number of components; Calculate the first Information entropy redundancy of the health status sequence of each component : Calculate the weight of the health status of each component. : The overall health of the equipment system is obtained by weighted summation of the current health status of each component. The calculation formula is as follows: in, For the first j The component in the first Health status at each point in time.

6. The system-level health status assessment method driven by multi-component state fusion according to claim 5, characterized in that, The first In the health sequence of component, the first... The proportion of each health level in the component's health level sequence The calculation formula is: in, For the first j The component in the first The health status at the 1st time point, i.e., the 1st The health sequence of the component is the first Health status.

7. The system-level health status assessment method driven by multi-component state fusion according to claim 1, characterized in that, The selection of the feature parameters used to characterize changes in equipment state is specifically as follows: Multiple sets of full lifecycle data are selected from historical system-level equipment status datasets, and parameters that can characterize the equipment degradation state and can be continuously monitored and recorded are selected as the feature parameters.

8. An electronic device, characterized in that, It includes one or more processors, memory, and one or more programs stored in the memory, said one or more programs including instructions for executing the system-level health status assessment method driven by the multi-component state fusion as described in any one of claims 1-7.

Images