A sub-health diagnosis method for rail vehicle door system
By preprocessing and feature extraction of the door opening and closing process of the rail vehicle door system, and clustering analysis using elastic network and DPC algorithm, the problem of difficulty in screening sub-health status in the existing technology is solved, and efficient and accurate Sub-health status diagnosis improves the operating reliability of the door system.
Patent Information
- Application Number
- CN202111575029.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2021-12-21
- Publication Date
- 2024-11-22
- Estimated Expiration
- 2041-12-21
AI Technical Summary
The prior art is difficult to efficiently and easily quickly screen out sub-health status from massive data of rail vehicle door systems, resulting in low fault prevention and maintenance efficiency.
By obtaining the door opening and closing operation status curves of the door system, data preprocessing and feature extraction are performed, and the characteristic data of the door opening and closing process are clustered and analyzed by using elastic network and DPC algorithm respectively to achieve the diagnosis of sub-health status.
It improves the efficiency and accuracy of sub-health status identification, and can quickly and accurately determine whether the door system is in a sub-health status, reducing operating costs and fault diagnosis pressure.
Smart Images

Figure CN114463151B_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the field of urban rail transit, and in particular relates to a sub-health diagnosis method for a rail vehicle door system. Background Art
[0002] With the development of the rail transit industry, the safety and reliability of urban rail door operation have received extensive attention. The high frequency of use of the door system, the complex vehicle operating environment and unknown human factors have caused the number of door failures to account for more than 30% of the total number of vehicle component failures. Therefore, improving the quality and reliability of the door system operation is of great significance to the efficient and safe operation of urban rail trains.
[0003] Sub-health of vehicle door system refers to an intermediate state of the door system. In this state, the system function can still be realized, but its state has deviated from the standard state to a certain extent during the realization process. Further development of this state will lead to one or more types of failures. Conducting sub-health state research can detect abnormalities before the failure actually occurs, effectively prevent the failure from occurring, and maintain the safety of system operation.
[0004] During the operation of the urban rail door system, a huge amount of data is generated every day. How to efficiently and easily quickly and accurately screen out possible sub-health data from a large amount of data needs to be studied. The method usually used in sub-health status research is to collect historical data or experimental data of door operation under different working conditions, extract characteristic values, build a fault classification model with the help of classification algorithms, and diagnose the real-time operation status of the door based on the fault classification model. However, the health status of the door is not a fixed state, but can be called "healthy" within a certain range; at the same time, the sub-health state is different from the fault state, and its deviation from the healthy state is smaller. Summary of the invention
[0005] Purpose of the invention: The purpose of the present invention is to provide a sub-health diagnosis method capable of distinguishing a sub-health state from a healthy state of a rail vehicle door system.
[0006] Technical solution: A method for diagnosing sub-health of a rail vehicle door system according to the present invention comprises the following steps:
[0007] (1) Obtain the operating status curves of the door system during the opening and closing processes, and perform data preprocessing;
[0008] (2) According to the difference between different sub-healthy operating states and healthy operating states, select the multi-component action time period that can reflect the difference between the sub-healthy state and the healthy state;
[0009] (3) Feature extraction is performed on the multi-component action periods of the door opening process and the door closing process respectively;
[0010] (4) Using the elastic network to filter the features extracted in step (3) to obtain a door opening feature set and a door closing feature set;
[0011] (5) Calculate the feature data of the door opening feature set and the door closing feature set of the data to be detected, combine them with the health status feature data, and use the DPC algorithm to perform cluster analysis on the door opening feature data and the door closing feature data to be detected, so as to complete the sub-health diagnosis of the data to be detected.
[0012] Preferably, in step (1), the door system opening and closing process is driven by a motor, and the obtained door system operation status curve includes the stroke, speed, and current status curves generated during the door opening and closing processes.
[0013] Preferably, in step (2), for each operating state curve, the physical mechanism of the sub-healthy operating state and the difference between the sub-healthy operating state curve and the healthy operating state curve are combined to obtain the time periods when different sub-healthy operating states are different from the healthy operating state; and the multi-component action time period is obtained by combining the time periods when different sub-healthy operating states are different from the healthy operating state.
[0014] Preferably, in step (3), the multi-component action period is divided into a number of sub-segments, and feature extraction is performed on each sub-segment respectively.
[0015] Furthermore, the features extracted from the multi-component action period during the door opening process and the door closing process in step (3) include: global information features of the multi-component action period, time domain features of each sub-segment of the multi-component action period and / or morphological features of the operating status curve of the multi-component action period.
[0016] Preferably, all the door opening characteristics of various operating state curves during the door opening process are obtained and formed into a matrix F open , obtain all the closing characteristics of various operating state curves during the closing process and form a matrix F colse ; Use elastic network to open door feature matrix F open and the closing feature matrix F colse Perform dimensionality reduction to obtain the door opening feature set F1 and door closing feature set F2 after dimensionality reduction, and complete feature screening.
[0017] Preferably, the door feature matrix F open and the closing feature matrix F close The loss function of the elastic net for dimensionality reduction is as follows:
[0018]
[0019] In the formula, y is the actual value of the training data, and sample For sub-health data, the value is 0. For healthy data, it is 1; m is the number of features, θ j is the weight of the jth feature; n is the number of samples; X is the normalized open feature matrix or closed feature matrix, x is the element in X; θ T Assign weights to each feature for the algorithm, is the fitting function, λ and ρ are adjustment parameters.
[0020] Preferably, the feature data of the door opening feature set or the door closing feature set of the data to be detected are combined with the health status feature data for cluster analysis using DPC. If the cluster analysis results of the health status data and the data to be detected are divided into two different categories, and each category includes samples of the health status data or the data to be detected that are not less than a set proportion, it is determined that a sub-healthy state exists; if the cluster analysis results cannot select two cluster centers, it is considered that a sub-healthy state does not occur.
[0021] Preferably, the sub-health diagnosis results of the door opening process and the sub-health diagnosis results of the door closing process of the data to be detected are obtained respectively. When any one of the sub-health diagnosis results of the door opening process and the sub-health diagnosis results of the door closing process is a sub-healthy state, it is determined that the door system that outputs the data to be detected is in a sub-healthy state.
[0022] Preferably, the steps of performing cluster analysis on the door opening process or the door closing process using DPC are as follows:
[0023] The feature set of the gate system's normal operation data to be tested and the feature set of the healthy data are stored as a feature matrix, and the local density ρ of each sample i is calculated. i and the distance δ from sample i to the sample j whose local density is larger than that of the sample i and closest to it i , the calculation formula is:
[0024] ρ i =∑χ(d ij -d c )
[0025]
[0026] Where, d ij is the Euclidean distance between data points i and j, d c is the pre-specified cutoff distance; χ is the discriminant function, when (d ij -d c )>0, χ=1, otherwise χ=0, according to the calculated ρ i and δ i , with ρ i is the horizontal axis, δ i Draw a decision diagram with as the vertical axis; and use the decision diagram to select the cluster center.
[0027] Furthermore, the sub-health diagnosis model established by this method is a common diagnosis model for a variety of common sub-health types, that is, different types of sub-health operation status data can be screened out by a model. Since each action of the door system includes two operations, opening and closing the door, and generates three door state curves of travel, speed, and current, the subsequent feature extraction is based on the above three door operation state curves.
[0028] When preprocessing the data, the collected door status data is corrected, and the data is aligned and the bad data is eliminated, including data with the number of sampling points obviously not within the normal range and data with a curve shape highly deviating from the normal state.
[0029] Before extracting features from the entire door movement process, first select the time period that can reflect the difference between health and sub-health during each door movement; analyze each sub-healthy working state one by one, and combine the physical mechanism of sub-health to determine the specific structure and position that the sub-healthy state will affect the door movement. At the same time, compare the sub-healthy state curve with the healthy curve, and judge the three door state curves of travel, speed, and current based on the above two points. Select some state curve data segments that can concentrate on reflecting the difference between sub-health and health, and define them as multi-component action segments. After completing the analysis of all sub-healthy states, the multi-component action segments of various sub-healthy states are merged according to the type of door state curve to obtain a time period that can include various sub-healthy states. Because some sub-healthy types have a greater impact on the door opening process, such as abnormal centering size; some sub-healthy types have a greater impact on the door closing process, such as the reduction of sealing size will increase the local resistance of door closing; and some sub-healthy types have a certain degree of impact on the door opening and closing process, such as abnormal door leaf height that will cause additional resistance to the door opening and closing. Therefore, consider modeling and analyzing the door opening data and door closing data separately.
[0030] When extracting features from the acquired state curve, feature extraction is performed on the multi-component action period of the door opening or closing process. Feature extraction can also be performed on the sub-segments of the state curve during the multi-component action period. The extracted feature types include: global information features during the multi-component action period, including door leaf opening, operating time, etc.; time domain features of each sub-segment during the multi-component action period, including the maximum value, minimum value, average value, standard deviation, peak-to-peak value, skewness, kurtosis, etc. of the state curve in the sub-segment; morphological features of the operating state curve during the multi-component action period, including the number of peaks, the number of troughs and their occurrence time, etc. After extraction, P door opening features in the door opening process and Q door closing features in the door closing process can be obtained, which are stored in two matrices F respectively. open and F close, where each row of the matrix represents all the features of a piece of data, and each column of the matrix represents a feature of all the data. Furthermore, the multi-component action time period of the door opening or closing process can be subdivided into different sub-segments based on time domain, frequency domain, image, etc.
[0031] Since the subhealth state is different from the fault state and its deviation from the healthy state is smaller, it is necessary to find feature data that can narrow the difference between health data and amplify the difference between subhealth and health. The feature data is selected based on the principle of maximizing the distance between classes and minimizing the distance within classes using the elastic network algorithm. The elastic network is a linear regression model that uses L1 and L2 priors as regularization matrices. Adding a complexity penalty factor, i.e., a regularization term, after the objective function J can be used to prevent overfitting. This combination allows learning a model in which only a small number of parameters are non-zero and sparse. By continuously iterating to minimize the residual sum of squares, features with multicollinearity can be effectively screened, thereby achieving the purpose of removing redundant features and obtaining a feature subset after dimensionality reduction optimization. By performing the gate feature matrix F open and the closing feature matrix F close After training, the available door opening feature set F1 and door closing feature set F2 are obtained. The trained feature set contains the features of different types of working state curves and is suitable for the identification of various sub-health states.
[0032] For the multiple door opening and closing operation data during the operation of the vehicle door, the DPC algorithm is used to perform cluster analysis on the filtered feature sets for the door opening feature data and the door closing feature data, and then the final diagnosis result is obtained based on the results obtained by both parties. DPC is a density-based clustering algorithm. Compared with the distance-based clustering algorithm, it is more robust to the shape of the cluster; and it is a rare clustering algorithm that does not require iteration. The clustering results can be obtained directly, saving time and being highly efficient. It is very suitable for scenarios such as mainline door diagnosis that require a large number of repeated executions.
[0033] When using the DPC algorithm for cluster analysis, the data to be clustered include the mainline door operation test data and standard health data. The standard health data is a mixture of multiple groups of data samples of health status after multiple door adjustments, that is, there are certain differences between health data. During the cluster analysis process, the decision diagram drawn can realize the selection of cluster centers. After the cluster center is selected, the remaining sample data is directly assigned to each class according to the algorithm rules. The allocation rule is that each remaining sample point is assigned to the cluster with higher density to which its nearest neighbor belongs; each data sample is the data corresponding to one door opening or closing.
[0034] The decision threshold is set to S, that is: if no less than S samples in the standard data and no less than S samples in the test data are respectively divided into two different categories, it is considered that sub-health has occurred; if two cluster centers cannot be selected, or after two cluster centers are selected, the standard data and test data are scattered into two categories below the threshold, it is considered that sub-health has not occurred and the patient is in a healthy state.
[0035] The door opening data diagnosis results and the door closing data diagnosis results of the main line door operation test data are used to obtain the final diagnosis results according to the logical or operation rules. That is, when the door opening data diagnosis results and the door closing data diagnosis results are both in a healthy state, the door system is determined to be in a healthy state.
[0036] During the model building and training stage, if there are obvious errors in the DPC clustering results, they are fed back to the elastic network iteration process to adjust the loss function parameters and optimize the elastic network's process of finding the door-opening feature set F1 and the door-closing feature set F2.
[0037] The present invention discloses a computer-readable storage medium, in which a computer program is stored. When the computer program is executed by a processor, the method steps for diagnosing sub-health of a rail vehicle door system are implemented.
[0038] Beneficial effects: Compared with the prior art, the present invention has the following significant advantages: Aiming at the sub-health characteristics of the door system, the data generated by the operation of the vehicle door is intercepted. Since the door opening and door closing stages are not symmetrical, the time domain features are individually extracted for the door opening and door closing stages according to the curve shapes, and feature selection is performed using elastic networks respectively. The features screened out by the elastic network can more accurately describe the sub-health state that deviates more slightly from the healthy state; the vehicle door opening data and door closing data are modeled and analyzed separately through DPC, which improves the efficiency and accuracy of sub-health state discrimination. The combined clustering analysis results of door opening and door closing can quickly and accurately determine whether the door system is in a sub-health state; this sub-health diagnosis method is used for a variety of sub-health types. BRIEF DESCRIPTION OF THE DRAWINGS
[0039] Figure 1 is a flow chart of the sub-health diagnosis method of the present invention;
[0040] Figure 2 It is a structural principle diagram of a plug sliding door in a specific embodiment of the present invention;
[0041] Figure 3 It is a decision diagram and a box-selected clustering center of the V-type abnormal sub-health state to be detected door opening test data and standard health data density peak clustering in a specific embodiment of the present invention;
[0042] Figure 4It is a decision diagram and a box-selected clustering center of the V-type abnormal sub-health state to be inspected closed-door test data and standard health data density peak clustering in a specific embodiment of the present invention;
[0043] Figure 5 It is a decision diagram and a box-selected clustering center for clustering the density peak values of the abnormal sub-health state to be detected door opening test data and standard health data in a specific embodiment of the present invention;
[0044] Figure 6 It is a decision diagram and a box-selected clustering center for clustering the density peak values of abnormal sub-health state closed-door test data to be inspected and standard health data in a specific embodiment of the present invention;
[0045] Figure 7 It is the clustering result of the V-type abnormal sub-health state to-be-detected door opening test data and the standard health data in the specific embodiment of the present invention;
[0046] Figure 8 It is the clustering result of the V-type abnormal sub-health state to-be-detected closed-door test data and the standard health data in the specific embodiment of the present invention;
[0047] Fig. 9 It is the clustering result of the abnormal sub-health state to be detected door opening test data and the standard health data in the specific embodiment of the present invention;
[0048] Fig.10 It is the clustering result of the abnormal sub-health state to be detected closed-door test data and standard health data in the specific embodiment of the present invention. DETAILED DESCRIPTION
[0049] The technical solution of the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments.
[0050] Taking the sub-health detection method of the plug sliding door system commonly used in urban rail subway as an example, the main structure of the plug sliding door is as follows Figure 2 As shown, according to its electromechanical structure, it can be roughly divided into five subsystems: electrical control system, drive locking system, load bearing guide system, internal and external operating system and basic components. The main components include: motor 1, door control system 2, connection device 3, front support 4, rear support 5, nut locking device 6, lock in place switch 7, middle support 8, door in place switch 9, screw rod 10, long guide column 11, end unlocking device 12, internal operating device 13, external operating device 14, left door leaf 15, right door leaf 16. The door control system is used to control the motor to drive the left door leaf and the right door leaf to move synchronously towards each other. One action includes door opening and closing. At the same time, the door control system can also collect the three door state curves of travel, speed and current generated during the door opening and closing process.
[0051] The specific steps of the sub-health diagnosis method based on the above-mentioned Sierra door system are as follows:
[0052] (1) Collect standard health data, which includes three door status curves of travel, speed, and current of the door system under healthy conditions. Standard health data is a mixture of multiple sets of health status data after multiple door adjustments, that is, there are certain differences between health data; collect different types of sub-health status data, including three door status curves of travel, speed, and current of the door system under each sub-health status.
[0053] (2) Preprocessing the collected data, correcting the collected door status data, aligning the data and eliminating bad data, including data whose number of sampling points is obviously not within the normal range and data whose curve shape deviates greatly from the normal state.
[0054] (3) The different types of door state curves under each sub-health state are analyzed separately. First, the physical mechanism of sub-health is combined to determine the specific location where the sub-health state will affect the door movement. At the same time, the sub-health state curve and the healthy curve are compared, and some time periods that can concentrate on reflecting the difference between sub-health and health are selected. Combined with the analysis results of all sub-health types, the time periods that can reflect various sub-health states in each door state curve are obtained to form a multi-component action section. In this embodiment, some sub-health types have a greater impact on the door opening process, such as abnormal centering size; some sub-health types have a greater impact on the door closing process, such as reduced sealing size will increase the local resistance to door closing; and some sub-health types have a certain degree of impact on the door opening and closing processes, such as abnormal door leaf height will cause additional resistance to opening and closing. Therefore, the door opening data and door closing data are modeled and analyzed separately, and the multi-component action section is specifically defined as the door opening start section and door opening speed overshoot section of the door opening data, and the door closing speed enhancement section of the door closing data.
[0055] (4) For the multi-component action segment, the segment is further subdivided into sub-segments according to the curve oscillation, the appearance of peaks and other conditions, and the feature type of each sub-segment and the information contained in each feature data are obtained. According to the inherent form and characteristics of the multi-component action segment curve, the feature extraction of the door opening data and the door closing data is performed respectively. A total of 62 door opening features and 44 door closing features are obtained, which are stored in two matrices F respectively. open and F close The extracted features include the global information features of the multi-component action period, including the door leaf opening, operating time, etc.; the time domain features of each sub-segment of the multi-component action period, including the maximum value, minimum value, average value, standard deviation, peak-to-peak value, skewness, kurtosis, etc. of the state curve in the sub-segment; the morphological features of the operating state curve of the multi-component action period, including the number of peaks, the number of troughs and their occurrence time, etc.
[0056] (5) Based on the principle of maximizing the distance between classes and minimizing the distance within classes, the elastic network algorithm is used to calculate the door feature matrix F open and the closing feature matrix F close Feature selection is performed to remove redundant features, and the door opening feature set F1 and door closing feature set F2 after dimensionality reduction optimization are obtained; in this embodiment, the loss function of the elastic network is as follows:
[0057]
[0058] In the formula, y is the actual value of the training data, and sample For sub-health data, the value is 0. For healthy data, it is 1; m is the number of features, θ j is the weight of the jth feature; n is the number of samples; X is the normalized open feature matrix or closed feature matrix, x is the element in X; θ T Assign weights to each feature for the algorithm, is the fitting function, λ and ρ are adjustment parameters.
[0059] (6) After the low-dimensional effective feature set is successfully extracted, the intelligent algorithm is used to construct a typical sub-health diagnosis model of the system. In the sub-health diagnosis process, the door opening process and the door closing process are diagnosed separately. Taking the door opening data diagnosis as an example, the door opening data of the main line door operation test data and the door opening feature set of the standard data are obtained and stored as a feature matrix f1. The local density ρ of each sample i is calculated. i and the distance δ from sample i to the sample j whose local density is larger than that of the sample i and closest to it i , the calculation formula is:
[0060] ρ i =∑χ(d ij -d c )
[0061]
[0062] Where, d ij is the Euclidean distance between data points i and j, d c is the pre-specified cutoff distance; χ is the discriminant function, when (d ij -d c )>0, χ=1, otherwise χ=0, and the feature set corresponding to each door opening data in the measured data and each door opening data in the standard data is a sample.
[0063] According to the calculated ρ i and δ i , with ρ i is the horizontal axis, δ i As the vertical axis, draw a decision diagram. Using the decision diagram, ρi and δ i The relatively high point, that is, the point in the upper right corner of the decision diagram, is marked as the cluster center. The cluster center of DPC can be flexibly selected. After the cluster center is selected, the remaining data is directly assigned to each class according to the algorithm rules. The allocation rule is that each remaining point is assigned to the cluster to which its nearest neighbor with higher density belongs.
[0064] (7) The decision threshold is set to 80%, that is, if no less than 80% of the samples in the standard data and no less than 80% of the samples in the test data are divided into two different categories, it is considered that sub-health has occurred; if two cluster centers cannot be selected, or after selecting two cluster centers, the standard data and test data are scattered below the threshold and enter two categories, it is considered that sub-health has not occurred and is in a healthy state. The sub-health diagnosis process of the closed door data of the main line door operation test data is the same as the sub-health diagnosis process of the open door data.
[0065] (8) Combining the door opening data diagnosis result and the door closing data diagnosis result, according to the logic or operation rule, the final diagnosis result is obtained, that is, when the door opening data diagnosis result and the door closing data diagnosis result are both in a healthy state, the door system is determined to be in a healthy state.
[0066] If there is an obvious error in the DPC clustering result, it is fed back to the elastic network iteration process to optimize the process of the elastic network searching for the door opening feature set F1 and the door closing feature set F2.
[0067] In this embodiment, Figure 3-Figure 6 As shown in the figure, they are the decision diagrams of the door opening data and door closing data of the V-type abnormal sub-health state and the medium abnormal sub-health state, respectively. i and δ i The relatively high points, i.e., those in the upper right corner of the decision diagram, are marked as cluster centers; Figure 7-10 It is the clustering result of the open and closed data of V-type abnormal sub-health state and medium abnormal sub-health state. Figure 7 and Figure 8 The V-shaped abnormal data and healthy data classification accuracy is 100%, Fig. 9 and Fig.10 The classification accuracy of abnormal data and healthy data is shown to be 98.84%, both of which are higher than the threshold. Combined with the sub-health diagnosis results of door opening and door closing, it can be determined that the Sierra door system is in a sub-healthy state.
[0068] In summary, this sub-health diagnosis method for rail vehicle door systems uses elastic network models for feature selection and DPC for cluster analysis of sub-health discrimination, and establishes a highly robust diagnostic model that can be used online for rail mainline door systems. It can successfully mine small and weak sub-health features and classify the common sub-health states and healthy states of urban rail vehicle doors. By mining the common differences between multiple types of sub-health states and healthy states, sub-health states can be diagnosed simply and quickly when applied, and are universal in the diagnosis of multiple common sub-health states. It is beneficial to improve the operational reliability of the door system, reduce the fault diagnosis pressure of the mainline system facing massive data, and reduce operating costs.
Claims
1. A method for diagnosing sub-health of a rail vehicle door system, characterized in that: The method comprises the following steps: (1) Obtain the operating status curves of the door system during the opening and closing processes, and perform data preprocessing; (2) According to the difference between different sub-healthy operating states and healthy operating states, select the multi-component action time period that can reflect the difference between the sub-healthy state and the healthy state; (3) Feature extraction is performed on the multi-component action periods of the door opening process and the door closing process respectively; (4) Using the elastic network to filter the features extracted in step (3) to obtain a door opening feature set and a door closing feature set; (5) Calculating the feature data of the door opening feature set and the door closing feature set of the data to be detected, and combining the health status feature data to perform cluster analysis on the door opening feature data and the door closing feature data to be detected, respectively, to complete the sub-health diagnosis of the data to be detected; Obtain all the door opening characteristics of various operating state curves during the door opening process and form a matrix F open , obtain all the closing characteristics of various operating state curves during the closing process and form a matrix F close ; Use elastic network to open door feature matrix F open and the closing feature matrix F close Perform dimensionality reduction to obtain the door opening feature set F1 and door closing feature set F2 after dimensionality reduction, and complete feature screening; Double door feature matrix F open and the closing feature matrix F close The loss function of the elastic network for dimensionality reduction is as follows: Where y is the actual value of the training data, 0 when sample i is sub-healthy data, and 1 when sample i is healthy data; m is the number of features, θ j is the weight of the jth feature; n is the number of samples; X is the normalized open feature matrix or closed feature matrix, x is the element in X; θ T Give the algorithm the weight of each feature, θ T x (i) is the fitting function, λ and ρ are adjustment parameters; The feature data of the door opening feature set or the door closing feature set of the data to be tested are combined with the health status feature data to perform cluster analysis using DPC. If the cluster analysis results of the health status data and the data to be tested are divided into two different categories, and each category includes samples of the health status data or the data to be tested that are not less than the set proportion, it is determined that a sub-healthy state exists; if the cluster analysis results cannot select two cluster centers, it is considered that a sub-healthy state does not exist; The steps for cluster analysis of the door opening process or door closing process using DPC are as follows: The feature set of the gate system's normal operation data to be tested and the feature set of the healthy data are stored as a feature matrix, and the local density ρ of each sample i is calculated. i and the distance δ from sample i to the sample j whose local density is larger than that of the sample i and closest to it i , the calculation formula is: r i =∑χ(d ij -d c ) Where, d ij is the Euclidean distance between data points i and j, d c is the pre-specified cutoff distance; χ is the discriminant function, when (d ij -d c )>0, χ=1, otherwise χ=0; according to the calculated ρ i and δ i , with ρ i is the horizontal axis, δ i As the vertical axis, draw a decision diagram; And use the decision diagram to select the cluster center.
2. The rail vehicle door system sub-health diagnosis method according to claim 1, characterized in that: In the step (1), the door system opening and closing process is driven by a motor, and the door system operation status curve obtained includes the stroke, speed, and current status curves generated during the door opening and closing processes.
3. The rail vehicle door system sub-health diagnosis method according to claim 2, characterized in that: In the step (2), for each operating state curve, combining the physical mechanism of the subhealthy operating state and the difference between the subhealthy operating state curve and the healthy operating state curve, obtaining the time period when different subhealthy operating states are different from the healthy operating state; The multi-component action time period is obtained by combining the time periods that are different between different sub-healthy operating states and healthy operating states.
4. The rail vehicle door system sub-health diagnosis method according to claim 3, characterized in that: In the step (3), the multi-component action period is divided into a number of sub-segments, and feature extraction is performed on each sub-segment.
5. The method for diagnosing sub-health of a rail vehicle door system according to claim 4, characterized in that: The features extracted from the multi-component action period during the door opening process and the door closing process in step (3) include: global information features of the multi-component action period, time domain features of each sub-segment of the multi-component action period and / or morphological features of the operating status curve of the multi-component action period.
6. The method for diagnosing sub-health of a rail vehicle door system according to claim 1, characterized in that: The sub-health diagnosis results of the door opening process and the sub-health diagnosis results of the door closing process of the data to be detected are respectively obtained. When any one of the sub-health diagnosis results of the door opening process and the sub-health diagnosis results of the door closing process is a sub-health state, it is determined that the door system outputting the data to be detected is in a sub-health state.
