Rail transit vehicle door PHM diagnosis method based on scene recognition interference removal

Abstract

The invention provides a rail transit vehicle door PHM diagnosis method based on scene recognition interference elimination, which aims at the condition that opening and closing of a vehicle door are influenced by a scene environment in an operation process, is divided into a main line operation mode, an in-warehouse maintenance mode and an in-warehouse standard mode aiming at train working conditions, and eliminates system maintenance working conditions by recognizing a vehicle door operation scene. According to the method, weighting is carried out on opening and closing doors in a main line and a library, an alarm is given through a fusion strategy, meanwhile, a vehicle door machine inspection report is generated to facilitate maintenance of maintainers, a new method is provided for rail vehicle door diagnosis, and effective door system parameters cannot be obtained through a main line diagnosis method for the vehicle door diagnosis problem influenced by the environment and working conditions; according to the method, the sub-health state of the vehicle door can be well recognized by establishing the scene recognition interference removal model and comparing and analyzing the long-term change trend of the characteristic values of the state parameters in the main line and the library, early warning is achieved, and the method has good practical application prospects.

Description

Technical Field

[0001] The present invention relates to the technical field of rail transit vehicle maintenance, and in particular to a rail transit vehicle door PHM diagnosis method based on scene recognition and interference removal. Background Art

[0002] Rail doors are the interface for people to enter and exit the vehicle. They are one of the most frequently used parts of the vehicle. As a key part for passengers to enter and exit the vehicle, their safety and reliability directly affect passenger safety and operational order. Existing door maintenance methods mainly include fault repair and planned repair. Fault repair means that after-sales personnel will handle the faulty door immediately after clearing passengers, immediately return it to the warehouse for processing after isolation, and return it to the warehouse for processing after the operation ends; planned repair means that maintenance personnel will perform frame repair, overhaul, third-level repair, fifth-level repair, etc. on the door according to the closed-loop management of fault information and the quality management system.

[0003] However, these methods are difficult to effectively predict and diagnose the sub-health status of car doors in the face of complex and changing operating environments. In order to carry out fault prediction and health management of car doors, advanced sensor technology is used to obtain the operating status information of the car door system and realize the intelligent operation and maintenance management of car doors. After the car door reports the sub-health status, the maintenance personnel will give different suggestions according to different sub-health status, that is, maintenance guidance: maintenance cycle, maintenance tools, maintenance steps, maintenance parts, and the manual inspection work can be greatly reduced through status maintenance.

[0004] The existing sub-health diagnosis is based on the data of the door opening and closing during the main operation, which has a good effect under the premise of stable and effective data. The actual operation scenario is complex, with temperature, wind pressure, system repair and lubrication, etc., which have a great impact on the real-time data collection of the motor and the IO signal parameters of the door controller, affecting the diagnosis effect. Summary of the invention

[0005] Purpose of the invention: The purpose of the present invention is to provide a PHM diagnostic method for rail transit doors based on scene recognition and interference removal, which can realize accurate diagnosis and early warning of door status by identifying the door operation scene, eliminating the interference of system maintenance conditions, and combining weighted analysis of mainline and in-library data.

[0006] Technical solution: The present invention provides a rail transit vehicle door PHM diagnostic method based on scene recognition and interference removal, comprising the following steps:

[0007] (1) The operating status information of the vehicle door system is collected through the sensors in the door system drive motor and the door controller, analyzed and pre-processed, and transmitted to the database; the operating status information of the vehicle door includes speed, angle, current, lock-in position switch jump position, auxiliary lock action time, door controller IO signal and station information; (2) The scene recognition module identifies the door operation scene according to the train operating conditions. If it is the mainline operation mode, the temperature, wind pressure, and maintenance information of the door operation are obtained. If it is the maintenance mode, the data in the diagnosis and filtering maintenance mode are obtained. The train operating conditions include the mainline operation mode, the in-depot maintenance mode, and the in-depot standard mode. (3) The mainline diagnosis module inputs temperature, wind pressure and maintenance information, diagnoses the mainline door opening and closing status data through the resistance threshold model and intelligent sub-health algorithm model, and records the diagnosis results into the database; (4) The in-depot secondary diagnosis module is used to perform in-depot secondary diagnosis on the doors of vehicles with sub-health events on the main line, and the in-depot door opening and closing is carried out under specified conditions; (5) Through the alarm push module, the results of the mainline and in-depot diagnosis points are weighted by fusion strategy to warn of sub-health of vehicle doors in real time; (6) Calculate the key feature values ​​of the door status information through the multi-door comparison module, and use the multi-door horizontal comparison algorithm model to mark abnormal doors; (7) The door inspection report module generates a door inspection report based on the fusion diagnosis results and multi-door comparison, and provides suggestions on maintenance cycle, tools, steps and parts as a basis for maintenance.

[0008] Furthermore, in step (3), the door switch data is judged, and the threshold model and intelligent algorithm model are diagnosed for the mainline operation data according to the input parameters and scene recognition, and the results are recorded, which specifically includes the following steps: (31) Filter, align, and remove redundant data from the door switch data, segment the speed, angle, and current data, and distinguish the eigenvalue data based on mathematical statistics; (32) Obtain input parameters such as temperature, wind pressure and maintenance information. If maintenance information exists, filter the switch diagnostic data. If temperature and wind pressure exist, automatically adjust the threshold of the threshold model according to the input parameters. (33) Identify the train operating conditions corresponding to the door switch data, including the mainline operation mode, the depot maintenance mode, and the depot standard mode, and diagnose the mainline operation mode; (34) Based on the change of eigenvalues, the maximum value, minimum value, mean value, standard deviation, skewness and kurtosis of each segment are calculated, and a threshold model is constructed to set the threshold of the item point eigenvalue. The formula is as follows: ); ); ; ; ; ; Where m is the number of eigenvalue samples, is the eigenvalue data size, is the mean eigenvalue, is the eigenvalue standard deviation, is the eigenvalue skewness, is the eigenvalue kurtosis; (35) Use intelligent algorithms to build models based on data-driven methods, calculate segmented eigenvalues ​​of normal and abnormal data and perform standardization, perform weighted clustering on eigenvalues, sort by cluster distance, select core eigenvalues ​​and change trends, establish corresponding sub-health model rules, and use the corresponding model rules for diagnosis; (36) Abnormal data diagnosed by the threshold model and intelligent algorithm model are marked and recorded as events in the database.

[0009] Furthermore, in step (4), the door opening and closing in the garage is carried out under specified conditions, and the door opening and closing in the garage requires normal voltage of the bow, no air conditioning startup, no wind pressure, no body deformation, and no passenger interference.

[0010] Furthermore, in step (6), the multi-door comparison module uses big data statistical analysis methods. When the standard data in the library is confirmed, the cross-sectional data such as the closed switch jump position, the locked switch jump position, the door opening and closing action time, and the opening degree are analyzed horizontally using box plots and z-score methods to locate the outlier doors.

[0011] Furthermore, in step (7), the door inspection report counts the alarm conditions of various sub-health points in each train cycle and compares multiple doors with key characteristic values, marks abnormal doors, lists the required maintenance parts, maintenance steps and tools, and generates a maintenance level based on the sub-health status of the door.

[0012] The present invention provides a rail transit door PHM diagnostic system based on scene recognition and interference removal, comprising: Data acquisition module: used to collect the operating status information of the vehicle door system through the sensors in the door system drive motor and the door controller, perform preprocessing after analysis, and transmit it to the database; the vehicle door operating status information includes speed, angle, current, lock-in position switch jump position, auxiliary lock action time, door controller IO signal and station information; Scene recognition module: used to identify the door operation scene according to the train working condition. If it is the mainline operation mode, the temperature, wind pressure and maintenance information of the door operation are obtained. If it is the maintenance mode, the data in the diagnosis and filtering maintenance mode are diagnosed. Among them, the train working conditions include the mainline operation mode, the in-depot maintenance mode and the in-depot standard mode. Mainline diagnostic module: used to input temperature, wind pressure and maintenance information, diagnose the mainline door opening and closing status data through the resistance threshold model and intelligent sub-health algorithm model, and record the diagnosis results into the database; In-depot secondary diagnosis module: used to perform in-depot secondary diagnosis on the doors of vehicles with sub-health events on the main line. The in-depot door opening and closing is carried out under specified conditions; Alarm push module: used to integrate the results of the mainline and in-depot diagnosis points and give real-time warnings of sub-health of vehicle doors; Multi-door comparison module: used to calculate the key feature values ​​of door status information and mark abnormal doors using a multi-door horizontal comparison algorithm model; Door inspection report module: used to generate door inspection reports based on fusion diagnosis results and multi-door comparisons, and provide suggestions on inspection cycles, tools, steps and locations as a basis for inspection.

[0013] Furthermore, in the mainline diagnosis module, the door switch data is judged, and the threshold model and intelligent algorithm model are diagnosed for the mainline operation data according to the input parameters and scene recognition, and the results are recorded, as follows: Filter, align, and remove redundant data from the door switch data, segment the speed, angle, and current data, and distinguish the eigenvalue data based on mathematical statistics. Obtain input parameters such as temperature, wind pressure and maintenance information. If maintenance information exists, filter the switch diagnostic data. If temperature and wind pressure exist, automatically adjust the threshold value of the threshold model according to the input parameters. Identify the train operating conditions corresponding to the door switch data, including the mainline operation mode, the depot maintenance mode, and the depot standard mode, and diagnose the mainline operation mode; Based on the change of eigenvalues, the maximum value, minimum value, mean value, standard deviation, skewness and kurtosis of each segment are calculated, and a threshold model is constructed to set the threshold of the item point eigenvalue. The formula is as follows: ) ) Where m is the number of eigenvalue samples, is the eigenvalue data size, is the mean eigenvalue, is the eigenvalue standard deviation, is the eigenvalue skewness, is the eigenvalue kurtosis; Using intelligent algorithms to build models based on data-driven methods, statistically segmenting eigenvalues ​​of normal and abnormal data and performing standardization, weighted clustering of eigenvalues, sorting by cluster distance, selecting core eigenvalues ​​and change trends, establishing corresponding sub-health model rules, and using corresponding model rules for diagnosis; Abnormal data diagnosed by the threshold model and intelligent algorithm model are marked and recorded as events in the database.

[0014] Furthermore, in the secondary diagnostic module inside the warehouse, the opening and closing of the doors inside the warehouse are carried out under specified conditions. The opening and closing of the doors inside the warehouse require normal voltage of the bow, no air conditioning started, no wind pressure, no body deformation, and no passenger interference.

[0015] Furthermore, in the multi-door comparison module, big data statistical analysis methods are used. When the standard data in the library is confirmed, the cross-sectional data such as the closed switch jump position, the locked switch jump position, the door opening and closing action time, and the opening degree are used to horizontally analyze and locate the outlier doors using methods such as box plots and z-score.

[0016] Furthermore, in the door machine inspection report module, the alarm conditions of various sub-health points in each train cycle and the multi-door comparison of key characteristic values ​​are counted, abnormal doors are marked, the required maintenance parts, maintenance steps and tools are listed, and the maintenance level is generated according to the sub-health status of the door.

[0017] Beneficial effects: Compared with the prior art, the present invention has the following advantages: in view of the scene interference problem in the diagnosis of sub-healthy doors, the system maintenance conditions and external environmental interference are eliminated through scene recognition; the mainline and in-depot fusion diagnosis strategy is used to achieve accurate diagnosis and improve the diagnostic accuracy of the sub-healthy state of the doors; at the same time, real-time monitoring and early warning of the door status are realized to reduce the probability of failure and ensure passenger safety and operational order; based on big data and intelligent algorithms, a scientific basis is provided for door maintenance, the inspection and maintenance strategy is optimized, and the maintenance cost is reduced; an intuitive and effective door machine inspection report is provided to facilitate maintenance personnel to quickly locate problems and implement maintenance, thereby achieving accurate diagnosis and health management of the door status and providing strong guarantees for the safe operation of rail transit vehicles. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 A flow chart of the rail transit door PHM diagnostic method based on scene recognition and interference removal provided by the present invention; Figure 2 A threshold diagnosis flow chart provided by the present invention; Figure 3 The intelligent algorithm diagnosis flow chart provided by the present invention; Figure 4 This is a multi-gate comparative diagnosis flow chart provided by the present invention. DETAILED DESCRIPTION

[0019] The present invention will be further described in detail below in conjunction with the accompanying drawings and specific embodiments.

[0020] like Figure 1 As shown, an embodiment of the present invention provides a rail transit vehicle door PHM diagnosis method based on scene recognition and interference removal, comprising the following steps: S1, collects the operating status information of the vehicle door system through the sensor in the door system drive motor and the door controller, analyzes it, pre-processes it, and transmits it to the database. The vehicle door operating status information includes speed, angle, current, lock-in position switch jump position, auxiliary lock action time, door controller IO signal and station information; S2, the scene recognition module identifies the door operation scene according to the train working condition. If it is the mainline operation mode, the temperature, wind pressure and maintenance information of the door operation are obtained. If it is the maintenance mode, the data in this mode is diagnosed and filtered. The train working condition includes the mainline operation mode, the in-depot maintenance mode and the in-depot standard mode. S3, the mainline diagnosis module inputs temperature, wind pressure and maintenance information, and diagnoses the mainline door switch status data through the resistance threshold model and the intelligent sub-health algorithm model, such as Figure 2 and Figure 3 As shown, recording the diagnosis results into a database specifically includes the following steps: (S31), filtering, aligning, and removing redundant acquisition of the door switch data, segmenting the speed, angle, and current data, and distinguishing each eigenvalue data based on mathematical statistics calculation; (S32), obtain input parameters such as temperature, wind pressure and maintenance information. If there is maintenance information, filter the switch diagnostic data. If there is temperature and wind pressure, automatically adjust the threshold model according to the input parameters. Under normal temperature and no wind pressure conditions, calculate the mean and standard deviation of the characteristic value. According to the normal distribution, the threshold is set as shown in formula (1). (1) in is the mean eigenvalue, is the eigenvalue standard deviation; (S33), identifying the train operating conditions corresponding to the door switch data, including the mainline operation mode, the in-depot maintenance mode, and the in-depot standard mode, and diagnosing the mainline operation mode; (S34), the threshold model calculates the maximum, minimum, mean, standard deviation, skewness, and kurtosis of each segment based on the change of the characteristic value, and sets the threshold for the characteristic value of the item point, as shown in formula (2) to formula (7). ) (2); ) (3); (4); (5); (6); (7) Where m is the number of eigenvalue samples, is the eigenvalue data size, is the mean eigenvalue, is the eigenvalue standard deviation, is the eigenvalue skewness, is the eigenvalue kurtosis; (S35), the intelligent algorithm is modeled based on a data-driven method, and the segmented eigenvalues ​​of normal data and abnormal data are statistically analyzed and standardized, and the eigenvalues ​​are weighted clustered, sorted according to the cluster distance, and the core eigenvalues ​​and change trends are selected to establish the corresponding sub-health model rules, such as the concepts of lower gear pin interference and pressure wheel overpressure, and the corresponding model rules are used for diagnosis; (S36), marking the abnormal data diagnosed by the threshold model and the intelligent algorithm model, and recording them as events into the database; S4, the in-depot secondary diagnosis module is used to perform in-depot secondary diagnosis on the doors of the main line with sub-health events. The in-depot door opening and closing is carried out under the specified conditions, that is, the in-depot door opening and closing requires the normal voltage of the bow, no air conditioning startup and no wind pressure, no body deformation, and no passenger interference; S5, through the alarm push module, the results of the diagnosis points on the main line and in the library are weighted by the fusion strategy, and the sub-health of the door is warned in real time. For example, the weights of the main line and in the library of the lower gear pin interference point are 0.3 and 0.7 respectively, and the final alarm probability is set to 0.6. When the main line confidence is 0.8 and the library confidence is 0.6, the alarm probability is 0.3*0.8+0.7*0.6=0.66, which is greater than 0.6, and an alarm is given; S6, through the multi-door comparison module, calculate the key feature values ​​of the door status information, use the multi-door horizontal comparison algorithm model to mark the abnormal doors, and use the big data statistical analysis method. When it is confirmed as the standard data in the library, the cross-sectional data such as the closed switch jump position, the locked switch jump position, the door opening and closing action time, and the opening degree are used to horizontally analyze and locate the outlier doors using methods such as box plots and z-score, such as Figure 4 As shown, the upper and lower limits of the box plot are shown in formulas (8) and (9). (8); (9); in, is the upper quartile, is the lower quartile, and IQR is the interquartile range - ; S7, through the door machine inspection report module, the door machine inspection report counts the alarm conditions of various sub-health points in each train cycle and compares multiple doors with key characteristic values, marks abnormal doors, lists the required maintenance parts, maintenance steps and tools, and generates maintenance levels according to the sub-health status of the doors, providing maintenance personnel with a basis for maintenance and greatly reducing maintenance hours.

[0021] The embodiment of the present invention further provides a rail transit vehicle door PHM diagnostic system based on scene recognition and interference removal, comprising: Data acquisition module: used to collect the operating status information of the vehicle door system through the sensors in the door system drive motor and the door controller, perform preprocessing after analysis, and transmit it to the database; the vehicle door operating status information includes speed, angle, current, lock-in position switch jump position, auxiliary lock action time, door controller IO signal and station information; Scene recognition module: used to identify the door operation scene according to the train working condition. If it is the mainline operation mode, the temperature, wind pressure and maintenance information of the door operation are obtained. If it is the maintenance mode, the data in the diagnosis and filtering maintenance mode are diagnosed. Among them, the train working conditions include the mainline operation mode, the in-depot maintenance mode and the in-depot standard mode. Mainline diagnostic module: used to input temperature, wind pressure and maintenance information, diagnose the mainline door switch status data through the resistance threshold model and intelligent sub-health algorithm model, and record the diagnosis results in the database; judge the door switch data, diagnose the mainline operation data with the threshold model and intelligent algorithm model according to the input parameters and scene recognition, and record the results, as follows: Filter, align, and remove redundant data from the door switch data, segment the speed, angle, and current data, and distinguish the eigenvalue data based on mathematical statistics. Obtain input parameters such as temperature, wind pressure and maintenance information. If maintenance information exists, filter the switch diagnostic data. If temperature and wind pressure exist, automatically adjust the threshold value of the threshold model according to the input parameters. Identify the train operating conditions corresponding to the door switch data, including the mainline operation mode, the depot maintenance mode, and the depot standard mode, and diagnose the mainline operation mode; Based on the change of eigenvalues, the maximum value, minimum value, mean value, standard deviation, skewness and kurtosis of each segment are calculated, and a threshold model is constructed to set the threshold of the item point eigenvalue. The formula is as follows: ); ); ; ; ; ; Where m is the number of eigenvalue samples, is the eigenvalue data size, is the mean eigenvalue, is the eigenvalue standard deviation, is the eigenvalue skewness, is the eigenvalue kurtosis; Using intelligent algorithms to build models based on data-driven methods, statistically segmenting eigenvalues ​​of normal and abnormal data and performing standardization, weighted clustering of eigenvalues, sorting by cluster distance, selecting core eigenvalues ​​and change trends, establishing corresponding sub-health model rules, and using corresponding model rules for diagnosis; Abnormal data diagnosed by the threshold model and intelligent algorithm model are marked and recorded as events in the database.

[0022] In-depot secondary diagnosis module: used for in-depot secondary diagnosis of car doors with sub-health events on the main line. The opening and closing of the car doors in the depot are carried out under specified conditions. The opening and closing of the car doors in the depot require normal voltage of the bow, no air conditioning started, no wind pressure, no body deformation, and no passenger interference.

[0023] Alarm push module: used to integrate the results of the mainline and in-depot diagnosis points and give real-time warnings of sub-health of vehicle doors; Multi-door comparison module: used to calculate the key feature values ​​of the door status information, and use the multi-door horizontal comparison algorithm model to mark abnormal doors; use big data statistical analysis methods, when confirmed as standard data in the library, through the cross-sectional data such as the closed switch jump position, the locked switch jump position, the door opening and closing action time, the opening degree, etc., use box plots and z-score methods to horizontally analyze and locate outlier doors.

[0024] Door inspection report module: used to generate door inspection reports based on fusion diagnosis results and multi-door comparison, and provide suggestions on inspection cycles, tools, steps and parts as the basis for inspection. Statistics are collected on the alarm conditions of sub-health points and multi-door comparison of key characteristic values ​​in each train cycle, abnormal doors are marked, and the required inspection parts, inspection steps and tools are listed, and maintenance levels are generated according to the sub-health status of the doors.

Claims

1. A PHM diagnosis method for rail transit doors based on scene recognition and interference removal. It is characterized in that The following steps are involved: (1) The operating status information of the vehicle door system is collected through the sensors in the door system drive motor and the door controller, analyzed and pre-processed, and transmitted to the database; the operating status information of the vehicle door includes speed, angle, current, lock-in position switch jump position, auxiliary lock action time, door controller IO signal and station information; (2) The scene recognition module identifies the door operation scene according to the train operating conditions. If it is the mainline operation mode, the temperature, wind pressure, and maintenance information of the door operation are obtained. If it is the maintenance mode, the data in the diagnosis and filtering maintenance mode are obtained. The train operating conditions include the mainline operation mode, the in-depot maintenance mode, and the in-depot standard mode. (3) The mainline diagnosis module inputs temperature, wind pressure and maintenance information, diagnoses the mainline door opening and closing status data through the resistance threshold model and intelligent sub-health algorithm model, and records the diagnosis results into the database; (4) The in-depot secondary diagnosis module is used to perform in-depot secondary diagnosis on the doors of vehicles with sub-health events on the main line, and the in-depot door opening and closing is carried out under specified conditions; (5) Through the alarm push module, the results of the mainline and in-depot diagnosis points are weighted by fusion strategy to warn of sub-health of vehicle doors in real time; (6) Calculate the key feature values ​​of the door status information through the multi-door comparison module, and use the multi-door horizontal comparison algorithm model to mark abnormal doors; (7) The vehicle door inspection report module generates a vehicle door inspection report based on the fusion diagnosis results and multi-door comparison, and provides maintenance cycle, tools, steps and location suggestions as a basis for maintenance.

2. According to claim 1, a rail transit door PHM diagnosis method based on scene recognition and interference removal, It is characterized in that In step (3), the door switch data is judged, and the threshold model and intelligent algorithm model are diagnosed for the mainline operation data according to the input parameters and scene recognition, and the results are recorded, which specifically includes the following steps: (31) Filter, align, and remove redundant data from the door switch data, segment the speed, angle, and current data, and distinguish the eigenvalue data based on mathematical statistics; (32) Obtaining temperature, wind pressure and maintenance information input parameters, if maintenance information exists, filtering the switch diagnostic data, if temperature and wind pressure exist, automatically adjusting the threshold of the threshold model according to the input parameters; (33) Identify the train operating conditions corresponding to the door switch data, including the mainline operation mode, the depot maintenance mode, and the depot standard mode, and diagnose the mainline operation mode; (34) Based on the change of eigenvalues, the maximum value, minimum value, mean value, standard deviation, skewness and kurtosis of each segment are calculated, and a threshold model is constructed to set the threshold of the item point eigenvalue. The formula is as follows: )； )； ； ； ； ； Where m is the number of eigenvalue samples, is the eigenvalue data size, is the mean eigenvalue, is the eigenvalue standard deviation, is the eigenvalue skewness, is the eigenvalue kurtosis; (35) Use intelligent algorithms to build models based on data-driven methods, calculate segmented eigenvalues ​​of normal and abnormal data and perform standardization, perform weighted clustering on eigenvalues, sort by cluster distance, select core eigenvalues ​​and change trends, establish corresponding sub-health model rules, and use the corresponding model rules for diagnosis; (36) Abnormal data diagnosed by the threshold model and intelligent algorithm model are marked and recorded as events in the database.

3. According to claim 1, a rail transit vehicle door PHM diagnosis method based on scene recognition and interference removal, It is characterized in that In step (4), the door opening and closing in the garage is carried out under specified conditions. The door opening and closing in the garage requires normal voltage of the bow, no air conditioning startup, no wind pressure, no body deformation, and no passenger interference.

4. According to claim 1, a rail transit door PHM diagnosis method based on scene recognition and interference removal, It is characterized in that In step (6), the multi-door comparison module uses big data statistical analysis methods. When the standard data in the library is confirmed, the outlier door is located by horizontal analysis using box plots and z-score methods through the closed switch jump position, locked switch jump position, door opening and closing action time, and opening cross-sectional data.

5. According to claim 1, a rail transit vehicle door PHM diagnosis method based on scene recognition and interference removal, It is characterized in that In step (7), the door inspection report counts the alarm conditions of various sub-health points in each train cycle and compares multiple doors with key characteristic values, marks abnormal doors, lists the required maintenance parts, maintenance steps and tools, and generates a maintenance level based on the sub-health status of the door.

6. A rail transit door PHM diagnostic system based on scene recognition and interference removal, It is characterized in that include: Data acquisition module: used to collect the operating status information of the vehicle door system through the sensors in the door system drive motor and the door controller, perform preprocessing after analysis, and transmit it to the database; the vehicle door operating status information includes speed, angle, current, lock-in position switch jump position, auxiliary lock action time, door controller IO signal and station information; Scene recognition module: used to identify the door operation scene according to the train working condition. If it is the mainline operation mode, the temperature, wind pressure and maintenance information of the door operation are obtained. If it is the maintenance mode, the data in the diagnosis and filtering maintenance mode are diagnosed. Among them, the train working conditions include the mainline operation mode, the in-depot maintenance mode and the in-depot standard mode. Mainline diagnostic module: used to input temperature, wind pressure and maintenance information, diagnose the mainline door opening and closing status data through the resistance threshold model and intelligent sub-health algorithm model, and record the diagnosis results into the database; In-depot secondary diagnosis module: used to perform in-depot secondary diagnosis on the doors of vehicles with sub-health events on the main line. The in-depot door opening and closing is carried out under specified conditions; Alarm push module: used to integrate the results of the mainline and in-depot diagnosis points and give real-time warnings of sub-health of vehicle doors; Multi-door comparison module: used to calculate the key feature values ​​of door status information and mark abnormal doors using a multi-door horizontal comparison algorithm model; Door inspection report module: used to generate door inspection reports based on fusion diagnosis results and multi-door comparisons, and provide inspection cycles, tools, steps and location recommendations as a basis for inspection.

7. A rail transit door PHM diagnostic system based on scene recognition and interference removal according to claim 6, It is characterized in that In the mainline diagnosis module, the door switch data is judged, and the threshold model and intelligent algorithm model are used to diagnose the mainline operation data according to the input parameters and scene recognition, and the results are recorded as follows: Filter, align, and remove redundant data from the door switch data, segment the speed, angle, and current data, and distinguish the eigenvalue data based on mathematical statistics. Obtain temperature, wind pressure and maintenance information input parameters. If maintenance information exists, filter the switch diagnostic data. If temperature and wind pressure exist, automatically adjust the threshold value of the threshold model according to the input parameters. Identify the train operating conditions corresponding to the door switch data, including the mainline operation mode, the depot maintenance mode, and the depot standard mode, and diagnose the mainline operation mode; Based on the change of eigenvalues, the maximum value, minimum value, mean value, standard deviation, skewness and kurtosis of each segment are calculated, and a threshold model is constructed to set the threshold of the item point eigenvalue. The formula is as follows: )； )； ； ； ； ； Where m is the number of eigenvalue samples, is the eigenvalue data size, is the mean eigenvalue, is the eigenvalue standard deviation, is the eigenvalue skewness, is the eigenvalue kurtosis; Using intelligent algorithms to build models based on data-driven methods, statistically segmenting eigenvalues ​​of normal and abnormal data and performing standardization, weighted clustering of eigenvalues, sorting by cluster distance, selecting core eigenvalues ​​and change trends, establishing corresponding sub-health model rules, and using corresponding model rules for diagnosis; Abnormal data diagnosed by the threshold model and intelligent algorithm model are marked and recorded as events in the database.

8. According to claim 6, a rail transit door PHM diagnostic system based on scene recognition and interference removal, It is characterized in that In the secondary diagnostic module inside the depot, the opening and closing of the doors inside the depot are carried out under specified conditions. The opening and closing of the doors inside the depot requires normal voltage of the bow, no air conditioning started, no wind pressure, no body deformation, and no passenger interference.

9. A rail transit door PHM diagnostic system based on scene recognition and interference removal according to claim 6, It is characterized in that In the multi-door comparison module, big data statistical analysis methods are used. When the standard data in the library is confirmed, the outlier doors are located through horizontal analysis using box plots and z-score methods through the closed switch jump position, locked switch jump position, door opening and closing action time, and opening cross-sectional data.

10. A rail transit door PHM diagnostic system based on scene recognition and interference removal according to claim 6, It is characterized in that In the door inspection report module, statistics are collected on the alarm conditions of various sub-health points in each train cycle and multi-door comparisons of key characteristic values ​​are performed. Abnormal doors are marked, and the required maintenance parts, maintenance steps and tools are listed, and maintenance levels are generated based on the sub-health status of the doors.

Images