Rail transit vehicle brake system adaptive switching fault diagnosis method

Abstract

This invention provides an adaptive switching fault diagnosis method for the braking system of rail transit vehicles. This invention integrates the main air duct, air compressor, valve components, and braking actuators in the braking system of rail transit vehicles into a unified diagnostic framework. It overcomes the problems of existing single-component early warning models being fragmented and having difficulty comprehensively judging multiple abnormal results. It enables unified analysis of associated anomalies across multiple components of the braking system. Through operational scenario identification, component-level candidate anomaly identification, Bayesian coupled reasoning, and anomaly coupling degree calculation, it can distinguish between single faults, coupled faults, fault propagation results, and scenario-induced anomalies when multiple candidate anomalies occur simultaneously, thereby improving the accuracy of locating the dominant fault source.

Description

Technical Field

[0001] This invention relates to the field of fault diagnosis and predictive health management technology for rail transit vehicles, and in particular to an adaptive switching fault diagnosis method for the braking system of rail transit vehicles. Background Technology

[0002] The braking system of rail transit vehicles includes air supply units, main air ducts, brake control valves, brake cylinders, and basic braking devices. These components are interconnected by air passages, pressure transmission, and timing relationships. With the development of vehicle PHM (Prognostics and Health Management) technology, the onboard system can collect data such as main air duct pressure, air compressor start signals, braking commands, train speed, valve pressure, and station distance, and transmit this data to a ground-based PHM platform for status alerts.

[0003] Existing PHM methods for braking systems mostly model individual components. For example, they might use the slope of the main duct pressure drop to determine air tightness anomalies, the slope of the pressure rise to determine air compressor blower capacity anomalies, and the pressure difference and duration of valves to determine average valve pressure response anomalies. In actual operation, main duct leaks, air compressor performance degradation, and valve pressure anomalies can all propagate and trigger warnings simultaneously. Existing methods struggle to identify the primary fault source and the fault propagation relationship.

[0004] Furthermore, scenarios such as depot parking for air replenishment, turnaround starts, pre-station braking, and frequent starts and stops can cause changes in pressure and operating time, easily leading to false alarms. Therefore, there is an urgent need to propose an adaptive switching diagnostic method for single and coupled faults in the braking system based on vehicle-to-ground cooperative PHM, in order to improve fault differentiation, source location, and maintenance guidance capabilities. Summary of the Invention

[0005] The purpose of this invention is to provide an adaptive switching fault diagnosis method for the braking system of rail transit vehicles, which solves the problem of difficulty in accurately locating the fault source when multiple components of the braking system are abnormal at the same time.

[0006] To achieve the above objectives, the present invention provides a method for diagnosing adaptive switching faults in the braking system of rail transit vehicles, the method comprising the following steps: S1. Acquire vehicle operating status data and braking system status data; S2: Identify vehicle operating scenarios based on vehicle operating parameters; S3: Based on the air circuit structure and functional relationship of the braking system, build a component-level candidate anomaly identification model and output multi-dimensional anomaly features; S4: Determine the candidate anomalies of the main duct air tightness, compare the main duct pressure drop rate with the historical baseline threshold, and output the candidate results of the main duct air tightness anomalies. S5: Determine candidate anomalies in the air compressor's pressure rise capability, compare with the baseline threshold of the pressure rise rate, and output candidate results for air compressor pressure rise capability anomalies. S6: Extract the multi-class duration features of the air compressor, compare them with the corresponding baseline thresholds, and output the candidate results of the reasonableness of the air compressor working duration anomaly. S7: Calculate valve pressure difference in the pre-station braking scenario. Output candidate results for abnormal valve pressure response when the pressure difference and duration meet the conditions. S8: Convert the output results of each component-level candidate anomaly identification model into a unified evidence set; S9: Construct a Bayesian network containing nodes for scene, component status, evidence, and diagnostic results, and construct directed edges of the Bayesian network based on gas path, pressure, and time sequence; S10: Calculate the probability of each type of diagnostic result using Bayesian inference; S11: Calculate the abnormal coupling degree of any two components with candidate abnormalities at the same time, and compare it with the preset coupling threshold to determine the fault transmission relationship or coupling relationship; S12: Based on the probability of the diagnostic results and the abnormal coupling degree of the components, perform a diagnostic mode switch.

[0007] Furthermore, the vehicle operation status data and braking system status data specifically include the collection of vehicle speed, stopping time, station distance, driver's cab activation status, braking command, emergency braking status, track conditions, main air duct pressure, air compressor start signal, air compressor working status, valve pressure, brake cylinder pressure, and related brake air supply, pressure regulation, and braking execution. The operation status data and braking system status data are collected by the on-board system, train network control system, or brake control unit and transmitted to the ground PHM platform or vehicle-to-ground collaborative diagnostic platform.

[0008] Furthermore, the operating parameters in step S2 include vehicle speed, parking time, station distance, braking command, emergency braking status, air compressor start status, and total duct pressure change.

[0009] Furthermore, the vehicle operation scenarios in step S2 include stationary pressure holding scenario, depot parking and air charging scenario, turnaround and start-up scenario, pre-station braking scenario, frequent start-stop scenario, continuous braking scenario, and normal operation scenario.

[0010] Furthermore, the component-level candidate anomaly identification model includes a main duct air tightness anomaly identification model, an air compressor pressure rise capability anomaly identification model, an air compressor working time rationality anomaly identification model, a valve pressure response anomaly identification model, and a brake execution anomaly identification model; the multi-dimensional anomaly features include candidate anomaly type, anomaly probability, anomaly confidence level, anomaly occurrence time, anomaly duration, and key anomaly features.

[0011] Furthermore, step S4 specifically includes: when the vehicle is stationary, there is no braking command, the air compressor is not started, and the parking time meets the preset conditions, extracting the pressure drop range of the main air duct and calculating the pressure drop rate of the main air duct. : in, This indicates the starting pressure of the pressure drop range; This indicates the pressure at the end of the pressure drop range; Indicates the duration of the pressure drop interval; Indicates the rate of decrease in total duct pressure; when If the pressure drop interval exceeds the historical baseline threshold for the corresponding vehicle and the duration exceeds the preset time, output the candidate result for the main duct airtightness anomaly. The historical baseline threshold is determined by the statistical values ​​of the historical valid samples of the same vehicle, the historical valid samples of the same model, or the slope of the sliding window according to preset statistical rules. The preset statistical rules include SPC control limits, the mean plus or minus a preset multiple of standard deviation, or quantile statistical methods.

[0012] Furthermore, step S5 specifically includes: Step S5.1: When the vehicle is stationary, there is no braking command, no significant air consumption, and the air compressor is running, extract the total duct pressure recovery range during the air compressor startup period and calculate the pressure rise rate. : in, Indicates the pressure at the start of the air compressor or the starting pressure of the pressure recovery range; This indicates the pressure at the moment the air compressor stops or at the end of the pressure recovery range; Indicates the duration of the pressure recovery interval; Indicates the rate of increase of the total duct pressure; Step S5.2: Calculate the mean_rise and standard deviation_rise of the pressure rise rate samples, and determine the baseline threshold for the pressure rise rate as follows: in, Indicates the baseline threshold for the rate of pressure rise; This represents the mean rate of pressure rise in historical normal samples; This represents the standard deviation of the rate of pressure rise in historical normal samples; k is a preset control coefficient with a value of 3. Step S5.3: When Less than At that time, the candidate result for abnormal air compressor pressure rise capability is output.

[0013] Furthermore, step S6 specifically includes: Step S6.1: Extract the continuous operating range of the air compressor based on the air compressor start signal, and calculate the air compressor operating time characteristics, including the single operating time. Cumulative working hours Startup frequency Duty cycle Unit pressure recovery time Among them, unit pressure recovery time Calculate as follows: in, This indicates the duration of continuous operation of the air compressor. This indicates the total duct pressure when the air compressor starts up. This indicates the total duct pressure when the air compressor stopped. Indicates the time required for unit pressure to recover; Step S6.2: When , , , or When at least one of the parameters exceeds the corresponding baseline threshold, output a candidate result for the reasonableness of the air compressor working time; Step S6.3: When the air compressor's operating time is abnormal and accompanied by an abnormal rate of pressure drop in the main duct, the coupling judgment of the abnormal air tightness of the main duct and the abnormal operation of the air compressor is initiated first.

[0014] Furthermore, step S7 specifically includes: Step S7.1: In the scenario of vehicle braking before entering the station, when the total air duct pressure is greater than the preset air supply pressure threshold and emergency braking is not applied, calculate the valve-related pressure difference. : in, and This represents two pressure signals related to valve components; This indicates the absolute pressure difference between the two. Step S7.2: When When the pressure difference exceeds the preset threshold and the duration exceeds the preset duration, output candidate results for abnormal valve pressure response.

[0015] Furthermore, step S7 also includes: in the identification of abnormal average valve pressure response, and The pressures are ASP1 and ASP2, respectively. When |ASP1−ASP2| is greater than 180 kPa and lasts for 30 s, the candidate results of the average valve pressure difference are output.

[0016] Furthermore, the evidence set includes abnormal pressure drop rate, abnormal pressure rise rate, abnormal single operating time of the air compressor, abnormal cumulative operating time of the air compressor, abnormal unit pressure recovery time, and abnormal valve pressure difference. Each piece of evidence includes an evidence value, the time of occurrence, the duration, the degree of deviation, and the confidence level; the degree of deviation is determined by the difference between the current feature value and the historical baseline, the baseline of the same vehicle, or the baseline of the same model.

[0017] Furthermore, step S9 specifically includes: Construct a Bayesian network that includes Scene nodes, Fault nodes, Evidence nodes, and Result nodes; The Scene is used to represent the current operating scenario of the vehicle; The Fault includes the airtightness of the main duct, the pressure rise capability of the air compressor, the working behavior of the air compressor, and the pressure response of the valves. The evidence is formed by the output of the candidate anomaly identification model at each component level; The Result includes single main duct air tightness abnormality, single air compressor air blowing capacity abnormality, single valve pressure response abnormality, main duct air tightness abnormality transmitted to air compressor malfunction, air compressor air blowing capacity abnormality transmitted to valve pressure response abnormality, scenario-induced abnormality, and multi-source independent abnormality. The directed edges of the Bayesian network are determined based on the air circuit structure of the braking system, pressure transmission relationship, and action timing relationship; the air tightness status of the main duct is associated with abnormal pressure drop rate and abnormal air compressor working time; the air compressor pressure rise capability status is associated with abnormal pressure rise rate and abnormal unit pressure recovery time; the valve pressure response status is associated with abnormal valve pressure difference; Scene is associated with Fault and Result to correct the prior probability of fault under different operating scenarios.

[0018] Furthermore, step S10 specifically includes: Step S10.1: For any diagnostic result node The probability calculation method for its diagnostic results is as follows: in, Show the diagnosis result of type i; Represents the set of candidate abnormal evidence; Indicates the current running scenario; This indicates the prior probability of the diagnostic result in the current operating scenario; This represents the conditional probability of the corresponding set of evidence appearing under the given diagnostic result and the current operating scenario; This indicates the probability of the evidence set appearing in the current operating scenario; Step 10.2: When When multiple evidentiary features are included, It can be obtained by combining the conditional probabilities of each evidentiary feature; Step 10.3: The prior probabilities and conditional probabilities mentioned above can be determined from historical operating data, fault work orders, maintenance records or model training results, and can be updated as ground PHM platform data accumulates.

[0019] Furthermore, step S11 specifically includes: For any two components that simultaneously have candidate anomalies and Calculate the abnormal coupling degree between the two. : in, For abnormal timing consistency, used to characterize and Does the sequence of anomalies conform to the fault propagation logic? The degree of matching between the air path or functional transmission relationship of the braking system is used to characterize... and Is there any connection between them in the air supply, pressure regulation, or braking execution links? The correlation of abnormal features is used to characterize whether there are synchronous or correlated changes between the abnormal features of two components. To ensure consistency across operating scenarios, this is used to characterize whether candidate anomalies conform to the fault judgment logic under the current operating scenario; , , , Weighting coefficients, and ; when When the value exceeds the preset coupling threshold, a determination is made. and There exists a fault propagation relationship or coupling relationship; when If the value is not greater than the preset coupling threshold, then determine and There is a lack of effective coupling between them.

[0020] Furthermore, step S12 specifically includes: When the posterior probability of the diagnostic result node corresponding to a single component failure is greater than the first threshold, and the abnormal probability of other components that have a preset association with the component corresponding to the single component failure is less than the second threshold, the single fault diagnosis mode is switched to output the corresponding single component failure. When the posterior probability of the diagnostic result node corresponding to the coupled fault is greater than the first threshold, and the abnormal coupling degree between related components is greater than the coupling threshold, the coupled fault diagnosis mode is switched to output the dominant fault source, affected components and fault propagation path. When the posterior probability of the diagnostic result node corresponding to the scene-induced anomaly is greater than the first threshold, and the deviation of the key anomaly features is less than the fault judgment threshold, switch to the scene-induced anomaly mode, reduce the alarm level or do not output the fault alarm. When the probability of diagnosis results corresponding to multiple single component failures is greater than the first threshold, but the abnormal coupling degree between related components is not greater than the coupling threshold, switch to multi-source independent anomaly mode and output the abnormal component and maintenance suggestions respectively. Finally, the diagnostic results are output, including the diagnostic mode, fault type, dominant fault source, affected components, fault propagation path, abnormal time period, key abnormal characteristics, diagnostic result probability, abnormal coupling degree, confidence level, and maintenance priority.

[0021] This invention integrates the main air duct, air compressor, valve components, and braking actuators in the braking system of rail transit vehicles into a unified diagnostic framework. This overcomes the problems of fragmented early warning models and the difficulty in comprehensively judging multiple abnormal results in existing single-component models, enabling unified analysis of interconnected anomalies across multiple components of the braking system. Through operational scenario identification, component-level candidate anomaly identification, Bayesian coupled inference, and anomaly coupling degree calculation, it can distinguish between single faults, coupled faults, fault propagation results, and scenario-induced anomalies when multiple candidate anomalies occur simultaneously, thereby improving the accuracy of locating the dominant fault source. For situations where main air duct air tightness anomalies, air compressor malfunctions, and valve pressure response anomalies interact, the affected components and fault propagation paths can be further output. Simultaneously, it introduces operational scenario correction mechanisms such as depot shutdown and charging, turnaround and start-up, pre-station braking, and frequent start-stop operations to reduce false alarms caused by changes in normal operating conditions. Furthermore, the diagnostic results can output fault type, key anomaly characteristics, confidence level, and maintenance priority, providing clearer technical basis for on-site maintenance location, fault investigation, and condition-based maintenance. Attached Figure Description

[0022] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0023] This invention collects vehicle operating status data and braking system status data through an on-board system, and the ground-based PHM platform or vehicle-to-ground collaborative diagnostic platform completes the operation scenario identification, component-level candidate anomaly identification, coupled fault reasoning, and diagnostic mode switching.

[0024] This invention provides an adaptive switching diagnostic method for single faults and coupled faults in the braking system of rail transit vehicles based on vehicle-ground cooperative PHM. The method includes the following steps: S1. Acquire vehicle operating status data and braking system status data; S2: Identify vehicle operating scenarios based on vehicle operating parameters; S3: Based on the air circuit structure and functional relationship of the braking system, build a component-level candidate anomaly identification model and output multi-dimensional anomaly features; S4: Determine the candidate anomalies of the main duct air tightness, compare the main duct pressure drop rate with the historical baseline threshold, and output the candidate results of the main duct air tightness anomalies. S5: Determine candidate anomalies in the air compressor's pressure rise capability, compare with the baseline threshold of the pressure rise rate, and output candidate results for air compressor pressure rise capability anomalies. S6: Extract the multi-class duration features of the air compressor, compare them with the corresponding baseline thresholds, and output the candidate results of the reasonableness of the air compressor working duration anomaly. S7: Calculate valve pressure difference in the pre-station braking scenario. Output candidate results for abnormal valve pressure response when the pressure difference and duration meet the conditions. S8: Convert the output results of each component-level candidate anomaly identification model into a unified evidence set. The evidence set also includes normal evidence. Normal evidence refers to the corresponding component-level candidate anomaly identification model not triggering anomalies and key features not exceeding the anomaly judgment threshold. Normal evidence of air compressor pressure rise capability refers to the pressure rise rate calculated in step S5 not being less than the pressure rise rate baseline threshold and no abnormal candidate results of air compressor pressure rise capability being output. S9: Construct a Bayesian network containing nodes for scene, component status, evidence, and diagnostic results, and construct directed edges of the Bayesian network based on gas path, pressure, and time sequence; S10: Calculate the probability of each type of diagnostic result using Bayesian inference; S11: Calculate the abnormal coupling degree of any two components with candidate abnormalities at the same time, and compare it with the preset coupling threshold to determine the fault transmission relationship or coupling relationship; S12: Based on the probability of the diagnostic results and the abnormal coupling degree of the components, perform a diagnostic mode switch; the total duct pressure recovery trend meets the normal air charging conditions, which means that in the case of air compressor shutdown and air charging scenario or turnaround start-up scenario, the total duct pressure continues to rise over time after the air compressor starts, the pressure rise rate is not less than the corresponding pressure rise rate baseline threshold, and no candidate result of total duct air tightness abnormality is triggered.

[0025] See Figure 1 The method steps provided by the present invention will be further explained as follows: The first step is to acquire vehicle operating status data and braking system status data. This includes collecting data on vehicle speed, stopping time, station distance, driver's cab activation status, braking commands, emergency braking status, track conditions, main duct pressure, air compressor start signal, air compressor operating status, valve pressure, brake cylinder pressure, and data related to brake air supply, pressure regulation, and brake execution. This data is collected by the onboard system, train network control system, or brake control unit and transmitted to the ground-based PHM platform or vehicle-to-ground collaborative diagnostic platform.

[0026] The second step is to identify the vehicle's operating scenario. Based on vehicle speed, parking duration, station distance, braking commands, emergency braking status, air compressor start status, and changes in main duct pressure, the current operating scenario of the vehicle is identified. These operating scenarios include static pressure holding scenario, depot stop and air charging scenario, turnaround and start scenario, pre-station braking scenario, frequent start-stop scenario, continuous braking scenario, and normal operation scenario. The results of the operating scenario identification are used to correct subsequent component-level anomaly judgment conditions and fault prior probabilities, reducing false alarms caused by normal air charging, short-term air consumption, and changes in braking response.

[0027] The third step is to establish component-level candidate anomaly identification models. Based on the air circuit structure and functional relationships of the braking system, the braking system is divided into air supply units, main air ducts, brake control valves, and brake actuators, and component-level candidate anomaly identification models are established for each. These models include a main air duct airtightness anomaly identification model, an air compressor pressure rise capability anomaly identification model, an air compressor operating time rationality anomaly identification model, and valve pressure response anomaly identification model. Each model outputs the candidate anomaly type, anomaly probability, anomaly confidence level, anomaly occurrence time, anomaly duration, and key anomaly characteristics.

[0028] The fourth step is to identify candidate anomalies in the main duct's airtightness. In a preferred embodiment, when the vehicle is stationary, there is no braking command, the air compressor is not started, and the parking time meets a preset condition, the pressure drop range of the main duct is extracted, and the pressure drop rate of the main duct is calculated. : in, This indicates the starting pressure of the pressure drop range; This indicates the pressure at the end of the pressure drop range; Indicates the duration of the pressure drop interval; Indicates the rate of decrease in total duct pressure; when If the pressure drop interval exceeds the historical baseline threshold for the corresponding vehicle and the duration exceeds the preset time, output the candidate result for the main duct airtightness anomaly. The historical baseline threshold is determined by the statistical values ​​of the historical valid samples of the same vehicle, the historical valid samples of the same model, or the slope of the sliding window according to preset statistical rules. The preset statistical rules include SPC control limits, the mean plus or minus a preset multiple of standard deviation, or quantile statistical methods.

[0029] The fifth step is to identify candidate anomalies in the air compressor's pressure rise capability. In a preferred embodiment, when the vehicle is stationary, there is no braking command, no significant air consumption, and the air compressor is running, the total duct pressure recovery range during the air compressor startup period is extracted, and the pressure rise rate is calculated. : in, Indicates the pressure at the start of the air compressor or the starting pressure of the pressure recovery range; This indicates the pressure at the moment the air compressor stops or at the end of the pressure recovery range; Indicates the duration of the pressure recovery interval; Indicates the rate of increase of the total duct pressure; when When the pressure rise rate is less than the baseline threshold for the corresponding vehicle, an abnormal candidate result for the air compressor pressure rise capability is output; the baseline threshold for the pressure rise rate is determined by historical normal samples that meet the conditions of being stationary, having no braking command, having no significant air consumption, and having the air compressor start-up; in a preferred embodiment, the mean_rise and standard deviation_rise of the pressure rise rate samples are calculated, and the baseline threshold for the pressure rise rate is determined as follows: in, Indicates the baseline threshold for the rate of pressure rise; This represents the mean rate of pressure rise in historical normal samples; This represents the standard deviation of the rate of pressure rise in historical normal samples; k is a preset control coefficient, preferably 3; when Less than At that time, the candidate result for abnormal air compressor pressure rise capability is output.

[0030] Step 6: Identify candidate anomalies in the reasonableness of air compressor operating time. In a preferred embodiment, the continuous operating range of the air compressor is extracted based on the air compressor start signal, and the air compressor operating time characteristics are calculated, including the single operating time. Cumulative working hours Startup frequency Duty cycle Unit pressure recovery time Among them, unit pressure recovery time Calculate as follows: in, This indicates the duration of continuous operation of the air compressor. This indicates the total duct pressure when the air compressor starts up. This indicates the total duct pressure when the air compressor stopped. Indicates the time required for unit pressure to recover; , , , and The corresponding baseline thresholds are all obtained from historical normal samples under the same operating scenarios and conditions. The baseline thresholds can be determined by adding a preset multiple of the standard deviation to the mean, or by using the P95 or P99 quantiles of historical normal samples. , , , and If at least one of the parameters exceeds the corresponding baseline threshold, output a candidate result indicating that the air compressor operating time is abnormal.

[0031] When the air compressor's operating time is abnormal and accompanied by an abnormal rate of pressure drop in the main duct, the system will prioritize the coupled judgment of the abnormal air tightness of the main duct and the abnormal operation of the air compressor.

[0032] Step 7: Identify candidate anomalies in valve pressure response. In a preferred embodiment, when a vehicle enters the station before braking, the total duct pressure is greater than a preset air supply pressure threshold, and emergency braking is not applied, calculate the valve-related pressure difference. : in, and This represents two pressure signals related to valve components; This indicates the absolute pressure difference between the two. when When the pressure difference exceeds the preset threshold and the duration exceeds the preset duration, output candidate results for abnormal valve pressure response.

[0033] In the identification of abnormal average valve pressure response and The pressures are ASP1 and ASP2, respectively. Preferably, when |ASP1−ASP2| is greater than 180 kPa and lasts for 30 s, the candidate result for the average valve pressure difference is output.

[0034] Step 8: Construct a candidate anomaly evidence set. Convert the outputs of each component-level candidate anomaly identification model into a unified evidence set.

[0035] The evidence includes abnormal pressure drop rate, abnormal pressure rise rate, abnormal single operating time of air compressor, abnormal cumulative operating time of air compressor, abnormal unit pressure recovery time, and abnormal valve pressure difference.

[0036] Each piece of evidence includes an evidence value, the time of occurrence, the duration, the degree of deviation, and the confidence level. The degree of deviation is determined by the difference between the current feature value and the historical baseline, the baseline of the same vehicle, or the baseline of the same model.

[0037] Step 9: Construct a coupled fault Bayesian inference model. In a preferred embodiment, a Bayesian network is established, comprising Scene node, Fault node, Evidence node, and Result node.

[0038] The Scene represents the current operating scenario of the vehicle; the Fault includes the air tightness status of the main duct, the air compressor pressure rise capability status, the air compressor operating behavior status, and the valve pressure response status; the Evidence is formed by the output of the candidate anomaly identification model at each component level; the Result includes single main duct air tightness anomaly, single air compressor air blowing capability anomaly, single valve pressure response anomaly, main duct air tightness anomaly transmitted to air compressor operating anomaly, air compressor air blowing capability anomaly transmitted to valve pressure response anomaly, scenario-induced anomaly, and multi-source independent anomaly; among them, the main duct air tightness anomaly transmitted to air compressor operating anomaly, the air compressor air blowing capability anomaly transmitted to valve pressure response anomaly, and other multi-component anomaly results that meet the conditions for judging the air circuit transmission relationship and abnormal coupling degree of the braking system constitute the coupling fault diagnosis result of the present invention; The directed edges of the Bayesian network are determined based on the air path structure, pressure transmission relationship, and action timing relationship of the braking system. The airtightness status of the main duct is associated with abnormal pressure drop rate and abnormal air compressor operating time; the air compressor pressure rise capability status is associated with abnormal pressure rise rate and abnormal unit pressure recovery time; the valve pressure response status is associated with abnormal valve pressure difference; Scene is associated with Fault and Result to correct the prior probability of fault under different operating scenarios.

[0039] Step 10: Calculate the probability of the diagnosis result.

[0040] In a preferred embodiment, Bayesian inference is used to calculate the probability of each type of diagnostic result. For any diagnostic result... The probability is calculated as follows: in, Show the diagnosis result of type i; Represents the set of candidate abnormal evidence; Indicates the current running scenario; This indicates the prior probability of the diagnostic result in the current operating scenario; This represents the conditional probability of the corresponding set of evidence appearing under the given diagnostic result and the current operating scenario; This indicates the probability of the evidence set appearing in the current operating scenario; when When multiple evidentiary features are included, It can be obtained by combining the conditional probabilities of each evidence feature.

[0041] The aforementioned prior probabilities and conditional probabilities can be determined from historical operational data, fault work orders, maintenance records, expert experience, or model training results, and can be updated as ground PHM platform data accumulates.

[0042] Step 11: Calculate the component anomalous coupling degree. In a preferred embodiment, for any two components that simultaneously exhibit candidate anomalous coupling... and Calculate the abnormal coupling degree between the two. : in, For abnormal timing consistency, used to characterize and Does the sequence of anomalies conform to the fault propagation logic? The degree of matching between the air path or functional transmission relationship of the braking system is used to characterize... and Is there any connection between them in the air supply, pressure regulation, or braking execution links? The correlation of abnormal features is used to characterize whether there are synchronous or correlated changes between the abnormal features of two components. To ensure consistency across operating scenarios, this is used to characterize whether candidate anomalies conform to the fault judgment logic under the current operating scenario; , , , Weighting coefficients, and ; when When the value exceeds the preset coupling threshold, a determination is made. and There exists a fault propagation relationship or coupling relationship; when If the value is not greater than the preset coupling threshold, then determine and There is a lack of effective coupling between them.

[0043] The , , and All values ​​can be represented by normalized values ​​between 0 and 1, and their values ​​can be determined by braking system mechanism rules, historical operating data, fault work order data, maintenance records, or expert experience calibration.

[0044] Step 12: Perform adaptive switching between single-fault and coupled-fault modes. Based on the diagnostic result probability and the degree of abnormal coupling of components, perform the diagnostic mode switching.

[0045] The probability of the diagnostic result is the diagnostic result node calculated in step S10. posterior probability The associated components refer to other components that have a preset association with the currently determined faulty component in terms of the braking system's air circuit structure, pressure transmission relationship, functional linkage relationship, or action timing relationship; the preset association relationship is formed by the Bayesian network directed edges in step S9 and the connections in step S11. Confirmed; For abnormal air tightness of the main duct, the associated components include the air compressor and components related to the recovery of air supply pressure; For abnormal air compressor pressure rise capability, the associated components include the main duct, valve pressure response components, and components related to pressure recovery; For abnormal valve pressure response, the associated components include the air supply unit, the main duct, and related braking control valve components. When the probability of a diagnosis result corresponding to a single component failure is greater than the first threshold and the probability of an abnormality of the associated component is less than the second threshold, switch to single fault diagnosis mode and output the corresponding single component failure. When the probability of the diagnosis result corresponding to the coupled fault is greater than the first threshold, and the abnormal coupling degree between related components is greater than the coupling threshold, the coupled fault diagnosis mode is switched to output the dominant fault source, affected components and fault propagation path. When the probability of the diagnosis result corresponding to the scene-induced anomaly is greater than the first threshold and the deviation of the key anomaly features is less than the fault judgment threshold, switch to the scene-induced anomaly mode, reduce the alarm level or do not output the fault alarm. When the probability of diagnosis results corresponding to multiple single component failures is greater than the threshold, but the abnormal coupling degree between related components is not greater than the coupling threshold, switch to multi-source independent anomaly mode and output the abnormal component and maintenance suggestions respectively. Finally, the diagnostic results are output. The output includes the diagnostic mode, fault type, dominant fault source, affected components, fault propagation path, abnormal time period, key abnormal characteristics, diagnostic result probability, abnormal coupling degree, confidence level, and maintenance priority.

[0046] In a preferred embodiment, when evidence of abnormal air tightness of the main duct, abnormal operating time of the air compressor, and normal pressure rise capability of the air compressor occur simultaneously, and the abnormal coupling degree between the main duct and the air compressor is greater than the coupling threshold, a diagnostic result is output indicating that the abnormal air tightness of the main duct has been transmitted to the abnormal operation of the air compressor.

[0047] In a preferred embodiment, when evidence of abnormal air compressor pressure rise capability, abnormal unit pressure recovery time, and abnormal valve pressure response occur simultaneously, and the abnormal coupling degree between the air compressor and valve components is greater than the coupling threshold, a diagnostic result is output indicating that the abnormal air compressor air blowing capability has been transmitted to the abnormal valve pressure response.

[0048] In a preferred embodiment, when evidence of abnormal air compressor operating time appears in a shutdown charging or turnaround startup scenario, and the recovery trend of the main air duct pressure meets the normal charging conditions, the output scenario induces abnormal results.

[0049] Through the above technical solution, the present invention combines operation scene recognition, component-level candidate anomaly recognition, Bayesian coupled reasoning and adaptive switching of diagnostic modes to achieve differential diagnosis of single faults, coupled faults, fault propagation results and scene-induced anomalies in the braking system.

[0050] This invention addresses five key issues: First, it solves the problem of fragmented early warning models for single components, making it difficult to comprehensively assess the fault correlation between the main duct, air compressor, and average valve. Second, it addresses the difficulty in distinguishing between single faults, coupled faults, and scenario-induced faults when multiple candidate anomalies are triggered simultaneously. Third, it addresses the problem of confusion between main duct airtightness anomalies and air compressor malfunctions, making it difficult to determine the dominant fault source. Fourth, it addresses the problem of a lack of correlation assessment between air supply anomalies and average valve pressure response anomalies, resulting in unclear fault propagation paths. Fifth, it addresses the problem of false alarms easily generated in scenarios such as silo shutdown and air charging, turnaround and start-up, pre-station braking, and frequent start-stop.

[0051] By solving the above-mentioned technical problems, the present invention can realize adaptive switching diagnosis of single faults and coupled faults in the braking system, and output the dominant fault source, related components and fault propagation path, thereby improving the accuracy of early warning results and maintenance guidance value.

Claims

1. A method for diagnosing adaptive switching faults in the braking system of rail transit vehicles, characterized in that, The method includes the following steps: S1. Acquire vehicle operating status data and braking system status data; S2: Identify vehicle operating scenarios based on vehicle operating parameters; S3: Based on the air circuit structure and functional relationship of the braking system, build a component-level candidate anomaly identification model and output multi-dimensional anomaly features; S4: Determine the candidate anomalies of the main duct air tightness, compare the main duct pressure drop rate with the historical baseline threshold, and output the candidate results of the main duct air tightness anomalies. S5: Determine candidate anomalies in the air compressor's pressure rise capability, compare with the baseline threshold of the pressure rise rate, and output candidate results for air compressor pressure rise capability anomalies. S6: Extract the multi-class duration features of the air compressor, compare them with the corresponding baseline thresholds, and output the candidate results of the reasonableness of the air compressor working duration anomaly. S7: Calculate valve pressure difference in the pre-station braking scenario. Output candidate results for abnormal valve pressure response if the pressure difference and duration meet the conditions. S8: Convert the output results of each component-level candidate anomaly identification model into a unified evidence set; S9: Construct a Bayesian network containing nodes for scene, component status, evidence, and diagnostic results, and construct directed edges of the Bayesian network based on gas path, pressure, and time sequence; S10: Calculate the probability of each type of diagnostic result using Bayesian inference; S11: Calculate the abnormal coupling degree of any two components with candidate abnormalities at the same time, and compare it with the preset coupling threshold to determine the fault transmission relationship or coupling relationship; S12: Based on the probability of the diagnostic results and the abnormal coupling degree of the components, perform a diagnostic mode switch.

2. The adaptive switching fault diagnosis method for a rail transit vehicle braking system according to claim 1, characterized in that, The vehicle operation status data and braking system status data specifically include the collection of vehicle speed, stopping time, station distance, driver's cab activation status, braking commands, emergency braking status, track conditions, main air duct pressure, air compressor start signal, air compressor working status, valve pressure, brake cylinder pressure, and related brake air supply, pressure regulation, and braking execution. The operation status data and braking system status data are collected by the on-board system, train network control system, or brake control unit and transmitted to the ground PHM platform or vehicle-to-ground collaborative diagnostic platform.

3. The adaptive switching fault diagnosis method for a rail transit vehicle braking system according to claim 1, characterized in that, The operating parameters in step S2 include vehicle speed, parking time, station distance, braking command, emergency braking status, air compressor start status, and main duct pressure change.

4. The adaptive switching fault diagnosis method for a rail transit vehicle braking system according to claim 1, characterized in that, The vehicle operation scenarios in step S2 include static pressure holding scenario, depot parking and air charging scenario, turnaround and start-up scenario, pre-station braking scenario, frequent start-stop scenario, continuous braking scenario, and normal operation scenario.

5. The adaptive switching fault diagnosis method for a rail transit vehicle braking system according to claim 1, characterized in that, The component-level candidate anomaly identification model includes a main duct air tightness anomaly identification model, an air compressor pressure rise capability anomaly identification model, an air compressor working time rationality anomaly identification model, a valve pressure response anomaly identification model, and a brake execution anomaly identification model; the multi-dimensional anomaly features include candidate anomaly type, anomaly probability, anomaly confidence, anomaly occurrence time, anomaly duration, and key anomaly features.

6. The adaptive switching fault diagnosis method for a rail transit vehicle braking system according to claim 1, characterized in that, Step S4 specifically includes: when the vehicle is stationary, there is no braking command, the air compressor is not started, and the parking time meets the preset conditions, extracting the pressure drop range of the main air duct and calculating the pressure drop rate of the main air duct. : in, This indicates the starting pressure of the pressure drop range; This indicates the pressure at the end of the pressure drop range; Indicates the duration of the pressure drop interval; Indicates the rate of decrease in total duct pressure; when If the pressure drop interval exceeds the historical baseline threshold for the corresponding vehicle and the duration exceeds the preset time, output the candidate result for the main duct airtightness anomaly. The historical baseline threshold is determined by the statistical values ​​of the historical valid samples of the same vehicle, the historical valid samples of the same model, or the slope of the sliding window according to preset statistical rules. The preset statistical rules include SPC control limits, the mean plus or minus a preset multiple of standard deviation, or quantile statistical methods.

7. The adaptive switching fault diagnosis method for a rail transit vehicle braking system according to claim 1, characterized in that, Step S5 specifically includes: Step S5.1: When the vehicle is stationary, there is no braking command, no significant air consumption, and the air compressor is running, extract the total duct pressure recovery range during the air compressor startup period and calculate the pressure rise rate. : in, Indicates the pressure at the start of the air compressor or the starting pressure of the pressure recovery range; This indicates the pressure at the moment the air compressor stops or at the end of the pressure recovery range; Indicates the duration of the pressure recovery interval; Indicates the rate of increase of the total duct pressure; Step S5.2: Calculate the mean_rise and standard deviation_rise of the pressure rise rate samples, and determine the baseline threshold for the pressure rise rate as follows: in, Indicates the baseline threshold for the rate of pressure rise; This represents the mean rate of pressure rise in historical normal samples; This represents the standard deviation of the rate of pressure rise in historical normal samples; k is a preset control coefficient with a value of 3. Step S5.3: When Less than At that time, the candidate result for abnormal air compressor pressure rise capability is output.

8. The adaptive switching fault diagnosis method for a rail transit vehicle braking system according to claim 1, characterized in that, Step S6 specifically includes: Step S6.1: Extract the continuous operating range of the air compressor based on the air compressor start signal, and calculate the air compressor operating time characteristics, including the single operating time. Cumulative working hours Startup frequency Duty cycle Unit pressure recovery time Among them, unit pressure recovery time Calculate as follows: in, This indicates the duration of continuous operation of the air compressor. This indicates the total duct pressure when the air compressor starts up. This indicates the total duct pressure when the air compressor stopped. Indicates the time required for unit pressure to recover; Step S6.2: When , , , or When at least one of the parameters exceeds the corresponding baseline threshold, output a candidate result for the reasonableness of the air compressor working time; Step S6.3: When the air compressor's operating time is abnormal and accompanied by an abnormal rate of pressure drop in the main duct, the coupling judgment of the abnormal air tightness of the main duct and the abnormal operation of the air compressor is initiated first.

9. The adaptive switching fault diagnosis method for a rail transit vehicle braking system according to claim 1, characterized in that, Step S7 specifically includes: Step S7.1: In the scenario of vehicle braking before entering the station, when the total air duct pressure is greater than the preset air supply pressure threshold and emergency braking is not applied, calculate the valve-related pressure difference. : in, and This represents two pressure signals related to valve components; This indicates the absolute pressure difference between the two. Step S7.2: When When the pressure difference exceeds the preset threshold and the duration exceeds the preset duration, output candidate results for abnormal valve pressure response.

10. A method for diagnosing adaptive switching faults in a rail transit vehicle braking system according to claim 9, characterized in that, Step S7 further includes: in the identification of abnormal average valve pressure response. and The pressures are ASP1 and ASP2, respectively. When |ASP1−ASP2| is greater than 180 kPa and lasts for 30 s, the candidate results of the average valve pressure difference are output.

11. The adaptive switching fault diagnosis method for a rail transit vehicle braking system according to claim 1, characterized in that, The evidence set includes abnormal pressure drop rate, abnormal pressure rise rate, abnormal single working time of air compressor, abnormal cumulative working time of air compressor, abnormal unit pressure recovery time, and abnormal valve pressure difference. Each piece of evidence includes an evidence value, the time of occurrence, the duration, the degree of deviation, and the confidence level; the degree of deviation is determined by the difference between the current feature value and the historical baseline, the baseline of the same vehicle, or the baseline of the same model.

12. The adaptive switching fault diagnosis method for a rail transit vehicle braking system according to claim 1, characterized in that, Step S9 specifically includes: Construct a Bayesian network that includes Scene nodes, Fault nodes, Evidence nodes, and Result nodes; The Scene is used to represent the current operating scenario of the vehicle; The Fault includes the airtightness of the main duct, the pressure rise capability of the air compressor, the working behavior of the air compressor, and the pressure response of the valves. The evidence is formed by the output of the candidate anomaly identification model at each component level; The Result includes single main duct air tightness abnormality, single air compressor air blowing capacity abnormality, single valve pressure response abnormality, main duct air tightness abnormality transmitted to air compressor malfunction, air compressor air blowing capacity abnormality transmitted to valve pressure response abnormality, scenario-induced abnormality, and multi-source independent abnormality. The directed edges of the Bayesian network are determined based on the air circuit structure of the braking system, pressure transmission relationship, and action timing relationship; the air tightness status of the main duct is associated with abnormal pressure drop rate and abnormal air compressor working time; the air compressor pressure rise capability status is associated with abnormal pressure rise rate and abnormal unit pressure recovery time; the valve pressure response status is associated with abnormal valve pressure difference; Scene is associated with Fault and Result to correct the prior probability of fault under different operating scenarios.

13. The adaptive switching fault diagnosis method for a rail transit vehicle braking system according to claim 1, characterized in that, Step S10 specifically includes: Step S10.1: For any diagnostic result node The probability calculation method for its diagnostic results is as follows: in, Show the diagnosis result of type i; Represents the set of candidate abnormal evidence; Indicates the current running scenario; This indicates the prior probability of the diagnostic result in the current operating scenario; This represents the conditional probability of the corresponding set of evidence appearing under the given diagnostic result and the current operating scenario; This indicates the probability of the evidence set appearing in the current operating scenario; Step 10.2: When When multiple evidentiary features are included, It can be obtained by combining the conditional probabilities of each evidentiary feature; Step 10.3: The prior probabilities and conditional probabilities mentioned above can be determined from historical operating data, fault work orders, maintenance records or model training results, and can be updated as ground PHM platform data accumulates.

14. The adaptive switching fault diagnosis method for a rail transit vehicle braking system according to claim 1, characterized in that, Step S11 specifically includes: For any two components that simultaneously have candidate anomalies and Calculate the abnormal coupling degree between the two. : in, For abnormal timing consistency, used to characterize and Does the sequence of anomalies conform to the fault propagation logic? The degree of matching between the air path or functional transmission relationship of the braking system is used to characterize... and Is there any connection between them in the air supply, pressure regulation, or braking execution links? The correlation of abnormal features is used to characterize whether there are synchronous or correlated changes between the abnormal features of two components. To ensure consistency across operating scenarios, this is used to characterize whether candidate anomalies conform to the fault judgment logic under the current operating scenario; , , , Weighting coefficients, and ; when When the value exceeds the preset coupling threshold, a determination is made. and There exists a fault propagation relationship or coupling relationship; when If the value is not greater than the preset coupling threshold, then determine and There is a lack of effective coupling between them.

15. The adaptive switching fault diagnosis method for a rail transit vehicle braking system according to claim 1, characterized in that, Step S12 specifically includes: When the posterior probability of the diagnostic result node corresponding to a single component failure is greater than the first threshold, and the abnormal probability of other components that have a preset association with the component corresponding to the single component failure is less than the second threshold, the single fault diagnosis mode is switched to output the corresponding single component failure. When the posterior probability of the diagnostic result node corresponding to the coupled fault is greater than the first threshold, and the abnormal coupling degree between related components is greater than the coupling threshold, the coupled fault diagnosis mode is switched to output the dominant fault source, affected components and fault propagation path. When the posterior probability of the diagnostic result node corresponding to the scene-induced anomaly is greater than the first threshold, and the deviation of the key anomaly features is less than the fault judgment threshold, switch to the scene-induced anomaly mode, reduce the alarm level or do not output the fault alarm. When the probability of diagnosis results corresponding to multiple single component failures is greater than the first threshold, but the abnormal coupling degree between related components is not greater than the coupling threshold, switch to multi-source independent anomaly mode and output the abnormal component and maintenance suggestions respectively. Finally, the diagnostic results are output, including the diagnostic mode, fault type, dominant fault source, affected components, fault propagation path, abnormal time period, key abnormal characteristics, diagnostic result probability, abnormal coupling degree, confidence level, and maintenance priority.

Images