Railway brake fault intelligent identification method based on pressure time sequence analysis

Abstract

The application provides a railway brake fault intelligent identification method based on pressure time sequence analysis, comprising: acquiring multi-channel pressure time sequence and test control state in railway brake test, determining first failure point and first anomaly, solidifying unqualified result and determining original air path boundary state; maintaining original air path boundary state when safety termination condition is not met, continuing original physical process of standard test, and forming natural pressure evidence segment after failure; determining candidate demand channel and upstream pressure channel based on the first anomaly, judging whether continuous aerodynamic demand is formed, and forming natural fault excitation, natural demand time sequence and synchronous upstream gas supply response time sequence when the continuous aerodynamic demand is formed; establishing gas supply response judgment benchmark according to normal test record, and forming single fault or composite fault identification result according to matching relationship between natural demand and upstream gas supply response; subsequently releasing the maintenance of the original air path boundary state, and executing confirmation test, test exit or fault disposal.

Description

Technical Field

[0001] This invention relates to the field of railway braking fault identification technology, and in particular to an intelligent identification method for railway braking faults based on pressure time sequence analysis. The applicable railway vehicles include electric vehicles that use electric power as traction, run on railway lines, and employ pneumatic braking systems. Background Technology

[0002] When conducting single-vehicle braking tests on railway vehicles, for electric vehicles that use electric power for traction, operate on railway lines, and employ pneumatic braking systems, it is also necessary to continuously collect pressure data from the brake pipes, air storage units, and brake cylinders. This data, combined with the vehicle's current state of charging, braking, pressure holding, or release, is used to determine whether each pressure level meets the test requirements. If a pressure channel experiences insufficient pressure, excessive pressure drop, or a response that does not meet regulations for the first time, the testing equipment will generally deem the current test unqualified and proceed to release, charging, reset, or the next test. This method can quickly detect braking abnormalities, but the detection process typically focuses on "what abnormality first occurred," while paying insufficient attention to how the pressure continues to change after the initial failure.

[0003] In reality, the initial anomaly does not necessarily indicate a single vehicle malfunction. For example, when a downstream pneumatic unit experiences continuous pressure loss, it continuously generates a demand for compressed air, keeping the upstream air storage unit and air supply path in a continuous air supply state. If the upstream air supply path also has faults such as partial obstruction, these faults may not be obvious when there is no continuous air supply demand. Only when the downstream experiences continuous pressure loss and continuous demand for air replenishment will the abnormal air supply response gradually become apparent. However, if the test equipment immediately performs relief, air replenishment, or reset after the initial failure, the original air supply, exhaust, isolation, and connectivity states will be altered. The pressure changes that the downstream anomaly could have continued to generate will be interrupted, and the upstream air supply path will also lose its continuous load-bearing condition. Subsequent pressure data will also be mixed with pressure changes caused by the test equipment's control actions, making it difficult to continue reflecting the natural development process of the original fault.

[0004] Therefore, the problem to be solved by this invention is: how to avoid the subsequent control actions from prematurely changing the original air circuit state after the first failure in the railway braking test, and to accurately determine whether the vehicle has a single fault or a coexisting fault that is further exposed by continuous pneumatic demand by using the pressure timing formed after the first failure. Summary of the Invention

[0005] The purpose of this invention is to address the shortcomings of existing technologies by providing an intelligent identification method for railway braking faults based on pressure time-series analysis, thereby solving the technical problems mentioned in the background section.

[0006] To achieve the above objectives, the present invention provides the following technical solution: A method for intelligent identification of railway braking faults based on pressure time series analysis includes the following steps: S1. Obtain the multi-channel pressure timing and test control status of the railway braking test, determine the first failure point and the first anomaly based on the non-compliance judgment conditions, solidify the non-compliance results, and determine the original gas path boundary state at the time of the first failure. S2. If the safety termination conditions are not met, maintain the original gas path boundary state, do not change the state and do not add new diagnostic disturbances, continue the original physical process of the standard test, and obtain the natural pressure evidence segment after failure. S3. Based on the first anomaly and the natural pressure evidence segment after failure, determine the candidate demand channel and the upstream pressure channel, determine whether the first anomaly forms a continuous gas demand, and determine it as a natural failure excitation when it forms, so as to form a natural demand timing sequence and a synchronous upstream gas supply response timing sequence. S4. Use normal test records of the same vehicle model, the same test items and the same original gas circuit boundary state to form the gas supply response judgment benchmark. Determine the gas supply response mismatch state based on the matching relationship between the natural demand time sequence and the upstream gas supply response time sequence, and form a single fault or compound fault identification result. S5. Based on the identification results, release the maintenance of the original gas path boundary state, perform confirmation test, exit test or brake component failure handling, and use the moment when the original gas path boundary state is changed as the control intervention boundary, and isolate the subsequent pressure data from the natural pressure evidence section after failure.

[0007] The beneficial effects of this invention are as follows: This invention solidifies the unqualified results after the initial failure and maintains the original gas path boundary state, preventing the pressure changes that continue to form due to the first anomaly from being mitigated and prematurely cutting off the air supply or reset actions. This preserves valid evidence for continuous pneumatic demand and coexisting fault identification, thereby improving the consistency and reliability of single and compound fault identification. By synchronously processing multi-channel pressure timing and test control states through edge computing nodes, and using a multi-agent system to determine the first failure point in conjunction with the test phases, the pressure anomaly and the corresponding gas path control state remain time-consistent, reducing identification errors introduced by communication delays, sampling misalignments, and phase misjudgments, thus improving the accuracy of first failure location.

[0008] This invention categorizes control actions following the initial failure into maintenance and modification actions, and prohibits actions that alter the original gas path boundary state under safe termination conditions. This separates the natural pressure evidence segment after failure from pressure changes caused by human control, reducing the contamination of fault pressure characteristics by control intervention and improving evidence stability. By identifying candidate demand channels and upstream pressure channels from the natural pressure evidence segment after failure, and converting the pneumatic demand that continuously forms the first anomaly into natural fault excitation, upstream gas supply anomalies that are not easily apparent under low gas supply loads gain actual load conditions, improving the identifiability of upstream gas supply path obstruction faults.

[0009] This invention establishes a gas supply response judgment benchmark based on normal test records of the same vehicle model, test items, and original gas path boundary conditions. By comparing the upstream gas supply response under continuous pneumatic demand conditions, it provides a clear reference for judging gas supply obstruction, reducing misjudgments caused by isolated pressure fluctuations, benchmark mismatches, and no-load conditions. By releasing the original gas path boundary holding state based on single or compound fault identification results and establishing a control intervention boundary based on the actual change in control state, it isolates subsequent control intervention data from natural fault evidence and verifies the handling execution status. This maintains consistency between fault identification results and the handling process, improving the reliability of subsequent confirmation and fault handling. Attached Figure Description

[0010] Figure 1 This is a flowchart of the intelligent identification method for railway braking faults based on pressure time sequence analysis according to the present invention. Detailed Implementation

[0011] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0012] Example: Figure 1 As shown, this embodiment provides a method for intelligent identification of railway braking faults based on pressure time series analysis, including the following steps: S1. Obtain the multi-channel pressure timing and test control status of the railway braking test, determine the first failure point and the first anomaly based on the non-compliance judgment conditions, solidify the non-compliance results, and determine the original gas path boundary state at the time of the first failure. S2. If the safety termination conditions are not met, maintain the original gas path boundary state, do not change the state and do not add new diagnostic disturbances, continue the original physical process of the standard test, and obtain the natural pressure evidence segment after failure. S3. Based on the first anomaly and the natural pressure evidence segment after failure, determine the candidate demand channel and the upstream pressure channel, determine whether the first anomaly forms a continuous gas demand, and determine it as a natural failure excitation when it forms, so as to form a natural demand timing sequence and a synchronous upstream gas supply response timing sequence. S4. Use normal test records of the same vehicle model, the same test items and the same original gas circuit boundary state to form the gas supply response judgment benchmark. Determine the gas supply response mismatch state based on the matching relationship between the natural demand time sequence and the upstream gas supply response time sequence, and form a single fault or compound fault identification result. S5. Based on the identification results, release the maintenance of the original gas path boundary state, perform confirmation test, exit test or brake component failure handling, and use the moment when the original gas path boundary state is changed as the control intervention boundary, and isolate the subsequent pressure data from the natural pressure evidence section after failure.

[0013] S1 specifically includes the following sub-steps: S110. Establish a time-synchronized multi-channel pressure timing sequence and test control state at the railway braking test site. The braking system of the railway vehicle currently undergoing single-vehicle testing is used as the processing object; when the tested railway vehicle is an electric vehicle using electric power for traction, running on the railway line, and employing a pneumatic braking system, the brake pipe, air storage unit, and brake cylinder of the electric vehicle are used as the corresponding pressure acquisition objects. Pressure sensors are connected to the pressure detection locations of the brake pipe, air storage unit, and brake cylinder corresponding to the existing test equipment. The edge computing node located on the single-vehicle test equipment site receives the pressure values ​​of each channel through the pressure acquisition interface. Among them, the air storage unit refers to the auxiliary air cylinder, working air cylinder, acceleration and relief air cylinder or other air storage air cylinder that actually participates in the braking process of this test according to the air circuit structure of the current vehicle model. Only the air cylinder that actually undertakes the function of air storage or air supply in the current test is included in the pressure sequence.

[0014] The edge computing node simultaneously reads the current test item, air charging, braking, pressure holding or release status, and the current air supply, exhaust, isolation and connectivity control status through the control interface of the single-vehicle test equipment. Internally, the edge computing node runs a multi-agent system, which includes at least a pressure determination agent, a test status agent, an evidence preservation agent, a demand determination agent, an air supply response agent, and a coordination agent. Each agent is a software execution unit that operates according to defined input fields, judgment rules, and output fields. Processing results are exchanged through a shared state table indexed by test identifiers and timestamps, and unpublished model inference results are not used to replace subsequent specific judgment rules.

[0015] The test identifier is a unique association identifier established by the edge computing node when the current single-vehicle test project is started, based on the current test vehicle, the current test project, and the current test task. It is used to associate the pressure data, control status, first anomaly, fault evidence, and handling results generated in the same test with the same test process.

[0016] Before the test begins, each pressure sensor completes zero-point and range calibration based on the verification results or the calibration results of the single-vehicle test equipment. The pressure value is uniformly expressed in kPa. The sampling period is determined according to the pressure sensor response time, the acquisition interface refresh cycle, and the requirement to cover the shortest effective pressure change process of the current test, and remains unchanged in the same test.

[0017] Edge computing nodes use a unified on-site clock to establish timestamps for each pressure channel and control status. Data with a time difference of no more than one sampling period are grouped into the same sampling index. Data that has not been obtained after more than one sampling period is marked as invalid. Interpolation is not performed across the test control status switching time.

[0018] Sampled values ​​exceeding the effective range of the pressure sensor are directly marked as out of bounds. For isolated outliers that occur at only one sampling point, exceed the allowable sampling jump variable determined by the equipment calibration, and whose preceding and following valid samples return to their original trend, a substitute value is obtained by linear interpolation of adjacent preceding and following valid sampled values. When consecutive missing values ​​reach the minimum effective duration required to cover the current test failure criterion, the initial failure identification is paused and an acquisition anomaly is output; the missing data is not interpreted as a braking failure.

[0019] The multi-channel pressure state at the kth valid sampling time is represented as:

[0020] in, Indicates time The multi-channel pressure status; k represents the valid sampling sequence number, which is a positive integer; This indicates the kth valid sampling time, in seconds; i represents the pressure channel number, which is a positive integer. Indicates that the i-th pressure channel is in The pressure value at any given time is expressed in kPa; m represents the number of effective pressure channels actually involved in the pressure determination during the current test, and is a positive integer.

[0021] If any of the current test-required pressure channels is in a continuously invalid state, a valid multi-channel pressure state for the corresponding time moment will not be formed. After completion, the edge computing node encapsulates the timestamp, multi-channel pressure state, test control state at the same time moment, and data validity marker into a synchronization data frame sequence, and inputs the synchronization data frame sequence into S120.

[0022] S120. Determine the first failure point and form the first anomaly based on the non-compliance criteria of the current test item. When the current test item is started, the edge computing node reads the non-compliance criteria corresponding to the current vehicle model and the current test item from the test parameter table of the single-vehicle test equipment. The non-compliance criteria adopt the test procedures already executed by the current single-vehicle test equipment, the calibrated upper pressure limit, lower pressure limit, pressure change within a specified time, allowable pressure drop during the pressure holding period, and braking or release response conditions. No additional empirical thresholds are set to form subsequent fault categories.

[0023] The pressure determination agent reads the multi-channel pressure status formed by S110 on a sample-by-sample basis. The test status agent reads the test control status at the same time simultaneously and first confirms whether each non-compliance criterion is applicable to the current air charging, braking, pressure holding or release state. If the pressure condition is met but the current test stage does not meet the eligibility of the criterion, it is not considered as the first failure.

[0024] Let the j-th non-conforming condition be represented as a failure determination function. The first failure point is as follows:

[0025] Determined; where j represents the number of the non-compliance condition, which is a positive integer. This indicates the time corresponding to the first failure point, in seconds. This represents the k-th valid sampling time, in seconds. Derived from S110; This indicates the test control status obtained through the test control interface at the same time. An output of 1 indicates that the j-th non-compliance condition is met in the corresponding test stage, while an output of 0 indicates that it is not met.

[0026] For criteria that require an abnormality to persist for a specified period before constituting a failure, the first failure is confirmed only after the persistence condition is fully met, and the start or end time is determined according to the start or end time specified in the test criterion itself. The time assignment shall not be changed by the edge computing node itself; for cases where the pressure value is exactly equal to the upper limit, lower limit, allowable pressure drop or allowable change boundary, the assignment shall be carried out according to the qualified or unqualified side specified in the original criterion of the current test equipment.

[0027] If multiple non-compliance criteria are met simultaneously at the same sampling time, all trigger criterion numbers are saved, but all criteria share the same first failure point. The anomaly that first meets the valid non-compliance condition is defined as the first anomaly. The first anomaly includes at least the trigger criterion number, the pressure channel involved, the first failure point, and the pressure change on which the criterion is based. Other pressure anomalies that occur after the first failure point are not included in the first anomaly. S120 outputs the first failure point, the first anomaly, the pressure sequence before the first failure, and the test control status at the time of the first failure, and sends them to S130.

[0028] S130: Solidify the unqualified result corresponding to the first failure and form the original gas path boundary state. After receiving the output of S120, the test state intelligent agent solidifies the qualified result of the current test item as unqualified. Solidification means that after the first failure has met the unqualified conditions of the current test, even if the subsequent pressure temporarily returns to the qualified range, the unqualified result will not be cancelled. However, at this time, the unqualified result will not immediately trigger early air charging, early relief, reset, or switch to the next test item.

[0029] Subsequently, the edge computing node extracts the control fields that determine the compressed air flow boundary from the test control state at the time of the first failure. These fields include the current test stage, air supply control state, exhaust control state, isolation state, connectivity state, and the current test actions that have been executed and have not yet been executed at the time of the first failure. The combination of states that determine the air supply, exhaust, isolation, or connectivity relationship after the first failure is defined as the original air path boundary state.

[0030] The difference between the original gas path boundary state and the complete test control state is that the complete test control state contains all the process information of the current equipment, while the original gas path boundary state only retains the control fields that determine the actual flow path of compressed air after the first failure, and serves as the direct basis for S210 to determine which actions can continue to be executed and which actions must be temporarily prohibited.

[0031] For control channels that can provide motion feedback for single-vehicle testing equipment, the existing motion feedback shall be used as the standard; for channels that do not have independent valve position feedback but whose controllers can read the output register status, the output status that has been written and confirmed by the control interface shall be used as the control status, without assuming the addition of a new valve position detection device.

[0032] If the key control channel that determines the gas path boundary cannot obtain the status, multiple control fields conflict with each other, or it cannot be confirmed whether the gas supply and exhaust path switching has occurred, then it is marked as the original gas path boundary status is not confirmed, and it will not enter the subsequent failure evidence preservation process. The current automatic identification will end according to the single vehicle test equipment abnormality handling.

[0033] Once completed, the edge computing node encapsulates the first failure point, the first anomaly, the pressure timing before the first failure, the solidified non-conforming results, and the original gas path boundary state into a unified first failure state information, saves it in the shared state table with the same test identifier, and uses the first failure state information as the input of S210.

[0034] S2 specifically includes the following sub-steps: S210: Based on the initial failure status information, determine the permitted test actions for maintaining the original gas path boundary and allowing continued execution. The edge computing node receives the initial failure status information generated in S130 and reads the control actions that have not yet been executed according to the current standard test procedure after the initial failure point through the single-vehicle test equipment control interface. The test state agent first classifies the unexecuted control actions into maintenance actions and modification actions based on the original gas path boundary status determined in S130.

[0035] Maintaining actions refer to actions that do not change the gas supply, exhaust, isolation, and connection relationships at the time of the first failure, but only maintain the test state that started before the first failure or complete the remaining process of the original test state; changing actions refer to actions that will cause premature air charging, premature relief, reset, switching of gas supply and exhaust paths, change of isolation state, change of connection state, or start of the next test item.

[0036] The evidence preservation agent further checks whether the holding actions are original actions determined by the current standard test before the first failure. Pressure reduction, air supply, air exhaust or valve switching added after the first failure to further distinguish the fault should not be included in the holding actions, so as to ensure that subsequent pressure changes still come from the original physical process of the current standard test and the continued effect of the first abnormality itself.

[0037] The coordinating agent synchronously reads the pressure safety boundaries, maximum permissible holding time, and protection conditions that must be safely mitigated or stopped in the single-vehicle test equipment for the current vehicle model and current test item. The safety conditions still adopt the existing protection parameters of the equipment and the requirements of the current test procedure, and will not be temporarily changed due to fault type or subsequent identification results. When the original gas path boundary state has been confirmed, the current standard test still has an original process that can continue to be executed without changing the original gas path boundary conditions, and the current pressure and equipment state have not yet met the safe termination conditions, the multi-agent system forms an original gas path boundary holding permission.

[0038] The edge computing node maintains the output state of the original gas path boundary determined by the controller, sets a temporary prohibition flag for change-type actions, and ensures that safety protection actions have a higher execution priority than boundary maintenance actions. If the device safety termination conditions are met simultaneously when the first failure occurs, no original gas path boundary maintenance permission is generated, and the process directly enters the safety termination process. S210 outputs the original gas path boundary maintenance permission, the set of maintenance-type actions, the set of change-type actions, and the current safety termination conditions, and inputs them into S220.

[0039] S220: During the period when the original gas path boundary is maintained, continue the original physical process of the current standard test and continuously collect the multi-channel pressure response after the first failure. After receiving the original gas path boundary maintenance permission formed by S210, the edge computing node allows the single-vehicle test control program to continue to execute maintenance-type actions, while maintaining the temporary prohibition state of change-type actions, and does not add new dedicated pressure disturbances for fault identification.

[0040] The pressure acquisition interface continues to receive pressure from the brake pipe, air storage unit, and brake cylinder according to the sampling period determined in S110. The pressure data continues to use the same calibration, synchronization, and outlier handling methods, and maintains the actual kPa value. It is not renormalized at the first failure point to preserve the pressure change amplitude and duration when the first anomaly continues to occur. Within each sampling period, the test state agent reads the current controller output state, and the evidence preservation agent compares the current air supply, exhaust, isolation, and connection states with the original air path boundary states formed in S130 item by item. The coordination agent simultaneously checks the pressure safety boundary and holding time.

[0041] Whether the original gas path boundary remains valid is indicated by:

[0042] in, Indicates time The boundaries remain effectively marked and are dimensionless; This indicates the multi-channel pressure safety domain that the current test can continue to maintain. The unit of each pressure boundary is kPa, and it is derived from the safety termination condition read from S210. This represents the hold time from the first failure point to the current sampling time, in seconds. This indicates the maximum time that the current test can be maintained, in seconds. Its value comes from the safety parameters of the single-vehicle test equipment or the applicable test procedures.

[0043] when Entering the end branch; whether the pressure safety boundary includes an equal sign is determined according to the existing boundary assignment of the corresponding equipment protection conditions. If pressure data is missing, making it impossible to confirm whether the current condition is... If the controller output state is different from the original gas path boundary state, or if the control interface reports an execution abnormality, then... .when Continue collecting data; first appearance Immediately revoke the original gas path boundary holding permission and allow the equipment to perform the original safety stop or mitigation actions. S220 outputs continuous multi-channel pressure data, synchronous control status sequence, boundary holding validity marker, and boundary holding end time after the first failure, and sends them to S230.

[0044] S230: Based on the actual maintenance of the original gas path boundary, extract the natural pressure evidence segment after failure and establish the control intervention boundary. The evidence preservation agent takes the continuous pressure data after the first failure, the synchronous control state sequence, the boundary maintenance validity marker, and the boundary maintenance end time output by S220 as input, and samples and verifies the gas path boundary at the time of each pressure data formation starting from the first failure point.

[0045] The continuous sampling interval that simultaneously satisfies the following conditions is as follows: the sampling time is no earlier than the first failure point and no later than the end time of the boundary maintenance output of S220; the boundary maintenance valid mark is established; the gas supply, exhaust, isolation and connection status is consistent with the original gas path boundary status of S130; and no change-type actions listed in S210 are actually executed during the period is determined as the valid interval of natural evidence; the multi-channel pressure timing in this interval is defined as the natural pressure evidence segment after failure.

[0046] The natural pressure evidence segment after failure is the pressure sequence formed by the original physical processes of the current standard test and the continued effect of the first anomaly itself after the first anomaly occurs, without the addition of special diagnostic disturbances and without the original gas path boundary being artificially altered. The difference between it and general pressure data after failure is that the latter may already include artificial pressure changes caused by relief, air filling, reset, or the next test action.

[0047] The control moment when the original gas path boundary state of S130 is first actually changed is defined as the control intervention boundary. For control channels with existing action feedback, the moment when the action feedback indicates that the original gas path boundary begins to change is used. For control channels that can only read the controller output state, the moment when the corresponding output state first confirms the change is used, and the moment when the control command is generated by software or queued is not used as a substitute. If no control action has actually changed the original gas path boundary state by the end of the boundary maintenance time formed in S220, the control intervention boundary is marked as not formed, and no control intervention moment is artificially assigned.

[0048] When the control intervention boundary has been formed, the multi-channel pressure data from the start of the control intervention boundary and thereafter are uniformly marked as control intervention data and are not incorporated into the post-failure natural pressure evidence segment; when the control intervention boundary has not yet been formed, no control intervention data is generated. If the safety termination condition is met immediately after the first failure but there is no continuous effective sampling interval that meets the above requirements, the state of insufficient natural evidence is output, and subsequent artificially controlled pressure data is not used to supplement it. S230 finally outputs the post-failure natural pressure evidence segment, the state of control intervention boundary formation, the control intervention boundary when it has been formed, and the state of insufficient natural evidence. After being associated with the first anomaly and the original gas path boundary state of S130 through the same test identifier, it is used as the input of S310.

[0049] S3 specifically includes the following sub-steps: S310: Based on the post-failure natural pressure evidence segment, determine the candidate demand channel, upstream pressure channel, and effective analysis interval for continuous pneumatic demand. The edge computing node receives the post-failure natural pressure evidence segment, control intervention boundary formation status, control intervention boundary when it has been formed, and natural evidence insufficiency status output by S230, and recalls the first anomaly and original gas path boundary status saved by S130 through the same test identifier.

[0050] When the state of insufficient natural evidence is established, the determination of candidate demand channels and upstream pressure channels will no longer be performed. The state of insufficient natural evidence and the first anomaly will be used as the input of the composite fault evidence insufficient branch of S430. When the state of insufficient natural evidence is not established, the demand determination agent reads the pressure channel involved in the first anomaly and determines whether the pneumatic unit where the first anomaly is located still has the physical conditions to consume compressed air and require the upstream path to continue supplying air under the current standard test conditions, based on the original gas path boundary state.

[0051] The pressure channel involved in the first anomaly, whose continuous pressure change indicates that the pneumatic unit still has an air supply demand, is identified as a candidate demand channel. Subsequently, along the air supply and connection relationship already confirmed in S130, a corresponding pressure channel is selected from the air storage unit or air supply branch currently actually participating in the air supply of the candidate demand channel as an upstream pressure channel, and the candidate demand channel and the corresponding upstream pressure channel form a demand-related channel set. The demand-related channel set is only used to limit which existing measurement points the subsequent natural pneumatic demand and air supply response are obtained from, without changing the braking system air circuit, and without including the pressure of other air cylinders that are not currently participating in the air supply of the candidate demand channel in the judgment.

[0052] The edge computing node starts at the point of first failure. When S230 has formed a control intervention boundary, it ends at that boundary. When S230 has not yet formed a control intervention boundary, it ends at the end of the natural pressure evidence segment after failure. The synchronous pressure sequence of the demand-related channel set is extracted from the natural pressure evidence segment after failure, and the time range between the starting point and the ending point is determined as the effective analysis interval. The synchronous pressure sequence continues to use the data validity markers, sampling period, and pressure units of S110, without interpolation across control intervention boundaries.

[0053] If the continuous data gaps in the candidate demand channel exceed the time range required for the continuity judgment in S320, an insufficient demand evidence state is output. If the original gas path boundary status shows that the candidate demand channel has been isolated from all upstream gas supply channels, a state indicating no continuous gas supply is available is output, and the subsequent downstream pressure drop is not interpreted as upstream gas supply demand. S310 outputs the demand-related channel set, the effective analysis interval, the corresponding synchronous pressure sequence, the insufficient demand evidence state, and the state indicating no continuous gas supply is available. When the insufficient demand evidence state is established, the combined fault judgment in S320 to S420 is not executed, and the insufficient demand evidence state is used as the input for the combined fault evidence insufficient branch in S430. When the condition of no continuous gas supply is established, no natural fault excitation is formed, and the state is input into S430 as a branch with no continuous gas demand; in other cases, the demand-related channel set, effective analysis interval and corresponding synchronous pressure sequence are input into S320.

[0054] S320: Based on the continuous pressure changes of candidate demand channels within the effective analysis interval, a natural demand sequence is formed, and it is determined whether the first anomaly results in continuous pneumatic demand. The demand determination agent reads the candidate demand channel pressure sequence output by S310, and determines the direction of pressure change of the candidate demand channel when continuously consuming compressed air based on the current standard test stage, the original air path boundary state of S130, and normal test records of the same vehicle model. For example, during the pressure holding process where the original air path boundary is maintained and the upstream air supply relationship still exists, if the continuous pressure loss of the downstream pneumatic unit can form a demand for continued air replenishment, the direction of downstream pressure drop is taken as the demand direction.

[0055] For the k-th valid sampling time, a natural aerodynamic demand characterization quantity is formed with w consecutive sampling intervals:

[0056] Where d represents the channel number of the current candidate demand channel. Indicates time The natural aerodynamic demand is a quantity characterized by kPa / s. and These represent the pressure values ​​of the candidate demand channel at two sampling times, in kPa, both from S310; represents the demand direction coefficient, which is dimensionless. It takes the value 1 when the pressure increases and corresponds to sustained demand, and -1 when the pressure decreases and corresponds to sustained demand. w represents the number of sampling intervals included in the calculation window, which is a positive integer.

[0057] w is determined based on the repeatability of the current pressure sensor, the sampling period, and the longest duration of instantaneous pressure fluctuations in normal tests of the same vehicle model, and remains unchanged until recalibration is completed for the same vehicle model and test items. When S230 has formed a control intervention boundary and the calculation window crosses the control intervention boundary, the natural aerodynamic demand characterization at the corresponding sampling time is not calculated; regardless of whether the control intervention boundary is formed, when the calculation window contains invalid samples or two timestamps are the same, the natural aerodynamic demand characterization at the corresponding sampling time is also not calculated, therefore the denominator of the formula is greater than 0 in effective calculations.

[0058] The threshold for determining continuous pneumatic demand is based on pressure records of the same vehicle model and test items, which have been confirmed by normal testing to have no continuous pressure loss faults, obtained under normal conditions. The effective upper bound of fluctuation is determined in conjunction with the measurement uncertainty of the pressure sensor; this threshold can be updated synchronously after the normal test benchmark is updated.

[0059] current When the threshold is met, it is classified as belonging to the demand-fulfilled side. Only... A continuous pneumatic demand is determined only when the threshold is continuously reached or exceeded, and the duration exceeds the longest continuous range of normal instantaneous pressure fluctuations. Isolated over-limit points are not considered as continuous demands.

[0060] The interval will continue to hold. Arrange the natural demand time series according to the original time index; if the continuity condition is not met by the end of the effective analysis interval, output no continuous aerodynamic demand and the first abnormal natural development evidence. S320 outputs the continuous aerodynamic demand determination result, the natural demand time series and the first abnormal natural development evidence, and inputs them into S330; among them, the first abnormal natural development evidence is only output when no continuous aerodynamic demand is formed.

[0061] S330: When continuous pneumatic demand is established, the natural demand timing is determined as a natural fault excitation, and an upstream gas supply response timing synchronized with it is formed. The edge computing node receives the output of S320; when the continuous pneumatic demand determination result is no continuous pneumatic demand, no natural fault excitation is artificially formed, and the result of no continuous pneumatic demand and the first abnormal natural development evidence are used as the input of the no natural excitation branch of S410.

[0062] When a continuous pneumatic demand is established, the demand-determining agent defines the natural demand sequence and the start and end times of the continuous demand interval as natural fault excitations. Natural fault excitations are continuous pneumatic demands that spontaneously form during the period when the first anomaly is maintained at the original gas path boundary and can cause the upstream gas supply path that actually undertakes the gas supply function to enter a loaded state. This is different from pressure reduction, air filling, exhaust, or valve switching that are performed separately to identify faults.

[0063] The air supply response agent reads the demand-related channel set of S310, extracts the actual pressure value of each upstream pressure channel within the time index that is exactly the same as the natural demand time sequence, and unifies the change direction of each upstream pressure channel to the normal air supply response direction according to the air supply action direction under the same air circuit boundary in normal tests of the same vehicle model.

[0064] For the i-th upstream pressure channel in the demand-related channel set, the pressure change rate per unit time along the normal gas supply direction is calculated using the same window width as S320, and this change rate is denoted as... The unit is kPa / s.

[0065] The upstream gas supply response timing sequence is composed of continuous valid timestamps, upstream channel numbers, pressure values, upstream gas supply response change rates, and valid data markers. This timing sequence uses only valid pressure data during the duration of the natural fault excitation. When S230 has already formed a control intervention boundary, it should also be ensured that the data used is located before the control intervention boundary. When S230 has not yet formed a control intervention boundary, the end time of the natural pressure evidence segment after the failure is used as the data endpoint, and the control intervention boundary restriction is not enforced.

[0066] If continuous data is missing for a certain upstream pressure channel, that channel is marked as having an undeterminable response; if all upstream pressure channels are undeterminable, an insufficient gas supply response evidence state is output. S330 ultimately outputs the continuous pneumatic demand determination result, natural fault excitation, natural demand timing, the upstream gas supply response timing synchronized with it, and the insufficient gas supply response evidence state. When the insufficient gas supply response evidence state is established, the gas supply response matching calculations of S410 and S420 are not performed; the insufficient gas supply response evidence state is used as the input to the composite fault evidence insufficient branch of S430; the remaining results are used as the direct input to S410.

[0067] S4 specifically includes the following sub-steps: S410: Establish an upstream gas supply response judgment benchmark under natural demand conditions when a natural fault stimulus is established. The edge computing node receives the output of S330 and first checks the continuous gas demand judgment result. If the result is no continuous gas demand, the upstream gas supply path gas supply obstruction fault judgment is not performed, and the first anomaly, the evidence of the natural development of the first anomaly, and the result of no continuous gas demand are directly used as the single fault branch input of S430.

[0068] If the continuous pneumatic demand is established, when S230 has formed the control intervention boundary, the demand determination agent verifies that all valid samples of the natural demand timing are located before the control intervention boundary; when S230 has not yet formed the control intervention boundary, all valid samples of the natural demand timing are located within the natural pressure evidence segment after the failure. The gas supply response agent simultaneously verifies whether the upstream gas supply response timing and the natural demand timing have the same time index; when the time indices of the two timing sequences are inconsistent, the state of insufficient evidence for the composite fault is directly determined, and the subsequent S420 is not executed.

[0069] For data that meets the requirements, the edge computing node reads the benchmark pressure record from the field benchmark library. This record must be of the same vehicle model, the same test item, and the same original gas path boundary state, and must have been confirmed by normal single-vehicle testing to have a fault-free gas supply path. The field benchmark library can be formed by normal tests during the acceptance or calibration phase of the test equipment, or it can be included after the equipment is put into use, based on records that have been confirmed by maintenance to have no corresponding gas supply faults and have completed normal single-vehicle testing. Records from different sources all use the same pressure calibration, sampling, and synchronization methods as S110.

[0070] The gas supply response agent forms demand intervals based on the values ​​of natural aerodynamic demand characteristics. The width of each demand interval must not be less than the demand variation range corresponding to the pressure measurement uncertainty, avoiding dividing the demand differences caused by the sensor's own resolution error into different reference intervals. For each demand interval covered by the normal reference record, the effective pressure response rate of change range along the normal gas supply direction of each upstream pressure channel is obtained, with the minimum effective value serving as the lower bound of the normal response and the maximum effective value serving as the upper bound of the normal response.

[0071] When the current natural demand value is exactly at the boundary of an adjacent demand interval, it is classified into the interval with the larger demand value; when the current natural demand value exceeds the effective range already covered by the field benchmark library, the output gas supply response benchmark is insufficient, and extrapolation values ​​are not used to complete the composite fault identification. Normal records with different vehicle models, different test items, or different original gas path boundary states are established separately for judgment, and are not mixed. The S410 outputs the effective natural demand timing sequence, synchronizes the upstream gas supply response timing sequence, the upper and lower bounds of the normal response of each upstream pressure channel in the corresponding demand interval, the benchmark validity mark, and the insufficient gas supply response benchmark status.

[0072] When the gas supply response baseline is insufficient, S420 is not executed, and the gas supply response baseline is insufficient is used as the input of the composite fault evidence insufficient branch of S430; only when the baseline valid flag is established and the natural demand timing is synchronized with the upstream gas supply response timing, the valid natural demand timing, the synchronized upstream gas supply response timing, and the corresponding normal response upper and lower bounds are input into S420.

[0073] S420. Determine the gas supply response mismatch state based on the conditional matching relationship between natural demand and upstream gas supply response. The edge computing node only executes this step when the baseline valid flag of S410 is established; the demand determination agent confirms by sampling that the current moment still belongs to the valid range of continuous pneumatic demand formed in S320, and the gas supply response agent simultaneously reads the response change rate of each upstream pressure channel that actually undertakes the gas supply function at the same moment, as well as the lower bound of the normal response corresponding to the range where the current natural demand is located.

[0074] The insufficient gas supply response of the i-th upstream pressure channel is represented as:

[0075] in, This indicates that the i-th upstream pressure channel is at time [time]. Insufficient gas supply response, in kPa / s; The quantity representing the natural aerodynamic demand generated by S320 is expressed in kPa / s. This represents the actual gas supply response change rate of the i-th upstream pressure channel formed by S330, in kPa / s; This represents the lower bound of the normal response corresponding to the interval of the current natural demand in S410, in kPa / s.

[0076] When the actual response reaches or exceeds the lower bound of the normal response When the actual response is lower than the lower bound of the normal response, If any of the natural demand value, actual response value, or lower bound of the normal response is missing, it will not be calculated. , , must not be filled with 0.

[0077] The gas supply response agent further determines the maximum effective duration required for the normal gas supply path to complete the response under the corresponding demand range based on the normal baseline record; only during the period when the natural gas demand is continuously established, does a certain upstream pressure channel responsible for the current gas supply continuously appear. An upstream pressure channel is marked as having a mismatched gas supply response only when the duration of insufficient response reaches or exceeds the maximum effective duration. If the response is equal to the time boundary, it is classified into a mismatch branch. Isolated insufficient response sampling points, response drops occurring after natural pneumatic demand has ended, missing data, or insufficient baselines do not constitute a conclusion of mismatched gas supply response. S420 outputs the response matching status, insufficient response duration range, and the judgment result of whether a mismatched gas supply response exists for each upstream pressure channel, and inputs this information into S430.

[0078] S430: Combine the first anomaly with the gas supply response judgment result or insufficient evidence state to form a single fault identification result, a compound fault identification result, or a compound fault evidence insufficient state, and solidify the corresponding fault evidence. For branches that normally execute to S420, the coordinating agent receives the response matching status of each upstream pressure channel, the insufficient response duration interval, and the gas supply response mismatch judgment result output by S420; for branches that do not execute to S420 due to insufficient evidence, the coordinating agent directly receives the natural evidence insufficient state output by S230, the demand evidence insufficient state output by S310, the gas supply response evidence insufficient state output by S330, and the gas supply response benchmark insufficient state output by S410, or the compound fault evidence insufficient state formed due to the inconsistency between the time index of the natural demand timing and the upstream gas supply response timing. Simultaneously, it reads the first anomaly formed by S120, the original gas path boundary state saved by S130, the natural pressure evidence segment and control intervention boundary formation state after failure formed by S230, and the natural fault excitation already formed by the corresponding branch.

[0079] The coordinating agent first checks for insufficient natural evidence, insufficient demand evidence, insufficient gas supply response evidence, insufficient gas supply response benchmark, and insufficient evidence of a composite fault directly formed by S410. If any of the above states are true, the composite fault evidence is directly determined, and it is no longer required to verify the natural demand timing, upstream gas supply response timing, or gas supply response judgment benchmark that has not yet been formed in the abnormal branch, nor is the incomplete data presumed to be a gas supply obstruction fault in the upstream gas supply path.

[0080] If none of the above-mentioned insufficient evidence conditions apply, and if S320 outputs no continuous pneumatic demand determination result or S310 outputs no continuous gas supply condition, then the unformed natural demand sequence, upstream gas supply response sequence, and gas supply response determination benchmark will no longer be verified, and the process will directly proceed to the subsequent single fault branch. If the continuous pneumatic demand is established, the coordinating agent will verify whether the natural demand sequence, upstream gas supply response sequence, and gas supply response determination benchmark belong to the same test identifier and the same original gas path boundary state. When S230 has already formed a control intervention boundary, it will also verify whether the above data are all located before the control intervention boundary. If there is inconsistency between the test identifier, the original gas path boundary state, or the control intervention boundary constraints, then the composite fault evidence condition will be determined as insufficient.

[0081] For cases where S320 outputs no continuous pneumatic demand judgment result, S310 outputs no continuous air supply condition state, or continuous pneumatic demand is established but all effective upstream air supply channels are in response matching state, the non-compliance state corresponding to the first anomaly is determined as a single fault identification result; the specific fault category of the first anomaly is determined based on the correspondence between the test items, non-compliance criteria and inspection objects currently used by the single vehicle test equipment. If the same non-compliance criterion still corresponds to multiple faults that need further confirmation, it is saved as the first fault candidate set, and further processed in S5 through existing confirmed items, instead of being reduced based on experience in S430.

[0082] If a continuous pneumatic demand is established, and at least one upstream pressure channel that actually undertakes the function of supplying gas to the candidate demand channel forms a continuous gas supply response mismatch, and there is no aforementioned insufficient evidence condition, the fault corresponding to the first anomaly and the gas supply obstruction fault of the upstream gas supply path are jointly identified as the composite fault identification result.

[0083] The blocked air supply is limited to the upstream air supply path where an effective response mismatch occurs, and is not extended to other air cylinders or air passages that are not involved in the current natural fault excitation air supply. The coordinating agent encapsulates the fault type, first failure point, pressure sequence before the first failure, first anomaly, natural pressure evidence segment after failure, natural fault excitation interval, upstream pressure channel where response mismatch occurs, insufficient response duration interval, and control intervention boundary into a fault identification result and corresponding fault evidence, and solidifies it on-site at the edge computing node. For natural fault excitation intervals, insufficient response duration intervals, control intervention boundaries, or other evidence fields that have not yet formed in the current judgment branch, the corresponding fields are marked as not formed, and no data is manually assigned. S430 outputs a single fault identification result, a compound fault identification result, or a compound fault evidence insufficiency status, and outputs the fault evidence or first anomaly evidence corresponding to the current result as the fault judgment input for S510.

[0084] S5 specifically includes the following sub-steps: S510: Based on the fault identification result, release the original air path boundary holding state and determine the subsequent confirmation test, test exit, or brake component fault handling. The edge computing node receives the single fault identification result, composite fault identification result, or insufficient composite fault evidence state output by S430, along with corresponding fault evidence or first anomaly evidence. Simultaneously, it reads the safety termination conditions saved in S210 and reads the current test control status through the single-vehicle test equipment control interface. The coordinating agent determines the handling branch according to the order of safety protection taking precedence over fault confirmation, and the already fixed fault identification result not being reversed by subsequent control actions.

[0085] When the current pressure or control state already meets the safety termination condition of S210, directly select the existing safety mitigation or exit the current test action of the single-vehicle test equipment, and do not continue to execute the fault confirmation. When the safety termination condition is not triggered and S430 outputs a state of insufficient evidence for a composite fault, do not form a conclusion of upstream air supply path air supply obstruction fault, and retain the first anomaly formed by S120 and the unqualified result already solidified by S130; if the current test item has already configured a confirmation item corresponding to the first anomaly, then enter that confirmation item; if there is no confirmation item that can be executed automatically, then output the fault object corresponding to the first anomaly as a pending confirmation object, and exit the current automatic composite fault identification process. When the safety termination condition is not triggered and S430 outputs a single fault identification result, determine whether to enter the corresponding confirmation test, exit the current test, or form a corresponding brake component inspection object according to the fault handling relationship already configured in the current test item; when S430 outputs a composite fault identification result, write the inspection object corresponding to the first fault and the upstream air supply path where continuous air supply response mismatch occurs into the disposal object set, so that the disposal object simultaneously covers the first fault and the upstream air supply path air supply obstruction fault.

[0086] If S430 outputs the first set of candidate faults, it calls the corresponding confirmation items already available on the single-vehicle test equipment for subsequent confirmation. If the equipment data does not specify the execution order of multiple confirmation items, the multi-agent system does not create a new test sequence on its own, but outputs the first set of candidate faults as objects to be confirmed.

[0087] After the coordinating agent determines the handling branch, if the original air path boundary maintenance permission is still valid, the test state agent will revoke the maintenance permission and the temporary prohibition mark for change-type actions; if the maintenance permission has been revoked due to safety termination conditions or boundary maintenance failure, its revoked state will remain. For handling branches that require mitigation, air filling, exit, or test item switching, the test state agent will send the corresponding handling command to the single-vehicle test equipment through the existing control interface; for handling branches that only form brake component inspection objects or objects to be confirmed, no control command to change the state of the original air path boundary will be sent.

[0088] For already sent disposal commands, the control execution status is confirmed by existing device action feedback or controller output status, without adding new execution detection hardware. If the control interface fails to write, the output status cannot be confirmed, or the execution feedback conflicts with the target action, subsequent automatic tests are stopped and a control execution exception is output. S510 outputs the determined disposal command, the set of disposal objects, and the disposal action execution status, and inputs them into S520. The disposal action execution status includes at least whether the disposal command has been sent, whether the control interface has executed successfully, whether the original gas path boundary state has changed, and the confirmation time when the change occurs.

[0089] S520: Establish the final control intervention boundary based on the actual control state of the handling action, and isolate natural fault evidence from control intervention data. After receiving the output of S510, the edge computing node continues to collect the pressure of the brake pipe, air storage unit and brake cylinder according to the sampling period determined by S110, without changing the pressure sensor calibration parameters, time base and data format.

[0090] The evidence preservation agent first reads the control intervention boundary formation status output by S230. If S230 has already formed a control intervention boundary, the control intervention boundary formed by S230 is directly determined as the final control intervention boundary, and it will not move backward due to subsequent actions in S510. If S230 has not yet formed a control intervention boundary, the actual execution status of the S510 action command is read, and the moment when it is first confirmed that the gas supply, exhaust, isolation, or connection status has deviated from the original gas path boundary status of S130 is taken as the final control intervention boundary. For control channels with existing action feedback, the moment when the action feedback indicates that the original gas path boundary has begun to change is used; for control channels that can only read the controller output status, the moment when the corresponding output status is first confirmed to have changed is used, and the moment when the action command is generated, queued, or sent is not used as a substitute.

[0091] When S230 has not yet formed a control intervention boundary and multiple control actions that change the gas path boundary occur in adjacent control cycles, the moment when the action that first confirmed the change to the original gas path boundary is taken as the final control intervention boundary. After the final control intervention boundary is formed, the natural pressure evidence segments that have met the conditions for valid natural evidence before it are preserved as natural fault evidence, and the multi-channel pressure data acquired from the final control intervention boundary are uniformly marked as control intervention data.

[0092] The purpose of control intervention data is limited to confirming whether the treatment action has been completed and whether the equipment has entered the state required for the next test. It is no longer input into S320 to calculate the natural demand timing, nor is it involved in the S420 gas supply response mismatch judgment, thereby avoiding pressure changes caused by relief, air filling or reset actions from re-entering the original fault identification process.

[0093] If S510 only generates a brake component inspection object or an object to be confirmed without actually executing a control action to change the air circuit state, then if S230 has already formed a control intervention boundary, it will continue to use that control intervention boundary. If S230 has not yet formed a control intervention boundary, the final control intervention boundary will remain in an unformed state, without artificially establishing a new pressure data boundary or generating control intervention data. If the final control intervention boundary has been formed but there are missing pressure samples near the boundary, then the actual state change time confirmed by the control interface will be used as the boundary, without inferring the time of the control action through pressure interpolation. S520 outputs the final control intervention boundary formation status, the final control intervention boundary when it has been formed, natural fault evidence, control intervention data, and the execution status of the handling action, and inputs them into S530; when the final control intervention boundary has not been formed, both the final control intervention boundary field and the control intervention data field are marked as unformed.

[0094] S530: Determine the handling verification method based on whether the handling branch actually executes the control action, and form the final fault handling record. The edge computing node receives the final control intervention boundary formation status, the final control intervention boundary when it has been formed, natural fault evidence, control intervention data, and handling action execution status output by S520, and reads the handling object set formed by S510.

[0095] When the S510 actually executes control actions such as mitigation, air supply, exit, or test item switching and generates control intervention data, the test state agent reads the existing target state criteria for the actual action from the single-vehicle test equipment according to the actual executed handling command: when mitigation is executed, it checks whether the braking system has completed mitigation according to the equipment's existing mitigation criteria; when exiting the test, it checks whether the current automatic test process has stopped; when entering the subsequent confirmation test, it checks whether the initial air supply, exhaust, isolation, and pressure states required by the confirmation item have been established. The pressure range, allowable time, and control state used for handling verification all adopt the equipment parameters or applicable test procedures already configured for the current vehicle model and current test item. When equal to the allowable boundary, the execution is performed according to the boundary of the original criteria itself, without setting additional empirical thresholds to form a successful handling result.

[0096] When the above-mentioned actual handling actions reach the target state, the coordinating agent combines the single fault identification result or compound fault identification result solidified in S430, the corresponding fault evidence, the set of handling objects in S510, the final control intervention boundary formation state in S520, the final control intervention boundary when it has been formed, and the handling execution result into a final fault handling record; for compound fault identification results, the first fault and the upstream gas supply path gas supply obstruction fault exposed by natural fault excitation and its associated upstream gas supply path are saved at the same time.

[0097] When S510 only forms a brake component inspection object or a pending confirmation object, does not actually execute a control action to change the air circuit state, and does not form control intervention data, the control action verification is not executed, the corresponding disposal object is recorded as a pending inspection state or a pending confirmation state, and the fault identification result or insufficient composite fault evidence formed by S430, the corresponding fault evidence or first abnormal evidence, the disposal object set of S510, and the final control intervention boundary not formed state are written together into the final fault disposal record.

[0098] If the actual execution of the handling instruction fails to achieve the target state, the handling is not completed and the automatic entry into the next test item is stopped. At the same time, the fault identification results and natural fault evidence that have been solidified by S430 are retained, and the fault classification is not re-executed due to handling failure. If a new equipment safety anomaly occurs during the handling phase, it is recorded separately as an anomaly during the handling phase and is not merged into the fault identification results that have been formed by S430.

[0099] The final fault handling record is stored on the edge computing node and can be sent to the upper-level test management system through the existing communication interface of the single-vehicle test equipment. The fault type, the first failure point, the natural fault excitation range when it has formed, the final control intervention boundary formation status, the final control intervention boundary when it has formed, the handling object, and the handling execution result are sent. The original high-frequency pressure timing is retained on the field side according to the data storage rules of the single-vehicle test equipment.

[0100] Thus, the multi-agent system completes a continuous technology chain within the edge computing node, including initial failure determination, preservation of the original gas path boundary, preservation of natural pressure evidence after failure, determination of continuous aerodynamic demand, formation of natural fault excitation, judgment of upstream gas supply response, identification of single or compound faults, and verification of fault handling.

[0101] The above formulas are calculated according to the physical quantities and units specified for each variable, where pressure is expressed in kPa, time in s, and pressure change rate in kPa / s. The calculation window, demand judgment threshold, pressure safety boundary, normal response upper and lower limits, and effective duration involved in the formulas are determined according to the pressure sensor performance, single-vehicle test equipment parameters, applicable test procedures, and normal test records of the same vehicle model as recorded in the aforementioned steps. No empirical parameters unrelated to the aforementioned determination methods are used.

[0102] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0103] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0104] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for intelligent identification of railway braking faults based on pressure time series analysis, characterized in that, Includes the following steps: S1. Obtain the multi-channel pressure timing and test control status of the railway braking test, determine the first failure point and the first anomaly based on the non-compliance judgment conditions, solidify the non-compliance results, and determine the original gas path boundary state at the time of the first failure. S2. If the safety termination conditions are not met, maintain the original gas path boundary state, do not change the state and do not add new diagnostic disturbances, continue the original physical process of the standard test, and obtain the natural pressure evidence segment after failure. S3. Based on the first anomaly and the natural pressure evidence segment after failure, determine the candidate demand channel and the upstream pressure channel, determine whether the first anomaly forms a continuous gas demand, and determine it as a natural failure excitation when it forms, so as to form a natural demand timing sequence and a synchronous upstream gas supply response timing sequence. S4. Use normal test records of the same vehicle model, the same test items, and the same original gas circuit boundary conditions to form the gas supply response judgment benchmark. Determine the gas supply response mismatch state based on the matching relationship between the natural demand time sequence and the upstream gas supply response time sequence, and form a single fault or compound fault identification result.

2. The intelligent identification method for railway braking faults based on pressure time series analysis according to claim 1, characterized in that, Also includes: S5. Based on the identification results, release the maintenance of the original gas path boundary state, perform confirmation test, exit test or brake component failure handling, and use the moment when the original gas path boundary state is changed as the control intervention boundary, and isolate the subsequent pressure data from the natural pressure evidence section after failure.

3. The intelligent identification method for railway braking faults based on pressure time series analysis according to claim 1, characterized in that, S1 specifically includes: The edge computing node obtains the pressure values ​​of the brake pipe, the gas storage unit that is actually supplying gas in the current test, and the brake cylinder through the pressure acquisition interface, and obtains the test control status through the control interface. Synchronization is performed according to a unified timestamp to form a sequence of synchronized data frames containing multi-channel pressure status, test control status, and valid data markers. Based on the non-compliance criteria corresponding to the current vehicle model and test items, the synchronous data frame sequence is sampled and judged one by one during the corresponding test phase. The moment when the valid non-compliance criteria are first met is determined as the first failure point, and the first anomaly is formed, which includes the trigger criteria, pressure channels, and pressure changes.

4. The intelligent identification method for railway braking faults based on pressure time series analysis according to claim 3, characterized in that, Also includes: Based on the first failure point, the unqualified results of the current test are solidified. The control state that determines the gas supply, exhaust, isolation and connection relationship is extracted from the test control state at the time of the first failure, forming the original gas path boundary state. This state, together with the first failure point, the first anomaly and the pressure sequence before the first failure, constitutes the first failure state information.

5. The intelligent identification method for railway braking faults based on pressure time series analysis according to claim 1, characterized in that, S2 specifically includes: Based on the initial failure status information, the control actions that have not yet been executed after the initial failure are divided into maintenance actions that do not change the original gas path boundary state and change actions that change the original gas path boundary state. When the current test safety termination conditions are not met, the original gas path boundary maintenance permission is formed, and the execution of change actions is prohibited. During the period when the original gas path boundary remains valid, only the maintenance actions already determined in the current standard test will continue to be performed. No new diagnostic disturbances will be added. Multi-channel pressure data will be continuously acquired according to the same sampling period as before the first failure. The boundary retention validity mark will be determined based on the pressure safety boundary, the allowable retention time, and whether the original gas path boundary state has changed. The continuous pressure sequence after the first failure point, during which the boundary remains valid and no change-type actions are performed, is defined as the natural pressure evidence segment after failure, and the moment when the original gas path boundary state is first actually changed is defined as the control intervention boundary.

6. The intelligent identification method for railway braking faults based on pressure time series analysis according to claim 1, characterized in that, S3 specifically includes: Based on the pressure channels involved in the first anomaly and the original gas path boundary status, the pressure channels that can reflect the continuous gas supply demand of the pneumatic unit where the first anomaly is located are identified as candidate demand channels, and the pressure channels that currently actually bear the gas supply function are identified as upstream pressure channels, forming a set of demand-related channels and corresponding effective analysis intervals. Within the effective analysis interval, a natural aerodynamic demand characterization quantity is formed based on the pressure change rate along the direction of continuous compressed air consumption in the candidate demand channel. A demand judgment threshold is formed by combining the effective fluctuation upper limit in the normal test of the same vehicle model and the same test item with the pressure measurement uncertainty. When the characterization quantity continuously reaches the threshold and exceeds the normal instantaneous pressure fluctuation duration range, a natural demand time series is formed.

7. The intelligent identification method for railway braking faults based on pressure time series analysis according to claim 6, characterized in that, Also includes: The natural demand timing sequence is determined as the natural fault excitation, and the pressure change rate of each upstream pressure channel along the normal gas supply direction is extracted within the same time index to form the upstream gas supply response timing sequence.

8. The intelligent identification method for railway braking faults based on pressure time series analysis according to claim 1, characterized in that, S4 specifically includes: When the continuous pneumatic demand is established, the lower and upper limits of the normal response of each upstream pressure channel are determined from the reference pressure records of the same model, the same test item, the same original air circuit boundary state, and the air supply path confirmed to be fault-free by normal testing, according to the demand interval where the natural pneumatic demand characterization quantity is located, to form the air supply response judgment benchmark. During the period when the natural gas demand is continuously established, the actual gas supply response change rate of each upstream pressure channel is compared with the normal response lower bound of the corresponding demand range. When the actual gas supply response is continuously lower than the normal response lower bound and the duration reaches the maximum effective duration for the normal gas supply path to complete the response in the demand range, it is determined that there is a gas supply response mismatch in the corresponding upstream pressure channel. In the absence of insufficient evidence or inconsistency in gas path boundaries, a single fault identification result or a composite fault identification result including the first fault and the corresponding upstream gas supply path gas supply obstruction fault is formed based on the mismatch between the first anomaly and the gas supply response.

9. The intelligent identification method for railway braking faults based on pressure time series analysis according to claim 2, characterized in that, S5 specifically includes: Based on the single fault identification result or the combined fault identification result and the current safe termination conditions, determine the confirmation test, exit test or brake component fault handling branch, revoke the original air circuit boundary maintenance permission, and execute the corresponding handling command through the control interface of the single vehicle test equipment; The moment when the gas supply, exhaust, isolation, or connection status is first confirmed to be outside the original gas path boundary is taken as the final control intervention boundary. The natural pressure evidence segment before the final control intervention boundary is saved as natural fault evidence, and the pressure data thereafter is marked as control intervention data. The control intervention data is no longer used to judge the mismatch between natural demand and gas supply response.

10. The intelligent identification method for railway braking faults based on pressure time series analysis according to claim 9, characterized in that, Also includes: Verify whether the handling action has been completed based on the control intervention data and the target state criteria corresponding to the single-vehicle test equipment, and form the final fault handling record by combining the fault identification results, fault evidence, handling object, final control intervention boundary and handling execution results.

Images