Intelligent networked automobile fault detection method, equipment and medium

Abstract

The invention discloses an intelligent networked automobile fault detection method and device and a medium, and relates to the technical field of fault diagnosis, and the method comprises the steps: obtaining control link data and networked state information of an intelligent networked automobile, segmenting the control link data according to a preset time window, and generating windowed data segments; performing consistency evaluation on the windowed data fragments according to a preset closed-loop consistency constraint, generating a residual vector, mapping the residual vector into a suspected fault set, and calculating an observability index based on the suspected fault set, the residual vector and the network connection state information; and when the drift index reaches an update judgment threshold, generating a check parameter update package, and updating the closed-loop consistency constraint and observation enhancement strategy by using the check parameter update package. According to the method, the closed-loop consistency constraint and the observation enhancement strategy are updated by applying the inspection parameter update package, so that the fault diagnosis reliability and continuity of the intelligent networked automobile are enhanced.

Description

Technical Field

[0001] This invention relates to the field of fault diagnosis technology, and in particular to a method, equipment and medium for fault detection in intelligent connected vehicles. Background Technology

[0002] As automotive electronic and electrical architecture evolves from distributed to domain-centralized and cross-domain integrated, intelligent connected vehicles are gradually forming a complex control closed loop with deep coupling of perception, decision-making, execution, and communication. The key execution links in the vehicle operation process not only rely on the real-time collaboration between on-board controllers, but are also affected by the communication status of vehicle-road-cloud, actuator constraint boundaries, and the accuracy of state estimation. Especially in the intelligent connected vehicle scenario, control link failures are no longer limited to traditional hardware failures, but also include complex anomalies such as communication disturbances, closed-loop mismatch, state drift, and constraint conflicts.

[0003] However, existing technologies still have the following shortcomings: they only detect single-type fault characteristics, resulting in insufficient ability to identify sequential faults, slowly changing faults, and coupled faults. Existing fault detection strategies often use fixed constraint parameters and fixed observation strategies, which are difficult to cope with parameter drift, operating condition migration, and control link degradation during long-term vehicle operation. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a fault detection method for intelligent connected vehicles to address the problems of parameter drift, operating condition shifts, and control link degradation that are difficult to handle during long-term vehicle operation.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, the present invention provides a fault detection method for intelligent connected vehicles, comprising, Acquire control link data and network status information of intelligent connected vehicles, segment the control link data according to a preset time window, and generate windowed data fragments; Consistency assessment is performed on windowed data segments according to preset closed-loop consistency constraints, residual vectors are generated and mapped to a set of suspected faults, and observability indicators are calculated based on the set of suspected faults, residual vectors and network status information. Based on the observability index, an observation enhancement strategy is selected and configured. The control link data is sampled and segmented through the observation enhancement strategy to obtain fault event packets. The residual vector is corrected based on the fault event packets, and the corrected residual vector is output. Fault identification is performed on the correction residual vector, the fault diagnosis result is output, and the correction residual vector is cached according to a preset time window to form a residual vector sequence. The drift index is calculated based on the fault diagnosis result and the residual vector sequence. When the drift index reaches the update decision threshold, an inspection parameter update package is generated and applied to update the closed-loop consistency constraints and observation enhancement strategies.

[0007] In a preferred embodiment of the intelligent connected vehicle fault detection method of the present invention, the specific steps for generating windowed data fragments are as follows: Acquire control link data and network status information of intelligent connected vehicles, unify the timestamp of control link data, and divide the control link data after unifying the timestamp according to a preset time window to obtain multiple window segments; Multiple window segments are bound to network status information one by one to generate windowed data segments.

[0008] In a preferred embodiment of the intelligent connected vehicle fault detection method of the present invention, the specific steps for generating the residual vector are as follows: Extract the control command sequence, feedback measurement sequence, state estimation sequence, and actuator constraint state from the windowed data fragments as closed-loop consistency evaluation data; Closed-loop consistency constraints are set based on vehicle control logic and actuator constraint logic. The closed-loop consistency constraints include control timing constraints, state consistency constraints, and actuator consistency constraints. The trigger time is extracted from the control command sequence, the arrival time is extracted from the feedback measurement sequence, the time difference between the trigger time and the arrival time is calculated, and the time difference is compared with the allowable delay range limited by the control timing constraints to generate the control timing consistency deviation. The state estimation sequence and the feedback measurement sequence are time-aligned and the difference is calculated to output the deviation sequence. The deviation sequence is then compared with the allowable deviation range defined by the state consistency constraint to generate the state consistency deviation. Extract the limiting interval and saturation interval from the actuator constraint state, extract the segments in the control command sequence and feedback measurement sequence that are located in the limiting interval and saturation interval, output the control command subsequence and feedback measurement subsequence, and perform difference calculation to obtain the actuator deviation sequence. Compare the actuator deviation sequence with the allowable deviation range defined by the actuator consistency constraint to generate the actuator consistency deviation. The control timing consistency deviation, state consistency deviation, and actuator consistency deviation are normalized and combined to generate a residual vector.

[0009] As a preferred embodiment of the intelligent connected vehicle fault detection method of the present invention, the specific steps for calculating the observability index based on the suspected fault set, residual vector, and connected status information are as follows: Perform component analysis on the residual vector to generate residual features; The residual features are matched with a pre-defined fault mode library to generate a set of suspected faults. Each candidate fault type in the suspected fault set is associated with a residual vector to generate a fault residual association table. Based on the network status information, the observability assessment rules are configured, and the observability assessment of the fault residual association table is performed through the observability assessment rules to generate observability indicators.

[0010] In a preferred embodiment of the intelligent connected vehicle fault detection method of the present invention, the specific steps for obtaining the fault event package are as follows: The suspected fault set is sorted by observability indicators to form a priority diagnosis order; The priority diagnosis order is hierarchically sampled and scheduled, and fragment caching rules are configured through network status information to form an observation enhancement strategy; By using observation enhancement strategies, control link data is sampled to generate highly observable data segments; Highly observable data fragments are cached and encapsulated according to fragment caching rules to obtain fault event packets.

[0011] As a preferred embodiment of the intelligent connected vehicle fault detection method of the present invention, the output correction residual vector refers to performing a consistency evaluation on the highly observable data segments in the fault event package to obtain the correction deviation component, updating the residual vector through the correction deviation component, and outputting the correction residual vector.

[0012] In a preferred embodiment of the intelligent connected vehicle fault detection method of the present invention, the specific steps for forming the residual vector sequence are as follows: The correction residual vector is matched with the candidate fault type to generate fault similarity. The fault similarity is then normalized, and the discrimination score is output. The fault type and faulty component are determined by the discrimination score, and the fault diagnosis result is output. The fault diagnosis results are bound one by one with the correction residual vector and timestamped to generate a correction residual time series record; The correction residual vector records are cached and collected according to the preset time window to form a residual vector sequence.

[0013] As a preferred embodiment of the intelligent connected vehicle fault detection method of the present invention, the specific steps of applying the check parameter update package to update the closed-loop consistency constraint and observation enhancement strategy are as follows: The drift index is compared with the update determination threshold to obtain the update trigger determination result; When the updated trigger judgment result representation needs to be updated, the parameters of the closed-loop consistency constraint and the parameters of the observation enhancement strategy are adjusted by the drift index and encapsulated into a check parameter update package. The parameters of the closed-loop consistency constraints and observation enhancement strategies are replaced by checking the parameter update package, and the updated closed-loop consistency constraints and observation enhancement strategies are output.

[0014] In a second aspect, the present invention provides a computer device including a memory and a processor, wherein the memory stores a computer program, wherein when the computer program is executed by the processor, it implements any step of the intelligent connected vehicle fault detection method as described in the first aspect of the present invention.

[0015] Thirdly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the intelligent connected vehicle fault detection method as described in the first aspect of the present invention.

[0016] The beneficial effects of this invention are as follows: By dividing the control link data after unified timestamps into multiple window segments according to a preset time window, time-series management of control signals and vehicle status is achieved, reducing data processing complexity and maintaining time correlation. By performing consistency evaluation on the windowed data segments and generating residual vectors, effective features can be extracted from complex vehicle control data, enabling fault diagnosis to specifically identify potential anomalies and reduce false positives and false negatives. By caching and encapsulating highly observable data segments through segment caching rules, fault event packages are obtained, significantly improving the real-time performance and accuracy of fault detection. By applying inspection parameter update packages to update closed-loop consistency constraints and observation enhancement strategies, the reliability and continuity of fault diagnosis for intelligent connected vehicles are enhanced. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 A flowchart for a fault detection method for intelligent connected vehicles; Figure 2A flowchart for generating windowed data fragments; Figure 3 A flowchart for generating a set of suspected faults; Figure 4 A flowchart for updating closed-loop consistency constraints and observation enhancement strategies. Detailed Implementation

[0019] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0020] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0021] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0022] Reference Figures 1-4 As one embodiment of the present invention, this embodiment provides a fault detection method for intelligent connected vehicles, including the following steps: S1: Obtain control link data and network status information of intelligent connected vehicles, segment the control link data according to a preset time window, and generate windowed data fragments.

[0023] S1.1: Obtain the control link data and network status information of the intelligent connected vehicle, unify the timestamp of the control link data, and divide the control link data after unifying the timestamp according to the preset time window to obtain multiple window segments.

[0024] Control link data and network status information of intelligent connected vehicles are collected through the vehicle bus interface, domain controller communication interface, and vehicle communication unit. Specifically, control command signals output from the vehicle controller, powertrain domain controller, chassis domain controller, body domain controller, or intelligent driving domain controller are collected through the domain controller communication interface; feedback measurement signals output from vehicle speed sensors, wheel speed sensors, steering angle sensors, pressure sensors, current / voltage sensors, temperature sensors, or displacement sensors are collected through the sensor acquisition interface; state estimation signals output from the battery management system, thermal management controller, powertrain controller, or vehicle controller are collected through the domain controller communication interface; and limiting state, saturation state, or fault protection state signals output from the actuator controller are collected through the domain controller communication interface as actuator constraint state signals. The control command signals, feedback measurement signals, state estimation signals, and actuator constraint state signals are then aggregated to obtain control link data, which is then transmitted through the vehicle bus interface. The communication unit is used to collect network connection status and its changes, and aggregate them to form network status information. The network connection status includes online status, offline status, connection in progress status, reconnection status, weak network status, packet loss status, high latency status, and communication recovery status. The control link data of the intelligent connected vehicle and the network status information are aggregated to form the original acquisition record. The sampling timestamps of the control link signals are read one by one from the original acquisition record and arranged from early to late to form a sampling timestamp sequence. The sampling timestamps in the sampling timestamp sequence are uniformly converted to the same time unit, such as ms, to form a unit-consistent sampling timestamp sequence. The smallest timestamp in the unit-consistent sampling timestamp sequence is selected as the starting time. Each timestamp in the unit-consistent sampling timestamp sequence is subtracted from the starting time to obtain the relative timestamp sequence. The relative timestamp sequence and the control link signal sampling values ​​are aggregated to form the control link data with unified timestamps.

[0025] A unified sampling time axis is established within the time range covered by the relative timestamp sequence. The unified sampling time axis starts with the minimum timestamp of the relative timestamp sequence and ends with the maximum timestamp. Starting from the start time, a fixed step size (e.g., 10ms) is incremented sequentially to generate time points. Each increment outputs one time point, and output stops when a time point exceeds the end time. All time points are collected in chronological order to form a time point sequence. Timestamp alignment is performed on each control link signal on the unified sampling time axis. Timestamp alignment uses a combination of finding the nearest sampled value in the time point sequence and linear interpolation to obtain an aligned sampled value sequence. This aligned sampled value sequence is then combined with the time point sequence to form a time... The control link data after a unified timestamp is consistent with the axis. The frequency of change of the control link signal is extracted from the relative timestamp sequence of the control link data after the unified timestamp. The time window length and time window step are determined according to the frequency of change of the control link signal. The time window length is used to cover the time span of a complete control action, and the time window step is used to control the sliding interval between window segments. For example, the time window length is 1 second and the time window step is 200 ms. The start and end time intervals are divided on the unified sampling time axis according to the time window length and time window step. The control link data after the unified timestamp covered by each start and end time interval is extracted to obtain a window segment, and multiple window segments are collected.

[0026] S1.2: Bind multiple window segments to network status information one by one to generate windowed data segments.

[0027] Each window segment is extracted one by one from multiple window segments, and the start and end times of each window segment are read. Among the network connection status change times, the network connection status change times between the start and end times of the window segment are selected. The network connection status change times and network status identifiers are combined to form a network status sequence within the window. When the network status sequence within the window is empty, the most recent network status identifier less than the start time of the window segment is selected as the network status identifier of the window segment. Otherwise, the network status identifiers of the network status sequence within the window are recorded in chronological order. The window segments are bound one by one with the network status identifiers of the window segments and combined to obtain windowed data segments.

[0028] S2: Perform consistency evaluation on the windowed data fragments according to the preset closed-loop consistency constraints, generate residual vectors, and map them to a set of suspected faults. Calculate the observability index based on the set of suspected faults, residual vectors, and network status information.

[0029] S2.1: Extract the control command sequence, feedback measurement sequence, state estimation sequence, and actuator constraint state from the windowed data fragment as closed-loop consistency evaluation data; set closed-loop consistency constraints based on vehicle control logic and actuator constraint logic, including control timing constraints, state consistency constraints, and actuator consistency constraints.

[0030] One windowed data segment is extracted from each windowed data segment, and the control link data is arranged in chronological order within the windowed data segment. Control command fields are then filtered from the chronologically arranged control link data, their values ​​are read, and aggregated in chronological order to form a control command sequence. Feedback measurement fields are then filtered from the chronologically arranged control link data, their values ​​are read, and aggregated in chronological order to form a feedback measurement sequence. State estimation fields are then filtered from the chronologically arranged control link data, their values ​​are read, and aggregated in chronological order to form a state estimation sequence. Actuator constraint fields are then filtered from the chronologically arranged control link data, their values ​​are read, and aggregated in chronological order to form an actuator constraint state. The control command sequence, feedback measurement sequence, state estimation sequence, and actuator constraint state are then bound according to the time range of the windowed data segment to form closed-loop consistency evaluation data.

[0031] In the vehicle control logic, the control cycle requirements and the effective order requirements of the control command sequence are extracted and written into the control timing constraints. The control timing constraints are used to limit the time sequence consistency and interval consistency of the control command sequence in the closed-loop consistency evaluation data. Similarly, the state update relationship between the feedback measurement sequence and the state estimation sequence is extracted in the vehicle control logic and written into the state consistency constraints. The state consistency constraints are used to limit the consistency of the change trend of the feedback measurement sequence and the change trend of the state estimation sequence in the closed-loop consistency evaluation data. In the actuator constraint logic, the amplitude limiting rule, speed limiting rule, and enabling rule of the actuator constraint state are extracted and written into the actuator consistency constraints. The actuator consistency constraints are used to limit the constraint consistency between the control command sequence and the actuator constraint state in the closed-loop consistency evaluation data. The control timing constraints, state consistency constraints, and actuator consistency constraints are combined to form the closed-loop consistency constraints.

[0032] It should be noted that closed-loop consistency constraints are a set of judgment rules established for the control command sequence, feedback measurement sequence, state estimation sequence, and actuator constraint state in the control link data. By extracting the control delay characteristics, state deviation characteristics, and actuator response deviation characteristics from the historical normal operation window segments, control timing constraints, state consistency constraints, and actuator consistency constraints are set respectively. Among them, the control timing constraints are represented by the allowable delay range, the state consistency constraints are represented by the allowable deviation range, and the actuator consistency constraints are represented by the actuator response deviation range.

[0033] S2.2: Extract the trigger time from the control command sequence, extract the arrival time from the feedback measurement sequence, calculate the time difference between the trigger time and the arrival time, and compare the time difference with the allowable delay range limited by the control timing constraints to generate the control timing consistency deviation.

[0034] In the control command sequence, control command values ​​are read sequentially over time, and the locations where control command values ​​change are identified. The timestamps of these changes are recorded as trigger moments. Multiple trigger moments are then aggregated into a trigger moment set. In the feedback measurement sequence, feedback measurement values ​​are read sequentially over time. Based on the response determination rules corresponding to the control command sequence, the corresponding response feature positions in the feedback measurement sequence are identified, and the timestamps of these response feature positions are recorded as response moments. Multiple response moments are then aggregated into a response moment set. Trigger moments are then extracted one by one from the trigger moment set and added to the response moment set. The minimum response time that is greater than the trigger time is selected. The minimum response time is subtracted from the trigger time to obtain the time difference. Multiple time differences are collected in chronological order to form a time difference set. The allowable delay range is read from the control timing constraints. The allowable delay range is the delay interval formed by the lower and upper limits of the delay. The value range of the allowable delay range is set according to the statistical results of the normal transmission delay from the control command to the feedback response, the vehicle control cycle, and the actuator response cycle. For example, it is 20ms-80ms. Each time difference in the time difference set is compared with the allowable delay range. Time differences that exceed the allowable delay range are recorded as control timing consistency deviations.

[0035] It should be noted that the response determination rule refers to the determination rule used to determine the response feature position in the feedback measurement sequence corresponding to the change of control command. The response determination rule is obtained by statistically analyzing the corresponding change relationship between the control command sequence and the feedback measurement sequence in the historical normal operation window segment and by the dynamic response characteristics of different control quantities. The response feature position includes the position where the feedback measurement sequence begins to respond to the change of control command, the position where the feedback measurement sequence enters the stable change range, or the position where the feedback measurement sequence reaches the effective response range.

[0036] S2.3: Time-align the state estimation sequence with the feedback measurement sequence, calculate the difference, output the deviation sequence, compare the deviation sequence with the allowable deviation range defined by the state consistency constraint, and generate the state consistency deviation.

[0037] In the state estimation sequence, each state estimation value and its corresponding timestamp are read in chronological order. Similarly, in the feedback measurement sequence, each feedback measurement value and its corresponding timestamp are read in chronological order. Based on the timestamp correspondence, each state estimation value in the state estimation sequence is paired with the most recent feedback measurement value in the feedback measurement sequence to generate time-aligned state estimation data pairs. The state quantity type corresponding to the state estimation value and the measurement quantity type corresponding to the feedback measurement value are read. It is then determined whether the state quantity type and the measurement quantity type are consistent. If they are consistent, the state estimation value is directly used as the mapping value. If they are inconsistent, the state mapping relationship table is called, and the mapping relationship corresponding to the state quantity type and the measurement quantity type is read. The state estimation value is then transformed according to the mapping relationship to generate a mapping value. Multiple mapping values ​​are aggregated in chronological order to generate a state mapping sequence.

[0038] The difference between each mapped value in the state mapping sequence and the corresponding feedback measurement value in the feedback measurement sequence is calculated to obtain the deviation value. Multiple deviation values ​​are collected in chronological order to output the deviation sequence. Each deviation value in the deviation sequence is compared with the allowable deviation range defined by the state consistency constraint. The allowable deviation range is the deviation interval consisting of the lower deviation limit and the upper deviation limit. The value range of the allowable deviation range is set according to the statistical distribution of the deviation between the state estimation sequence and the feedback measurement sequence in the historical normal operation window segment and the sensor measurement error range, for example, -3% to 3%. Deviation values ​​that exceed the allowable deviation range are recorded as state consistency deviations.

[0039] It should be noted that the state mapping table refers to the set of correspondences between state quantity types and measurement quantity types. It is obtained by statistically analyzing the correspondence between state estimation sequences and feedback measurement sequences in historical normal operation window segments, and by understanding the physical meaning of state quantities and measurement quantities.

[0040] S2.4: Extract the limiting interval and saturation interval from the actuator constraint state, extract the segments in the control command sequence and feedback measurement sequence that are located in the limiting interval and saturation interval, output the control command subsequence and feedback measurement subsequence, and perform difference calculation to obtain the actuator deviation sequence. Compare the actuator deviation sequence with the allowable deviation range limited by the actuator consistency constraint to generate the actuator consistency deviation.

[0041] In the actuator constraint state, the numerical ranges of the limiting interval and saturation interval are read one by one, and the start and end values ​​of the limiting interval and saturation interval are recorded. The values ​​of each control instruction in the control instruction sequence are filtered through the limiting interval and saturation interval, and the control instructions located within the limiting interval and saturation interval are extracted to form a control instruction subsequence. At the same time, the measurement values ​​at the corresponding time points in the feedback measurement sequence are filtered in the same way to form a feedback measurement subsequence. The actuator type corresponding to the control instruction subsequence and the response type corresponding to the feedback measurement subsequence are read. It is determined whether the instruction quantity of the control instruction subsequence and the response quantity of the feedback measurement subsequence belong to the same response representation type. When the instruction quantity of the control instruction subsequence and the response quantity of the feedback measurement subsequence belong to the same response representation type, the control instruction value in the control instruction subsequence is used as the reference response value. Otherwise, the mapping relationship corresponding to the actuator type and response type in the actuator response mapping relationship table is read, and the control instruction value in the control instruction subsequence is transformed in dimension through the mapping relationship to generate a reference response value. Multiple reference response values ​​are collected in chronological order to generate a response reference sequence.

[0042] The difference between each reference response value in the response reference sequence and the corresponding feedback response value in the feedback measurement subsequence is calculated, and the absolute value of the difference is taken as the actuator deviation value. Multiple actuator deviation values ​​are aggregated in chronological order to obtain an actuator deviation sequence. Each actuator deviation value in the sequence is compared with the actuator response deviation range defined by the actuator consistency constraint. The actuator response deviation range is the response deviation interval consisting of the lower and upper limits of the response deviation, based on the changes in control commands and feedback measurements within the limiting and saturation intervals during historical normal operation window segments. The statistical distribution of the deviation of the response change, the dynamic response characteristics of the actuator, and the boundary setting of the actuator protection action set the range of the actuator response deviation, for example, -5% to 5%. The actuator deviation value that exceeds the range of the actuator response deviation is recorded as the actuator consistency deviation. The control timing consistency deviation, state consistency deviation and actuator consistency deviation are normalized. The normalization is performed by the maximum and minimum value normalization process to obtain the normalized control timing deviation value, the normalized state consistency deviation value and the normalized actuator consistency deviation value. They are then combined in time order to form a residual vector.

[0043] It should be noted that the actuator response mapping table refers to the set of correspondences between actuator type, control command type, and feedback response type, which is obtained by statistical analysis of the corresponding changes in the control command sequence and feedback measurement sequence in historical normal operation window segments.

[0044] S2.5: Perform component analysis on the residual vector to generate residual features; match the residual features with a preset fault mode library to generate a set of suspected faults.

[0045] Each component of the residual vector is read, and the meaning of each component in different types of deviations is labeled, including control timing deviation components, state consistency deviation components, and actuator consistency deviation components. Statistical characteristics of each component, such as absolute value, trend, and relative magnitude, are extracted to form residual features. These residual features are then matched with the features of each fault mode in the fault mode library. The cosine of the angle between all components of the residual features and all vectors in the fault mode features is calculated and used as the similarity score. A similarity threshold is set based on historical fault data statistical analysis and engineering experience. The range of values ​​for the similarity threshold is... The similarity threshold is set to a range of 0.7-0.95. Based on the minimum distinguishable limit and the maximum tolerance limit obtained from the distribution analysis of residual characteristics of different fault types in the control link, the similarity threshold is set. The range of 0.7-0.95 can take into account the false negative rate and the false positive rate, ensuring that the selected suspected fault modes include real faults while minimizing false positives. When the similarity is greater than the similarity threshold, the fault mode corresponding to the similarity is recorded. Conversely, the fault modes corresponding to the similarity less than or equal to the similarity threshold are removed. All fault modes with similarity greater than the similarity threshold are summarized to obtain a complete set of suspected faults.

[0046] It should be noted that the fault mode library refers to a collection of known control link fault modes, obtained through statistical analysis and fault mechanism analysis of historical fault data. It includes residual feature templates and abnormal performance features corresponding to various faults, which are used to match with the residual features extracted in real time to filter possible fault types. The faults covered by the fault mode library include control timing abnormality faults, state estimation deviation faults, actuator limiting faults, actuator saturation faults, communication delay abnormality faults, communication packet loss faults, closed-loop mismatch faults, state drift faults, and constraint conflict faults.

[0047] S2.6: Associate each candidate fault type in the suspected fault set with the residual vector to generate a fault residual association table; configure observability assessment rules based on the network status information, and perform observability assessment on the fault residual association table through the observability assessment rules to generate observability indicators.

[0048] Each candidate fault type is extracted one by one from the suspected fault set. The residual value corresponding to the candidate fault type is compared with the residual vector item by item. The component and index in the residual vector that matches the residual value of the candidate fault type are recorded to complete the association and correspondence of fault residuals and form a fault residual association table.

[0049] Candidate fault types are retrieved one by one from the fault residual association table, and the indexes associated with all candidate fault types are aggregated to form a residual component index set. Network status identifiers are read sequentially in the windowed data segment, and multiple network status intervals are divided based on the change time of the network status identifiers. Each network status interval has a network status identifier, an interval start time, and an interval end time. Observability evaluation rules are configured based on the network status information. The observability evaluation rules include interval division rules based on the change time of the network status identifier. Within each network status interval, the residual component values ​​indicated by the residual component index set are read, and... The rules for extracting interval residual amplitudes are: the maximum absolute value of residual component values; the effective residual judgment rule is: the interval residual amplitude is greater than the effective residual judgment threshold; the ratio of the number of effective residual intervals to the total number of network status intervals is: the residual occurrence frequency; the sum of the time lengths of effective residual intervals is: the length of the continuous residual time period; and the ratio of the number of network status identifier types with effective residuals to the total number of network status identifier types is: the statistical rule for status coverage integrity. The rules for normalizing and combining the residual occurrence frequency, the length of the continuous residual time period, and the status coverage integrity are then used to form observable indicators.

[0050] According to the observability assessment rules, the residual vector is read sequentially within each network status interval. The residual component values ​​indicated by the residual component index set are read from the residual vector. The maximum absolute value of the residual component value is taken as the interval residual amplitude of the network status interval. The interval residual amplitude and the time length from the start time to the end time of the interval are recorded to form an interval assessment record. Statistical analysis is performed based on the residual component distribution in the historical normal operation window segment. The effective residual judgment threshold is set by the residual fluctuation range and noise level. The effective residual judgment threshold range is 0.05~0.2. The effective residual judgment threshold range is set according to the residual distribution discrimination between normal samples and abnormal samples and the balance requirement between false alarm rate and false negative rate. The range of 0.05~0.2 can filter out small residual fluctuations caused by random noise and short-term disturbances, while retaining continuous residual changes caused by faults and anomalies, thereby improving the discrimination of residual occurrence frequency, residual continuous time period length and status coverage integrity.

[0051] In the interval evaluation record, the number of network status intervals with interval residual amplitudes greater than the effective residual judgment threshold is counted. The number of network status intervals with interval residual amplitudes greater than the effective residual judgment threshold is compared with the total number of network status intervals to obtain the residual occurrence frequency. The time lengths of network status intervals with interval residual amplitudes greater than the effective residual judgment threshold are summed to obtain the length of the continuous residual time period. The number of network status identifier types with interval residual amplitudes greater than the effective residual judgment threshold is read. The ratio of the number of network status identifier types with interval residual amplitudes greater than the effective residual judgment threshold to the total number of network status identifier types is used as the state coverage integrity. The residual occurrence frequency, the length of the continuous residual time period, and the state coverage integrity are arranged and combined to form an observability index. The observability index is written into the record position of the corresponding candidate fault type in the fault residual association table.

[0052] Furthermore, the observability index is used to characterize the degree to which candidate fault types are effectively observed and distinguished under the current network conditions. It is jointly formed by the residual occurrence frequency, the length of the residual continuous time period, and the state coverage integrity. Specifically, the residual occurrence frequency, the length of the residual continuous time period, and the state coverage integrity are mapped to the range of 0-1 using an interval mapping method to obtain the frequency normalized value, the duration normalized value, and the coverage integrity normalized value. Based on the degree of influence of the residual occurrence frequency on fault identifiability, the degree of influence of the residual continuous time period length on fault stability, and the degree of influence of the state coverage integrity on the sufficiency of observation, frequency weight, duration weight, and coverage integrity weight are set respectively. The frequency normalized value, duration normalized value, and coverage integrity normalized value are weighted with the frequency weight, duration weight, and coverage integrity weight respectively to generate the observability index.

[0053] S3: Select and configure observation enhancement strategies based on observability indicators, sample and cache control link data through observation enhancement strategies to obtain fault event packets, correct residual vectors based on fault event packets, and output corrected residual vectors.

[0054] S3.1: Sort the suspected fault set by observability indicators to form a priority diagnosis order; perform hierarchical sampling scheduling on the priority diagnosis order, and configure fragment caching rules through network status information to form an observation enhancement strategy.

[0055] Candidate fault types are extracted one by one from the suspected fault set, and the observability indicators corresponding to the candidate fault types are read from the fault residual association table. Each candidate fault type and observability indicator in the suspected fault set are combined into a sorting record. The sorting record is sorted from largest to smallest by residual occurrence frequency. If the residual occurrence frequency is the same, it is sorted from largest to smallest by the length of the continuous residual time period. If the length of the continuous residual time period is the same, it is sorted from largest to smallest by the state coverage integrity to obtain the priority diagnosis order.

[0056] Based on the residual occurrence frequency, residual continuous time period length, and state coverage integrity corresponding to each candidate fault type in the priority diagnosis sequence, the priority diagnosis sequence is divided into multiple diagnosis levels, including high-priority diagnosis level, medium-priority diagnosis level, and low-priority diagnosis level. Sampling scheduling rules are set for each of the high-priority, medium-priority, and low-priority diagnosis levels. The sampling scheduling rules include sampling frequency and sampling window length. For example, the high-priority diagnosis level uses a sampling frequency of once per second and a sampling window length of 1 second, the medium-priority diagnosis level uses a sampling frequency of once every two seconds and a sampling window length of 2 seconds, and the low-priority diagnosis level uses a sampling frequency of once every five seconds and a sampling window length of 5 seconds.

[0057] The network connection status interval is divided according to the time of network connection status change. The network connection status interval is used to configure fragment caching rules. The fragment caching rules include caching trigger conditions, caching time span, and caching retention quantity. The caching trigger conditions adopt the network connection status identifier change trigger method or the network connection status change trigger method. The caching time span is used to limit the time range of windowed data fragments that need to be cached. The caching retention quantity is used to limit the number of windowed data fragments that need to be retained. The sampling scheduling rules of hierarchical sampling scheduling are combined with the fragment caching rules to obtain the observation enhancement strategy.

[0058] S3.2: Sampling of control link data through observation enhancement strategy to generate highly observable data segments; caching and encapsulating the highly observable data segments according to the segment caching rules to obtain fault event packets.

[0059] The sampling trigger time and sampling window length are determined by sampling scheduling rules. These two parameters determine the time range of control link data to be sampled. Control link data segments are extracted according to the sampling trigger time and sampling window length, and the extracted control link data segments are bound to network status information within the same time range. The bound result is recorded as a high observability data segment. The high observability data segments that need to be cached are identified by the cache trigger conditions of the segment caching rules. When the cache trigger conditions are met, the high observability data segments are written into the cache queue. Based on the cache time span of the segment caching rules, high observability data segments that fall within the cache time span are retained from the cache queue, and high observability data segments exceeding the cache retention limit are deleted, forming a cached segment set. The cached segment set is encapsulated with the candidate fault type identifier, network status identifier, and timestamp in the priority diagnosis order to obtain a fault event packet.

[0060] S3.3: Perform a consistency assessment on the highly observable data segments in the fault event package to obtain the correction bias component, update the residual vector through the correction bias component, and output the correction residual vector.

[0061] Extract the highly observable data segment from the fault event packet. Align the control timing deviation, state consistency deviation, and actuator consistency deviation in the highly observable data segment with the corresponding deviation components in the residual vector according to the time order. Calculate the difference between the deviation value in the highly observable data segment and the corresponding deviation component in the residual vector. Use the difference as the correction deviation value. Collect all correction deviation values ​​according to the order of the residual vector deviation components to form correction deviation components. Replace the corresponding components in the residual vector with the correction deviation components according to the component order of the residual vector to obtain the updated correction residual vector.

[0062] S4: Perform fault identification on the correction residual vector, output the fault diagnosis result, and cache the correction residual vector according to the preset time window to form a residual vector sequence. Calculate the drift index based on the fault diagnosis result and the residual vector sequence.

[0063] S4.1: Match the corrected residual vector with the candidate fault types to generate fault similarity, normalize the fault similarity, and output the discrimination score; determine the fault type and faulty component through the discrimination score, and output the fault diagnosis result.

[0064] The control timing deviation, state consistency deviation, and actuator consistency deviation in the correction residual vector are matched with the residual features of each candidate fault type in the fault mode library. The cosine value of the angle between the candidate fault type and the correction residual vector is calculated to generate the fault similarity. The fault similarity is normalized using the maximum-minimum normalization method to obtain the discrimination score. Based on the discrimination score, the candidate fault type with the highest score and the corresponding fault component are selected as the fault diagnosis result.

[0065] S4.2: Bind the fault diagnosis results to the correction residual vectors one by one and mark them with timestamps to generate correction residual time sequence records; cache and collect the correction residual vector records according to the preset time window to form a residual vector sequence.

[0066] The fault diagnosis results are associated with the correction residual vectors in chronological order, and the acquisition timestamp is marked on each associated item to form a correction residual time series record. The correction residual time series record is divided into multiple time segments according to a fixed time window length, such as 200ms. The multiple time segments are cached and aggregated one by one to generate a continuous residual vector sequence. The correction residual time series record and the residual vector sequence are read. At each time point, the absolute value of each component of the residual vector is read. The absolute values ​​of all components in the residual vector are averaged to obtain the residual amplitude at each time point. Then, the residual amplitudes at all time points are used as a continuous sequence to form a drift index. The drift index reflects the trend of the residual vector change over time.

[0067] Furthermore, the drift index does not directly classify arbitrary residual fluctuations as parameter drift. Instead, it makes a judgment based on the continuous cross-window change characteristics of the residual vector sequence and fault diagnosis results. Residual fluctuations caused by changes in operating conditions usually occur simultaneously with driving mode switching, control mode switching, and network status switching. After the operating conditions stabilize, the residual vector sequence can recover to the original fluctuation range. Such changes are judged as operating condition change disturbances. Residual fluctuations caused by noise disturbances usually exhibit discrete suddenness, short duration, low cross-window repeatability, and no stable change trend. Such changes are judged as noise disturbances. Residual changes caused by parameter drift usually exhibit continuous existence under the same operating conditions, slow accumulation across multiple consecutive time windows, and the formation of a stable offset trend. At the same time, the distribution of the corresponding fault diagnosis results undergoes continuous changes. Such changes are judged as parameter drift.

[0068] S5: When the drift index reaches the update decision threshold, generate the check parameter update package and apply the check parameter update package to update the closed-loop consistency constraint and observation enhancement strategy.

[0069] S5.1: Compare the drift index with the update decision threshold to obtain the update trigger decision result; when the update trigger decision result indicates that an update is needed, adjust the parameters of the closed-loop consistency constraint and the parameters of the observation enhancement strategy through the drift index, and encapsulate them into a check parameter update package.

[0070] Based on the statistical analysis and operational patterns of historical control link data, an update trigger judgment threshold is set, with a value range of 0.1 to 0.3. The value range of the update trigger judgment threshold is also set based on the distribution boundaries of drift indicators under normal and abnormal operating conditions. Using a value range of 0.1 to 0.3 effectively balances update frequency and system stability. Drift indicators are read one by one, and each drift indicator is compared with the update trigger judgment threshold. When the drift indicator is greater than the update trigger judgment threshold, it is determined that an update needs to be triggered at the corresponding time point; otherwise, it is determined that no update is needed, forming an update trigger judgment result. When the update trigger judgment result indicates that an update is needed, the parameters of the control timing constraints, state consistency constraints, and actuator consistency constraints of the closed-loop consistency constraint are incrementally adjusted using the drift indicator. Simultaneously, the sampling ratio of data segments and the caching strategy parameters in the observation enhancement strategy are adjusted based on the drift indicator, forming the adjusted parameters of the closed-loop consistency constraint and the observation enhancement strategy, and then encapsulated into a check parameter update package.

[0071] It should be noted that, in order to avoid false updates triggered by instantaneous drift fluctuations, a short delay and multi-point confirmation mechanism can be set when the update trigger determination result indicates that an update is required. That is, the encapsulation of the check parameter update package is only executed when it is determined that an update is required at several consecutive time points and no abnormal lock occurs. When the network connection is interrupted, the sensor is abnormal, or the vehicle is in a safety-sensitive control state, the update can be locked and the encapsulation of the check parameter update package can be suspended.

[0072] S5.2: Replace the parameters of the closed-loop consistency constraints and observation enhancement strategies by checking the parameter update package, and output the updated closed-loop consistency constraints and the updated observation enhancement strategies.

[0073] The system reads the parameters of the closed-loop consistency constraints and the observation enhancement strategy from the parameter update package. It then divides the closed-loop consistency constraint parameters into control timing constraint parameters, state consistency constraint parameters, and actuator consistency constraint parameters, and the observation enhancement strategy parameters into sampling scheduling parameters and fragment buffer parameters. Within the closed-loop consistency constraints, it locates the original positions of the control timing constraint parameters, state consistency constraint parameters, and actuator consistency constraint parameters, and writes these parameters from the parameter update package into their original positions, thus completing the parameter update for the closed-loop consistency constraints. Similarly, within the observation enhancement strategy, it locates the original positions of the sampling scheduling parameters and fragment buffer parameters, and writes these parameters from the parameter update package into their original positions, completing the parameter update for the observation enhancement strategy. Finally, it saves the updated closed-loop consistency constraints and the updated observation enhancement strategy for use in the consistency evaluation of control link data and the sampling scheduling of highly observable data fragments.

[0074] It should be noted that after updating the closed-loop consistency constraints and observation enhancement strategies using the application parameter update package, the updated closed-loop consistency constraints and observation enhancement strategies are used as temporary parameters. Consistency assessment and fault diagnosis are then performed again using the control link data of the next time window. If the residual vector formed by the temporary parameters and the fault diagnosis results do not show abnormal fluctuations, the updated closed-loop consistency constraints and observation enhancement strategies are retained; otherwise, the closed-loop consistency constraints and observation enhancement strategies before the update are restored.

[0075] This embodiment also provides a computer device applicable to the fault detection method for intelligent connected vehicles, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the fault detection method for intelligent connected vehicles as proposed in the above embodiment.

[0076] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0077] This embodiment also provides a storage medium storing a computer program, which, when executed by a processor, implements the intelligent connected vehicle fault detection method proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0078] In summary, this invention achieves time-series management of control signals and vehicle states by dividing control link data with a unified timestamp into multiple window segments according to a preset time window. This reduces data processing complexity and maintains time relevance. By performing consistency evaluation on the windowed data segments and generating residual vectors, effective features can be extracted from complex vehicle control data, enabling fault diagnosis to specifically identify potential anomalies and reduce false positives and false negatives. By caching and encapsulating highly observable data segments through segment caching rules, fault event packets are obtained, significantly improving the real-time performance and accuracy of fault detection. By applying check parameter update packets to update closed-loop consistency constraints and observation enhancement strategies, the reliability and continuity of fault diagnosis for intelligent connected vehicles are enhanced.

[0079] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A fault detection method for intelligent connected vehicles, characterized in that: include, Acquire control link data and network status information of intelligent connected vehicles, segment the control link data according to a preset time window, and generate windowed data fragments; Consistency assessment is performed on windowed data segments according to preset closed-loop consistency constraints, residual vectors are generated and mapped to a set of suspected faults, and observability indicators are calculated based on the set of suspected faults, residual vectors and network status information. Based on the observability index, an observation enhancement strategy is selected and configured. The control link data is sampled and segmented through the observation enhancement strategy to obtain fault event packets. The residual vector is corrected based on the fault event packets, and the corrected residual vector is output. Fault identification is performed on the correction residual vector, the fault diagnosis result is output, and the correction residual vector is cached according to a preset time window to form a residual vector sequence. The drift index is calculated based on the fault diagnosis result and the residual vector sequence. When the drift index reaches the update decision threshold, an inspection parameter update package is generated and applied to update the closed-loop consistency constraints and observation enhancement strategies.

2. The intelligent connected vehicle fault detection method according to claim 1, characterized in that: The specific steps for generating the windowed data fragment are as follows: Acquire control link data and network status information of intelligent connected vehicles, unify the timestamp of control link data, and divide the control link data after unifying the timestamp according to a preset time window to obtain multiple window segments; Multiple window segments are bound to network status information one by one to generate windowed data segments.

3. The intelligent connected vehicle fault detection method according to claim 2, characterized in that: The specific steps for generating the residual vector are as follows: Extract the control command sequence, feedback measurement sequence, state estimation sequence, and actuator constraint state from the windowed data fragments as closed-loop consistency evaluation data; Closed-loop consistency constraints are set based on vehicle control logic and actuator constraint logic. The closed-loop consistency constraints include control timing constraints, state consistency constraints, and actuator consistency constraints. The trigger time is extracted from the control command sequence, the arrival time is extracted from the feedback measurement sequence, the time difference between the trigger time and the arrival time is calculated, and the time difference is compared with the allowable delay range limited by the control timing constraints to generate the control timing consistency deviation. The state estimation sequence and the feedback measurement sequence are time-aligned and the difference is calculated to output the deviation sequence. The deviation sequence is then compared with the allowable deviation range defined by the state consistency constraint to generate the state consistency deviation. Extract the limiting interval and saturation interval from the actuator constraint state, extract the segments in the control command sequence and feedback measurement sequence that are located in the limiting interval and saturation interval, output the control command subsequence and feedback measurement subsequence, and perform difference calculation to obtain the actuator deviation sequence. Compare the actuator deviation sequence with the allowable deviation range defined by the actuator consistency constraint to generate the actuator consistency deviation. The control timing consistency deviation, state consistency deviation, and actuator consistency deviation are normalized and combined to generate a residual vector.

4. The intelligent connected vehicle fault detection method according to claim 3, characterized in that: The specific steps for calculating the observability index based on the suspected fault set, residual vector, and network status information are as follows. Perform component analysis on the residual vector to generate residual features; The residual features are matched with a pre-defined fault mode library to generate a set of suspected faults. Each candidate fault type in the suspected fault set is associated with a residual vector to generate a fault residual association table. Based on the network status information, the observability assessment rules are configured, and the observability assessment of the fault residual association table is performed through the observability assessment rules to generate observability indicators.

5. The intelligent connected vehicle fault detection method according to claim 4, characterized in that: The specific steps for obtaining the fault event packet are as follows: The suspected fault set is sorted by observability indicators to form a priority diagnosis order; The priority diagnosis order is hierarchically sampled and scheduled, and fragment caching rules are configured through network status information to form an observation enhancement strategy; By using observation enhancement strategies, control link data is sampled to generate highly observable data segments; Highly observable data fragments are cached and encapsulated according to fragment caching rules to obtain fault event packets.

6. The intelligent connected vehicle fault detection method according to claim 5, characterized in that: The output correction residual vector refers to performing a consistency assessment on the highly observable data segments in the fault event packet to obtain the correction deviation component, updating the residual vector through the correction deviation component, and outputting the correction residual vector.

7. The intelligent connected vehicle fault detection method according to claim 6, characterized in that: The specific steps for forming the residual vector sequence are as follows: The correction residual vector is matched with the candidate fault type to generate fault similarity. The fault similarity is then normalized, and the discrimination score is output. The fault type and faulty component are determined by the discrimination score, and the fault diagnosis result is output. The fault diagnosis results are bound one by one with the correction residual vector and timestamped to generate a correction residual time series record; The correction residual vector records are cached and collected according to the preset time window to form a residual vector sequence.

8. The intelligent connected vehicle fault detection method according to claim 7, characterized in that: The application checks the parameter update package to update the closed-loop consistency constraints and observation enhancement strategies. The specific steps are as follows: The drift index is compared with the update determination threshold to obtain the update trigger determination result; When the updated trigger judgment result representation needs to be updated, the parameters of the closed-loop consistency constraint and the parameters of the observation enhancement strategy are adjusted by the drift index and encapsulated into a check parameter update package. The parameters of the closed-loop consistency constraints and observation enhancement strategies are replaced by checking the parameter update package, and the updated closed-loop consistency constraints and observation enhancement strategies are output.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the intelligent connected vehicle fault detection method according to any one of claims 1 to 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the intelligent connected vehicle fault detection method according to any one of claims 1 to 8.

Images