Multi-path load intelligent diagnosis and protection method of vehicle body controller

Abstract

The invention discloses a multipath load intelligent diagnosis and protection method for a vehicle body controller, and relates to the technical field of vehicle control units. The method comprises the following steps: synchronously acquiring current and voltage signals of a target load and a bus voltage value of a vehicle-mounted power supply to obtain a current and voltage sequence and a bus voltage sequence; load cooperative working condition decoupling analysis is carried out on the current and voltage sequence to obtain a load feature vector; comparing the target load with a preset normal characteristic threshold value, and judging whether the target load has abnormal change or not; and analyzing the bus voltage sequence to obtain power feature vectors, inputting the two feature vectors into a load power coupling noise recognition model to obtain an association judgment result, and executing a protection action of a target load according to the association judgment result. According to the method, through correlation analysis of load and power supply characteristics, power supply coupling noise and load independent faults can be accurately distinguished, error protection caused by power supply fluctuation is avoided, corresponding protection actions can be matched for real faults, and the accuracy of vehicle-mounted load diagnosis is improved.

Description

Technical Field

[0001] This invention belongs to the field of vehicle controller technology, specifically relating to a method for intelligent diagnosis and protection of multi-channel loads in a vehicle body controller. Background Technology

[0002] As the complexity of vehicle electrical systems increases, the multi-load diagnostic and protection technology of body controllers has gradually evolved from single fault detection to multi-load collaborative monitoring. In the early stages, it focused on overcurrent and overvoltage protection of hardware circuits. Later, it incorporated software judgment logic based on voltage and current thresholds, realizing the identification of typical faults such as load open circuit and short circuit. In recent years, it has further combined simple timing feature analysis to try to adapt to dynamic scenarios such as vehicle start-stop and load switching. Overall, it is developing towards multi-parameter fusion and automated protection, and is gradually becoming one of the basic technologies to ensure the stable operation of vehicle electrical systems.

[0003] However, in the complex and dynamic real vehicle electrical environment, it is impossible to effectively distinguish between system transient noise and real load faults, which leads to high false alarm rates and low diagnostic accuracy. When the vehicle starts, the air conditioner or power steering is turned on and off under heavy loads, the drastic fluctuations in the power network are easily misjudged by the diagnostic system as open circuit or short circuit faults, resulting in frequent false faults and ineffective repairs. For gray faults such as increased contact resistance and slow degradation of load performance, where current changes are not significant, the accuracy of on-board load diagnosis is low. Summary of the Invention

[0004] The purpose of this invention is to solve the problem that transient noise and real load faults cannot be effectively distinguished in complex and dynamic real vehicle electrical environments, resulting in low accuracy of vehicle load diagnosis. Therefore, this invention proposes a multi-channel load intelligent diagnosis and protection method for vehicle body controllers.

[0005] This invention proposes a multi-channel load intelligent diagnosis and protection method for a vehicle body controller, the method comprising: The current and voltage signals of the target load, as well as the voltage value of the vehicle power bus, are acquired synchronously to obtain the current-voltage sequence and the bus voltage sequence; the target load is any one of the loads. The load characteristic vector is obtained by performing load coordination decoupling analysis on the current-voltage sequence; The target load feature vector is compared with a preset normal feature threshold to determine whether there is an abnormal change in the target load; If abnormal changes are found, the bus voltage sequence is analyzed in multiple dimensions to obtain a power supply feature vector; The load feature vector and the power supply feature vector are input into a preset load-power supply coupling noise identification model to obtain the correlation determination result; Based on the correlation determination result, a protection action is performed on the target load.

[0006] Optionally, the load feature vector obtained by performing load-coordinated operating condition decoupling analysis on the current-voltage sequence includes: The current-voltage sequence is clustered by load operating mode to obtain a sub-current-voltage sequence set consisting of sub-current-voltage sequences corresponding to each load-specific operating mode; Construct a modal dynamic constraint function for the target sub-current-voltage sequence; the target sub-current-voltage sequence is any one of the sub-current-voltage sequences in the set of sub-current-voltage sequences. Based on the modal dynamic constraint function, the target sub-current and voltage sequence is analyzed for time-series features and parameters are extracted to obtain the modal constraint feature sequence. Cross-modal correlation features are obtained by calculating the trend matching degree of the modal constraint feature sequences corresponding to different load-specific operating modes; The load feature vector is obtained by weighting and fusing the modal constraint feature sequence and the cross-modal correlation feature based on the proportion of the sub-current and voltage sequence corresponding to each load-specific operating mode in the current and voltage sequence.

[0007] Optionally, the modal dynamic constraint functions for constructing the target sub-current and voltage sequences include: The modal dynamic constraint function is: ;in, Let t be the instantaneous rate of change of current with respect to voltage. Let be the voltage value at time t. This is the reference voltage for the specific operating mode of this load.

[0008] Optionally, the modal constraint feature sequence obtained by performing time-series feature analysis and parameter extraction on the target sub-current-voltage sequence based on the modal dynamic constraint function includes: Substituting the time-series parameters of the target sub-current and voltage sequence into the modal dynamic constraint function yields the constraint function value sequence corresponding to the target sub-current and voltage sequence; The constraint function value sequence is time-series segmented to obtain a constraint function segmented sequence set; The fluctuation extreme value characteristics are obtained by traversing the maximum and minimum values ​​of the segmented sequence of the target constraint function; the segmented sequence of the target constraint function is any one of the segmented sequences of the constraint function in the set of segmented sequences of the constraint function. Calculate the difference sequence of adjacent function values ​​in the segmented sequence of the objective constraint function, and obtain the time series stationarity characteristics by statistically analyzing the proportion of fluctuation intervals in the difference sequence; Linear fitting is performed on the piecewise sequence of the objective constraint function to obtain trend change characteristics based on the change of the fitting slope; The characteristic parameters corresponding to the segmented sequence of the target constraint function are obtained by integrating the fluctuation extreme value characteristics, time series stationarity characteristics and trend change characteristics. The modal constraint feature sequence is obtained by arranging the feature parameters of all constraint function segment sequences in temporal order.

[0009] Optionally, cross-modal correlation features are obtained by calculating the trend matching degree of the modal constraint feature sequences corresponding to different load-specific operating modes, including: Trend fitting is performed on the modal constraint feature sequences corresponding to each load-specific operating mode to obtain a set of trend fitting curves composed of the trend fitting curves corresponding to each modal constraint feature sequence; Calculate the trend similarity between any two trend fitting curves in the trend fitting curve set to obtain a similarity coefficient set; A similarity matrix is ​​constructed based on the set of similarity coefficients; the matrix elements of the similarity matrix are the similarity coefficients corresponding to the two load-specific operating modes. The similarity matrix is ​​subjected to feature dimensionality reduction processing to obtain cross-modal association features.

[0010] Optionally, performing multi-dimensional feature analysis on the bus voltage sequence to obtain the power supply feature vector includes: Perform transient and steady-state timing decoupling on the bus voltage sequence to obtain transient voltage subsequences and steady-state voltage subsequences; The transient voltage subsequence is subjected to rate of change and convergence trend features to obtain the transient voltage dynamic features; The steady-state voltage subsequence is traversed through a preset sliding window, and the average voltage within each window is calculated and used as the voltage baseline for that window to obtain the window baseline sequence. The difference between adjacent voltage baselines in the window baseline sequence is calculated to obtain the baseline deviation sequence; The steady-state voltage dynamic characteristics are obtained by extracting the fluctuation amplitude characteristics and overall offset trend characteristics of the baseline deviation sequence; The temporal coupling characteristics are obtained by calculating the correlation between the transient voltage subsequence and the steady-state voltage subsequence in the time dimension; The transient voltage dynamic characteristics, steady-state voltage dynamic characteristics, and time-domain coupling characteristics are fused to obtain the power supply feature vector.

[0011] Optionally, performing protection actions on the target load based on the association determination result includes: If the correlation determination result is a correlation, the abnormal change is determined to be coupling noise caused by power fluctuation, and a noise suppression command is generated to terminate the fault determination of the target load. If the correlation determination result is non-correlated, the abnormal change is determined to be an independent fault of the target load or its power supply circuit, and fault handling and load protection actions are performed according to the fault type matched by the load feature vector.

[0012] The beneficial effects of this invention are as follows: This invention proposes a multi-channel load intelligent diagnosis and protection method for vehicle body controllers. It obtains corresponding sequence data by synchronously acquiring the current and voltage signals of the target load and the vehicle power bus voltage. Then, it performs load-coordinated operating condition decoupling analysis on the current and voltage sequences to generate load feature vectors, which are then compared with normal thresholds to determine if there are any anomalies in the load. Once an anomaly is detected, the bus voltage sequence is parsed to obtain power feature vectors. These two types of feature vectors are input into a load-power coupling noise identification model to obtain a correlation judgment result. Finally, protection actions are executed on the target load based on this result. This method, through correlation analysis of load and power characteristics, can not only accurately distinguish between power coupling noise and independent load faults, avoiding false protection caused by power fluctuations, but also match corresponding protection measures to actual faults, effectively improving the accuracy of vehicle load diagnosis. Attached Figure Description

[0013] The present invention will now be further described with reference to the accompanying drawings.

[0014] Figure 1 A flowchart of a multi-channel load intelligent diagnosis and protection method for a vehicle body controller provided in an embodiment of the present invention; Figure 2 A flowchart illustrating the load feature vector generation process provided in an embodiment of the present invention. Detailed Implementation

[0015] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0016] 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.

[0017] This invention provides a method for intelligent diagnosis and protection of multiple loads in a vehicle body controller. See also... Figure 1 , Figure 1 A flowchart illustrating a multi-channel load intelligent diagnosis and protection method for a vehicle body controller, provided in an embodiment of the present invention. The method includes the following steps: S101, synchronously acquire the current and voltage signals of the target load and the voltage value of the vehicle power bus to obtain the current-voltage sequence and the bus voltage sequence; S102, Perform load-coordinated decoupling analysis on the current-voltage sequence to obtain the load feature vector; S103, compare the target load feature vector with the preset normal feature threshold to determine whether there is an abnormal change in the target load; S104, if there are abnormal changes, perform multi-dimensional feature analysis on the bus voltage sequence to obtain the power supply feature vector; S105, Input the load feature vector and power supply feature vector into the preset load power supply coupling noise identification model to obtain the correlation determination result; S106, Perform protection actions on the target load based on the correlation determination result; The target load is any one of the loads.

[0018] This invention provides a method for intelligent diagnosis and protection of multi-channel loads in a vehicle body controller. The method first synchronously acquires the current and voltage signals of the target load and the vehicle power bus voltage to form a corresponding sequence. Then, it performs load-coordinated operating condition decoupling analysis on the current and voltage sequences to generate load feature vectors, which are compared with normal thresholds to determine if the load is abnormal. If an abnormality is detected, the bus voltage sequence is parsed to obtain power feature vectors. These two types of feature vectors are input into a load power coupling noise identification model to obtain a correlation judgment result. Finally, load protection actions are executed based on this result. This process, through the correlation analysis of load and power characteristics, can accurately distinguish between power coupling noise and independent load faults, avoiding false protection caused by power fluctuations, and can match corresponding protection measures to actual faults, effectively improving the accuracy of vehicle load diagnosis.

[0019] In one implementation, the preset normal characteristic threshold is obtained through statistical analysis of sample data of the target load under normal operating conditions: First, the current and voltage sequences of the load under various typical normal operating scenarios are collected, and a large number of normal load feature vectors are generated through load collaborative operating condition decoupling analysis; then, the parameters of each dimension of these feature vectors are statistically analyzed, and their mean and standard deviation are calculated. Usually, the mean ± 2 to 3 times the standard deviation is taken as the normal characteristic threshold range of the corresponding dimension; for example, the current characteristic threshold of vehicle lighting load may be 0.5A ± 0.1A, and the voltage characteristic threshold may be 12V ± 0.5V; while the current characteristic threshold of vehicle motor load may be 5A ± 1A, and the voltage characteristic threshold may be 12V ± 0.8V.

[0020] In one implementation, the pre-defined load-power supply coupling noise identification model processes as follows: After receiving the load feature vector and the power supply feature vector, firstly, using the time sequence labels of each modal constraint feature sequence in the load feature vector as a reference, the transient voltage dynamic features and steady-state voltage dynamic features in the power supply feature vector are aligned in time sequence to generate a modal-power supply time sequence correlation matrix; then, the power supply feature segments corresponding to each load-specific operating mode in the matrix are extracted and matched with the dynamic constraint function threshold of that mode, and the time difference and trend similarity between load modal abnormal fluctuations and power supply feature abnormalities are calculated to obtain the constraint matching coefficient; subsequently, the load is determined based on the cross-modal correlation features in the load feature vector through a dynamic weight allocation network. The system adaptively adjusts the weights of two types of features based on their type and operating condition. Next, it calculates the coupling degree using a coupling degree evaluation function that takes constraint matching coefficients, temporal alignment similarity, and feature mapping weight matching degree as variables, and then weights the results with dynamically adjusted coefficients. Simultaneously, it references a load fault probability library and estimates the probability of independent faults based on fault association parameters in the load feature vector. Finally, it performs a dual-threshold decision based on preset coupling degree and fault probability thresholds. If the coupling degree meets the threshold but the fault probability does not, the fault is determined to be coupling noise caused by power fluctuations. If the coupling degree does not meet the threshold but the fault probability does, the fault is determined to be an independent fault of the load or its circuit. In critical cases, the system backtracks to the original data and recalculates the relevant features to ensure accurate judgment results.

[0021] In one embodiment, see Figure 2 , Figure 2 This is a flowchart of the load feature vector generation process provided in an embodiment of the present invention. The load feature vector is obtained by performing load-coordinated operating condition decoupling analysis on the current-voltage sequence, including: S201, Perform load operation mode clustering on the current and voltage sequences to obtain a set of sub-current and voltage sequences consisting of sub-current and voltage sequences corresponding to each load-specific operation mode; S202, construct modal dynamic constraint functions for the target sub-current-voltage sequence; S203, based on the modal dynamic constraint function, perform time-series feature analysis and parameter extraction on the target sub-current and voltage sequence to obtain the modal constraint feature sequence; S204, calculate the trend matching degree of the modal constraint feature sequence corresponding to different load-specific operating modes to obtain cross-modal correlation features; S205, based on the proportion of the sub-current-voltage sequence corresponding to each load-specific operating mode in the current-voltage sequence, the mode constraint feature sequence and cross-mode correlation feature are weighted and fused to obtain the load feature vector; The target sub-current voltage sequence is any one of the sub-current voltage sequences in the set of sub-current voltage sequences.

[0022] In one implementation, the load operation mode clustering process is as follows: First, the collected current and voltage sequences are encoded in both the time domain and the feature domain, mapping the current and voltage values ​​of each sampling point and the changing trends of the preceding sampling points of a preset length into a high-dimensional feature vector. Then, prior constraints on load conditions are introduced, and the initial distribution range of cluster centers is preset based on the typical operating scenarios of vehicle loads. Next, a constrained density clustering algorithm is used, taking the current-voltage high-dimensional feature vector as input and combining it with the preset operating condition center range, to aggregate sampling points whose feature vector similarity meets the threshold and falls within the same operating condition range into one class. Finally, the continuity of the time series segments corresponding to each class is checked, and after removing discrete and isolated segments, the continuous time series segments are merged into subsequences, with each subsequence corresponding to a load-specific operating mode.

[0023] In one embodiment, constructing a modal dynamic constraint function for the target sub-current-voltage sequence includes: Modal dynamic constraint function is ;in, Let t be the instantaneous rate of change of current with respect to voltage. Let be the voltage value at time t. This is the reference voltage for the specific operating mode of this load.

[0024] In one implementation, the modal dynamic constraint function is constructed by combining the instantaneous rate of change of current with voltage with the degree of voltage deviation relative to a reference value. The core logic lies in accurately adapting to the current-voltage coupling characteristics of vehicle load operation. The instantaneous rate of change term describes the real-time correlation between current and voltage dynamic response during load operation, while the voltage deviation term applies a reference voltage constraint to this response relationship. This avoids the problem that a single rate of change cannot distinguish between normal dynamic response and fault-type deviation, and can specifically capture the dynamic matching law of current and voltage under the specific operating mode of the load. The uniqueness of this design lies in the fact that it is not a general current-voltage feature function, but a technical solution that fully fits the premise of the specific dynamic correlation between current and voltage under different modes after load operating mode clustering. It can directly provide feature dimensions that reflect dynamic response and anchor modal reference for subsequent modal constraint feature extraction, solving the pain point that traditional feature functions cannot simultaneously adapt to intramodal dynamics and intermodal distinguishability. It is the exclusive core link connecting modal clustering and feature extraction in this technical solution.

[0025] In one embodiment, the modal constraint feature sequence obtained by performing time-series feature analysis and parameter extraction on the target sub-current-voltage sequence based on the modal dynamic constraint function includes: Substituting the time-series parameters of the target sub-current and voltage sequence into the modal dynamic constraint function yields the constraint function value sequence corresponding to the target sub-current and voltage sequence; The constraint function value sequence is segmented over time to obtain a set of constraint function segment sequences; The fluctuation extreme value characteristics are obtained by traversing the maximum and minimum values ​​of the piecewise sequence of the objective constraint function; the piecewise sequence of the objective constraint function is any one of the constraint function piecewise sequences in the set of constraint function piecewise sequences. Calculate the difference sequence of adjacent function values ​​in the segmented sequence of the objective constraint function, and obtain the time series stationarity characteristics by statistically analyzing the proportion of fluctuation intervals in the difference sequence; Linear fitting is performed on the piecewise sequence of the objective constraint function to obtain trend change characteristics based on the change of the fitting slope; The characteristic parameters corresponding to the piecewise sequence of the objective constraint function are obtained by integrating the characteristics of fluctuation extreme values, time series stationarity, and trend change. The modal constraint feature sequence is obtained by arranging the feature parameters of all constraint function segment sequences in temporal order.

[0026] In one implementation, the timing segmentation uses a fixed time interval as the segmentation interval, which is set by the technicians.

[0027] In one implementation, the linear fitting process involves segmenting the target constraint function into a series. First, the constraint function values ​​in the series are selected as the dependent variable, and the corresponding time-series sampling points are selected as the independent variables. The least squares method is used to perform linear fitting on the segmented series to obtain the slope of the fitted line. Then, the segmented series is divided into several continuous sub-segments, and the above linear fitting operation is repeated for each sub-segment to obtain the fitting slope corresponding to each sub-segment. Finally, the magnitude and direction of the changes in the fitting slopes of these sub-segments are statistically analyzed to determine the trend characteristics of the segmented sequence of the target constraint function. For example, a continuously increasing slope corresponds to an accelerating upward trend in the constraint function value, while frequent slope fluctuations correspond to poor trend stability of the constraint function value.

[0028] In one implementation, the process first divides the continuous sequence of constraint function values ​​into several local segments through time-series segmentation. This adapts to the intermittent dynamic characteristics of vehicle load operation and avoids information confusion in long-series feature extraction. Then, for each segment sequence, fluctuation extremes, time-series stationarity, and trend change features are extracted to characterize the numerical fluctuation range, stability, and trend of the segment, achieving a multi-dimensional characterization of the local dynamics of the constraint function. Finally, the features are integrated and arranged in time sequence to restore the complete dynamic law of the constraint function under different load operation modes. This process refines the local features of the constraint function while preserving the global correlation in the time dimension, solving the problem that traditional single feature extraction cannot take into account both local dynamic details and global time-series laws. Based on the premise design of load operation mode clustering and the exclusive constraint function corresponding to each mode, the segmented-feature extraction method accurately captures the exclusive dynamic features of the constraint function under different load modes, providing a feature foundation for subsequent cross-modal correlation analysis that distinguishes modes while preserving dynamic details.

[0029] In one embodiment, calculating the trend matching degree of the modal constraint feature sequences corresponding to different load-specific operating modes to obtain cross-modal correlation features includes: Trend fitting is performed on the modal constraint feature sequences corresponding to each load-specific operating mode to obtain a set of trend fitting curves composed of the trend fitting curves corresponding to each modal constraint feature sequence; Calculate the trend similarity between any two trend fitting curves in the trend fitting curve set to obtain the similarity coefficient set; A similarity matrix is ​​constructed based on the set of similarity coefficients; the elements of the similarity matrix are the similarity coefficients of the two load-specific operating modes. Cross-modal association features are obtained by performing feature dimensionality reduction on the similarity matrix.

[0030] In one implementation, the trend fitting process involves taking the modal constraint feature sequence corresponding to each load-specific operating mode, using the feature parameter values ​​in the sequence as dependent variables and the corresponding time-series sampling points as independent variables, and performing trend fitting on the sequence using a multinomial fitting or exponential fitting method to obtain a fitting curve that can characterize the overall change trend of the sequence; after all modal constraint feature sequences corresponding to all load-specific operating modes have been fitted, the fitting curves corresponding to each sequence are summarized and integrated to obtain a trend fitting curve set composed of these trend fitting curves.

[0031] In one implementation, the trend similarity calculation process involves traversing any two trend fitting curves in the trend fitting curve set, first extracting segments of the two curves within the same time interval, calculating the mean absolute error of the function values ​​at corresponding time points of the two curves within that interval, simultaneously extracting trend parameters such as the fitting slope and curvature of the two curves and calculating the parameter difference, then weighting and fusing the mean error of the function values ​​and the parameter difference according to a preset weight to obtain the trend similarity coefficients corresponding to these two curves; after all pairwise combinations of curves in the curve set have completed the similarity calculation, these coefficients are summarized to obtain the trend similarity coefficient set.

[0032] In one implementation, the feature dimensionality reduction process involves first determining the target dimension for feature dimensionality reduction, then using principal component analysis (PCA) to process the similarity matrix, calculating the eigenvalues ​​and eigenvectors of the matrix, and selecting the top few principal components whose cumulative variance contribution reaches a preset threshold. Subsequently, the similarity matrix is ​​projected onto the feature space corresponding to the selected principal components to obtain a low-dimensional matrix projection result. Finally, the projection result is converted into a one-dimensional feature vector according to a preset rule, thus obtaining cross-modal correlation features that can characterize the correlation between different load-specific operating modes.

[0033] In one implementation, discrete modal constraint feature sequences are first transformed into continuous curves through trend fitting, eliminating the interference of time-series fluctuations on trend judgment. Then, the correlation degree of different load modal features is quantified by calculating pairwise curve similarity. Subsequently, a similarity matrix is ​​constructed and its dimensionality is reduced, preserving the correlation between modes while compressing the high-dimensional matrix into feature vectors that can be directly fused. This process perfectly aligns with the premise that load-specific operating modes have differentiated dynamic characteristics. By matching at the trend level rather than comparing single numerical values, it accurately captures the correlation between different load modes in their dynamic change patterns. It is suitable for the multimodal and dynamic operating characteristics of vehicle loads and provides cross-modal correlation dimension support for the subsequent fusion of load feature vectors. This solves the pain point of traditional feature extraction, which only focuses on a single mode and cannot reflect the correlation between loads. It is a dedicated technical path connecting multi-load modal features with global load correlation analysis.

[0034] In one embodiment, obtaining a power supply feature vector by performing multi-dimensional feature analysis on the bus voltage sequence includes: Perform transient and steady-state timing decoupling on the bus voltage sequence to obtain transient voltage subsequences and steady-state voltage subsequences; The dynamic characteristics of transient voltage are obtained by extracting the rate of change and convergence trend characteristics of the transient voltage subsequence. By traversing the steady-state voltage subsequence through a preset sliding window, the average voltage within each window is calculated and used as the voltage baseline for that window to obtain the window baseline sequence. The baseline deviation sequence is obtained by calculating the difference between adjacent voltage baselines in the window baseline sequence; The steady-state voltage dynamic characteristics are obtained by extracting the fluctuation amplitude characteristics and overall offset trend characteristics of the baseline deviation sequence; The temporal coupling characteristics are obtained by calculating the correlation between transient voltage subsequences and steady-state voltage subsequences in the time dimension; The power supply feature vector is obtained by fusing transient voltage dynamic characteristics, steady-state voltage dynamic characteristics, and time-domain coupling characteristics.

[0035] In one implementation, the length of the preset sliding window is set by the technician and can be 0.1-0.3 seconds.

[0036] In one implementation, the transient and steady-state timing decoupling process involves first selecting a preset filtering threshold, then using a high-pass filter to process the bus voltage sequence, retaining the high-frequency fluctuations in the sequence whose rate of change exceeds the threshold, and using this part of the data as the transient voltage subsequence; simultaneously, using a low-pass filter to process the same bus voltage sequence, retaining the smooth changes in the sequence whose rate of change is below the threshold, and using this part of the data as the steady-state voltage subsequence, thereby completing the transient and steady-state timing decoupling of the bus voltage sequence.

[0037] In one implementation, the extraction process of transient voltage dynamic features involves first traversing the transient voltage subsequence, calculating the ratio of the voltage difference between adjacent sampling points to the sampling time interval, and statistically analyzing the maximum, minimum, and average values ​​of these ratios to obtain the rate of change feature; then selecting continuously fluctuating segments in the subsequence, tracking the process of the voltage value in each segment approaching the center of fluctuation, calculating the convergence speed of the voltage value in the segment and the reduction in the fluctuation range to obtain the convergence trend feature; finally, integrating the rate of change feature and the convergence trend feature to obtain the transient voltage dynamic feature.

[0038] In one implementation, the extraction process of steady-state voltage dynamic characteristics involves first reading all deviation values ​​in the baseline deviation sequence one by one, recording the maximum and minimum values, and calculating the absolute difference between them. This difference is used as the fluctuation amplitude feature of the sequence to characterize the magnitude of the baseline deviation fluctuation within the time range. Next, the baseline deviation sequence is divided into multiple continuous non-overlapping segments according to a fixed time length. The arithmetic mean of all deviation values ​​in each segment is calculated, and these means are arranged in chronological order of the segments. The numerical trend of the mean sequence is observed—if the mean continues to increase, it is determined to be a positive offset trend; if the mean continues to decrease, it is determined to be a negative offset trend. At the same time, the mean of the absolute value of the difference between the means of adjacent segments is calculated to characterize the rate of change of the offset trend. These trends and rate information together constitute the overall offset trend feature. Finally, the obtained fluctuation amplitude feature and the overall offset trend feature are combined into a feature set, which yields the steady-state voltage dynamic characteristics.

[0039] In one embodiment, protecting the target load based on the association determination result includes: If the correlation determination result is correlation, the abnormal change is determined to be coupled noise caused by power fluctuation, and a noise suppression command is generated to terminate the fault determination of the target load. If the correlation determination result is non-correlated, the abnormal change will be determined as an independent fault of the target load or its power supply circuit, and fault handling and load protection actions will be performed according to the fault type matched by the load feature vector.

[0040] In one implementation, when the correlation determination result is positive, the correlation analysis record of the load feature vector and the power supply feature vector corresponding to the target load is first retrieved to confirm that the matching degree of their dynamic change trends reaches a preset correlation threshold. Based on this, the abnormal change of the target load is determined to be coupled noise formed by power fluctuations transmitted to the load circuit. Subsequently, a noise suppression command is sent to the fault determination module of the body controller. This command triggers the module to pause the fault detection process for the current target load, and records this anomaly as a coupled noise event and marks it as non-fault data, thereby terminating the fault determination for the target load. When the correlation determination result is negative, the target load is first checked... The dynamic change trend matching degree between the load characteristic vector and the power supply characteristic vector of the target load is confirmed to be below the preset correlation threshold. Based on this, the abnormal change of the target load is judged as an independent fault of the target load itself or its power supply circuit. Subsequently, the preset fault type matching library is retrieved, and the characteristic parameters such as the fluctuation extreme value and timing stability corresponding to the load characteristic vector are compared with the characteristic templates of the fault types in the library to determine the matching fault type. Finally, the corresponding fault handling process is triggered according to the matching result, such as initiating a load circuit power-off command for short circuit faults, sending a circuit detection signal and recording fault information for poor contact faults, and completing the corresponding load protection action.

[0041] The foregoing has described one embodiment of the present invention in detail, but this content is merely a preferred embodiment and should not be considered as limiting the scope of the present invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the scope of the claims of this invention.

Claims

1. A method for intelligent diagnosis and protection of multi-channel loads in a vehicle body controller, characterized in that, The method includes: The current and voltage signals of the target load and the voltage value of the vehicle power bus are acquired simultaneously to obtain the current-voltage sequence and the bus voltage sequence; the target load is any one of the loads. The load characteristic vector is obtained by performing load coordination condition decoupling analysis on the current-voltage sequence; The target load feature vector is compared with a preset normal feature threshold to determine whether there is an abnormal change in the target load; If abnormal changes are found, the bus voltage sequence is analyzed in multiple dimensions to obtain a power supply feature vector; The load feature vector and the power supply feature vector are input into a preset load-power supply coupling noise identification model to obtain the correlation determination result; Based on the correlation determination result, a protection action is performed on the target load.

2. The method for intelligent diagnosis and protection of multi-channel loads in a vehicle body controller according to claim 1, characterized in that, The load feature vector obtained by performing load-coordinated operating condition decoupling analysis on the current-voltage sequence includes: The current-voltage sequence is clustered by load operating mode to obtain a sub-current-voltage sequence set consisting of sub-current-voltage sequences corresponding to each load-specific operating mode; Construct modal dynamic constraint functions for the target sub-current-voltage sequence; the target sub-current-voltage sequence is any one of the sub-current-voltage sequences in the set of sub-current-voltage sequences. Based on the modal dynamic constraint function, the target sub-current and voltage sequence is analyzed for time-series features and parameters are extracted to obtain the modal constraint feature sequence. Cross-modal correlation features are obtained by calculating the trend matching degree of the modal constraint feature sequences corresponding to different load-specific operating modes; The load feature vector is obtained by weighting and fusing the modal constraint feature sequence and the cross-modal correlation feature based on the proportion of the sub-current and voltage sequence corresponding to each load-specific operating mode in the current and voltage sequence.

3. The method for intelligent diagnosis and protection of multi-channel loads in a vehicle body controller according to claim 2, characterized in that, The modal dynamic constraint functions for constructing the target sub-current and voltage sequences include: The modal dynamic constraint function is: ;in, Let t be the instantaneous rate of change of current with respect to voltage. Let be the voltage value at time t. This is the reference voltage for the specific operating mode of this load.

4. The method for intelligent diagnosis and protection of multi-channel loads in a vehicle body controller according to claim 2, characterized in that, Based on the modal dynamic constraint function, the target sub-current-voltage sequence is analyzed for time-series features and parameters are extracted to obtain the modal constraint feature sequence, which includes: Substituting the time-series parameters of the target sub-current and voltage sequence into the modal dynamic constraint function yields the constraint function value sequence corresponding to the target sub-current and voltage sequence; The constraint function value sequence is time-series segmented to obtain a constraint function segmented sequence set; The fluctuation extreme value characteristics are obtained by traversing the maximum and minimum values ​​of the segmented sequence of the target constraint function; the segmented sequence of the target constraint function is any one of the segmented sequences of the constraint function in the set of segmented sequences of the constraint function. Calculate the difference sequence of adjacent function values ​​in the segmented sequence of the objective constraint function, and obtain the time series stationarity characteristics by statistically analyzing the proportion of fluctuation intervals in the difference sequence; Linear fitting is performed on the piecewise sequence of the objective constraint function to obtain trend change characteristics based on the change of the fitting slope; The characteristic parameters corresponding to the segmented sequence of the target constraint function are obtained by integrating the fluctuation extreme value characteristics, time series stationarity characteristics and trend change characteristics. The modal constraint feature sequence is obtained by arranging the feature parameters of all constraint function segment sequences in temporal order.

5. A method for intelligent diagnosis and protection of multi-channel loads in a vehicle body controller according to claim 2, characterized in that, The cross-modal correlation features are obtained by calculating the trend matching degree of the modal constraint feature sequences corresponding to different load-specific operating modes, including: Trend fitting is performed on the modal constraint feature sequences corresponding to each load-specific operating mode to obtain a set of trend fitting curves composed of the trend fitting curves corresponding to each modal constraint feature sequence; Calculate the trend similarity between any two trend fitting curves in the trend fitting curve set to obtain a similarity coefficient set; A similarity matrix is ​​constructed based on the set of similarity coefficients; the matrix elements of the similarity matrix are the similarity coefficients corresponding to the two load-specific operating modes. The similarity matrix is ​​subjected to feature dimensionality reduction processing to obtain cross-modal association features.

6. The method for intelligent diagnosis and protection of multi-channel loads in a vehicle body controller according to claim 1, characterized in that, The power supply feature vector obtained by performing multi-dimensional feature analysis on the bus voltage sequence includes: Perform transient and steady-state timing decoupling on the bus voltage sequence to obtain transient voltage subsequences and steady-state voltage subsequences; The transient voltage subsequence is subjected to rate of change and convergence trend features to obtain the transient voltage dynamic features; The steady-state voltage subsequence is traversed through a preset sliding window, and the average voltage within each window is calculated and used as the voltage baseline for that window to obtain the window baseline sequence. The difference between adjacent voltage baselines in the window baseline sequence is calculated to obtain the baseline deviation sequence; The steady-state voltage dynamic characteristics are obtained by extracting the fluctuation amplitude characteristics and overall offset trend characteristics of the baseline deviation sequence; The temporal coupling characteristics are obtained by calculating the correlation between the transient voltage subsequence and the steady-state voltage subsequence in the time dimension; The transient voltage dynamic characteristics, steady-state voltage dynamic characteristics, and time-domain coupling characteristics are fused to obtain the power supply feature vector.

7. The method for intelligent diagnosis and protection of multi-channel loads in a vehicle body controller according to claim 1, characterized in that, The protection action for the target load based on the association determination result includes: If the correlation determination result is a correlation, the abnormal change is determined to be coupling noise caused by power fluctuation, and a noise suppression command is generated to terminate the fault determination of the target load. If the correlation determination result is non-correlated, the abnormal change is determined to be an independent fault of the target load or its power supply circuit, and fault handling and load protection actions are performed according to the fault type matched by the load feature vector.

Images