A method for identifying the cause of electric vehicle speed loss control failure

Abstract

The present invention relates to the technical field of electric vehicle fault analysis, and discloses a method for identifying the cause of an electric vehicle speed runaway fault, comprising the following steps: S1, collecting and parsing message data during vehicle operation, extracting signal parameters therefrom, and obtaining vehicle operation data T after preprocessing; S2, segmenting T, and establishing a training set and a test set using the accelerator pedal opening value in T as a feature and the torque value in T as a target; S3, establishing a regression model based on and, and predicting the target accelerator pedal opening value based on the regression model, and obtaining a predicted target torque value; S4, calculating an offset from the actual torque value; S5, confirming high-risk points based on the offset, and outputting a multidimensional feature map; S6, performing fault determination and cause analysis based on the multidimensional feature map. The present invention can accurately identify and locate the cause of a speed runaway fault, and the identification is accurate and the analysis is reliable.

Description

Technical Field

[0001] The present invention relates to the technical field of electric vehicle fault analysis, and in particular to a method for identifying the cause of an electric vehicle speed out-of-control fault. Background Art

[0002] Since speed loss of control may involve multiple issues such as battery BMS misjudgment, motor controller IGBT junction temperature protection failure, or autonomous driving perception fusion algorithm defects, the causes of accidents are complex, sudden loss of control accidents are highly sporadic, and effective data on accident cases is relatively scarce. Therefore, there is currently no clear research on the identification and location methods of speed loss of control faults. At present, there is still an urgent need for an effective method to identify the causes of electric vehicle speed loss of control faults to respond to the latest development trends of the new energy vehicle industry and promote the reduction of new energy vehicle driving risks. Summary of the Invention

[0003] The present invention aims to provide a method for identifying the cause of a speed out-of-control fault in an electric vehicle, which can accurately identify and locate the cause of the speed out-of-control fault, and has accurate identification and reliable analysis.

[0004] The basic solution provided by the present invention is: a method for identifying the cause of an electric vehicle speed out-of-control fault, comprising the following steps:

[0005] S1, collects and analyzes the message data during vehicle operation, extracts signal parameters from it, and obtains vehicle operation data T after preprocessing;

[0006] S2, the vehicle operation data T is divided and the accelerator pedal opening value in the vehicle operation data T is used As a feature, the torque value in the vehicle operation data T As a goal, establish training sets and test sets;

[0007] S3, built on and The target accelerator pedal opening value is predicted based on the regression model , and get the predicted target torque value ;

[0008] S4, calculate the target torque value and the actual torque value Offset ;

[0009] S5, according to the offset , confirm high-risk points and output multi-dimensional feature maps;

[0010] S6, perform fault determination and cause analysis based on the multi-dimensional feature graph.

[0011] The working principle and advantages of the present invention are:

[0012] The present invention provides a method for identifying the cause of electric vehicle speed loss control failures. This method constructs an anomaly identification and location method in a data-driven manner. It can quickly locate high-risk locations in complex and redundant vehicle operation history data and automatically output a multidimensional feature map. Furthermore, combined with image feature analysis and verification, this method implements knowledge-driven identification and determination of the cause of electric vehicle speed loss control failures. The method achieves accurate identification and reliable analysis, helping to promote the paradigm shift of new energy vehicle safety engineering from "post-event disposal" to "pre-event prevention."

[0013] The key point is this: This solution specifically establishes a dynamic regression model based on accelerator pedal opening and torque output. Accelerator pedal opening directly reflects the driver's power demand command, and torque is the core parameter of the motor's power output. The two have a strong physical correlation (driver intention → control signal → power output). By establishing a regression model for the two, the expected response relationship under normal operating conditions can be quantified, and the core contradictions in human-vehicle interaction can be effectively captured. Furthermore, by collaboratively analyzing driving intention (corresponding to accelerator pedal opening) and execution results (corresponding to torque output), and capturing risk points, it is possible to accurately locate faults and intuitively present fault clustering characteristics with multi-dimensional feature maps. This analysis method has a strong feature-target correlation and is based on a data-driven approach. It avoids the limitations of traditional FMEA methods that require pre-defined failure modes. It is particularly suitable for detecting occasional anomalies caused by the complex logic of smart car software and can more comprehensively detect faults.

[0014] Specifically, this solution addresses the complex, sporadic, and difficult-to-identify fault causes by decoupling complex control strategies (such as regenerative braking) from driver input. This data-driven approach establishes a dynamic baseline model, enhancing anomaly detection sensitivity and enabling the introduction of a mechanistic model to trace the fault's cause. During data preprocessing, this solution eliminates the vulnerability of traditional threshold methods to control strategy interference by eliminating torque fluctuations caused by regenerative braking / single-pedal operation when the accelerator is released. Retaining only data from the effective accelerator pedal movement interval ensures the model reflects the true relationship between acceleration demand and power response. Furthermore, the use of linear regression rather than fixed thresholds allows for adaptive adaptation to the dynamic characteristics of different vehicle models (e.g., torque response differences between high-performance and family vehicles). By minimizing prediction error and dynamically fitting optimal parameters, this approach overcomes the limitations of traditional static threshold methods. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 This is a schematic diagram of the method flow structure of a first embodiment of a method for identifying the cause of an electric vehicle speed out-of-control fault according to the present invention;

[0016] Figure 2This is a schematic diagram of the visualization effect of a regression model of a first embodiment of a method for identifying causes of an electric vehicle speed out-of-control fault according to the present invention;

[0017] Figure 3 This is an example diagram of the first characteristic diagram of Embodiment 1 of a method for identifying the cause of a speed out-of-control fault of an electric vehicle according to the present invention;

[0018] Figure 4 This is an example diagram of the second characteristic diagram of Embodiment 1 of a method for identifying the cause of an electric vehicle speed out-of-control fault according to the present invention;

[0019] Figure 5 This is a signal analysis schematic diagram of a first embodiment of a method for identifying the cause of an electric vehicle speed out-of-control fault according to the present invention;

[0020] Figure 6 This is an example diagram of the first characteristic diagram of Embodiment 2 of a method for identifying the cause of a speed out-of-control fault of an electric vehicle according to the present invention;

[0021] Figure 7 This is a first example diagram of the second characteristic diagram of a second embodiment of a method for identifying the cause of a speed out-of-control fault of an electric vehicle according to the present invention;

[0022] Figure 8 This is a second example diagram of the second characteristic diagram of Embodiment 2 of a method for identifying the cause of a speed out-of-control fault of an electric vehicle according to the present invention;

[0023] Figure 9 This is a signal analysis diagram of a second embodiment of a method for identifying the cause of an electric vehicle speed out-of-control fault according to the present invention;

[0024] Figure 10 This is an example diagram of the first characteristic diagram of Embodiment 3 of a method for identifying the cause of a speed out-of-control fault of an electric vehicle according to the present invention;

[0025] Figure 11 This is an example diagram of the second characteristic diagram of Embodiment 3 of a method for identifying the cause of a speed out-of-control fault of an electric vehicle according to the present invention;

[0026] Figure 12 This is a signal analysis schematic diagram of a third embodiment of a method for identifying the cause of an electric vehicle speed out-of-control fault according to the present invention. DETAILED DESCRIPTION

[0027] The following is a further detailed description through specific implementation methods:

[0028] Example 1

[0029] The embodiment is basically as shown in the attached Figure 1 A method for identifying the cause of an electric vehicle speed out-of-control fault is shown, comprising the following steps:

[0030] S1, collects and analyzes the message data during vehicle operation, extracts signal parameters from it, and obtains vehicle operation data T after preprocessing.

[0031] The signal parameters include: data acquisition time, accelerator pedal opening, brake pedal opening, torque, voltage, speed, etc.

[0032] Furthermore, the accelerator pedal opening value in the vehicle operation data is extracted and torque values , ; ; and process the data according to the following processing strategy to filter out abnormal data:

[0033] Filter the data with both accelerator pedal opening value and torque value being positive, i.e. filter data, and exclude In a downward trend and The rate of change of the data exceeds the torque mutation threshold.

[0034] In this embodiment, the torque mutation threshold is set to This operation can eliminate abnormal data caused by sudden changes in torque data due to control strategies such as energy recovery or single-pedal braking when the accelerator pedal is gradually released during vehicle operation, thereby reducing interference with subsequent data analysis.

[0035] Calculate the difference in accelerator pedal opening values ; ;

[0036] Where, represents the difference value of the accelerator pedal opening value of the i-th frame data, and N represents the total number of frames of data in T;

[0037] filter The part of the data that is smaller than the accelerator pedal mutation threshold t0 is used to filter the vehicle operation data T by the index of this part of the data: ;

[0038] Where, Indicates the index of the position in the vector that satisfies the condition.

[0039] In this embodiment, the accelerator pedal sudden change threshold Set to 2. This operation can eliminate the interference of abnormal data caused by sudden acceleration or sudden release on fault judgment.

[0040] S2, the vehicle operation data T is divided and the accelerator pedal opening value in the vehicle operation data T is used As a feature, the torque value in the vehicle operation data T As a goal, create a training set and a test set.

[0041] S3, built on and The target accelerator pedal opening value is predicted based on the regression model , and get the predicted target torque value .

[0042] The regression model is ;

[0043] Where w and b are model parameters, which are determined as follows: ;

[0044] Where N is the total number of frames of data in T, is the torque value of the i-th frame data, is the accelerator pedal opening value of the i-th frame data. The visualization effect of the regression model is as follows Figure 2 shown.

[0045] S4, calculate the target torque value and the actual torque value Offset .

[0046] in, .

[0047] S5, according to the offset , confirm high-risk points, and output multi-dimensional feature maps.

[0048] The high-risk point refers to the offset The corresponding data points exceeding the threshold t1; the high-risk points and their nearby m frames of data points constitute a multi-dimensional feature map.

[0049] In this embodiment, the value of t1 is 30. In actual application, this value can be adjusted according to the power performance of different vehicle models. The value of m is 50. In actual application, this value can be set according to data quality and fault type.

[0050] The multi-dimensional characteristic diagram includes: a first characteristic diagram - a characteristic diagram of the changing relationship between the accelerator pedal opening and the torque value; and a second characteristic diagram - a characteristic diagram of the changing relationship between the speed, the accelerator pedal opening and the voltage.

[0051] S6, perform fault determination and cause analysis based on the multi-dimensional feature graph.

[0052] When conducting fault determination and cause analysis, the following situations are included:

[0053] Observe the numerical fluctuation characteristics in the multidimensional feature graph, such as Figure 3As shown, when the numerical fluctuation feature in the first characteristic map shows that the accelerator pedal opening value increases and the drive motor torque suddenly drops to 0, the synchronization verification based on the second characteristic map is triggered - as shown in Figure 4 As shown, if the total voltage drops below the safety threshold within the preset time threshold, the message data of the same period is retrieved. If the message data contains BMS total voltage loss alarm, DC-DC alarm and drive motor status abnormal signal at the same time, such as Figure 5 As shown, it is determined that the high voltage signal acquisition fails.

[0054] The BMS total voltage loss alarm is a fault warning signal triggered when the vehicle's BMS (battery management system) detects an abnormal drop in the total voltage of the power battery pack or a complete loss of voltage. The DC-DC alarm is a fault prompt signal triggered when the DC-DC converter (the DC-DC converter, responsible for converting the DC power of the high-voltage power battery into low-voltage DC power to power the vehicle's electronic equipment) detects an abnormal state. The drive motor status abnormality signal is a fault indication signal triggered when the vehicle's motor controller (MCU) or vehicle control unit (VCU) detects that the drive motor's operating parameters have exceeded a safe range or have failed.

[0055] The high-voltage signal acquisition failure includes at least one of: a voltage acquisition board hardware failure, poor contact of the high-voltage signal harness, or an abnormality in the control unit signal receiving channel.

[0056] The preset time threshold for synchronization verification is the time window from the accelerator pedal opening value rising to the drive motor torque suddenly dropping to 0, and this time window does not exceed 200ms. In this embodiment, the safety threshold can be set to 40% of the rated value.

[0057] This embodiment provides a method for identifying the cause of an electric vehicle speed loss fault. This method constructs an anomaly identification and location method in a data-driven manner. This method can quickly locate high-risk locations in complex and redundant vehicle operation history data and automatically output a multi-dimensional feature map. Furthermore, combined with image feature analysis and verification, this method achieves knowledge-driven identification and determination of the cause of an electric vehicle speed loss fault, achieving accurate identification and reliable analysis.

[0058] The key point is: Compared with traditional detection methods based on rules or fixed thresholds, this solution establishes a correlation benchmark between driving intention and power output through dynamic modeling. It can effectively remove the interference of control strategies such as energy recovery and single-pedal mode on the torque signal, and focus on the actual "acceleration demand-power response" link anomaly detection.

[0059] At the data processing level, a data screening mechanism based on the effective accelerator pedal action interval is employed to mitigate torque fluctuations under non-acceleration commands. At the model building level, linear regression is used to establish dynamic response benchmarks for different driving scenarios, preserving the personalized parameter space for the dynamic characteristics of different vehicle models while establishing a unified quantitative evaluation standard for abnormalities. This dual innovation in data processing and modeling enables the system to capture unexpected deviations in torque output on a millisecond timescale. Furthermore, through the visualization of multidimensional feature maps, fault modes implicit in high-dimensional data space are projected into low-dimensional space, providing engineers with an analytical interface that combines data support with physical interpretability, significantly improving the efficiency of tracing complex faults.

[0060] Secondly, the regression model construction logic of this solution has strong advantages - the accelerator pedal opening is a direct quantitative representation of the driver's intention, and it constitutes the core control link of the vehicle power system with the motor torque output. The two have a clear physical mapping relationship under normal working conditions. By establishing a dynamic regression model of the two, it is not only possible to quantify the expected response relationship, but also to effectively distinguish between normal control strategy intervention and torque deviation caused by abnormal faults. Specifically, in the preprocessing stage, by filtering out the data fragments when the pedal is released, the torque reverse output interference caused by the active intervention of the energy recovery system can be avoided; focusing on the effective travel range of the pedal ensures that the model only analyzes the response anomalies of clear acceleration commands, thereby enabling the model to accurately identify real faults such as loss of electronic control commands and actuator response delays, while avoiding misjudging normal control logic such as brake energy recovery and thermal management torque limiting as abnormalities.

[0061] Compared to the limitations of traditional Failure Mode and Effects Analysis (FMEA) methods, which require pre-defined fault types, this solution automatically learns normal response patterns through data-driven analysis. This approach is adaptable to scenarios characterized by frequent software iterations and complex control logic in smart cars. It can detect unknown faults such as electronic control command loss and execution delays, avoiding the missed detections often associated with incomplete fault libraries in traditional methods. Furthermore, by minimizing prediction error using a linear regression model, the system automatically adjusts parameters based on the dynamic characteristics of different vehicle models (e.g., the high torque and rapid response of high-performance vehicles versus the smooth output characteristics of family cars), overcoming the strong dependence of fixed threshold methods on vehicle platforms and driving scenarios. Furthermore, this dynamic nature eliminates the need for complex calibration for each vehicle model, enabling cross-platform deployment. Furthermore, by visually analyzing multidimensional feature maps, abstract data anomalies are transformed into spatial clustering features. Cross-validation is then performed using physical knowledge, such as the motor electromagnetic torque equation and battery degradation models, to form a complete analytical chain: data anomaly discovery, physical mechanism tracing, and failure mode confirmation. This provides a viable technical path for transitioning new energy vehicle safety engineering from reactive accident response to proactive risk prevention.

[0062] Example 2

[0063] A method for identifying the cause of an electric vehicle speed out-of-control fault is provided, based on the first embodiment, with the following adjustments.

[0064] When conducting fault determination and cause analysis, the following situations are included:

[0065] Observe the numerical fluctuation characteristics in the multidimensional feature graph, such as Figure 6 As shown, when the numerical fluctuation feature in the first characteristic map shows that the accelerator pedal opening value increases and the drive motor torque suddenly drops to 0, the synchronization verification based on the second characteristic map is triggered - as shown in Figure 7 、 Figure 8 As shown in the figure, if the total voltage drops below the corresponding threshold and this situation (i.e., the total voltage drops) occurs twice, the message data of the same period is retrieved and multi-stage judgment is performed:

[0066] If the total voltage drops when the vehicle is stationary and there is no driving activity, if the total voltage drops by more than the static threshold (V1) and recovers automatically within 10-30 minutes, and there is no associated system alarm, it is marked as an occasional voltage fluctuation event;

[0067] If the total voltage drops below the driving threshold (V2) within the preset time (5 seconds) while the vehicle is in motion, and the following conditions are met at the same time:

[0068] The contradictory output of the accelerator pedal travel value increasing and the torque of at least one drive motor suddenly dropping to 0;

[0069] Triggering the composite signal of BMS total pressure loss alarm, DC-DC alarm and abnormal drive motor status; such as Figure 9 As shown;

[0070] It is determined that the high-voltage signal acquisition link fails.

[0071] According to the triggering scenarios of the high-voltage signal acquisition link failure, it is divided into:

[0072] If there is a record of abnormal voltage in static state, it is determined to be an intermittent poor contact fault;

[0073] If it appears for the first time and there is no static mark, it is determined to be a hardware failure of the voltage acquisition board.

[0074] This embodiment provides a method for identifying the cause of an electric vehicle's speed runaway fault. Compared to the first embodiment, it offers a more comprehensive fault determination and cause analysis strategy. Specifically, through differentiated diagnosis between stationary and driving states, combined with the conflicting phenomenon of acceleration demand and torque drop, and multi-system composite alarm verification, it can accurately trace the source of high-voltage system faults (distinguishing between occasional poor contact and hardware failure). Furthermore, the use of state-adaptive dynamic threshold management (different voltage thresholds and time windows for stationary and driving) ensures driving safety while reducing the rate of false positives.

[0075] Example 3

[0076] A method for identifying the cause of an electric vehicle speed out-of-control fault is provided, based on the first or second embodiment, with the following adjustments.

[0077] When conducting fault determination and cause analysis, the following situations are included:

[0078] Observe the numerical fluctuation characteristics in the multidimensional feature graph, such as Figure 10 As shown, when the numerical fluctuation feature in the first characteristic map shows that the accelerator pedal opening value increases and the drive motor torque suddenly drops to 0, the synchronization verification based on the second characteristic map is triggered - as shown in Figure 11 As shown, if the accelerator pedal opening value continues to rise and the total voltage drops below the safety threshold within the preset time threshold, the message data of the same period is retrieved and further judgment is made:

[0079] If the following conditions are met at the same time: the drive motor controller reports a torque output failure alarm, the BMS triggers a total voltage loss fault code, and the DC-DC converter output is abnormal (i.e., the message data contains BMS total voltage loss alarm, DC-DC alarm, and drive motor status abnormality signal); and the vehicle high-voltage bus current drops below the safety threshold, such as Figure 12 As shown; it is determined that the high-voltage signal acquisition link fails.

[0080] This embodiment provides a method for identifying the cause of an electric vehicle's speed runaway failure. Compared to the first embodiment, it offers a more comprehensive fault determination and cause analysis strategy. Specifically, by real-time monitoring of the conflict between acceleration demand and high-voltage sag, combined with comprehensive verification of alarms from multiple systems, including the drive motor, BMS, and DC-DC converter, it can accurately identify failures in the high-voltage signal acquisition link (such as a faulty voltage acquisition board or wiring harness), effectively eliminating false alarms from a single sensor.

[0081] The above is only an embodiment of the present invention. Common knowledge such as the specific structure and characteristics of the scheme is not described in detail here. Ordinary technicians in the relevant field are aware of all common technical knowledge in the technical field of the invention before the application date or priority date, can obtain all existing technologies in the field, and have the ability to apply conventional experimental means before that date. Ordinary technicians in the relevant field can improve and implement this scheme in combination with their own abilities under the guidance of this application. Some typical well-known structures or well-known methods should not become obstacles for ordinary technicians in the relevant field to implement this application. It should be pointed out that for those skilled in the art, without departing from the structure of the present invention, several variations and improvements can be made, which should also be regarded as the scope of protection of the present invention. These will not affect the effect of the implementation of the present invention and the practicality of the patent.

Claims

1. A method for identifying the cause of an electric vehicle speed out-of-control fault, characterized in that: The following steps are involved: S1, collects and analyzes the message data during vehicle operation, extracts signal parameters from it, and obtains vehicle operation data T after preprocessing; S2, the vehicle operation data T is divided and the accelerator pedal opening value in the vehicle operation data T is used As a feature, the torque value in the vehicle operation data T As a goal, establish training sets and test sets; S3, built on and The target accelerator pedal opening value is predicted based on the regression model , and get the predicted target torque value ; S4, calculate the target torque value and the actual torque value Offset ; S5, according to the offset , confirm high-risk points and output multi-dimensional feature maps; The multi-dimensional characteristic diagram includes: a first characteristic diagram - a characteristic diagram of the relationship between the accelerator pedal opening and the torque value; a second characteristic diagram - a characteristic diagram of the relationship between the speed, the accelerator pedal opening and the voltage; S6, perform fault determination and cause analysis based on the multi-dimensional feature graph; When conducting fault determination and cause analysis, the following situations are included: Observe the numerical fluctuation characteristics in the multidimensional feature map. When the numerical fluctuation characteristics in the first feature map show an increase in the accelerator pedal opening value and a sudden drop to 0 in the drive motor torque, trigger synchronization verification based on the second feature map. If the total voltage drops below the safety threshold within the preset time threshold, retrieve the message data for the same period. If the message data contains BMS total pressure loss alarm, DC-DC alarm and drive motor status abnormality signals, it is determined that the high-voltage signal acquisition has failed. The high-voltage signal acquisition failure includes at least one of: a voltage acquisition board hardware failure, poor contact of the high-voltage signal harness, or an abnormality in the control unit signal receiving channel.

2. The method for identifying the cause of an electric vehicle speed out-of-control fault according to claim 1, characterized in that: The signal parameters include: data acquisition time, accelerator pedal opening, brake pedal opening, torque, voltage, and speed.

3. The method for identifying the cause of an electric vehicle speed out-of-control fault according to claim 1, characterized in that: Before step 2, the process also includes: extracting the accelerator pedal opening value from the vehicle operation data; and torque values , ; ; and process the data in accordance with the following processing strategy: Filter the data with both accelerator pedal opening value and torque value being positive, i.e. filter data, and exclude In a downward trend and The rate of change of the data exceeds the torque mutation threshold.

4. The method for identifying the cause of an electric vehicle speed out-of-control fault according to claim 3, characterized in that: The processing strategy also includes: Calculate the difference in accelerator pedal opening values ; Where, represents the difference value of the accelerator pedal opening value of the i-th frame data, and N represents the total number of frames of data in T; filter The part of the data that is smaller than the accelerator pedal mutation threshold t0 is used to filter the vehicle operation data T by the index of this part of the data: ; Where, Indicates the index of the position in the vector that satisfies the condition.

5. The method for identifying the cause of an electric vehicle speed out-of-control fault according to claim 1, characterized in that: The regression model is ; Where w and b are model parameters, which are determined as follows: ; Where N is the total number of frames of data in T, is the torque value of the i-th frame data, is the accelerator pedal opening value of the i-th frame data.

6. The method for identifying the cause of an electric vehicle speed out-of-control fault according to claim 1, characterized in that: The high-risk point refers to the offset The corresponding data points exceeding the threshold t1; A multi-dimensional feature map is constructed using high-risk points and their nearby m frames of data points.

7. The method for identifying the cause of an electric vehicle speed out-of-control fault according to claim 6, characterized in that: The value of t1 is 30, and the value of m is 50.

8. The method for identifying the cause of an electric vehicle speed out-of-control fault according to claim 1, characterized in that: The preset time threshold for the synchronization verification is a time window from the accelerator pedal opening value rising to the drive motor torque suddenly dropping to 0, and the time window does not exceed 200ms.

Images