Intelligent test bench and method for SAS steering sensor and joint debugging test method

Abstract

This invention relates to an intelligent test bench, intelligent testing method, and integrated testing method for SAS steering sensors. The intelligent test bench includes a test bench body, a host computer, a servo motor system, and an EMB controller hardware-in-the-loop interactive system. The SAS steering sensor to be tested is mounted on the test bench body. The EMB controller hardware-in-the-loop interactive system integrates a CAN / LIN bus communication interface, a fault injection module, and a signal conditioning module. The fault injection module is used to modify the SAS message signals on the bus side using individual injection or timing coupling to simulate various abnormal fault conditions of the SAS steering sensor. The host computer is used to perform data parsing, index calculation, fault injection control, and test report generation. The intelligent test bench of this invention can test the accuracy of the SAS steering sensor and its compatibility with the EMB system, and further verify the EMB system's ability to automatically diagnose and enter a safe state under SAS steering sensor failure.

Description

Technical Field

[0001] This invention relates to the field of automotive control, and specifically to an intelligent test bench, intelligent testing method, and integrated testing method for SAS steering sensors. Background Technology

[0002] Electromechanical braking (EMB), as a core technology in the field of brake-by-wire for intelligent connected vehicles, is a crucial core actuator in the vehicle chassis domain. The various active safety functions integrated into the EMB system, such as Electronic Stability Control (ESC), Anti-lock Braking System (ABS), Traction Control System (TCS), and Active Collision Avoidance, all rely on the steering wheel angle and steering rate signals output by the steering angle sensor (SAS) as core input parameters. These signals enable closed-loop control of vehicle dynamics, directly determining the safety and stability of the vehicle's braking.

[0003] Unlike traditional hydraulic braking systems, the EMB system eliminates the mechanical hydraulic pipeline structure and has no hydraulic buffer redundancy structure. The electrical signal output by the SAS steering sensor can be quickly applied to the braking actuator. The sensor signal and the EMB braking control logic are strongly coupled and correlated. The accuracy, real-time performance, and stability of the signal directly determine the execution effect of each active safety function of the EMB system, playing a decisive role in the driving safety of the vehicle under extreme conditions.

[0004] In traditional hydraulic braking systems, the transmission delay of hydraulic lines and mechanical damping can naturally offset some of the signal deviations of the SAS steering sensor, resulting in a high fault tolerance threshold. A typical ±3° angle error only causes about 2% deviation in braking force distribution, which is generally sufficient for the vehicle's normal driving braking needs. However, the EMB system lacks a mechanical buffer structure. Problems such as sensor signal accuracy deviations, transmission delays, signal jumps, and frame drops are directly transmitted to the four-wheel brake motor control terminal without buffering or attenuation. Actual testing has verified that under extreme vehicle conditions such as high-speed obstacle avoidance, an angle error of only ±0.5° in the SAS steering sensor can lead to a yaw rate calculation deviation exceeding 8% and a single-wheel braking force adjustment deviation exceeding 15%. This can easily cause ESC function to mis-trigger, trigger lag, or even complete malfunction, resulting in serious safety accidents such as understeer, oversteer, skidding, and rollover.

[0005] Based on the high safety and high response control characteristics of the EMB system, in order to meet the highest functional safety level requirements of vehicles ASIL-D, the industry has put forward stringent quantitative testing standards for SAS steering sensors adapted to the EMB system: requiring the static angle error of the sensor to be controlled within ±0.5° and the dynamic following error to be less than ±0.8°. At the same time, strict limits are set for signal response delay, all-condition adaptability, and fault tolerance. These testing standards are far higher than the sensor testing requirements of traditional hydraulic braking systems.

[0006] Currently, most industry-standard SAS steering sensor testing solutions are designed for traditional hydraulic braking systems, exhibiting significant technical limitations and failing to meet the high-precision, full-condition, and closed-loop testing requirements of EMB systems. Existing testing methods only perform static factory calibration tests on individual SAS steering sensors, considering only basic individual parameters such as the sensor's gear ring tolerance, speed, and ambient temperature, completely detached from the actual operating conditions of the EMB system. This testing solution cannot reproduce sensor signal delays, dynamic errors, and signal anomalies under scenarios such as CAN bus load fluctuations, ESC dynamic braking intervention, and complex driving condition coupling during actual vehicle operation. It not only suffers from time-consuming testing and large system-level testing errors but also fails to verify the compatibility of sensor error compensation schemes with the EMB system control logic, making it difficult to guarantee the stable operation of the EMB system under complex conditions.

[0007] Meanwhile, existing general-purpose test benches can only perform basic performance qualification tests on individual SAS steering sensors. The test dimensions are limited, and they lack the ability to perform hardware-in-the-loop closed-loop joint debugging tests with the EMB controller. They cannot complete the collaborative matching test between the sensor and the EMB system, and it is difficult to cover the testing and verification needs of the entire life cycle of EMB system development, calibration, and mass production. This greatly restricts the research and development iteration and mass production of high-safety EMB systems. Summary of the Invention

[0008] The purpose of this invention is to overcome the shortcomings of the prior art and provide an intelligent test bench for SAS steering sensors. The intelligent test bench can not only test the accuracy of SAS steering sensors and their compatibility with EMB systems, but also verify the EMB system's ability to automatically diagnose and enter a safe state in the event of SAS steering sensor failure, providing reliable test support for the development, calibration, verification, and mass production testing of EMB systems.

[0009] The second objective of this invention is to provide an intelligent testing method for SAS steering sensors.

[0010] The third objective of this invention is to provide a method for joint testing of SAS steering sensors.

[0011] The technical solution of the present invention to solve the above-mentioned technical problems is:

[0012] A smart test bench for SAS steering sensors includes a test bench body, and a host computer, a servo motor system, and an EMB controller hardware-in-the-loop interactive system mounted on the test bench body.

[0013] The SAS steering sensor to be tested is mounted on the test bench body. The gear ring of the SAS steering sensor is rigidly connected to the servo motor shaft in the servo motor system. The servo motor outputs set angle and speed excitation to drive the SAS steering sensor to operate, in order to simulate different EMB real vehicle working conditions. The servo motor is equipped with an encoder to output reference angle signal and reference speed signal.

[0014] The EMB controller hardware-in-the-loop interactive system integrates a CAN / LIN bus communication interface, a fault injection module, and a signal conditioning module.

[0015] The CAN / LIN bus communication interface is used to establish a real-time bus data interaction channel between the host computer, the SAS steering sensor, and the EMB controller. The reference angle signal and reference speed signal output by the servo motor are first uploaded to the host computer. The host computer then transmits the measured angle signal and measured speed signal output by the SAS steering sensor in real time, along with a timestamp, to the EMB controller synchronously via the CAN / LIN bus communication interface. The CAN / LIN bus communication interface also collects the four-wheel braking control commands and system fault diagnosis information output by the EMB controller and sends them back to the host computer.

[0016] The fault injection module is used to modify the SAS message signals on the bus side by means of individual injection or time coupling to simulate various abnormal fault conditions of the SAS steering sensor.

[0017] The signal conditioning module is used to perform level conversion and noise filtering of the SAS steering sensor output signal to achieve signal adaptation.

[0018] The host computer is used to receive the reference angle signal and reference speed signal output by the encoder in the servo motor, the measured angle signal and measured speed signal output by the SAS steering sensor, and the feedback information from the EMB controller, and to complete data parsing, index calculation, fault injection control and test report generation.

[0019] Preferably, the fault injection module adopts a composite fault simulation method that combines CAN message modification with hardware electrical conditioning to simulate various abnormal operating conditions of the SAS steering sensor. Specifically, for signal communication faults, online modification of CAN messages is used to simulate signal jumps, message frame drops, numerical deviations, and bus interference faults, supporting customizable fault amplitude, duration, and triggering timing. For sensor power supply faults, dynamic voltage regulation of the programmable power supply is used to simulate abnormal electrical operating conditions.

[0020] Preferably, the EMB actual vehicle operating conditions include steady-state operating conditions, transition operating conditions, and high-dynamic operating conditions; wherein, when the steering angle rate is ≤50° / s, it is a steady-state operating condition; when 50° / s < steering angle rate <200° / s, it is a transition operating condition; and when the steering angle rate is ≥200° / s, it is a high-dynamic operating condition.

[0021] A smart testing method for SAS steering sensors, applied to the aforementioned smart test bench for SAS steering sensors, includes the following steps:

[0022] S101: Install the SAS steering sensor on the intelligent test bench and rigidly fix the gear ring of the SAS steering sensor to the shaft of the servo motor.

[0023] S102: Collect multiple sets of bench calibration sample data under different EMB real vehicle operating conditions, construct a sample dataset, divide the sample dataset into a steady-state sample set and a high-dynamic sample set, train a multi-parameter dynamic intelligent compensation model for each, obtain the compensation weight set corresponding to the steady-state operating condition and the high-dynamic operating condition, and calibrate the filtering strategy matching each operating condition. Store the compensation weight set and the corresponding filtering strategy in pairs; the filtering strategy is obtained by calibration, and the filtering strategy corresponding to each operating condition contains the filtering delay calibration constant obtained by bench testing; the steady-state operating condition and the high-dynamic operating condition correspond to independent filtering strategies.

[0024] S103: The host computer loads the preset test angle sequence and speed parameters, drives the servo motor shaft to rotate to the target angle at the specified speed through the motion control board, and simultaneously sends the corresponding EMB real vehicle condition label. The angle and speed messages output by the SAS steering sensor via the CAN bus are monitored in real time throughout the test process.

[0025] S104: The multi-parameter dynamic intelligent compensation model identifies the current vehicle operating condition based on real-time collected data. If it is identified as a steady-state condition or a high-dynamic condition, it calls the compensation weight set and filtering strategy corresponding to the current condition. If it is identified as a transitional condition, it performs condition weight smoothing adaptation processing. It performs predictive compensation and dynamic correction on the angle signal output by the SAS steering sensor in real time, and obtains the speed signal from the differential of the compensated angle signal, and outputs the optimal measurement value adapted to the current EMB real vehicle operating condition.

[0026] S105: After the test, the host computer completes a multi-dimensional quantitative performance evaluation based on the collected full test data and generates a full-condition performance test report of the SAS steering sensor adapted to the EMB system.

[0027] Preferably, in step S102, the multi-parameter dynamic intelligent compensation model adopts a linear regression prediction model, and the specific training steps are as follows:

[0028] Step 1: Collect multiple sets of bench calibration sample data under the corresponding EMB real vehicle conditions. Use the number of teeth on the positive and negative gear rings of the SAS steering sensor, the speed of the servo motor, the running time of the SAS steering sensor, the first-order filter coefficient of the SAS steering sensor, the EMB real vehicle condition label, braking stage characteristics, and the CAN bus load rate as model input parameters, and use the reference angle value output by the servo motor's built-in encoder as the model's true value output parameter to construct a model training sample dataset. The braking stage characteristics include no braking, conventional braking, ABS anti-lock braking, and ESC vehicle stability intervention braking.

[0029] Step 2: Use the least squares method to perform preliminary fitting on the model training sample dataset, and solve for the initial weight values ​​of each input parameter as the initial starting point for model iterative training; use the preset loss function as the optimization objective, and adopt the batch gradient descent algorithm. In each iteration, use the whole batch of training samples to calculate the gradient, and iteratively update the weights and model bias terms corresponding to all model input parameters along the gradient in the opposite direction.

[0030] Step 3: After each iteration, the model loss value is calculated using the validation set. When the change in the loss value is less than the preset convergence threshold for N consecutive iterations, the model training is determined to be converged, the iteration operation is terminated, and the final compensation weight set corresponding to the current EMB real vehicle working condition is output.

[0031] Step 4: Use test samples to verify the compensation accuracy of the trained multi-parameter dynamic intelligent compensation model; if the model compensation accuracy does not meet the preset performance index requirements, supplement the bench calibration sample data and re-iterate the training until the model compensation accuracy meets the requirements; if the preset performance index requirements are met, the model training is completed.

[0032] Preferably, in the smooth adaptation process of the working condition weights in step S104, the parameter adaptation process for the transition working condition is specifically as follows:

[0033] The initial fusion coefficient of the weights is solved based on the real-time turning rate as the basic variable. The initial fusion coefficient of the weights is dynamically corrected in combination with the ESC working stage and CAN bus load rate to obtain the final fusion coefficient of the transition condition. Based on the final fusion coefficient, a linear interpolation algorithm is used to smoothly transition the compensation weights between the steady-state condition and the high-dynamic condition. The same linear interpolation is performed on the filter delay calibration constants corresponding to the steady-state condition and the high-dynamic condition. At the same time, the parameter gradual change period is set to limit the single update amplitude of both the compensation weights and the filter delay calibration constant.

[0034] Preferably, in step S104, the step of predictive compensation and dynamic correction of the angle signal output in real time by the SAS steering sensor is as follows:

[0035] The multi-parameter dynamic intelligent compensation model calculates and accumulates the product of each model input parameter and the corresponding weight coefficient obtained through training to obtain the angle compensation amount; the angle compensation amount is then superimposed with the original angle output by the SAS steering sensor to obtain the first angle correction value.

[0036] To address the signal phase lag caused by CAN bus transmission and internal sensor filtering, a first-order phase lead filtering algorithm is used for delay correction. Specifically, the first angle correction value is superimposed on the product of the turning rate and the corresponding EMB vehicle-level filtering delay calibration constant to output the final second angle correction value. The filtering delay calibration constant is determined by the operating condition. In steady-state and high-dynamic conditions, the filtering delay calibration constant is taken from the filtering strategy of the corresponding operating condition. In transitional conditions, the filtering delay calibration constant is obtained in real time by linear interpolation of the final fusion coefficient with respect to the filtering delay calibration constant within the filtering strategy of steady-state and high-dynamic conditions, and then constrained by the parameter gradual change period. The turning rate is the rate of change obtained by differentiating the first angle correction value with respect to time.

[0037] Preferably, in step S105, the evaluation indicators of the multi-dimensional quantification performance include static angle error, dynamic tracking error, smoothness of operating condition switching, signal delay time, and message integrity rate.

[0038] A method for joint debugging and testing of SAS steering sensors, applied to the aforementioned intelligent test bench for SAS steering sensors, includes the following steps:

[0039] S201: Install the SAS steering sensor on the intelligent test bench and rigidly fix the gear ring of the SAS steering sensor to the servo motor shaft; connect the EMB controller to the EMB controller hardware-in-the-loop interactive system, complete the bus communication configuration and power supply wiring, and verify that the communication link between the EMB controller, the host computer, and the SAS steering sensor is smooth; the servo motor's built-in encoder outputs a reference speed signal and a reference angle signal as a reference signal, and establishes a time stamp synchronization acquisition mechanism between the reference signal and the SAS output message;

[0040] S202: The host computer loads a preset test angle sequence and speed parameters, and drives the servo motor shaft to rotate to the target angle at a specified speed through the motion control board. The host computer collects angle and speed messages output by the SAS steering sensor via the CAN bus in real time, adds a timestamp to each message frame, and forwards the time-stamped angle and speed messages to the EMB controller. The EMB controller completes the braking control logic calculation based on the received angle and speed messages, outputs four-wheel braking control commands, and simultaneously feeds back the four-wheel braking control commands, internal system function status, and fault flag bits to the host computer in real time. The host computer uses the reference angle and reference speed output by the servo motor's built-in encoder as the truth benchmark to evaluate the signal transmission quality of the SAS steering sensor and the CAN bus. At this stage, the response of the EMB controller is quantitatively evaluated, and the consistency of the four-wheel braking control commands between each wheel is analyzed. The four-wheel braking control commands output by the EMB controller are compared with the theoretical control results under preset working conditions to calculate the signal transmission response deviation and system response time, and to evaluate the accuracy and real-time performance of the EMB control logic.

[0041] S203: The host computer performs fault injection, simulates the abnormal SAS signal scenario, collects and records the response status of the EMB controller, changes in braking commands and fault diagnosis information, and simultaneously completes the accuracy evaluation of ESC function triggering.

[0042] S204: After the test, the host computer collects all valid test data under all test conditions and conducts quantitative fusion analysis from three independent dimensions: SAS sensor output performance, EMB braking control response performance, and ESC fault safety triggering performance. Finally, a test report on the coordinated closed-loop joint debugging of the sensor and EMB controller is generated.

[0043] Preferably, in step S203, four typical faults are injected: SAS signal jump, frame loss, out-of-tolerance, and signal delay. Among them, signal jump is an instantaneous step change in the SAS output angle and rotation speed; signal frame loss is the periodic loss of CAN bus SAS messages; signal out-of-tolerance is the SAS output signal exceeding the sensor's rated output range; and signal delay is the superimposed fixed or variable transmission delay during the SAS message transmission process.

[0044] Compared with the prior art, the present invention has the following advantages:

[0045] 1. This invention, based on a traditional SAS steering sensor test bench, adds an EMB controller hardware-in-the-loop interactive system, establishing a closed-loop test architecture for collaborative operation between the SAS sensor and the EMB controller. Compared to traditional test benches that can only perform single-unit sensor performance testing, this invention can accurately test the output accuracy and operating condition adaptability of the SAS steering sensor itself, and fully verify the fault diagnosis and safety degradation control capabilities of the EMB brake-by-wire system under extreme operating conditions such as SAS sensor signal anomalies and functional failures. It can comprehensively support the algorithm development, operating condition calibration, performance verification, and mass production testing of the EMB system, providing a reliable test platform for the functional safety development of brake-by-wire systems.

[0046] 2. The multi-parameter dynamic intelligent compensation model adopted in this invention uses a pre-calibrated EMB operating condition dataset as the training basis and relies on an independent training mechanism for different operating conditions to adapt to different driving scenarios. The model can identify the current EMB operating condition in real time, dynamically match the compensation weight set and filtering strategy under the corresponding operating condition, predict and compensate for the angle output by the SAS steering sensor in real time, and output the optimal measurement adapted to the current operating condition. Therefore, the multi-parameter dynamic intelligent compensation model in this invention can accurately correct the static deviation, dynamic following error, and signal transmission delay deviation of the SAS sensor under different EMB real vehicle operating conditions, completely solving the technical problems of traditional fixed compensation algorithms being unable to adapt to dynamic error correction under all operating conditions and having poor scene adaptability, ensuring the accuracy, stability, and real-time performance of the SAS steering sensor output signal, and meeting the ASIL-D high functional safety level requirements of the EMB brake-by-wire system.

[0047] 3. This invention timestamps the measured output messages of the SAS sensor by the host computer and forwards them to the EMB controller. It also collects the four-wheel braking control commands and fault diagnosis status information output by the EMB controller based on the measured SAS messages in real time. At the same time, relying on the fault injection module, it realizes single-mode or time-series coupled injection of various types of faults such as SAS signal jumps, message frame drops, numerical deviations, and power supply fluctuations. The system analyzes the impact of various SAS signal anomalies on the EMB braking control accuracy, four-wheel coordination consistency, system response real-time performance, and fault safety logic. It effectively makes up for the technical defects of traditional test equipment that can only carry out single-machine sensor testing and cannot realize closed-loop coordinated debugging of SAS steering sensors and EMB systems, and greatly improves the completeness and effectiveness of EMB brake-by-wire system testing. Attached Figure Description

[0048] Figure 1 This is a structural block diagram of the intelligent test bench for SAS steering sensors according to the present invention.

[0049] Figure 2This is a flowchart illustrating the intelligent testing method for SAS steering sensors according to the present invention.

[0050] Figure 3 This is a flowchart of the intelligent testing method for SAS steering sensors according to the present invention.

[0051] Figure 4 This is a flowchart of the joint testing method for SAS steering sensors according to the present invention. Detailed Implementation

[0052] The present invention will be further described in detail below with reference to the embodiments and accompanying drawings, but the embodiments of the present invention are not limited thereto.

[0053] Example 1

[0054] See Figure 1 The intelligent test bench for SAS steering sensors of the present invention includes a test bench body and a host computer, a servo motor system, and an EMB controller hardware-in-the-loop interactive system integrated on the test bench body.

[0055] See Figure 1 The host computer, serving as the control center of the intelligent test bench of this invention, primarily undertakes core functions such as controlling the motion control board, acquiring CAN bus data, and intelligent analysis and logical judgment of test data. It coordinates and manages the entire operation process and data interaction of the intelligent test bench. Specifically, in terms of motion control, the host computer establishes communication with the motion control board via the industrial PCI bus, issuing precise motor control commands and collecting and reading real-time status information such as angle and speed feedback during servo motor operation, achieving closed-loop precise control of the SAS steering sensor under test. Regarding CAN data interaction and intelligent analysis, the host computer completes real-time data acquisition from multiple devices via the CAN bus, simultaneously acquiring angle and speed messages output by the SAS steering sensor, as well as core information such as four-wheel braking control commands, ESC operating status, and fault codes fed back by the EMB controller. This establishes a full-duplex real-time communication link between the intelligent test bench of this invention, the sensor under test (SAS steering sensor), and the EMB controller, providing complete and real-time data support for subsequent test data analysis, performance judgment, and fault verification.

[0056] See Figure 1The host computer is equipped with an intelligent compensation and data processing module. This module has a built-in multi-parameter dynamic intelligent compensation model adapted to EMB functional safety requirements. The multi-parameter dynamic intelligent compensation model uses a pre-calibrated EMB real-vehicle operating condition dataset as its training basis. Through training, it obtains the compensation weight set and matching filtering strategy corresponding to different EMB real-vehicle operating conditions. During actual testing, the multi-parameter dynamic intelligent compensation model identifies the current EMB real-vehicle operating condition in real time, adaptively matches and calls the corresponding compensation weight set and filtering strategy for the EMB real-vehicle operating condition, performs predictive compensation and dynamic correction on the raw angle signal output by the SAS steering sensor in real time, and obtains the speed signal by differential solution of the corrected angle signal, outputting the optimal measurement value adapted to the current EMB real-vehicle operating condition. Therefore, this invention can accurately correct the static deviation, dynamic following error, and signal phase lag caused by bus transmission and inherent sensor filtering of the SAS steering sensor under different EMB real vehicle operating conditions. It overcomes the technical defects of traditional fixed parameter compensation algorithms that cannot adapt to dynamic error correction under all operating conditions, and enables the accuracy and real-time performance of the angle and speed signals output by the SAS steering sensor to meet the high-precision usage requirements of the ASIL-D functional safety level of the EMB system.

[0057] In this embodiment, the EMB actual vehicle operating conditions include steady-state operating conditions, transitional operating conditions, and high-dynamic operating conditions; wherein, when the steering angle rate is ≤50° / s, it is a steady-state operating condition; when 50° / s < steering angle rate <200° / s, it is a transitional operating condition; and when the steering angle rate is ≥200° / s, it is a high-dynamic operating condition.

[0058] See Figure 1 The servo motor system operates in position pulse mode, with pulse signals and direction control signals output from the motion control board. The SAS steering sensor under test is mounted on the test bench body, and the gear ring of the SAS steering sensor is rigidly connected to the servo motor shaft of the servo motor system. The motion control board drives the servo motor to rotate to the target angle at a set speed, outputting set angle and speed excitation to drive the SAS steering sensor to operate, simulating various EMB real-vehicle operating conditions. The servo motor has a built-in encoder, which, on the one hand, cooperates with the servo motor to realize closed-loop motion execution, and on the other hand, outputs reference angle and reference speed signals. The real-time angle and speed information of the servo motor is collected and uploaded to the host computer as the true value basis for test comparison.

[0059] In this embodiment, the SAS steering sensor acquires the raw sensing signal output by the internal magnetoelectric effect sensor. After reading the raw sensing value, the sensor's internal processing unit performs algorithm calculation and signal filtering to remove signal noise and interference glitches, and calculates the angle and rotational speed parameters corresponding to the gear ring. The processed gear ring angle and rotational speed data are encapsulated into a CAN message, and then output to the outside via the CAN bus. This CAN message is transmitted to the host computer on one hand, and sent to the EMB controller hardware-in-the-loop interaction system on the other hand, to receive the subsequent benchtop data interaction content.

[0060] See Figure 1 The EMB controller hardware in the loop-in-the-loop interactive system integrates a CAN / LIN bus communication interface, a fault injection module, and a signal conditioning module.

[0061] The CAN / LIN bus communication interface is used to establish a real-time bus data interaction channel between the host computer, the SAS steering sensor, and the EMB controller. The reference angle signal and reference speed signal output by the servo motor are first uploaded to the host computer. The host computer then transmits the measured angle signal and measured speed signal output in real time by the SAS steering sensor, along with a timestamp, to the EMB controller synchronously via the CAN / LIN bus communication interface. The CAN / LIN bus communication interface also collects the four-wheel braking control commands and system fault diagnosis information output by the EMB controller and sends them back to the host computer.

[0062] The fault injection module is used to modify the SAS message signals on the bus side using either individual injection or time-coupled methods to simulate various abnormal fault conditions of the SAS steering sensor. This fault injection module employs a composite fault simulation method combining CAN message modification and hardware electrical conditioning, enabling full-dimensional simulation of abnormal operating conditions of the SAS steering sensor. Specifically, for signal communication faults, online modification of CAN messages simulates signal jumps, message frame drops, numerical deviations, and bus interference faults, supporting customizable fault amplitude, duration, and trigger timing. For sensor power supply faults, dynamic voltage regulation of the programmable power supply simulates electrical abnormal operating conditions such as voltage drops and voltage ripple fluctuations. The modified abnormal bus messages are transmitted to the EMB controller via the CAN bus to evaluate the EMB brake-by-wire system's fault diagnosis logic, four-wheel brake distribution strategy, and system degradation control performance under abnormal SAS steering signal conditions.

[0063] The signal conditioning module preprocesses the raw electrical signal output by the SAS steering sensor, performing level conversion and high-frequency noise filtering. This ensures electrical compatibility and signal normalization between the sensor output signal and the CAN bus communication link and EMB controller hardware interface, guaranteeing the stability and integrity of the raw signal transmission. In this embodiment, the signal conditioning module employs a mature signal conditioning circuit architecture, effectively avoiding signal distortion caused by transmission interference and providing a clean raw signal base for subsequent message parsing, fault injection, and algorithm compensation.

[0064] See Figures 2-3 The intelligent testing method for SAS steering sensors of the present invention includes the following steps:

[0065] S101: Install the SAS steering sensor on the intelligent test bench and rigidly fix the gear ring of the SAS steering sensor to the shaft of the servo motor.

[0066] S102: Collect multiple sets of bench calibration sample data under different EMB real vehicle conditions, construct a sample dataset, divide the sample dataset into a steady-state sample set and a high-dynamic sample set, train a multi-parameter dynamic intelligent compensation model for each, obtain the compensation weight set corresponding to the steady-state condition and the high-dynamic condition, and calibrate the filtering strategy matching each condition. Store the compensation weight set and the corresponding filtering strategy in pairs.

[0067] This invention partitions the vehicle's operating conditions based on the measured steering angle rate, classifying the vehicle's operating state into three categories of EMB real-vehicle operating conditions: steady-state operating conditions, transitional operating conditions, and high-dynamic operating conditions. Differentiated compensation weight sets and filtering strategies are matched for different operating conditions to achieve adaptive error compensation for the SAS steering sensor under all operating conditions. This effectively solves the shortcomings of traditional fixed compensation methods that cannot simultaneously take into account steady-state accuracy and dynamic following performance.

[0068] The steady-state operating condition corresponds to low-dynamic operating scenarios with a steering angle rate ≤50° / s, covering conventional driving conditions such as normal straight driving, constant speed gentle steering, and smooth cornering. Under steady-state conditions, the EMB system mainly relies on high-precision steady-state angle signals to achieve smooth braking force distribution, requiring high signal stability and absolute measurement accuracy. Therefore, the multi-parameter dynamic intelligent compensation model takes static system error correction as its core optimization objective, adaptively increasing the correction weights for static error terms such as temperature drift, long-term working zero-point drift, and SAS steering sensor gear ring indexing error. At the same time, a steady-state adaptive filtering strategy (e.g., using a second-order low-pass filtering algorithm) is configured to smooth the original acquired signal, enhance high-frequency noise suppression capabilities, effectively filter out minor signal fluctuations caused by circuit interference and mechanical micro-vibrations, eliminate the problem of frequent EMB braking force fine-tuning caused by angle signal jitter under steady-state conditions, and ensure the absolute accuracy and output stability of steady-state angle measurement.

[0069] High-dynamic operating conditions correspond to high-dynamic operating scenarios with a steering angle rate ≥200° / s, covering extreme safety control scenarios such as vehicle ESC active intervention, high-speed emergency obstacle avoidance, rapid large-angle steering, and extreme condition correction. Under high-dynamic conditions, the EMB system has extremely high requirements for signal real-time performance and dynamic tracking performance; phase lag and response delay can directly cause braking control failure. To address this, the multi-parameter dynamic intelligent compensation model adopts a dynamic tracking performance-prioritized optimization strategy, focusing on increasing the correction weight of dynamic error parameters such as motor real-time speed, angular acceleration, and CAN bus transmission delay. Simultaneously, a first-order phase lead filtering mechanism is activated to appropriately weaken the low-frequency noise reduction intensity, specifically compensating for system phase lag, significantly improving the dynamic response speed and trajectory tracking accuracy of the angle signal. The speed signal is obtained by differential solution of the compensated angle signal, ensuring the overall tracking accuracy of the angle and speed signals. This solves the response lag and tracking distortion problems commonly found in traditional fixed-parameter compensation models under high-dynamic conditions, ensuring the effective triggering and precise control of EMB active safety functions under extreme conditions.

[0070] The transition condition corresponds to a medium-dynamic operating scenario with a turning rate between 50° / s and 200° / s, covering most daily driving conditions such as regular lane changes, medium-speed steering, and ABS braking triggering. It serves as a buffer zone between steady-state and high-dynamic conditions. To avoid abrupt changes in compensation logic and signal jumps caused by direct switching between steady-state and high-dynamic parameters, this invention adopts a linear interpolation smooth transition strategy. First, the initial fusion coefficient of the weights is calculated based on the real-time collected steering rate values. Then, the initial fusion coefficient is dynamically corrected by combining the ESC working stage and CAN bus load rate to obtain the final fusion coefficient of the transition condition. Based on this final fusion coefficient, a linear interpolation algorithm is used to smoothly transition the compensation weights between the compensation weight set in the steady-state condition and the compensation weight set in the high-dynamic condition. On the other hand, the same linear interpolation is performed on the filter delay calibration constant in the filter strategy in the steady-state condition and the filter delay calibration constant in the filter strategy in the high-dynamic condition. At the same time, a parameter gradual update cycle of 50ms is set to limit the single update amplitude of the compensation weights, filter parameters, and filter delay calibration constant, so as to realize the gradual and smooth switching of each parameter, eliminate the potential for parameter jumps at the moment of switching between different EMB vehicle conditions, and ensure the continuity, stability, and consistency of the SAS steering sensor output angle signal and speed signal during the condition switching process, thereby avoiding abnormal braking intervention of the EMB system due to sudden changes in the compensation strategy. The 50ms parameter gradual update cycle is the preferred setting in this embodiment, and can be adaptively adjusted according to the model of the sensor under test and the platform requirements of the EMB system.

[0071] For steady-state and high-dynamic operating conditions, the multi-parameter dynamic intelligent compensation model adopts a linear regression prediction model based on gradient descent. The training steps for this multi-parameter dynamic intelligent compensation model are as follows:

[0072] S1021: Collect multiple sets of bench calibration sample data under the corresponding EMB real vehicle conditions. Use the number of teeth on the positive and negative gear rings of the SAS steering sensor, the speed of the servo motor, the running time of the SAS steering sensor, the first-order filter coefficient of the SAS steering sensor, the EMB real vehicle condition label, braking stage characteristics, and the CAN bus load rate as model input parameters. Among them, the braking stage characteristics are identification parameters characterizing the working state of the EMB braking system. They are taken from the EMB controller bus messages and include four states: no braking, conventional braking, ABS anti-lock braking, and ESC vehicle stability intervention braking. They are used to identify whether the braking control logic intervenes and the specific intervention type. After the discrete braking stage characteristics are transformed by one-hot encoding, they are input into the model together with the other continuous input parameters (i.e., "multi-parameter dynamic intelligent compensation" below). Use the reference angle value output by the servo motor's built-in encoder as the model ground truth label to construct the model training sample dataset.

[0073] S1022: The least squares method is used to perform preliminary fitting on the model training sample dataset to obtain the initial weight values ​​of each input parameter, which serve as the initial starting point for iterative training of the model. With the preset loss function as the optimization objective, the batch gradient descent algorithm is used. In each iteration, the gradient is calculated using the entire batch of training samples, and the weights and model bias terms corresponding to all model input parameters are iteratively updated along the reverse direction of the gradient.

[0074] The expression for the preset loss function is:

[0075]

[0076] in, The number of samples used in training; For the first Angle correction values ​​for each sample after model compensation; For the first The reference angle for each sample For the first Signal delay time of each sample The delay error weight hyperparameter is given in advance by the bench calibration. It is used to normalize the magnitude of the angle error term and the signal delay constraint term and adjust the loss weighting, to compensate for the difference in the dimensions between the angle error term and the signal delay constraint term, so that the two losses can participate in the iterative optimization together. The target expected value for signal delay is set to meet EMB functional safety requirements.

[0077] The aforementioned preset loss function consists of two parts: an angle fitting error term and a signal delay constraint term. The angle fitting error term minimizes the deviation between the compensated angle correction value and the reference angle to ensure angle compensation accuracy. The signal delay constraint term constrains the signal delay of the model output, preventing the model from exceeding the signal delay limit simply by pursuing angle accuracy. Since the angle fitting error term outputs in angle dimensions and the signal delay constraint term outputs in time dimensions, their dimensions are inconsistent. The delay error weighting coefficients are obtained through pre-calibrated bench settings. On the one hand, the magnitude of the two errors is normalized to offset the numerical imbalance caused by the difference in dimensions. On the other hand, the proportion of the angle fitting error term and the signal delay constraint term is adjusted. Then, the batch gradient descent method is used to iteratively update the weights of each model input parameter along the gradient in the opposite direction with the preset loss function as the optimization objective, so as to complete the minimization solution of the preset loss function.

[0078] S1023: After each iteration, the loss value is calculated using the validation set. When the change in the loss value is less than the preset convergence threshold (e.g., 0.001) for multiple consecutive iterations (e.g., 100 times), the model training is determined to be converged, the iteration operation is terminated, and the final compensation weight set corresponding to the current EMB real vehicle working condition is output.

[0079] S1024: Use test samples to verify the compensation accuracy of the trained multi-parameter dynamic intelligent compensation model; if the model compensation accuracy does not meet the preset performance index requirements, supplement the bench calibration sample data and re-iterate the training until the model compensation accuracy meets the requirements; if the preset performance index requirements are met, the model training is completed.

[0080] For steady-state and high-dynamic operating conditions, the above independent training process is executed separately to generate mutually independent and non-interfering compensation weight sets.

[0081] S103: The host computer loads the preset test angle sequence and speed parameters, drives the servo motor shaft to rotate to the target angle at the specified speed through the motion control board, and simultaneously sends the corresponding EMB real vehicle condition label. The angle and speed messages output by the SAS steering sensor via the CAN bus are monitored in real time throughout the test process.

[0082] In this embodiment, the test is initiated via a host computer, and a standardized test condition sequence covering the entire EMB scenario is preset, specifically as follows:

[0083] Steady-state accuracy test: Six test points were selected: 0°, 30°, 90°, 180°, 270°, and 330°. The angle range from 0° to 330° was covered with a step size of 30°, and each angle point was held for 5 seconds.

[0084] Dynamic tracking test: includes 60° and 120° step angle conditions, 90° amplitude and 0.5Hz frequency sinusoidal angle conditions, and 30° amplitude and 2Hz frequency sinusoidal angle conditions.

[0085] Operating condition switching test: The slope turning condition, which involves a constant speed increase from 0° / s to 300° / s and then back to 0° / s, is used to verify the output smoothness of the operating condition switching process.

[0086] S104: The multi-parameter dynamic intelligent compensation model identifies the current vehicle operating condition based on real-time acquired data, i.e., determines the category of the current operating condition (steady-state / transitional / high dynamic) by using the steering rate threshold. If identified as a steady-state or high-dynamic condition, the compensation weight set and filtering strategy corresponding to the current operating condition are invoked. The compensation weight set consists of the weight coefficients of each model input parameter obtained from model training, and the filtering strategy includes the filtering delay calibration constant corresponding to each EMB real-vehicle operating condition and the sensor's inherent first-order filtering configuration parameters (i.e., filtering parameters). If identified as a transitional condition, the operating condition weight smoothing adaptation process is performed. The angle signal output by the SAS steering sensor in real time is predicted, compensated, and dynamically corrected. The rotational speed signal is obtained by differential analysis of the compensated angle signal, and the optimal measurement value adapted to the current EMB real-vehicle operating condition is output.

[0087] In the process of smooth adaptation of working condition weights, the parameter adaptation process for transitional working conditions is specifically as follows: the initial fusion coefficient of the weights with a value range of [0,1] is solved based on the real-time angular rate as the basic variable. The initial weighting fusion coefficient is determined by combining the ESC working phase and CAN bus load rate. Dynamic correction is performed: When the EMB system enters the braking intervention phase, or the CAN bus load rate exceeds the load rate threshold pre-calibrated on the test bench, it is determined that the vehicle's dynamic control requirements have increased. Based on the braking intervention intensity and the extent to which the CAN bus load rate exceeds the limit, the initial weighting fusion coefficients are adjusted. A corresponding magnitude offset is made towards the direction of high dynamic operating conditions to enhance the dynamic tracking performance weight; after the offset is completed, the initial fusion coefficients of the weights are adjusted. Perform amplitude limiting processing within the [0,1] interval to prevent parameters from exceeding the limit, and obtain the final fusion coefficient for the transition condition. Based on the final fusion coefficient A linear interpolation algorithm is adopted to smoothly transition the compensation weights between the steady-state and high-dynamic conditions. The same linear interpolation is performed on the filter delay calibration constants corresponding to the steady-state and high-dynamic conditions simultaneously. At the same time, a 50ms parameter change period is set to limit the single update amplitude of the compensation weights and filter delay calibration constants, so as to avoid sensor signal jumps caused by sudden changes in compensation parameters during the EMB real vehicle condition switching process.

[0088] Among them, the initial fusion coefficient of the weights for:

[0089]

[0090] In the formula: Current real-time rotation rate , The threshold for the working condition boundary, when hour, ;when hour, ;

[0091] The linear interpolation formula (taking weights as an example) is as follows:

[0092] In the formula: For the first steady-state condition The weights of the input parameters of the item model, For the first under high dynamic working conditions The weights of the input parameters of the model are used to avoid parameter step jumps through weight smoothing interpolation;

[0093] The compensated SAS sensor output signal is compared frame by frame with the reference angle and reference speed output by the servo motor's built-in encoder to output a real-time error curve. At the same time, quantitative evaluation indicators such as the average absolute deviation of the angle, the maximum angle deviation, and the speed deviation are statistically analyzed to verify the correction effect of the multi-parameter dynamic intelligent compensation model. Through this multi-parameter dynamic intelligent compensation model, it is possible to effectively suppress and correct the dynamic errors of working conditions that are difficult to cover by traditional fixed parameter compensation models.

[0094] After repeated testing and verification under various EMB real-vehicle operating conditions, the following results were obtained: Under steady-state conditions, the maximum static angle error is ≤ ±0.15°, and the root mean square angle error is ≤ 0.15°; under high-dynamic conditions, the maximum following error for step and sinusoidal conditions is ≤ ±0.5°; and under condition switching (transitional conditions), the signal jump is < 0.2°. The overall performance of the intelligent control method in this invention is superior to that of traditional fixed parameter compensation schemes, and it can meet the high-precision testing requirements of the EMB system for the SAS steering sensor under all operating conditions.

[0095] In this embodiment, the angle and speed values ​​output by the SAS steering sensor are input into a multi-parameter dynamic intelligent compensation model that has been trained and converged. The multi-parameter dynamic intelligent compensation model determines the corresponding EMB real-vehicle operating condition based on the speed value, and combines this with the number of teeth on the positive and negative gear rings of the SAS steering sensor, the servo motor speed, the SAS steering sensor's operating time, the SAS steering sensor's first-order filter coefficient, the EMB real-vehicle operating condition label, braking stage characteristics, and the CAN bus load rate to output the current angle compensation value and steering rate compensation value. The steering rate compensation value is obtained by performing a time derivative operation on the angle compensation value, thereby completing the error correction and delay compensation of the original output signal of the SAS steering sensor. Specifically, the angle error correction is implemented by the trained multi-parameter dynamic intelligent compensation model, and the specific calculation logic is as follows:

[0096]

[0097] In the formula: The weight coefficients of each input parameter of the multi-parameter dynamic intelligent compensation model obtained during training; Input parameters for the model acquired in real time; The original angle output by the SAS steering sensor; The output angle of the SAS sensor after model compensation correction (i.e., the first angle correction value). An angle compensation correction increment is used to correct the deviation of the original output angle of the SAS steering sensor.

[0098] For the compensated angle By performing time derivative calculations, the steering rate compensation value is obtained, which together completes the error correction and delay compensation of the original signal of the SAS steering sensor.

[0099] Phase delay compensation employs an independent first-order phase lead filtering algorithm, calling the filter delay calibration constant corresponding to the EMB real-vehicle operating condition within the aforementioned filtering strategy. The calculation is completed, and the specific calculation logic is as follows:

[0100]

[0101] In the formula: For the first The filter delay calibration constant under the test conditions is the same as the filter delay calibration constant corresponding to the current EMB real vehicle condition. Under steady-state and high-dynamic conditions, the filter delay calibration constant is taken from the corresponding filtering strategy. Under transition conditions, the filter delay calibration constant is determined by the aforementioned final fusion coefficient. The filter delay calibration constants for steady-state and high-dynamic conditions are obtained by linear interpolation. The interpolated filter delay calibration constants need to be constrained by a 50ms parameter gradual update cycle to obtain the filter delay calibration constants that actually participate in the calculation at the current moment. To compensate for the differential of the angle, i.e., the angular rate; The final output SAS corrected angle signal (i.e., the second angle correction value) after error correction and phase delay compensation is achieved; the signal transmission and processing delay is offset by the differential lead; and the high-frequency noise introduced by the differential operation is suppressed by the aforementioned filtering strategy and parameter gradual constraint.

[0102] S105: After the test, the host computer completes a multi-dimensional quantitative performance evaluation based on the collected full test data and generates a full-condition performance test report of the SAS steering sensor adapted to the EMB system.

[0103] In this embodiment, the evaluation indicators for multi-dimensional quantitative performance include static angle error, dynamic tracking error, operational condition switching smoothness, signal delay time, and message integrity rate. The maximum deviation, root mean square error, tracking deviation, and phase lag time involved in static angle error and dynamic tracking error can all be calculated using existing methods. Operational condition switching smoothness is measured by the maximum fluctuation of the angle output per unit time, a method that belongs to existing signal time-domain statistical methods. The specific definitions of each indicator are as follows:

[0104] Static angle error: Under steady-state conditions, obtain the compensated angle (i.e., the second angle correction value) and the reference angle, and calculate the maximum deviation and root mean square error between the two.

[0105] Dynamic tracking error: Under step and sinusoidal excitation conditions, obtain the compensated angle (i.e., the second angle correction value) and the reference angle, and calculate the tracking deviation and phase lag time;

[0106] Smoothness of operation switching: Under the slope turning angle condition, the maximum fluctuation of the angle output per unit time is statistically used to characterize the smoothness of the output signal during the operation switching process.

[0107] Signal delay time and message integrity rate: Signal delay time is the time difference between the CAN message output time and the encoder reference signal output time; message integrity rate is the ratio of the number of valid received messages to the total number of messages that should be received within the test period.

[0108] The model training in this invention is based on an improved batch gradient descent method, and the following improvements are made by combining it with actual EMB linear control braking test scenarios:

[0109] First, it abandons the training method of uniform fitting with a single model and adopts an independent iterative training mechanism for different working conditions. For steady-state working conditions and high-dynamic working conditions, it constructs a dedicated training sample set and a corresponding compensation weight set, which effectively solves the defects of low fitting accuracy and insufficient dynamic adaptability caused by a single model adapting to multiple working conditions.

[0110] Second, based on traditional sensor parameters, EMB system-specific features such as operating condition labels, CAN bus load rate, and braking intervention stages are introduced as model input parameters to enrich the error correlation dimensions and improve the model's ability to fit the error generation mechanism under complex braking conditions.

[0111] Third, the traditional loss function is structurally optimized to construct an improved loss function that adapts to multiple feature inputs and fits different working conditions. This enhances the model's ability to accurately fit angle deviations and dynamic hysteresis errors, and better meets the dynamic error compensation requirements under the coupled working conditions of the SAS steering sensor and EMB controller.

[0112] Example 2

[0113] See Figure 4 The present invention provides a method for joint testing of the SAS steering sensor for EMB brake-by-wire systems, comprising the following steps:

[0114] S201: Assemble the SAS steering sensor onto the intelligent test bench and rigidly connect the gear ring of the SAS steering sensor to the servo motor shaft; connect the EMB controller to the EMB controller hardware-in-the-loop interactive system, complete the bus communication configuration and power supply wiring, and verify the smooth communication link between the EMB controller, the host computer, and the SAS steering sensor. The servo motor outputs a reference angle signal and a reference speed signal through its built-in encoder as reference signals, and establishes a time-stamp synchronization acquisition mechanism for the reference signals and SAS output messages: using the global unified clock of the intelligent test bench as the time reference, the global timestamp is simultaneously marked while acquiring the reference angle signal output by the servo motor encoder; when the EMB controller hardware-in-the-loop interactive system receives the CAN message output by the SAS steering sensor, the timestamp under the same global clock is simultaneously marked on the CAN message, realizing high-precision synchronous acquisition of the reference angle signal and the SAS output message under a unified time axis.

[0115] S202: The host computer loads a preset test angle sequence and speed parameters, drives the servo motor shaft to rotate to the target angle at a specified speed via the motion control board, and collects angle and speed messages output by the SAS steering sensor via the CAN bus in real time, marking each message with a high-precision timestamp; the angle and speed messages are synchronously transmitted to the EMB controller via the CAN / LIN bus communication interface. The EMB controller executes braking control logic operations based on the received angle and speed messages, outputs four-wheel braking control commands, and feeds back the four-wheel braking control commands, the internal system function status of the controller, and fault flag bits to the host computer in real time. The host computer uses the reference angle and reference speed output by the servo motor's built-in encoder as the truth benchmark, on the one hand, to complete the signal transmission quality assessment of the SAS steering sensor and CAN bus, and on the other hand, to quantitatively evaluate the response performance of the EMB controller and analyze the consistency of the four-wheel braking control commands; by comparing the four-wheel braking control commands output by the EMB controller with the theoretical control results corresponding to the preset working conditions, the signal transmission response deviation and system response time are calculated, realizing a comprehensive evaluation of the accuracy and real-time performance of the EMB control logic.

[0116] The host computer compares the measured angle and speed output by the SAS steering sensor with the reference angle and speed output by the servo motor encoder to calculate the absolute angle deviation and absolute speed deviation, thereby evaluating the transmission accuracy of the SAS steering sensor and the CAN bus link sensor signals. Specifically, the absolute angle deviation is the absolute value of the difference between the SAS steering sensor output angle and the encoder reference angle in a single frame, and the absolute speed deviation is the absolute value of the difference between the SAS steering sensor output speed and the encoder reference speed in a single frame. By statistically analyzing all test samples, the average absolute angle deviation, average absolute speed deviation, maximum angle deviation, and maximum speed deviation are further obtained, effectively characterizing the degree of distortion of the SAS steering sensor output signal during CAN bus transmission.

[0117] After completing the signal quality assessment on the signal input side, the response performance of the EMB controller is quantitatively evaluated. The system dynamic response delay is calculated based on the high-precision timestamp difference of CAN bus messages: the timestamp of the EMB controller receiving valid SAS steering sensor angle and speed messages is taken as the signal input time, and the timestamp of the EMB controller outputting four-wheel braking control command messages is taken as the command output time. The time difference between the command output time and the signal input time is calculated to obtain the single system response time. The average system response time and the maximum response delay are obtained through multiple sets of sampling statistics, thereby quantitatively characterizing the real-time performance of the EMB controller's computation and processing.

[0118] The specific implementation method for the quantitative evaluation of the consistency of four-wheel braking control commands is as follows: Under the normal steering test condition without differential braking and ESP active intervention, the braking pressure command values ​​of the left front, right front, left rear, and right rear are simultaneously collected at the same time under the same condition. The dispersion of the four-wheel pressure commands is calculated using the sample standard deviation (denominator is the number of samples minus one), and this is used as the evaluation index of four-wheel braking coordination consistency. The larger the standard deviation value, the greater the difference in the four-wheel braking command output and the worse the four-wheel coordination consistency; the smaller the standard deviation value, the better the synchronization and consistency of the four-wheel braking output. All sampling times of all EMB real vehicle conditions are traversed, and the mean and maximum standard deviation values ​​are statistically analyzed to complete the comprehensive quantitative evaluation of four-wheel braking coordination consistency. When the test condition triggers differential braking and ESP active correction logic, the four-wheel braking pressure is the result of the controller's active differential allocation, and the data at this time is not included in the statistical calculation of four-wheel coordination consistency. The host computer retrieves the pre-stored theoretical four-wheel braking control results under the current EMB test condition. These theoretical four-wheel braking control results are pre-calculated using the EMB braking allocation algorithm, combined with the steering angle and speed parameters of the corresponding test condition, through full-condition offline simulation. This generates a four-wheel target pressure sequence with complete timestamp information and pre-stores it in the host computer. During the test, the host computer synchronously acquires the real-time output of the four-wheel braking control commands and their sampling timestamps from the EMB controller's hardware-in-the-loop interactive system. A globally unified clock is used to precisely align the timestamps of the actual output data with the pre-stored theoretical pressure sequence, achieving one-to-one matching of time-series data. After time-series matching is completed, the actual four-wheel braking control commands output by the EMB controller are compared frame-by-frame with the theoretical control results at the corresponding time. The absolute deviation between the actual braking pressure command and the corresponding theoretical pressure value of each wheel at a single frame is calculated. Based on the deviation data of all sampled frames under all EMB test conditions, the average absolute deviation of the commands under all conditions is statistically obtained. This invention defines the maximum overshoot as the maximum deviation of the actual braking pressure command from the theoretical pressure range. This maximum deviation includes both the positive deviation when the actual braking pressure command exceeds the upper limit of the theoretical pressure and the negative deviation when the actual braking pressure command is below the lower limit of the theoretical pressure, thus achieving a complete statistical analysis of the overall deviation characteristics of the braking command.

[0119] In summary, the host computer retrieves the pre-stored target pressure sequence of all four wheels under the current test conditions, which includes timestamp information. After precisely aligning the timestamps of the real-time braking commands output by the EMB controller with the pre-stored theoretical pressure sequence using a globally unified clock, frame-by-frame and wheel-by-wheel data comparison is achieved under time-series matching. The absolute deviation between the braking command of each frame and the corresponding theoretical pressure value is calculated, and the average absolute deviation and maximum overshoot are statistically obtained based on the full-condition sampled data. This invention quantifies and evaluates the precision and accuracy of EMB braking control through the average absolute deviation and maximum overshoot, quantifies and evaluates the consistency of four-wheel braking output through the discrete measurement of four-wheel pressure command samples, and quantifies and evaluates the real-time performance of EMB braking control through system response time. These multi-dimensional evaluation indicators are combined and mutually corroborate each other, achieving a quantifiable, reproducible, and comprehensive performance evaluation of the accuracy, output consistency, and dynamic response real-time performance of the EMB control logic.

[0120] S203: The host computer performs fault injection, simulating SAS signal anomaly scenarios, collecting and recording the response status of the EMB controller, changes in braking commands, and fault diagnosis information, and simultaneously completing the accuracy evaluation of ESC function triggering. This embodiment injects four typical faults: SAS signal jump, frame loss, out-of-tolerance, and signal delay. Specifically, signal jump refers to a sudden step change in the SAS output angle or speed; frame loss refers to the periodic or random loss of SAS CAN messages; out-of-tolerance refers to the SAS output angle or speed signal exceeding the sensor's rated output range; and signal delay refers to a fixed or variable delay superimposed during SAS message transmission.

[0121] During the testing process, the success rate of ESC function triggering, ESC trigger response time, and ESC false triggering rate were statistically analyzed. The specific calculation methods are as follows: Test cases that should theoretically trigger ESC intervention after fault injection are taken as valid test samples. The success rate of ESC function triggering is the ratio of the number of samples that actually trigger ESC function normally to the total number of valid test samples. The ESC trigger response time is taken as the starting point of the timing when the fault is injected and the timestamp of the ESC activation command message output by the EMB controller is taken as the ending point. The time difference between the two times is the ESC trigger response time under a single operating condition. The ESC false triggering rate is calculated by selecting samples of normal control operating conditions without faults, which is the ratio of the number of abnormally activated ESC samples under the condition without SAS faults to the total number of samples in the control group.

[0122] In addition, a comprehensive verification of the correction results of the original output signal of the SAS steering sensor is carried out by combining the multi-parameter dynamic intelligent compensation model: the signal correction value output by the multi-parameter dynamic intelligent compensation model, the original signal value output by the SAS steering sensor under fault conditions, and the actual received signal value of the EMB controller are compared frame by frame to verify the correction capability and accuracy of the multi-parameter dynamic intelligent compensation model for various abnormal SAS signals. The model can be used to distinguish between signal deviation scenarios that can be compensated by the model and hardware faults and serious bus abnormalities that cannot be compensated. The influence of abnormal SAS signals on the output accuracy and control logic stability of EMB braking control commands is quantitatively obtained.

[0123] Synchronous calculation of system safety degradation response time under fault conditions: The timing starts with the message timestamp of the EMB controller successfully detecting and identifying the SAS signal fault, and ends with the message timestamp of the EMB controller completing the control strategy switch, setting the corresponding fault diagnostic code, and entering the preset safety degradation working mode. The difference between the two timestamps is the safety degradation response time of the EMB system for the SAS signal fault.

[0124] S204: After the test, the host computer collects all valid test data under all test conditions and performs quantitative fusion analysis from three independent dimensions: SAS sensor output performance, EMB braking control response performance, and ESC fault safety triggering performance. Finally, a collaborative closed-loop joint debugging test report of SAS sensor and EMB controller is generated.

[0125] The first dimension is the evaluation of the output performance of the SAS steering sensor itself: Based on the collected full-condition test data, the average absolute deviation of angle, average absolute deviation of speed, maximum angle deviation, and maximum speed deviation are used as the core evaluation indicators to quantitatively evaluate the signal output accuracy and CAN bus transmission stability of the SAS steering sensor under static conditions, high dynamic conditions, and fault injection conditions. It also determines whether there are abnormal states such as distortion, phase lag, and abnormal jumps in the original SAS output signal, and completes the accuracy rating and performance evaluation of the test input link.

[0126] The second dimension is the evaluation of the EMB controller's braking control logic response performance: based on the average absolute deviation of braking commands, maximum overshoot, average system response delay, maximum response delay, and the mean and maximum standard deviations of the four-wheel braking control commands obtained from frame-by-frame comparison, the closed-loop control performance of the EMB controller is quantitatively evaluated from three aspects: control accuracy, real-time performance, and consistency. Specifically, the braking control accuracy is assessed using command deviation and overshoot, the system's dynamic real-time response capability is assessed using the response delay parameter, and the consistency of multi-wheel braking coordinated control is assessed using the four-wheel pressure dispersion.

[0127] The third dimension is the performance evaluation of ESC function triggering and fault safety response: Based on the ESC function triggering success rate, average ESC triggering time, ESC false triggering rate, and system safety degradation response time obtained from fault injection operating conditions, the system quantitatively evaluates the accuracy of function triggering, anti-interference capability, and functional safety degradation efficiency of the EMB system under single or coupled fault scenarios such as SAS signal jumps, message frame loss, numerical deviations, and transmission delays. It also verifies whether the system's fault identification, fault tolerance, and safety degradation control logic meet the functional safety design requirements of the brake-by-wire system.

[0128] Based on the independent quantitative evaluation of the above three dimensions, the host computer performs multi-index coupled fusion analysis: the signal accuracy deviation of the SAS steering sensor is regarded as the source of system input error, and the influence weight of abnormal input signal on EMB braking command deviation, system response delay, and four-wheel braking consistency is analyzed; at the same time, the signal correction effect of the multi-parameter dynamic intelligent compensation model is combined, and the difference in braking control accuracy and ESC trigger stability before and after compensation is compared, and the optimization capability of the multi-parameter dynamic intelligent compensation model for the dynamic control performance of the EMB system is quantitatively evaluated.

[0129] This invention employs a hierarchical difference quantification comparison method to achieve quantitative output of fusion evaluation. The specific process is as follows: First, the average deviation of sensor signals, system response delay, four-wheel pressure dispersion, ESC trigger success rate, ESC false trigger rate, and safety degradation response time are statistically analyzed across all operating conditions before and after model compensation, calculating the improvement amount and improvement rate of each individual performance indicator. Second, using the data of each indicator under fault-free standard operating conditions as benchmark thresholds, the test indicators under high-dynamic operating conditions and fault-injected operating conditions are normalized and the deviation is calculated against the benchmark thresholds to obtain the performance deviation of each dimension. By comprehensively considering three types of performance deviation indicators—sensor signal input accuracy, braking control real-time performance and consistency, and fault safety trigger reliability—a multi-dimensional performance coupling quantitative evaluation is completed. Through the performance difference before and after compensation, the overall optimization effect of the multi-parameter dynamic intelligent compensation model on the collaborative control performance of the EMB brake-by-wire system is quantified, outputting quantifiable and comparable closed-loop test evaluation results.

[0130] Finally, the host computer integrates all EMB test conditions, various quantitative indicators, comparison data before and after compensation, and fault response results to automatically generate a SAS-EMB closed-loop joint debugging test report. The test report covers raw test data, statistics of individual performance indicators, multi-dimensional error characteristic analysis, fault response timing analysis, and comparison of intelligent model compensation effects, realizing full-link quantitative testing and comprehensive performance evaluation from sensor signal acquisition, CAN bus signal transmission, controller logic operation, four-wheel braking command output to fault safety response.

[0131] The above are preferred embodiments of the present invention, but the embodiments of the present invention are not limited to the above content. Any changes, modifications, substitutions, combinations, or simplifications made without departing from the spirit and principle of the present invention shall be considered equivalent substitutions and shall be included within the protection scope of the present invention.

Claims

1. A smart test bench for SAS steering sensors, characterized in that, It includes a test bench body, and a host computer, servo motor system, and EMB controller hardware-in-the-loop interaction system mounted on the test bench body; The SAS steering sensor to be tested is mounted on the test bench body. The gear ring of the SAS steering sensor is rigidly connected to the servo motor shaft in the servo motor system. The servo motor outputs set angle and speed excitation to drive the SAS steering sensor to operate, in order to simulate different EMB real vehicle working conditions. The servo motor is equipped with an encoder to output reference angle signal and reference speed signal. The EMB controller hardware-in-the-loop interactive system integrates a CAN / LIN bus communication interface, a fault injection module, and a signal conditioning module. The CAN / LIN bus communication interface is used to establish a real-time bus data interaction channel between the host computer, the SAS steering sensor, and the EMB controller. The reference angle signal and reference speed signal output by the servo motor are first uploaded to the host computer. The host computer then transmits the measured angle signal and measured speed signal output by the SAS steering sensor in real time, along with a timestamp, to the EMB controller synchronously via the CAN / LIN bus communication interface. The CAN / LIN bus communication interface also collects the four-wheel braking control commands and system fault diagnosis information output by the EMB controller and sends them back to the host computer. The fault injection module is used to modify the SAS message signals on the bus side by means of individual injection or time coupling to simulate various abnormal fault conditions of the SAS steering sensor. The signal conditioning module is used to perform level conversion and noise filtering of the SAS steering sensor output signal to achieve signal adaptation. The host computer is used to receive the reference angle signal and reference speed signal output by the encoder in the servo motor, the measured angle signal and measured speed signal output by the SAS steering sensor, and the feedback information from the EMB controller, and to complete data parsing, index calculation, fault injection control and test report generation.

2. The intelligent test bench for SAS steering sensors according to claim 1, characterized in that, The fault injection module employs a composite fault simulation method combining CAN message modification and hardware electrical conditioning to simulate various abnormal operating conditions of the SAS steering sensor. Specifically, for signal communication faults, online modification of CAN messages is used to simulate signal jumps, message frame drops, numerical deviations, and bus interference faults, supporting customizable fault amplitude, duration, and triggering timing. For sensor power supply faults, dynamic voltage regulation of the programmable power supply is used to simulate abnormal electrical operating conditions.

3. The intelligent test bench for SAS steering sensors according to claim 1, characterized in that, The EMB real-vehicle operating conditions include steady-state operating conditions, transition operating conditions, and high-dynamic operating conditions; among them, when the steering angle rate is ≤50° / s, it is a steady-state operating condition; when 50° / s < steering angle rate <200° / s, it is a transition operating condition; and when the steering angle rate is ≥200° / s, it is a high-dynamic operating condition.

4. A smart testing method for SAS steering sensors, applied to the smart test bench for SAS steering sensors as described in claim 3, characterized in that, Includes the following steps: S101: Install the SAS steering sensor on the intelligent test bench and rigidly fix the gear ring of the SAS steering sensor to the shaft of the servo motor. S102: Collect multiple sets of bench calibration sample data under different EMB real vehicle conditions, construct a sample dataset, divide the sample dataset into a steady-state sample set and a high-dynamic sample set, train a multi-parameter dynamic intelligent compensation model for each, obtain the compensation weight set corresponding to the steady-state condition and the high-dynamic condition, and calibrate the filtering strategy matching each condition. Store the compensation weight set and the corresponding filtering strategy in pairs. S103: The host computer loads the preset test angle sequence and speed parameters, drives the servo motor shaft to rotate to the target angle at the specified speed through the motion control board, and simultaneously sends the corresponding EMB real vehicle condition label. The angle and speed messages output by the SAS steering sensor via the CAN bus are monitored in real time throughout the test process. S104: The multi-parameter dynamic intelligent compensation model identifies the current vehicle operating condition based on real-time collected data. If it is identified as a steady-state condition or a high-dynamic condition, it calls the compensation weight set and filtering strategy corresponding to the current condition. If it is identified as a transitional condition, it performs condition weight smoothing adaptation processing. It performs predictive compensation and dynamic correction on the angle signal output by the SAS steering sensor in real time, and obtains the speed signal from the differential of the compensated angle signal, and outputs the optimal measurement value adapted to the current EMB real vehicle operating condition. S105: After the test, the host computer completes a multi-dimensional quantitative performance evaluation based on the collected full test data and generates a full-condition performance test report of the SAS steering sensor adapted to the EMB system.

5. The intelligent testing method for SAS steering sensors according to claim 4, characterized in that, In step S102, the multi-parameter dynamic intelligent compensation model adopts a linear regression prediction model, and the specific training steps are as follows: Step 1: Collect multiple sets of bench calibration sample data under the corresponding EMB real vehicle conditions. Use the number of teeth on the positive and negative gear rings of the SAS steering sensor, the speed of the servo motor, the running time of the SAS steering sensor, the first-order filter coefficient of the SAS steering sensor, the EMB real vehicle condition label, braking stage characteristics, and the CAN bus load rate as model input parameters, and use the reference angle value output by the servo motor's built-in encoder as the model's true value output parameter to construct a model training sample dataset. The braking stage characteristics include no braking, conventional braking, ABS anti-lock braking, and ESC vehicle stability intervention braking. Step 2: Use the least squares method to perform preliminary fitting on the model training sample dataset, and solve for the initial weight values ​​of each input parameter as the initial starting point for model iterative training; use the preset loss function as the optimization objective, and adopt the batch gradient descent algorithm. In each iteration, use the whole batch of training samples to calculate the gradient, and iteratively update the weights and model bias terms corresponding to all model input parameters along the gradient in the opposite direction. Step 3: After each iteration, the model loss value is calculated using the validation set. When the change in the loss value is less than the preset convergence threshold for N consecutive iterations, the model training is determined to be converged, the iteration operation is terminated, and the final compensation weight set corresponding to the current EMB real vehicle working condition is output. Step 4: Use test samples to verify the compensation accuracy of the trained multi-parameter dynamic intelligent compensation model; if the model compensation accuracy does not meet the preset performance index requirements, supplement the bench calibration sample data and re-iterate the training until the model compensation accuracy meets the requirements; if the preset performance index requirements are met, the model training is completed.

6. The intelligent testing method for SAS steering sensors according to claim 4, characterized in that, In the smooth adaptation process of working condition weights in step S104, the parameter adaptation process for the transition working condition is as follows: The initial fusion coefficient of the weight is solved based on the real-time turning rate as the basic variable. The initial fusion coefficient of the weight is dynamically corrected by combining the ESC working stage and the CAN bus load rate to obtain the final fusion coefficient of the transition condition. Based on the final fusion coefficient, a linear interpolation algorithm is used to smoothly transition the compensation weights between the steady-state and high-dynamic conditions. Simultaneously, the same linear interpolation is performed on the filter delay calibration constants corresponding to the steady-state and high-dynamic conditions. At the same time, a parameter gradient period is set to limit the single update amplitude of both the compensation weights and the filter delay calibration constants.

7. The intelligent testing method for SAS steering sensors according to claim 4, characterized in that, In step S104, the steps for predictive compensation and dynamic correction of the angle signal output in real time by the SAS steering sensor are as follows: The multi-parameter dynamic intelligent compensation model calculates and accumulates the product of each model input parameter and the corresponding weight coefficient obtained through training to obtain the angle compensation amount; the angle compensation amount is then superimposed with the original angle output by the SAS steering sensor to obtain the first angle correction value. To address the signal phase lag caused by CAN bus transmission and internal sensor filtering, a first-order phase lead filtering algorithm is used to correct the delay. Specifically, the first angle correction value is superimposed on the product of the turning rate and the corresponding EMB vehicle operating condition filtering delay calibration constant, and the final second angle correction value is output. The turning rate is the rate of change obtained by differentiating the first angle correction value with respect to time.

8. The intelligent testing method for SAS steering sensors according to claim 4, characterized in that, In step S105, the evaluation indicators of the multi-dimensional quantification performance include static angle error, dynamic tracking error, smoothness of operating condition switching, signal delay time, and message integrity rate.

9. A method for joint debugging and testing of a SAS steering sensor, applied to the intelligent test bench for the SAS steering sensor as described in claim 3, characterized in that, Includes the following steps: S201: Install the SAS steering sensor on the intelligent test bench and rigidly fix the gear ring of the SAS steering sensor to the servo motor shaft; connect the EMB controller to the EMB controller hardware-in-the-loop interactive system, complete the bus communication configuration and power supply wiring, and verify that the communication link between the EMB controller, the host computer, and the SAS steering sensor is smooth; the servo motor's built-in encoder outputs a reference speed signal and a reference angle signal as a reference signal, and establishes a time stamp synchronization acquisition mechanism between the reference signal and the SAS output message; S202: The host computer loads the preset test angle sequence and speed parameters, and drives the servo motor shaft to rotate to the target angle at the specified speed through the motion control board; the host computer collects the angle and speed messages output by the SAS steering sensor via the CAN bus in real time, adds a timestamp to each message, and forwards the angle and speed messages with timestamps to the EMB controller; The EMB controller completes braking control logic calculations based on received angle and speed messages, outputs four-wheel braking control commands, and simultaneously feeds back the four-wheel braking control commands, internal system function status, and fault flags to the host computer in real time. The host computer uses the reference angle and reference speed output by the servo motor's built-in encoder as the truth benchmark to evaluate the signal transmission quality of the SAS steering sensor and CAN bus. During this stage, the response of the EMB controller is quantitatively evaluated, and the consistency of the four-wheel braking control commands between each wheel is analyzed. The four-wheel braking control commands output by the EMB controller are compared with the theoretical control results under preset operating conditions to calculate the signal transmission response deviation and system response time, and to evaluate the accuracy and real-time performance of the EMB control logic. S203: The host computer performs fault injection, simulates the abnormal SAS signal scenario, collects and records the response status of the EMB controller, changes in braking commands and fault diagnosis information, and simultaneously completes the accuracy evaluation of ESC function triggering. S204: After the test, the host computer collects all valid test data under all test conditions and conducts quantitative fusion analysis from three independent dimensions: SAS sensor output performance, EMB braking control response performance, and ESC fault safety triggering performance. Finally, a test report on the coordinated closed-loop joint debugging of the sensor and EMB controller is generated.

10. The method for joint debugging and testing of SAS steering sensors according to claim 9, characterized in that, In step S203, four typical faults are injected: SAS signal jump, frame loss, out-of-tolerance, and signal delay. Among them, signal jump is an instantaneous step change in the SAS output angle and speed; signal frame loss is the periodic loss of CAN bus SAS messages; signal out-of-tolerance is the SAS output signal exceeding the sensor's rated output range; and signal delay is the superposition of fixed or variable transmission delay during the SAS message transmission process.

Images