Dual-electric brake power-assisted system for heavy-duty vehicle and coordinated control method of dual-electric brake power-assisted system

Abstract

The invention discloses a dual-electric brake boosting system for a heavy-duty vehicle and a coordinated control method of the dual-electric brake boosting system, and relates to the technical field of vehicle brake control. The invention discloses a dual-electric brake power-assisted system for a heavy-duty vehicle and a coordinated control method thereof. The coordinated control method comprises the following steps that S1, a multi-source sensor synchronously collects and preprocesses wheel speed, current, suspension and pedal signals; s2, estimating the load in real time through Kalman filtering, and outputting a dynamic load coefficient to guide distribution; s3, triggering interruption when the pedal force change rate exceeds the threshold value, and directly outputting the maximum braking force; s4, dynamically distributing the braking force, predicting the temperature rise, and adjusting the output weights of the double boosters; s5, outputting a digital twinborn verification theory, and starting a compensation algorithm after deviation overrun; s6, encrypting an instruction by a TSN protocol, and ensuring low-delay transmission and communication security; s7, a closed-loop optimization link is formed based on the actual braking data calibration model; the method has the beneficial effect that the real-time performance, safety and reliability of the braking system are comprehensively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of vehicle brake control, and in particular to a dual electric brake booster system for a heavy-load vehicle and a coordinated control method thereof. Background Art

[0002] Heavy-loaded vehicles (such as trucks, buses, construction machinery, etc.) have extremely high requirements for the safety, response speed and stability of the braking system due to their heavy loads and complex working conditions. Traditional hydraulic braking systems rely on mechanical structures, and have problems such as response delays, uneven braking force distribution, and easy overheating; and although the existing electric braking system has improved in energy saving and controllability, it still has significant deficiencies in dynamic load adaptation, emergency braking priority processing, temperature balance management and communication security. Inaccurate dynamic load estimation: Traditional methods rely on static parameters or a single sensor, and cannot adapt to load changes in real time, resulting in braking force distribution deviations; emergency braking response lag: conventional systems need to use complex algorithms to optimize the process, and cannot quickly output the maximum braking force in an emergency; temperature management lag: lack of forward-looking temperature rise prediction, easy to cause brake failure due to local overheating; communication security risks: CAN bus protocol is vulnerable to attack, and multi-node communication latency is high, affecting control real-time performance.

[0003] The dual electric brake assist system and coordinated control method thereof of the present invention systematically solve the above-mentioned problems through multi-source data fusion, intelligent algorithm collaboration and high-security communication protocol, and significantly improve the braking performance and reliability of heavy-loaded vehicles. Summary of the invention

[0004] The present invention provides a dual electric brake assist system for heavy-loaded vehicles and a coordinated control method thereof. Through multi-source sensor data fusion, dynamic load estimation, temperature rise prediction, emergency brake priority response, digital twin verification and high-security communication protocol, intelligent distribution of braking force, temperature balance management, fault-tolerant control and closed-loop optimization are realized, thereby comprehensively improving the real-time, safety and reliability of the braking system.

[0005] Technical solution: A dual electric brake booster system for heavy-duty vehicles and a coordinated control method thereof, comprising the following steps:

[0006] S1. Synchronously collect data from wheel speed sensors, motor current sensors, and suspension displacement sensors, as well as driver's brake pedal travel and pedal force signals through the vehicle-mounted CAN bus and hard-wire signals as a unified input source for subsequent steps;

[0007] S2. Input the wheel speed change rate, motor current fluctuation and suspension displacement data obtained in S1 into the pre-trained Kalman filter model, estimate the vehicle load in real time and output the dynamic load coefficient, which is used for braking force distribution and temperature prediction in subsequent steps;

[0008] S3. Based on the driver's pedaling force signal in S1, the pedaling force change rate is calculated. When the change rate exceeds the preset threshold, the highest priority interrupt is triggered, the S4-S5 algorithm optimization and verification process is skipped, and the maximum braking force is output directly through the S6 encrypted instruction to S7; if the interrupt is not triggered, it enters S4;

[0009] S4. Combine the dynamic load coefficient output by S2 with the real-time motor data in S1, calculate the target braking force ratio of the dual electric boosters through the load distribution formula, input the historical temperature and current data into the LSTM neural network, predict the temperature rise trend in the next 10 seconds, and dynamically adjust the output weight of the master / slave boosters to ensure temperature balance;

[0010] S5. Based on the target braking force ratio calculated in S4, the digital twin model is driven to generate a theoretical output value, and compared with the actual output force of the dual booster in S1 in real time; if the deviation exceeds 15%, the sensor or motor is judged to be faulty, and the reinforcement learning compensation algorithm is started to correct the target braking force command to maintain the braking force not less than 80% of the theoretical value;

[0011] S6. Encapsulate the braking force command verified by S5 through the time-sensitive network TSN protocol to ensure that the transmission delay of the dual booster control signal is less than 1ms, and use the national secret SM4 algorithm to encrypt the command to prevent CAN bus attacks;

[0012] S7. Send the encrypted braking force command of S6 to the dual electric boosters for execution. At the same time, according to the wheel speed change rate of S1 and the actual braking distance, reversely calibrate the Kalman filter model of S2 and the LSTM temperature rise prediction parameters of S4 to form a closed-loop optimization link.

[0013] Preferably, the S1 specifically comprises the following steps:

[0014] S1-1. Wheel speed data collection and preprocessing: The wheel speed is collected by the wheel speed sensor, and the change in wheel speed per unit time is calculated to provide kinematic basic data for the dynamic load estimation in step S2;

[0015] S1-2. Dual motor current synchronous acquisition: obtain the real-time working current of the dual booster motor through the motor current sensor, including the real-time working current of the first motor and the real-time working current of the second motor, and the current data is used for load distribution and temperature rise prediction in step S4;

[0016] S1-3. Suspension deformation data fusion: The suspension deformation of each axle is measured by the suspension displacement sensor, and the average deformation value is calculated as an indirect load signal, which is input together with the wheel speed change in step S1-1 into the load estimation model in step S2;

[0017] S1-4. Multi-source data hardware synchronization: Use hardware synchronization trigger mechanism to align the acquisition timestamps of wheel speed, current, and suspension data to ensure that the time error is less than one millisecond and eliminate the timing deviation of multi-sensor data to ensure the consistency of algorithm input in steps S2 to S4;

[0018] S1-5. Pedal signal noise reduction processing: The pedal travel signal is subjected to sliding average filtering to eliminate noise interference caused by mechanical vibration. The purified pedal signal is used for interruption determination and braking force mapping in step S3.

[0019] Preferably, the S2 specifically comprises the following steps:

[0020] S2-1. Kalman filter modeling: Construct the Kalman filter state equation, take the suspension deformation variable in step S1-3 as the observed variable and the vehicle load as the hidden state variable, and establish the core model of dynamic load estimation;

[0021] S2-2. Acceleration back-calculation correction: Based on the motor current fluctuation in step S1-2, the vehicle acceleration is back-calculated, and the load estimation value is corrected in combination with Newton's second law to improve the estimation accuracy of the model in step S2-1;

[0022] S2-3. Online learning mechanism triggering: When the load estimation value exceeds 20% of the historical record range for five consecutive seconds, the model online learning function is triggered to dynamically adapt to the load mutation scenario and update the Kalman filter gain matrix to prevent model failure;

[0023] S2-4. Standardized output of dynamic coefficient: Output the dynamic load coefficient, which is calculated by the formula of the difference between the estimated load and the empty mass divided by the difference between the maximum allowable load and the empty mass, as the core parameter of the load distribution in step S4.

[0024] Preferably, the S3 specifically includes the following steps:

[0025] S3-1. Dynamic monitoring of pedal force: Based on the pedal signal filtered in step S1-5, the rate of change of the pedal force per unit time is calculated, using the first-order difference method and a sampling window of one hundred milliseconds to provide input for emergency braking determination;

[0026] S3-2. Emergency braking double determination: If the pedal force change rate is greater than or equal to 100 Newtons per second and the pedal stroke exceeds 50%, it is determined as emergency braking and triggers the highest priority interrupt, and the algorithm optimization process of steps S4 to S5 is skipped directly;

[0027] S3-3. Interruption event recording and warning: After the interrupt is triggered, the current control parameters are saved to the non-volatile memory, and the red warning light on the instrument panel is lit to provide data support for fault backtracking and warn the driver that the system has entered manual mode;

[0028] S3-4. Manual mode switching and authority transfer: In manual mode, all algorithm optimization modules are disabled, and the pedal travel and braking force are mapped into a linear relationship to ensure the driver's direct control and connect the braking force output of step S7.

[0029] Preferably, the S4 specifically comprises the following steps:

[0030] S4-1. Dynamic load distribution calculation: Calculate the dual booster target force through the load distribution formula, the dynamic load coefficient in the formula comes from step S2-4, and the temperature weight factor is dynamically adjusted by the temperature rise prediction result of step S4-2 to achieve optimal distribution of braking force under heavy load scenarios;

[0031] S4-2. LSTM temperature rise prediction modeling: Input the motor current and temperature time series data of the first thirty seconds collected in step S1-2 into the long short-term memory network to predict the temperature value in the next ten seconds to guide the thermal management strategy of step S4-3;

[0032] S4-3. Overheat protection strategy activation: When the predicted temperature exceeds 80% of the maximum allowable temperature of the motor, the output weight of the high-load side is dynamically reduced and fed back to the formula of step S4-1 to prevent overheating faults;

[0033] S4-4. Dynamic switching of master-slave roles: The master-slave booster roles are switched every ten seconds based on the predicted temperature in step S4-2, with the side with lower temperature being given priority to bear the main load, thus achieving life balance of the dual systems.

[0034] Preferably, the S5 specifically includes the following steps:

[0035] S5-1. Real-time verification of digital twins: Based on the target braking force ratio of step S4-1, the digital twin model is driven to generate a theoretical output value, wherein the model includes the electromagnetic characteristic equation of the motor and is compared with the actual output force in real time;

[0036] S5-2. Reinforcement learning fault-tolerant compensation: If the deviation between the actual output force and the theoretical value exceeds 15% for three seconds, it is judged as a sensor or motor failure and the reinforcement learning compensation algorithm is started to dynamically adjust the current step size to maintain the braking force;

[0037] S5-3. Output limit safety strategy: Limit the output of a single booster to no more than 85% of the maximum value during compensation, ensure that the combined force of the dual systems is not less than 80% of the theoretical value, and ensure the braking safety of step S7;

[0038] S5-4. End-to-end command transmission: The modified braking force command is transmitted to the encryption module in step S6 to form an end-to-end fault-tolerant control link.

[0039] Preferably, the S6 specifically comprises the following steps:

[0040] S6-1.TSN network priority configuration: Assign the highest priority queue of the time-sensitive network to the braking force instruction of step S5-4, and reserve bandwidth of not less than 30% to ensure the real-time requirements of step S7;

[0041] S6-2. SM4 dynamic encryption implementation: Use the CTR mode encryption instruction of the national secret SM4 algorithm, bind the initialization vector to the vehicle VIN code to prevent replay attacks, and ensure the communication security of steps S5 to S7;

[0042] S6-3. Controller access whitelist: Set the whitelist MAC address on the ECU side, allowing only the dual booster controller to receive instructions, isolating the interference of unauthorized devices on the control process of steps S4 to S5;

[0043] S6-4. CRC check retransmission mechanism: Attach a CRC-32 check code to each frame of instructions. When the check fails, start the retransmission mechanism within two milliseconds to ensure the integrity of the instructions executed in step S7.

[0044] Preferably, the S7 specifically includes the following steps:

[0045] S7-1. Braking distance calibration algorithm: Calculate the actual braking distance based on the wheel speed data of step S1-1, and compare the Kalman filter model parameters of step S2 with the theoretical value to improve the dynamic load estimation accuracy;

[0046] S7-2. Adaptive adjustment of model parameters: When the deviation between the actual braking distance and the theoretical value exceeds 10%, the update step size of the noise covariance matrix of the Kalman filter process is increased to dynamically adapt to complex working conditions;

[0047] S7-3.LSTM continuous learning mechanism: fine-tune the long short-term memory network parameters of step S4-2 every 24 hours, retain the data of the last seven days to optimize the temperature rise prediction model, and feed back to the thermal management strategy of step S4;

[0048] S7-4. Cloud parameter synchronization and closed-loop update: The calibrated parameters are synchronized to the cloud digital twin model through OTA, forming a continuous learning closed loop of fault-tolerant control in step S5.

[0049] Compared with the prior art, the advantages of the present invention are:

[0050] (1) Dynamic load and temperature collaborative optimization: Combine Kalman filtering to estimate load in real time, and use LSTM neural network to predict future temperature rise trends, dynamically adjust the dual booster output weights, and achieve optimal braking force distribution under load-temperature joint constraints, avoiding local overheating and extending system life;

[0051] (2) Intelligent priority for emergency braking: Based on the dual judgment mechanism of pedal force change rate and travel, the highest priority interrupt is triggered, skipping the optimization process and directly outputting the maximum braking force, ensuring the response speed in emergency scenarios is less than 100ms, significantly improving safety;

[0052] (3) Fault-tolerant control and secure communication: Use a digital twin model to verify output force deviation in real time, and use a reinforcement learning compensation algorithm to maintain braking force stability; combine the TSN protocol (latency <1ms) with the national SM4 encryption technology to ensure the real-time and anti-attack capabilities of command transmission; DETAILED DESCRIPTION

[0053] Example

[0054] A dual electric brake booster system for a heavy-duty vehicle and a coordinated control method thereof, characterized in that it comprises the following steps:

[0055] S1. Synchronously collect data from wheel speed sensors, motor current sensors, and suspension displacement sensors, as well as driver's brake pedal travel and pedal force signals through the vehicle-mounted CAN bus and hard-wire signals as a unified input source for subsequent steps;

[0056] The S1 specifically includes the following steps:

[0057] S1-1. Wheel speed data collection and preprocessing: The wheel speed is collected by the wheel speed sensor, and the change in wheel speed per unit time is calculated to provide kinematic basic data for the dynamic load estimation in step S2;

[0058] S1-2. Dual motor current synchronous acquisition: obtain the real-time working current of the dual booster motor through the motor current sensor, including the real-time working current of the first motor and the real-time working current of the second motor, and the current data is used for load distribution and temperature rise prediction in step S4;

[0059] S1-3. Suspension deformation data fusion: The suspension deformation of each axle is measured by the suspension displacement sensor, and the average deformation value is calculated as an indirect load signal, which is input together with the wheel speed change in step S1-1 into the load estimation model in step S2;

[0060] S1-4. Multi-source data hardware synchronization: Use hardware synchronization trigger mechanism to align the acquisition timestamps of wheel speed, current, and suspension data to ensure that the time error is less than one millisecond and eliminate the timing deviation of multi-sensor data to ensure the consistency of algorithm inputs in steps S2 to S4. Specifically, it includes the following contents:

[0061] S1-4-1. Hardware synchronization trigger protocol: PTP (Precision Time Protocol) is used to synchronize the local clock of each sensor node, and the master clock signal is broadcast through the vehicle Ethernet to ensure that the clock deviation of each node is ≤50μs; at the beginning of each acquisition cycle (for example, every 10ms), the central controller sends a synchronization pulse signal to all sensors to force trigger synchronous data sampling;

[0062] S1-4-2. Timestamp calibration mechanism: A hardware timestamp unit (HTSU) is configured for each sensor, and a local timestamp is immediately stamped after data collection is completed. The sensor's internal processing delay is compensated through the FPGA logic circuit (the compensation value is pre-written through laboratory calibration). After the central controller receives the data, it aligns the timestamp based on the PTP protocol, and performs interpolation or discarding processing on data packets with a delay of more than 1ms to avoid confusion in the algorithm input timing.

[0063] S1-4-3. Abnormal synchronization detection and recovery: Real-time monitoring of the clock offset of each sensor. If a node has a clock offset of more than 100μs for three consecutive times, it is judged as clock desynchronization, and the backup sensor is switched to trigger the whitelist isolation mechanism of S6-3. After the desynchronized node is recovered, progressive clock calibration is performed: 1μs is adjusted every 10ms until the deviation is ≤50μs and then it is re-integrated into the synchronization network.

[0064] S1-4-4. Redundant data buffer design: set up double buffer queues in the central controller, the main queue stores synchronous data, and the backup queue stores asynchronous data. When the main queue data is missing, the backup queue data is called for linear prediction filling to ensure the continuity of the algorithm input;

[0065] S1-5. Pedal signal noise reduction processing: The pedal stroke signal is subjected to sliding average filtering to eliminate noise interference caused by mechanical vibration. The purified pedal signal is used for interrupt determination and braking force mapping in step S3;

[0066] S2. Input the wheel speed change rate, motor current fluctuation and suspension displacement data obtained in S1 into the pre-trained Kalman filter model, estimate the vehicle load in real time and output the dynamic load coefficient, which is used for braking force distribution and temperature prediction in subsequent steps;

[0067] The S2 specifically includes the following steps:

[0068] S2-1. Kalman filter modeling: Construct the Kalman filter state equation, take the suspension deformation variable in step S1-3 as the observed variable and the vehicle load as the hidden state variable, and establish the core model of dynamic load estimation;

[0069] S2-2. Acceleration back-calculation correction: Based on the motor current fluctuation in step S1-2, the vehicle acceleration is back-calculated, and the load estimation value is corrected in combination with Newton's second law to improve the estimation accuracy of the model in step S2-1. Specifically, it includes the following contents:

[0070] S2-2-1. Current-acceleration conversion model: Establish the relationship equation between motor current and torque: Torque = motor constant × current, and calculate the actual driving force based on wheel radius and transmission efficiency; Based on Newton's second law (F = ma), combined with the total mass of the vehicle (empty mass + estimated load), reverse the acceleration

[0071] S2-2-2. Acceleration correction algorithm: Compare the reverse thrust acceleration with the actual acceleration of the wheel speed sensor. If the deviation exceeds 5%, the load estimation value is corrected: Deviation direction determination: If the reverse thrust acceleration > the actual value, the load estimation value is adjusted upward (+Δm); otherwise, it is adjusted downward (-Δm); correction amount Δm = deviation percentage × current load estimation value × 0.1 (empirical coefficient);

[0072] S2-2-3. Dynamic smoothing filter: The corrected load estimate value needs to be smoothed by a first-order low-pass filter (time constant τ = 2 seconds) to avoid sudden changes that may cause model oscillation;

[0073] S2-3. Online learning mechanism triggering: When the load estimation value exceeds 20% of the historical record range for five consecutive seconds, the model online learning function is triggered to dynamically adapt to the load mutation scenario and update the Kalman filter gain matrix to prevent model failure;

[0074] S2-4. Dynamic coefficient standardization output: Output the dynamic load coefficient, which is calculated by the formula of the difference between the estimated load and the empty mass divided by the difference between the maximum allowable load and the empty mass, as the core parameter of the load distribution in step S4

[0075] S3. Based on the driver's pedaling force signal in S1, the pedaling force change rate is calculated. When the change rate exceeds the preset threshold, the highest priority interrupt is triggered, the S4-S5 algorithm optimization and verification process is skipped, and the maximum braking force is output directly through the S6 encrypted instruction to S7; if the interrupt is not triggered, it enters S4;

[0076] The S3 specifically includes the following steps:

[0077] S3-1. Dynamic monitoring of pedal force: Based on the pedal signal filtered in step S1-5, the rate of change of the pedal force per unit time is calculated, using the first-order difference method and a sampling window of one hundred milliseconds to provide input for emergency braking determination;

[0078] S3-2. Emergency braking double determination: If the pedal force change rate is greater than or equal to 100 Newtons per second and the pedal stroke exceeds 50%, it is determined as emergency braking and triggers the highest priority interrupt, and the algorithm optimization process of steps S4 to S5 is skipped directly;

[0079] S3-3. Interruption event recording and warning: After the interrupt is triggered, the current control parameters are saved to the non-volatile memory, and the red warning light on the instrument panel is lit to provide data support for fault backtracking and warn the driver that the system has entered manual mode;

[0080] S3-4. Manual mode switching and authority transfer: In manual mode, all algorithm optimization modules are disabled, and the pedal travel and braking force are mapped into a linear relationship to ensure the driver's direct control and connect the braking force output of step S7, which specifically includes the following contents:

[0081] S3-4-1. Security verification for authority transfer: Before switching to manual mode, the driver's identity must be verified (such as fingerprint recognition or specific brake pedal pedaling mode) to prevent false triggering; during the transfer process, the output weight of the algorithm optimization module is gradually reduced (reduced by 10% every 100ms), and the manual control weight is linearly increased to ensure a smooth transition of the braking force;

[0082] S3-4-2. Linear mapping parameter design: The braking force range corresponding to the pedal stroke of 0% to 100% is 30% to 100% of the theoretical maximum value, and the slope is adjustable to adapt to different driver habits; set the dead zone (no braking force output when the stroke is 0% to 5%) and the saturation zone (the maximum braking force is locked when the stroke exceeds 95%) to avoid misoperation;

[0083] S3-4-1. Manual mode monitoring and switching back mechanism: If the braking force fluctuation exceeds ±10% for 5 seconds in manual mode, the driver's control is considered abnormal, and the system automatically switches back to the algorithm mode and triggers an alarm;

[0084] S4. Combine the dynamic load coefficient output by S2 with the real-time motor data in S1, calculate the target braking force ratio of the dual electric boosters through the load distribution formula, input the historical temperature and current data into the LSTM neural network, predict the temperature rise trend in the next 10 seconds, and dynamically adjust the output weight of the master / slave boosters to ensure temperature balance;

[0085] The S4 specifically comprises the following steps:

[0086] S4-1. Dynamic load distribution calculation: Calculate the dual booster target force through the load distribution formula, the dynamic load coefficient in the formula comes from step S2-4, and the temperature weight factor is dynamically adjusted by the temperature rise prediction result of step S4-2 to achieve optimal distribution of braking force under heavy load scenarios;

[0087] S4-2. LSTM temperature rise prediction modeling: Input the motor current and temperature time series data of the first thirty seconds collected in step S1-2 into the long short-term memory network to predict the temperature value in the next ten seconds to guide the thermal management strategy of step S4-3;

[0088] S4-3. Overheat protection strategy activation: When the predicted temperature exceeds 80% of the maximum allowable temperature of the motor, the output weight of the high-load side is dynamically reduced and fed back to the formula of step S4-1 to prevent overheating faults. Specifically, it includes the following contents:

[0089] S4-3-1. Dynamic adjustment of temperature threshold: Correct the upper limit of the motor's allowable temperature according to the ambient temperature (obtained through the vehicle-mounted temperature and humidity sensor): For every 10°C increase in ambient temperature, the upper limit of the allowable temperature is lowered by 5°C; when the predicted temperature exceeds 80% of the dynamic upper limit, overheating protection is triggered;

[0090] S4-3-2. Output weight adjustment algorithm: High load side output weight The lower limit of the weight is 50% of the original value; the adjusted weight is fed back to the load distribution formula in real time: F1=W×F total ,F2=(1-W)×F total ;

[0091] S4-3-3. Cooling system collaborative control: After the overheat protection is triggered, the motor cooling fan is started synchronously (the speed is increased to 80% of the maximum value), and the fan speed is gradually restored after the temperature drops below the threshold;

[0092] S4-4. Dynamic switching of master and slave roles: every ten seconds, the master and slave booster roles are switched based on the predicted temperature in step S4-2, and the side with lower temperature is preferentially selected to bear the main load to achieve life balance of the dual systems, which specifically includes the following contents:

[0093] S4-4-1. Temperature difference threshold determination: The master-slave switching trigger condition is that the predicted temperature difference of the boosters on both sides exceeds 15°C. If the difference does not reach the threshold, the current master-slave allocation is maintained; if the difference exceeds the threshold, the switching process is immediately started;

[0094] S4-4-2. Switching frequency optimization: To avoid mechanical wear caused by frequent switching, the shortest interval between two switches is ten seconds, and the temperature prediction difference within a single switching cycle must exceed the threshold for more than three seconds;

[0095] S4-4-3. Load smooth transition strategy: During the switching process, the output weight of the main booster decreases linearly at a rate of 5% per millisecond, and the output weight of the slave booster increases synchronously to ensure that the total fluctuation of the braking force does not exceed 2% of the theoretical value to avoid brake tremor;

[0096] S4-4-4. Life balance algorithm: record the historical cumulative working time and temperature load of the dual boosters. If the cumulative working time of one side exceeds that of the other side by 20%, it will be switched to the low-load side first to force the dual system losses to be balanced;

[0097] S4-4-5. Abnormal scenario fault tolerance: If the booster temperature sensor on one side fails, the switching logic will reverse the temperature based on the current data by default, and simulate the temperature value in combination with the digital twin model to ensure that the switching function continues to be effective;

[0098] S4-4-6. Dynamic weight compensation mechanism: After switching, the initial output weight of the new main booster is increased by an additional 3% to compensate for the loss of braking force during the transition period, and gradually returns to the target ratio within five seconds;

[0099] S5. Based on the target braking force ratio calculated in S4, the digital twin model is driven to generate a theoretical output value, and compared with the actual output force of the dual booster in S1 in real time; if the deviation exceeds 15%, the sensor or motor is judged to be faulty, and the reinforcement learning compensation algorithm is started to correct the target braking force command to maintain the braking force not less than 80% of the theoretical value;

[0100] The S5 specifically includes the following steps:

[0101] S5-1. Real-time verification of digital twins: Based on the target braking force ratio of step S4-1, the digital twin model is driven to generate a theoretical output value, wherein the model includes the electromagnetic characteristic equation of the motor and is compared with the actual output force in real time;

[0102] S5-2. Reinforcement learning fault-tolerance compensation: If the deviation between the actual output force and the theoretical value exceeds 15% for three seconds, it is judged as a sensor or motor failure and the reinforcement learning compensation algorithm is started to dynamically adjust the current step size to maintain the braking force. Specifically, it includes the following:

[0103] S5-2-1. Reinforcement learning framework design: Adopt the deep deterministic policy gradient (DDPG) algorithm, define the state space as the current deviation value, motor temperature, and historical output weight, the action space as the current step adjustment amount (within ±5%), and the reward function as the weighted sum of the braking force deviation reduction rate and temperature stability; update the policy network parameters every 100ms, and store the data of the last 10 minutes through the experience replay pool to avoid overfitting;

[0104] S5-2-2. Dynamic adjustment strategy of current step length: In the initial compensation stage, the step length is adjusted according to the deviation ratio: when the deviation is 15% to 20%, the step length adjustment amount is 1% / second, when it is 20% to 25%, it is 2% / second, and when it exceeds 25%, the emergency degradation mode is triggered; introduce the inertia attenuation factor, and the step length change after each adjustment is exponentially attenuated (attenuation coefficient 0.9) to avoid overshoot oscillation;

[0105] S5-2-3. Safety constraint mechanism: Real-time monitoring of the temperature of the dual boosters during compensation. If the temperature on one side exceeds the safety threshold (such as 120°C), the step adjustment on that side will be immediately frozen and forced to switch to the backup booster; set the lower limit of the braking force: the braking force after compensation shall not be less than 80% of the theoretical value, and the fluctuation range of the combined force of the dual systems shall be controlled within ±3%;

[0106] S5-2-4. Fault type classification and adaptive compensation: Distinguish sensor drift (low frequency and small deviation) and motor stall (high frequency and large deviation) through the fault feature library, and adopt different compensation strategies accordingly:

[0107] Sensor drift: Prioritize calling the digital twin model to simulate the output force instead of the actual sensor data;

[0108] Motor stall: trigger pulse current shock (lasting 10ms, intensity is 150% of rated current) to try to release mechanical jam;

[0109] S5-3. Output limit safety strategy: Limit the output of a single booster to no more than 85% of the maximum value during compensation, ensure that the combined force of the dual systems is not less than 80% of the theoretical value, and ensure the braking safety of step S7;

[0110] S5-4. End-to-end command transmission: The modified braking force command is transmitted to the encryption module in step S6 to form an end-to-end fault-tolerant control link, which specifically includes the following contents:

[0111] S5-4-1. Instruction encapsulation protocol: adopts the PDU (protocol data unit) format in the AUTOSAR standard, which includes fields such as instruction type, timestamp, checksum, etc. to ensure data parsing consistency; each frame of instruction is attached with a serial number, and the receiving end verifies the continuity of the serial number and detects packet loss or repeated transmission;

[0112] S5-4-2. End-to-end verification mechanism: The sender calculates the command hash value (SHA-256) and embeds it at the end of the command. The receiver verifies the hash value match and requests retransmission if they are inconsistent. Each frame of command must receive a confirmation response within 10ms. If it times out, the CRC retransmission mechanism of S6-4 is triggered.

[0113] S6. Encapsulate the braking force command verified by S5 through the time-sensitive network TSN protocol to ensure that the transmission delay of the dual booster control signal is less than 1ms, and encrypt the command using the national secret SM4 algorithm to prevent CAN bus attacks; S6 specifically includes the following steps:

[0114] S6-1.TSN network priority configuration: Assign the highest priority queue of the time-sensitive network to the braking force instruction of step S5-4, and reserve bandwidth of not less than 30% to ensure the real-time requirements of step S7, specifically including the following contents;

[0115] S6-1-1. Dynamic bandwidth allocation mechanism: Real-time monitoring of TSN network load. If the current bandwidth occupancy rate exceeds 70%, the bandwidth dynamic adjustment algorithm is automatically triggered to compress the bandwidth share of non-critical data (such as log upload) step by step according to priority, ensuring that the reserved bandwidth of the braking force command is always ≥ 30%; the reserved bandwidth is divided into two parts: the basic bandwidth (20%) is used for regular command transmission, and the elastic bandwidth (10%) is used to compensate for network jitter or burst traffic, and the delay fluctuation is suppressed through the traffic shaping algorithm;

[0116] S6-1-2. Multi-queue hierarchical management: Divide the TSN network into four levels of queues:

[0117] Queue 0 (highest priority): braking force command, sensor heartbeat signal;

[0118] Queue 1: Temperature rise prediction model parameter update;

[0119] Queue 2: historical data synchronization; Queue 3: diagnostic information reporting;

[0120] Queue 0 adopts a preemptive scheduling strategy, allowing interruption of low-priority queue transmission to ensure that instruction latency is ≤1ms;

[0121] S6-1-3. Network abnormality fault tolerance strategy: If the transmission delay of three consecutive frames of instructions exceeds 1ms, the redundant link switching mechanism is activated, the instructions are transmitted through the vehicle Ethernet backup channel, and the CRC check retransmission mechanism of S6-4 is triggered; when the network is congested, the "instruction compression" mode is enabled, and the floating point precision of the braking force instruction is reduced from 32 bits to 16 bits, sacrificing some precision to reduce the data volume and ensure real-time performance;

[0122] S6-1-4. Time synchronization calibration: Based on the IEEE 802.1AS protocol, the TSN switch timestamp is periodically synchronized with the vehicle clock source, with a calibration error of ≤10μs, eliminating the impact of multi-node clock drift on timing consistency; a network topology self-check is performed every five minutes to identify link interruptions or switch failures, and the repair path is simulated through a digital twin model;

[0123] S6-2. SM4 dynamic encryption implementation: Use the CTR mode encryption instruction of the national secret SM4 algorithm, bind the initialization vector to the vehicle VIN code to prevent replay attacks, and ensure the communication security of steps S5 to S7;

[0124] S6-3. Controller access whitelist: Set the whitelist MAC address on the ECU side to only allow the dual booster controller to receive instructions, isolating the interference of unauthorized devices on the control process of steps S4 to S5, specifically including the following:

[0125] S6-3-1. Dynamic whitelist update mechanism: The whitelist is dynamically updated through the on-board diagnostic interface (OBD) or OTA remote management platform. New devices need to provide a digital certificate signed by the SM2 algorithm, and the main controller verifies the legitimacy of the certificate before temporarily authorizing access; the whitelist is automatically refreshed every 24 hours, and MAC addresses that have not communicated for more than 72 hours are cleared to prevent zombie devices from occupying resources;

[0126] S6-3-2. MAC address binding and obfuscation: Bind the MAC address of the dual booster controller to the vehicle VIN code, and dynamically generate a virtual MAC (based on the hash chain algorithm) during communication to prevent external devices from obtaining the real address through sniffing; enable the MAC address randomization function and change the virtual MAC every 5 minutes to increase the difficulty of reverse engineering by attackers;

[0127] S6-3-3. Intrusion detection and active defense: Deploy lightweight IDS (intrusion detection system) to monitor abnormal command frequency (such as more than 10 requests within 1 second), immediately block the source MAC address and report to the cloud security center after triggering; return false response data packets (such as fake braking force feedback) to non-whitelisted devices for connection attempts, mislead attackers and record attack characteristics;

[0128] S6-3-4. Physical layer isolation reinforcement: Deploy a hardware firewall between the ECU and the TSN switch to allow only physical port communications with whitelisted MAC addresses, blocking the possibility of physical access by unauthorized devices; use a shielded twisted pair design for the control command bus, and apply a common-mode interference suppression circuit to reduce the risk of bus monitoring;

[0129] S6-4. CRC check retransmission mechanism: CRC-32 check code is attached to each frame instruction. When the check fails, the retransmission mechanism is started within two milliseconds to ensure the integrity of the instruction executed in step S7;

[0130] S7. Send the encrypted braking force command of S6 to the dual electric boosters for execution. Meanwhile, according to the wheel speed change rate and actual braking distance in S1, reversely calibrate the Kalman filter model of S2 and the LSTM temperature rise prediction parameters of S4 to form a closed-loop optimization link.

[0131] The S7 specifically comprises the following steps:

[0132] S7-1. Braking distance calibration algorithm: Calculate the actual braking distance based on the wheel speed data of step S1-1, and compare the Kalman filter model parameters of step S2 with the theoretical value to improve the dynamic load estimation accuracy;

[0133] S7-2. Adaptive adjustment of model parameters: When the deviation between the actual braking distance and the theoretical value exceeds 10%, the update step size of the noise covariance matrix of the Kalman filter process is increased to dynamically adapt to complex working conditions. Specifically, it includes the following contents:

[0134] S7-2-1. Noise covariance adjustment algorithm: The update step size of the process noise covariance matrix Q is ΔQ = deviation percentage × reference Q × 0.5 (reference Q is determined by calibration test); after adjustment, Q new =Q old +ΔQ, while limiting Q new The diagonal elements of do not exceed 200% of the benchmark value to prevent the model from diverging;

[0135] S7-2-2. Real-time verification and rollback: After the parameters are adjusted, the model accuracy is verified using the next three braking data. If the deviation still exceeds 8%, the model is rolled back to the parameters before adjustment and the LSTM model optimization of S7-3 is triggered.

[0136] S7-3.LSTM continuous learning mechanism: fine-tune the long short-term memory network parameters of step S4-2 every 24 hours, retain the data of the last seven days to optimize the temperature rise prediction model, and feed back to the thermal management strategy of step S4;

[0137] S7-4. Cloud parameter synchronization and closed-loop update: The calibrated parameters are synchronized to the cloud digital twin model through OTA, forming a continuous learning closed loop of fault-tolerant control in step S5.

[0138] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments. The above embodiments and descriptions are only preferred examples of the present invention and are not intended to limit the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. A dual electric brake booster system for heavy-duty vehicles and a coordinated control method thereof, characterized in that: The following steps are involved: S1. synchronously collect data from wheel speed sensor, motor current sensor and suspension displacement sensor, as well as driver's brake pedal travel and pedal force signals through the vehicle CAN bus and hard-wire signal, as a unified input source for subsequent steps; S2. Input the wheel speed change rate, motor current fluctuation and suspension displacement data obtained in S1 into the pre-trained Kalman filter model, estimate the vehicle load in real time and output the dynamic load coefficient, which is used for braking force distribution and temperature prediction in subsequent steps; S3. Based on the driver's pedaling force signal in S1, the pedaling force change rate is calculated. When the change rate exceeds the preset threshold, the highest priority interrupt is triggered, the S4-S5 algorithm optimization and verification process is skipped, and the maximum braking force is output directly through the S6 encrypted instruction into S7; If the interrupt is not triggered, enter S4; S4. Combine the dynamic load coefficient output by S2 with the real-time motor data in S1, calculate the target braking force ratio of the dual electric boosters through the load distribution formula, input the historical temperature and current data into the LSTM neural network, predict the temperature rise trend in the next 10 seconds, and dynamically adjust the output weight of the master / slave boosters to ensure temperature balance; S5. Based on the target braking force ratio calculated in S4, the digital twin model is driven to generate a theoretical output value, and compared with the actual output force of the dual booster in S1 in real time; if the deviation exceeds 15%, the sensor or motor is judged to be faulty, and the reinforcement learning compensation algorithm is started to correct the target braking force command to maintain the braking force not less than 80% of the theoretical value; S6. Encapsulate the braking force command verified by S5 through the time-sensitive network TSN protocol to ensure that the transmission delay of the dual booster control signal is less than 1ms, and use the national secret SM4 algorithm to encrypt the command to prevent CAN bus attacks; S7. Send the encrypted braking force command of S6 to the dual electric boosters for execution. At the same time, according to the wheel speed change rate and actual braking distance in S1, reversely calibrate the Kalman filter model of S2 and the LSTM temperature rise prediction parameters of S4 to form a closed-loop optimization link.

2. A dual electric brake booster system for heavy-duty vehicles and a coordinated control method thereof according to claim 1, characterized in that: The S1 specifically includes the following steps: S1-1. Wheel speed data collection and preprocessing: The wheel speed is collected by the wheel speed sensor, and the change in wheel speed per unit time is calculated to provide kinematic basic data for the dynamic load estimation in step S2; S1-2. Dual motor current synchronous acquisition: obtain the real-time working current of the dual booster motor through the motor current sensor, including the real-time working current of the first motor and the real-time working current of the second motor, and the current data is used for load distribution and temperature rise prediction in step S4; S1-3. Suspension deformation data fusion: The suspension deformation of each axle is measured by the suspension displacement sensor, and the average deformation value is calculated as an indirect load signal, which is input together with the wheel speed change in step S1-1 into the load estimation model in step S2; S1-4. Hardware synchronization of multi-source data: Use hardware synchronization trigger mechanism to align the acquisition timestamps of wheel speed, current, and suspension data to ensure that the time error is less than one millisecond and eliminate the timing deviation of multi-sensor data to ensure the consistency of algorithm input in steps S2 to S4; S1-5. Pedal signal noise reduction processing: The pedal travel signal is subjected to sliding average filtering to eliminate noise interference caused by mechanical vibration. The purified pedal signal is used for interruption determination and braking force mapping in step S3.

3. The dual electric brake booster system for heavy-duty vehicles and the coordinated control method thereof according to claim 1, characterized in that: The S2 specifically includes the following steps: S2-1. Kalman filter modeling: Construct the Kalman filter state equation, take the suspension deformation variable in step S1-3 as the observed variable and the vehicle load as the hidden state variable, and establish the core model of dynamic load estimation; S2-2. Acceleration back-calculation correction: Based on the motor current fluctuation in step S1-2, the vehicle acceleration is back-calculated, and the load estimate is corrected in combination with Newton's second law to improve the estimation accuracy of the model in step S2-1; S2-3. Online learning mechanism triggering: When the load estimation value exceeds 20% of the historical record range for five consecutive seconds, the model online learning function is triggered to dynamically adapt to the load mutation scenario and update the Kalman filter gain matrix to prevent model failure; S2-4. Standardized output of dynamic coefficient: Output the dynamic load coefficient, which is calculated by the formula of the difference between the estimated load and the empty mass divided by the difference between the maximum allowable load and the empty mass, as the core parameter of the load distribution in step S4.

4. The dual electric brake booster system for heavy-duty vehicles and the coordinated control method thereof according to claim 1, characterized in that: The S3 specifically includes the following steps: S3-1. Dynamic monitoring of pedal force: Based on the pedal signal filtered in step S1-5, the rate of change of the pedal force per unit time is calculated, using the first-order difference method and a sampling window of one hundred milliseconds to provide input for emergency braking determination; S3-2. Emergency braking double determination: If the pedal force change rate is greater than or equal to 100 Newtons per second and the pedal stroke exceeds 50%, it is determined as emergency braking and triggers the highest priority interrupt, directly skipping steps S4 to S5 of the algorithm optimization process; S3-3. Interruption event recording and warning: After the interrupt is triggered, the current control parameters are saved to the non-volatile memory, and the red warning light on the instrument panel is lit to provide data support for fault backtracking and warn the driver that the system has entered the manual mode; S3-4. Manual mode switching and authority transfer: In manual mode, all algorithm optimization modules are disabled, and the pedal travel and braking force are mapped into a linear relationship to ensure direct control by the driver and connect the braking force output of step S7.

5. The dual electric brake booster system for heavy-duty vehicles and the coordinated control method thereof according to claim 1, characterized in that: The S4 specifically comprises the following steps: S4-1. Dynamic load distribution calculation: Calculate the dual booster target force through the load distribution formula, the dynamic load coefficient in the formula comes from step S2-4, and the temperature weight factor is dynamically adjusted by the temperature rise prediction result of step S4-2 to achieve optimal distribution of braking force under heavy load scenarios; S4-2. LSTM temperature rise prediction modeling: Input the motor current and temperature time series data of the first thirty seconds collected in step S1-2 into the long short-term memory network to predict the temperature value in the next ten seconds to guide the thermal management strategy of step S4-3; S4-3. Overheat protection strategy activation: When the predicted temperature exceeds 80% of the maximum allowable temperature of the motor, the output weight of the high-load side is dynamically reduced and fed back to the formula of step S4-1 to prevent overheating faults; S4-4. Dynamic switching of master-slave roles: The master-slave booster roles are switched every ten seconds based on the predicted temperature in step S4-2, with the side with lower temperature being given priority to bear the main load, thus achieving life balance of the dual systems.

6. A dual electric brake booster system for heavy-duty vehicles and a coordinated control method thereof according to claim 1, characterized in that: The S5 specifically includes the following steps: S5-1. Real-time verification of digital twins: Based on the target braking force ratio of step S4-1, the digital twin model is driven to generate a theoretical output value, wherein the model includes the electromagnetic characteristic equation of the motor and is compared with the actual output force in real time; S5-2. Reinforcement learning fault-tolerant compensation: If the deviation between the actual output force and the theoretical value exceeds 15% for three seconds, it is judged as a sensor or motor failure and the reinforcement learning compensation algorithm is started to dynamically adjust the current step size to maintain the braking force; S5-3. Output limit safety strategy: Limit the output of a single booster to no more than 85% of the maximum value during compensation, ensure that the combined force of the dual systems is not less than 80% of the theoretical value, and ensure the braking safety of step S7; S5-4. End-to-end command transmission: The modified braking force command is transmitted to the encryption module in step S6 to form an end-to-end fault-tolerant control link.

7. A dual electric brake booster system for heavy-duty vehicles and a coordinated control method thereof according to claim 1, characterized in that: The S6 specifically comprises the following steps: S6-1.TSN network priority configuration: Assign the highest priority queue of the time-sensitive network to the braking force instruction of step S5-4, and reserve bandwidth of not less than 30% to ensure the real-time requirements of step S7; S6-2. SM4 dynamic encryption implementation: Use the CTR mode encryption instruction of the national secret SM4 algorithm, bind the initialization vector to the vehicle VIN code to prevent replay attacks, and ensure the communication security of steps S5 to S7; S6-3. Controller access whitelist: Set the whitelist MAC address on the ECU side, allowing only the dual booster controller to receive instructions, isolating the interference of unauthorized devices on the control process of steps S4 to S5; S6-4. CRC check retransmission mechanism: Attach a CRC-32 check code to each frame of instructions. When the check fails, start the retransmission mechanism within two milliseconds to ensure the integrity of the instructions executed in step S7.

8. The dual electric brake booster system for heavy-duty vehicles and the coordinated control method thereof according to claim 1, characterized in that: The S7 specifically comprises the following steps: S7-1. Braking distance calibration algorithm: Calculate the actual braking distance based on the wheel speed data of step S1-1, and compare the Kalman filter model parameters of step S2 with the theoretical value to improve the dynamic load estimation accuracy; S7-2. Adaptive adjustment of model parameters: When the deviation between the actual braking distance and the theoretical value exceeds 10%, the update step size of the noise covariance matrix of the Kalman filter process is increased to dynamically adapt to complex working conditions; S7-3.LSTM continuous learning mechanism: fine-tune the long short-term memory network parameters of step S4-2 every 24 hours, retain the data of the last seven days to optimize the temperature rise prediction model, and feed back to the thermal management strategy of step S4; S7-4. Cloud parameter synchronization and closed-loop update: The calibrated parameters are synchronized to the cloud digital twin model through OTA, forming a continuous learning closed loop of fault-tolerant control in step S5.