Suspension adjustment method based on air spring damping device of vehicle intelligent control system

Abstract

The application provides a suspension adjustment method based on an air spring damping device of a vehicle intelligent control system, comprising: an intelligent control unit receiving a pretreatment signal, combining a preset multi-mode control strategy to identify a working condition result of a vehicle operation condition; the intelligent control unit calling a vehicle dynamics model, combining the pretreatment signal and the working condition result to predict a key parameter of a suspension system to obtain a predicted parameter; a fuzzy neural network controller receiving the pretreatment signal, the working condition result and the predicted parameter to output an adjustment instruction, the adjustment instruction comprising air spring inflation and deflation control instructions and shock absorber damping coefficient adjustment instructions; an actuator receiving the adjustment instruction to control air spring inflation and deflation and shock absorber damping coefficient through a high-speed on-off electromagnetic valve and a pneumatic actuator; a sensor collects suspension system feedback parameters, compares the feedback parameters with the predicted parameters, and corrects the adjustment instruction through the intelligent control unit.

Description

Technical Field

[0001] This invention relates to the field of information technology, and in particular to a suspension adjustment method based on an air spring damping device in a vehicle intelligent control system. Background Technology

[0002] As a crucial component of automotive engineering, the vehicle suspension system directly impacts driving stability and safety. Its research and optimization hold undeniable value in enhancing vehicle performance and user experience. The suspension system must balance vehicle comfort and handling stability under various road conditions and driving circumstances, placing extremely high demands on technological innovation. Especially when facing complex environments, the dynamic adaptability of the suspension system becomes paramount.

[0003] Currently, although various suspension adjustment methods exist on the market, these methods often suffer from slow response speed and insufficient adjustment precision when dealing with complex driving scenarios. Especially during vehicle operation, factors such as road conditions, vehicle speed, and load distribution change rapidly, making it difficult for existing technologies to accurately judge and make corresponding adjustments in a short time. As a result, when the vehicle is on bumpy roads or making sharp turns, the suspension system cannot effectively mitigate impacts or maintain stability, affecting the driving experience and safety.

[0004] A deeper technical challenge lies in achieving the suspension system's ability to rapidly and dynamically adapt to different operating conditions. This adaptability goes beyond simply adjusting stiffness or height; it requires precisely controlling the performance parameters of suspension components based on changes in vehicle condition and the external environment within a very short time. For example, the inflation / deflation volume of air springs and the damping of shock absorbers need to be matched to the current driving conditions in real time; otherwise, there will be adjustment lag or over-adjustment. For instance, when a vehicle suddenly enters a bumpy section of road from a smooth surface, if the suspension system cannot quickly increase its damping effect, the vehicle will experience severe jolts; conversely, if the response is too slow or the adjustment is inappropriate, the vehicle may become too stiff on subsequent smooth sections, affecting comfort. This lack of rapid dynamic adaptation is the core obstacle to optimizing the performance of the suspension system under complex conditions.

[0005] Therefore, how to achieve rapid and precise adjustment of the suspension system in response to rapidly changing road conditions and driving states during vehicle operation, in order to balance comfort and stability, has become a key problem that this study urgently needs to solve. Summary of the Invention

[0006] This invention provides a suspension adjustment method based on an air spring damping device in a vehicle intelligent control system, mainly comprising: Sensors collect real-time vehicle driving status parameters, including vehicle speed, longitudinal and lateral acceleration, steering angle, road surface excitation, vehicle posture, and load changes. The collected signals are preprocessed to obtain a preprocessed signal, which is then transmitted to the intelligent control unit. The intelligent control unit receives the preprocessed signal and, in conjunction with a preset multi-mode control strategy, identifies the vehicle's operating conditions to obtain operating condition results. The intelligent control unit then uses a vehicle dynamics model, combined with the preprocessed signal and the operating condition results, to predict key parameters of the suspension system, obtaining predicted parameters. A fuzzy neural network controller receives the preprocessed signal, the operating condition results, and the predicted parameters, and outputs adjustment commands, including air spring inflation / deflation control commands and shock absorber damping coefficient adjustment commands. The actuator receives the adjustment commands and controls the air spring inflation / deflation and shock absorber damping coefficient via a high-speed switching solenoid valve and a pneumatic actuator. Sensors collect suspension system feedback parameters, compare these feedback parameters with the predicted parameters, and the intelligent control unit corrects the adjustment commands.

[0007] Furthermore, the sensors collect vehicle driving status parameters in real time, including vehicle speed, longitudinal and lateral acceleration, steering angle, road surface excitation, vehicle body posture, and load changes. The collected signals are preprocessed, including: the vehicle speed sensor collects vehicle speed, the acceleration sensor collects longitudinal and lateral acceleration, the steering angle sensor collects steering angle, the road surface sensor collects road surface excitation, the posture sensor collects vehicle body posture, and the load sensor collects load changes. The collected signals are then filtered, denoised, normalized, and abnormal data is removed using a Kalman filter algorithm to obtain the preprocessed signal, which is then transmitted to the intelligent control unit.

[0008] Furthermore, the intelligent control unit receives the preprocessed signal and combines it with a preset multi-mode control strategy to identify the vehicle's operating conditions, including: extracting feature values ​​of each parameter in the preprocessed signal, including vehicle speed threshold, acceleration change rate, steering angle range, and road excitation intensity; performing pattern matching between the feature values ​​and a preset operating condition feature library to determine the operating condition result; and dynamically adjusting the identification threshold according to load changes.

[0009] Furthermore, the intelligent control unit calls the vehicle dynamics model and combines the preprocessed signal and the working condition results to predict the key parameters of the suspension system, including: the vehicle dynamics model covers the dynamic characteristics of the vehicle body, wheels, air springs, and shock absorbers, taking into account nonlinear stiffness and damping characteristics and load effects; the radial basis function neural network compensates the model parameters to predict the target working air pressure shock absorber target damping coefficient, the trend of vehicle body attitude change, and subsequent changes in road surface excitation to obtain the predicted parameters.

[0010] Furthermore, the fuzzy neural network controller receives the preprocessed signal, the operating condition result, and the predicted parameters, and outputs adjustment instructions. The adjustment instructions include air spring inflation / deflation control instructions and shock absorber damping coefficient adjustment instructions, comprising: processing the input through fuzzy inference in combination with preset fuzzy rules; optimizing the fuzzy rules and network weights through neural network self-learning, and outputting the air spring inflation / deflation control instructions and the shock absorber damping coefficient adjustment instructions according to the operating condition requirements.

[0011] Furthermore, the actuator receives the adjustment command and controls the air spring inflation / deflation and the damper damping coefficient through a high-speed switching solenoid valve and a pneumatic actuator, including: the high-speed switching solenoid valve uses pulse width modulation technology to control the inflation / deflation rate and time according to the air spring inflation / deflation control command to adjust the air spring working pressure; the pneumatic actuator adjusts the damping force of the variable damping damper according to the damper damping coefficient adjustment command; when the air pressure sensor monitors that the air pressure in the air tank is lower than a preset threshold, the air compressor starts to replenish the air, which is then processed by a filter and dryer.

[0012] Furthermore, the sensor collects feedback parameters from the suspension system, compares the feedback parameters with the predicted parameters, and corrects the adjustment command through the intelligent control unit. This includes: the air pressure sensor collects the actual working air pressure of the air spring, collects the actual damping coefficient of the shock absorber, the actual value of the vehicle body posture, and the tire adhesion coefficient to obtain the feedback parameters; the deviation between the feedback parameters and the predicted parameters is calculated, and the adjustment command is corrected by a fuzzy neural network combined with a PID algorithm to adjust the action parameters of the actuator.

[0013] Furthermore, the vehicle dynamics model combines the preprocessed signal and the operating condition results to predict key parameters of the suspension system, including: the vehicle dynamics model establishes a description of the dynamic response law of the suspension system based on classical mechanics methods; and provides advance adjustment command by predicting and obtaining the dynamic demand of the suspension in advance.

[0014] The technical solutions provided by the embodiments of the present invention may include the following beneficial effects: This invention discloses an intelligent vehicle suspension adjustment method based on a multi-mode control strategy. By fusing real-time vehicle driving state parameters collected by sensors, including vehicle speed, acceleration, steering angle, and road surface excitation, and combining multi-mode operating condition recognition and vehicle dynamics model prediction by an intelligent control unit, it solves the problems of untimely dynamic response and insufficient adjustment precision of the suspension system under complex operating conditions. This invention outputs precise air spring inflation / deflation and shock absorber damping coefficient adjustment commands through a fuzzy neural network controller, and achieves rapid response using high-speed switching solenoid valves and pneumatic actuators. Simultaneously, it ensures the accuracy of the adjustment commands by correcting the deviation between feedback parameters and predicted parameters. This invention enables the suspension system to dynamically adapt to different operating conditions, improving vehicle ride stability and safety. Especially under complex road conditions and varying load scenarios, it significantly optimizes suspension performance, demonstrating high technological advancement and practical value. Attached Figure Description

[0015] Figure 1 This is a flowchart of a suspension adjustment method based on an air spring damping device of a vehicle intelligent control system according to the present invention.

[0016] Figure 2 This is a schematic diagram of a suspension adjustment method based on an air spring damping device of a vehicle intelligent control system according to the present invention.

[0017] Figure 3 This is another schematic diagram of a suspension adjustment method based on an air spring damping device of a vehicle intelligent control system according to the present invention.

[0018] Figure 4 This is another schematic diagram of a suspension adjustment method based on an air spring damping device of a vehicle intelligent control system according to the present invention.

[0019] Figure 5 This is another schematic diagram of a suspension adjustment method based on an air spring damping device of a vehicle intelligent control system according to the present invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and thoroughly described below with reference to the accompanying drawings. The described embodiments are merely some embodiments of the present invention.

[0021] like Figures 1-5 This embodiment of a suspension adjustment method based on an air spring damping device of a vehicle intelligent control system may specifically include: S1, the sensor collects vehicle driving status parameters in real time, including vehicle speed, longitudinal and lateral acceleration, steering angle, road surface excitation, vehicle body posture and load changes, and preprocesses the collected signals to obtain preprocessed signals which are then transmitted to the intelligent control unit.

[0022] Step S1: The sensor collects vehicle driving status parameters in real time, including vehicle speed, longitudinal and lateral acceleration, steering angle, road surface excitation, vehicle body posture and load changes. The collected signals are preprocessed to obtain preprocessed signals, which are then transmitted to the intelligent control unit.

[0023] Step S11: Real-time vehicle driving status parameters are collected using various sensors installed on the vehicle. These parameters include vehicle speed, longitudinal and lateral acceleration, steering angle, road surface excitation, vehicle body posture, and load changes.

[0024] In one embodiment, the sensor module begins to operate continuously after the vehicle is started.

[0025] Specifically, the vehicle speed sensor collects the vehicle's speed in real time, the acceleration sensor collects longitudinal and lateral acceleration, the steering angle sensor collects the steering wheel angle, the road surface sensor collects road surface excitation signals, the attitude sensor collects the vehicle roll angle, pitch angle, and vehicle height, and the load sensor collects changes in vehicle load. Road surface excitation is indirectly collected through the road surface sensor or the vehicle body vibration sensor, load changes are collected through the load sensor, and vehicle attitude is collected through the attitude sensor.

[0026] For example, taking a passenger vehicle, the vehicle speed sensor collects real-time vehicle speeds ranging from 0 to 120 kilometers per hour, the acceleration sensor collects longitudinal and lateral accelerations ranging from -5 to 5 meters per second squared, and the steering angle sensor collects steering wheel angles ranging from -90 degrees to 90 degrees. By comprehensively collecting these multi-dimensional state parameters, a rich data foundation can be provided for subsequent accurate identification and pattern judgment of vehicle operating conditions, avoiding the inaccurate operating condition identification problem caused by traditional systems that only collect partial parameters.

[0027] Step S12: The collected vehicle driving state parameter signals are processed by Kalman filtering algorithm to remove interference noise from the signals.

[0028] Understandably, when vehicles travel under complex road conditions, the raw signals collected by sensors often contain a significant amount of electromagnetic interference and mechanical vibration interference. This noise can severely affect the judgment accuracy of the control system. The Kalman filter algorithm is an algorithm that uses the state equations of a linear system to make an optimal estimate of the system state using the system's input and output observation data. Since the observation data includes the influence of noise and interference in the system, the optimal estimation can also be regarded as a filtering process.

[0029] Specifically, the Kalman filter algorithm comprises two core stages: prediction and update. In the prediction stage, the algorithm uses the vehicle's driving state estimate from the previous moment and the vehicle's kinematic model to calculate the predicted state value for the current moment, and also calculates the error covariance of this prediction. In the update stage, the algorithm compares the actual sensor observations with the predicted value at the current moment, and calculates the Kalman gain. Subsequently, the Kalman gain is used to correct the predicted value, obtaining the optimal state estimate for the current moment, and the error covariance is updated, preparing for filtering at the next moment.

[0030] For example, when a passenger vehicle travels on a bumpy road, the road excitation signal collected by the vehicle vibration sensor will be superimposed with a large amount of high-frequency mechanical vibration noise caused by wheel bounce. If this signal is directly used for suspension adjustment, it will cause frequent and irregular changes in shock absorber damping, exacerbating vehicle vibration. By introducing a Kalman filter algorithm, the system can accurately identify and filter out these high-frequency noises based on the continuous motion characteristics of the vehicle, extracting the low-frequency excitation signal that truly reflects the road surface undulations. This noise reduction process effectively suppresses noise interference, ensures the accuracy and stability of the signal, and significantly improves the adjustment precision of the suspension system and ride comfort under complex road conditions.

[0031] Step S13: Normalize the denoised signal to standardize parameters of different magnitudes and units, identify and remove abnormal data, and obtain the preprocessed signal.

[0032] In one embodiment, parameters collected by various sensors have different physical meanings and dimensions. For example, vehicle speed is measured in kilometers per hour and may have a value exceeding 100, while acceleration is measured in meters per second squared and typically has a value in the single digits. If these non-standardized parameters are directly input into the subsequent fuzzy neural network controller, parameters with larger values ​​will dominate the network's weight updates, causing smaller but equally crucial parameters to be ignored, thus severely reducing the accuracy of the control algorithm. Therefore, normalization processing of the denoised signal is necessary. The normalization process uses a maximum-minimum mapping method, scaling all parameters proportionally to convert them into standardized digital signals between 0 and 1. Specifically, the calculation process involves subtracting the minimum value of the parameter under preset operating conditions from the current parameter value, and then dividing by the difference between the maximum and minimum values ​​of the parameter. Simultaneously, during preprocessing, the system also identifies and removes abnormal data collected by the sensors. Abnormal data is usually caused by momentary short circuits in the sensors or external extreme electromagnetic pulses, manifesting as values ​​momentarily exceeding physical limits.

[0033] For example, if the vehicle roll angle output by the attitude sensor suddenly reaches 80 degrees within a certain sampling period, while the values ​​in the preceding and following periods are both around 5 degrees, the system will determine that the 80-degree value is abnormal data and discard it, replacing it with the interpolated value from the preceding and following periods. Through normalization processing and abnormal data removal, parameters of different magnitudes and units are uniformly converted into standardized digital signals that the intelligent control unit can recognize and process. This avoids abnormal data affecting the accuracy of adjustment decisions, further improving the reliability and robustness of the system.

[0034] Step S14: The preprocessed signal is transmitted to the intelligent control unit.

[0035] Preferably, the preprocessed signal obtained after the above processing is transmitted to the intelligent control unit in real time through the vehicle's internal data bus, providing reliable data support for subsequent parameter prediction and effectively avoiding the lag of adjustment commands.

[0036] S2, the intelligent control unit receives the preprocessed signal and combines it with a preset multi-mode control strategy to identify the vehicle's operating conditions and obtain the operating condition results.

[0037] Step S2: The intelligent control unit receives the preprocessed signal and combines it with a preset multi-mode control strategy to identify the vehicle's operating conditions and obtain the operating condition results.

[0038] In step S21, the intelligent control unit receives the preprocessed signal and extracts the feature values ​​of each driving state parameter.

[0039] Step S211: Extract vehicle speed threshold, acceleration change rate, steering angle range, and road excitation intensity as feature values ​​from the preprocessed signal.

[0040] Specifically, after acquiring the standardized digital signals that have undergone filtering and normalization, the intelligent control unit performs dimensionality reduction and feature extraction on these continuous time-series signals. The vehicle speed threshold reflects the vehicle's current absolute speed, the rate of change of acceleration reflects the intensity of the vehicle's dynamic acceleration and deceleration in the longitudinal and lateral directions, the steering angle range characterizes the driver's steering intention and the vehicle's yaw tendency, and the road surface excitation intensity directly reflects the unevenness of the tire contact surface. By extracting these feature values, massive amounts of low-level sensor data are transformed into high-dimensional feature vectors that can be directly used for logical judgment, providing fundamental data support for subsequent pattern matching.

[0041] Step S22: Dynamically adjust the threshold parameters for working condition identification based on load changes.

[0042] Step S221: Obtain the load change amount. When the load change amount exceeds the preset benchmark value, correct the judgment boundary of the vehicle speed threshold and the acceleration change rate according to the load ratio coefficient.

[0043] In one embodiment, the vehicle's total mass and center of gravity position change significantly with the number of passengers or the amount of cargo loaded during actual driving. If a fixed threshold is used, the vehicle's inertia increases under full load or overload conditions, and the same rate of change of acceleration or steering angle may cause more severe body roll or pitch than when unloaded. Therefore, a dynamic threshold adjustment mechanism is needed. The intelligent control unit monitors the load change returned by the load sensor in real time and sets a reference load state. When the actual load deviates from the reference load and exceeds a preset reference value, the system calculates a load proportionality coefficient. This load proportionality coefficient is obtained by dividing the difference between the actual load and the reference load by the reference load. Subsequently, this load proportionality coefficient is used to linearly or nonlinearly correct the original vehicle speed threshold and the judgment boundary of the rate of change of acceleration.

[0044] For example, assuming the baseline load is unloaded, the longitudinal acceleration threshold for emergency braking is -3 m / s². When the vehicle is fully loaded, the load change reaches 500 kg, exceeding the preset baseline value of 100 kg. At this point, the system calculates the load proportionality coefficient and corrects the longitudinal acceleration threshold for emergency braking to -2.5 m / s². This means that under full load, due to increased vehicle inertia and longer braking distance, the system can identify emergency braking earlier and more sensitively, thus outputting adjustment commands in advance to avoid severe nose-diving. This dynamic adjustment mechanism greatly improves the accuracy of load condition identification and environmental adaptability.

[0045] Step S23: Perform pattern matching between the extracted feature values ​​and the preset operating condition feature library to determine the current operating condition of the vehicle.

[0046] Step S231: Input the corrected feature value into the working condition feature library, and output the corresponding working condition result by comparing the feature value with the numerical range of the preset working condition. The working condition result includes low-speed stable driving condition, high-speed driving condition, turning condition, bumpy road driving condition, emergency braking condition, and load change condition.

[0047] Understandably, the operating condition feature library pre-stores feature value ranges for various typical driving scenarios. The intelligent control unit compares the feature values ​​corrected in step S22 with these value ranges one by one. When the extracted vehicle speed feature value is less than 30 km / h, the absolute value of acceleration is less than 1 m / s², the absolute value of steering angle is less than 5 degrees, and the road excitation intensity is less than 0.5 times the gravitational acceleration, the feature value falls entirely within the value range for low-speed, stable driving, and the system determines that the current condition is low-speed, stable driving. When the vehicle speed feature value is greater than or equal to 80 km / h and the absolute value of acceleration is less than 0.5 m / s², it is determined to be a high-speed driving condition. When the absolute value of steering angle is greater than or equal to 15 degrees, regardless of the vehicle speed, the system prioritizes determining it as a turning condition to suppress vehicle roll. When the road excitation intensity is greater than or equal to 1 times the gravitational acceleration, it is determined to be a bumpy road driving condition. When the longitudinal acceleration is less than or equal to the corrected emergency braking threshold, it is determined to be an emergency braking condition. When the load change is greater than or equal to 500 kg within a short period of time, it is determined to be a load change condition. Through this multi-dimensional feature extraction and pattern matching, the system can accurately identify the current operating state of the vehicle and output accurate operating condition results. This provides a reliable decision-making basis for subsequent parameter prediction based on the vehicle dynamics model and command output of the fuzzy neural network controller, ultimately realizing the adaptive matching of the vehicle suspension adjustment method under different driving conditions.

[0048] S3, the intelligent control unit calls the vehicle dynamics model and combines the preprocessed signal and the working condition result to predict the key parameters of the suspension system to obtain the predicted parameters.

[0049] S3, the intelligent control unit calls the vehicle dynamics model and combines the preprocessed signal and the working condition result to predict the key parameters of the suspension system to obtain the predicted parameters; specifically, S31, the intelligent control unit calls the preset vehicle dynamics model, which is a mathematical analysis framework based on classical mechanics methods that covers the vehicle body mass, wheel mass, nonlinear stiffness of air springs and damping characteristics of shock absorbers.

[0050] In one possible implementation, a vehicle dynamics model can accurately describe the dynamic response of the vehicle suspension system under different road surface excitations and vehicle motion states. Because the stiffness of air springs exhibits nonlinear characteristics with air pressure changes, and the damping force of shock absorbers also exhibits nonlinearity at different speeds, traditional linear models cannot accurately reflect the true state of the suspension system. By establishing a vehicle dynamics model that incorporates these nonlinear characteristics, the intelligent control unit can extrapolate and calculate the force conditions and motion trends of the suspension system in the near future based on the current vehicle state.

[0051] S32, the preprocessed signal and the working condition result are input into the vehicle dynamics model, and the model parameters are dynamically compensated by combining the radial basis function neural network.

[0052] Specifically, the preprocessed signal includes full-dimensional state parameters such as vehicle speed, longitudinal and lateral acceleration, steering angle, road excitation, vehicle attitude, and load changes. The operating condition results indicate the specific driving scenario the vehicle is currently in, such as high-speed driving or driving on bumpy roads. The radial basis function neural network (RBN) is a feedforward neural network with a single hidden layer. It uses radial basis functions as activation functions for hidden layer neurons, enabling it to approximate any continuous function with arbitrary precision. During vehicle operation, deviations occur between the theoretical vehicle dynamics model and the actual physical system due to factors such as mechanical wear, ambient temperature changes, and sensor measurement errors. After inputting the preprocessed signal and operating condition results into the model, the RBN calculates compensation values ​​for the model parameters based on real-time input data and historical learning experience, correcting the internal parameters of the vehicle dynamics model in real time, thereby significantly improving the model's prediction accuracy for the dynamic response of complex nonlinear systems.

[0053] S33, the predicted parameters of the suspension system are calculated through the compensated vehicle dynamics model. The predicted parameters include the target working air pressure of the air spring and the target damping coefficient of the shock absorber.

[0054] For example, parameter prediction can provide advance information on the dynamic requirements of the suspension system, allowing for precise adjustments and effectively avoiding the lag in adjustment command output found in traditional control methods. In actual passenger vehicle applications, when the operating condition is identified as a high-speed driving condition and the preprocessed signal indicates a vehicle speed of 100 km / h, the compensated vehicle dynamics model predicts the target working air pressure of the air springs to be 0.35 MPa, based on the downforce and anti-roll moment required for vehicle stability, and simultaneously predicts a standardized target damping coefficient of 0.8 for the shock absorbers. This combination of high damping and low air pressure effectively lowers the vehicle's center of gravity, enhancing its road grip and driving stability at high speeds.

[0055] In one embodiment, when the driving condition is identified as a bumpy road driving condition, and the road excitation intensity in the preprocessed signal exceeds a preset severe bump threshold, the vehicle dynamics model focuses on buffering road impacts and attenuating vehicle body vibrations. At this time, the model, combined with compensation calculations using a radial basis function neural network, predicts that the target operating air pressure of the air spring needs to be increased to 0.45 MPa to increase suspension travel and prevent the suspension system from bottoming out. Simultaneously, it predicts that the target damping coefficient of the shock absorber needs to be reduced to a standardized value of 0.3, making the suspension system softer and thus maximizing the absorption of high-frequency vibration impacts from the road surface. Through this parameter prediction mechanism based on a precise model and neural network compensation, the intelligent control unit can ensure that the suspension system adapts promptly to drastic changes in vehicle driving conditions, significantly improving passenger comfort and reducing mechanical wear on suspension system components.

[0056] S4, the fuzzy neural network controller receives the preprocessed signal, the operating condition result, and the predicted parameters and outputs adjustment instructions, the adjustment instructions including air spring inflation / deflation control instructions and shock absorber damping coefficient adjustment instructions.

[0057] S4, the fuzzy neural network controller receives the preprocessed signal, the operating condition result, and the predicted parameters, and outputs adjustment instructions. These adjustment instructions include air spring inflation / deflation control instructions and shock absorber damping coefficient adjustment instructions; specifically including... S41, the fuzzy neural network controller receives the preprocessed signal, the operating condition result, and the prediction parameters as input to the controller.

[0058] It should be noted that the preprocessed signal refers to the driving state parameters, such as vehicle speed, longitudinal and lateral acceleration, steering angle, road excitation, vehicle posture, and load changes, collected in real time by various sensors installed on the vehicle. These parameters are then processed using a Kalman filter algorithm to remove electromagnetic interference and mechanical vibration noise, and normalized to form a standardized digital signal. The operating condition result refers to the current vehicle operating state determined by the intelligent control unit based on a multi-mode control strategy through feature extraction and pattern matching, such as low-speed stable driving, high-speed driving, turning, and driving on bumpy roads. The predicted parameters refer to the target working air pressure of the air springs and the target damping coefficient of the shock absorbers, calculated based on a vehicle dynamics model covering the dynamic characteristics of the body, wheels, air springs, and shock absorbers, combined with radial basis function neural network compensation. The above-mentioned multi-dimensional standardized digital signals, operating state classifications, and target predicted values ​​are input into the fuzzy neural network controller to provide a comprehensive data foundation for subsequent precise adjustments.

[0059] S42, the fuzzy neural network controller combines preset fuzzy rules and the self-learning capability of the neural network to analyze and process the input, thereby realizing the fusion of fuzzy reasoning and neural network.

[0060] In one possible implementation, the fuzzy neural network controller is an intelligent control algorithm that combines the qualitative reasoning capability of fuzzy logic with the quantitative learning capability of neural networks. The controller internally includes a fuzzification layer, a fuzzy inference layer, and a defuzzification layer, with the connection weights between each layer dynamically adjusted through the backpropagation algorithm of the neural network. The preset fuzzy rules are conditional statements formulated based on expert control experience of the vehicle suspension system under different operating conditions. For example, when the operating condition is high-speed driving, the corresponding fuzzy rule is to decrease the air spring pressure and increase the shock absorber damping; when the operating condition is a bumpy road surface, the corresponding fuzzy rule is to increase the air spring pressure and decrease the shock absorber damping. The self-learning capability of the neural network refers to the controller continuously calculating the error between the actual output and the ideal target based on the input preprocessed signal and predicted parameters during operation, and automatically updating the network weights in the fuzzy inference layer using this error, thereby overcoming the shortcomings of low accuracy in traditional fuzzy control and poor generalization ability of single neural networks. Through this fusion processing, the controller can calculate the optimal adjustment parameters for complex and ever-changing driving scenarios.

[0061] S43, the fuzzy neural network controller outputs precise adjustment commands based on the analysis and processing results. The adjustment commands include air spring inflation / deflation control commands and shock absorber damping coefficient adjustment commands.

[0062] For example, the air spring inflation / deflation control command specifically includes the pulse width modulation duty cycle and inflation / deflation duration for controlling the opening and closing state of the high-speed switching solenoid valve, wherein the pulse width modulation duty cycle determines the inflation / deflation rate. The shock absorber damping coefficient adjustment command specifically includes a standardized value of the target damping coefficient for controlling the action of the pneumatic actuator. When the vehicle is in high-speed driving condition, the fuzzy neural network controller receives a preprocessed signal with a vehicle speed greater than or equal to 80 km / h, the high-speed driving condition result, and the predicted target air spring working air pressure of 0.35 MPa and target shock absorber damping coefficient of 0.8. After internal fuzzy inference and weight optimization, the controller outputs an air spring deflation command, specifically a pulse width modulation duty cycle of 30% and a deflation time of 0.5 seconds, while simultaneously outputting an adjustment command to increase the shock absorber damping coefficient to 0.8. After this command is transmitted to the actuator, it can effectively reduce the vehicle height and increase suspension damping, thereby significantly improving the vehicle's stability at high speeds.

[0063] In one possible implementation, when the vehicle encounters complex road conditions, the fuzzy neural network controller receives a preprocessed signal indicating that the road surface excitation intensity is greater than or equal to one standard gravitational acceleration, the operating conditions of driving on bumpy roads, and the predicted target working air spring pressure of 0.45 MPa and target damper damping coefficient of 0.3. Based on the corresponding fuzzy rules and combined with self-learning optimized network weights, the controller outputs an air spring inflation command: a pulse width modulation duty cycle of 50% and an inflation time of 0.8 seconds. Simultaneously, it outputs an adjustment command to reduce the damper damping coefficient to 0.3. By outputting these high-precision adjustment commands, the actuator can rapidly increase the air spring pressure to buffer road impacts and reduce the damping coefficient to attenuate vehicle body vibrations, effectively reducing the bumpy feeling during vehicle operation and significantly improving passenger comfort. This command output mechanism based on multi-dimensional input and intelligent algorithms completely solves the problems of rigid adjustment methods, lag in response, and inaccurate control in existing technologies, achieving intelligent adaptive matching of suspension system stiffness and damping.

[0064] S5, the actuator receives the adjustment command and controls the air spring inflation and deflation and the damping coefficient of the shock absorber through the high-speed switching solenoid valve and the pneumatic actuator.

[0065] In step S5, the actuator receives the adjustment command and controls the inflation and deflation of the air spring and the damping coefficient of the shock absorber through the high-speed switching solenoid valve and the pneumatic actuator.

[0066] In step S51, the actuator receives the adjustment command and parses it into a pulse width modulation signal and a pneumatic control signal.

[0067] In one possible implementation, a microprocessor within the actuator receives adjustment commands output by a fuzzy neural network controller. These commands include air spring inflation / deflation control commands and damper damping coefficient adjustment commands. The microprocessor parses the air spring inflation / deflation control commands into pulse-width modulation (PWM) signals with specific duty cycles and periods. Simultaneously, the microprocessor parses the damper damping coefficient adjustment commands into pneumatic control signals for driving the pneumatic actuator. The duty cycle of the PWM signal determines the opening degree of the high-speed switching solenoid valve, thus determining the inflation / deflation rate. The voltage amplitude of the pneumatic control signal corresponds to the displacement of the pneumatic actuator, thus determining the damping force of the damper.

[0068] In step S52, the high-speed switching solenoid valve controls the inflation / deflation rate and inflation / deflation time of the air spring according to the pulse width modulation signal.

[0069] For example, the response time of the high-speed switching solenoid valve is less than or equal to 4 milliseconds. When a pulse width modulation (PWM) signal acts on the solenoid coil of the high-speed switching solenoid valve, the coil generates an electromagnetic force that overcomes the spring force, driving the valve core to move and opening the air passage. The larger the duty cycle of the PWM signal, the longer the valve core remains open within one cycle, the larger the average flow area of ​​the air passage, and the higher the inflation / deflation rate of the air spring. The duration of the PWM signal is the inflation / deflation time. By precisely controlling the duty cycle and duration, precise control of the air spring's inflation / deflation volume can be achieved, thereby adjusting the air spring's working air pressure and changing the stiffness of the suspension system.

[0070] Specifically, under high-speed driving conditions, the vehicle requires lower suspension stiffness to maintain stability. At this time, the analyzed pulse width modulation signal has a duty cycle of 30% and a duration of 0.5 seconds. The high-speed switching solenoid valve executes a deflation action based on this signal, causing the air spring's operating air pressure to smoothly decrease to 0.35 MPa. This low duty cycle deflation method avoids sudden changes in vehicle posture caused by excessively rapid air pressure drops, ensuring stability at high speeds.

[0071] In one embodiment, when the vehicle is traveling on a bumpy road, higher suspension stiffness is required to cushion road impacts. In this case, the analyzed pulse width modulation (PWM) signal has a 50% duty cycle and a duration of 0.8 seconds. The high-speed switching solenoid valve executes an inflation action based on this signal, rapidly increasing the air spring's working pressure to 0.45 MPa. The high duty cycle ensures a high inflation rate, allowing the suspension system to complete stiffness adjustment quickly, effectively absorbing the energy from road bumps and significantly improving passenger comfort. By controlling the high-speed switching solenoid valve with pulse width modulation technology, the shortcomings of traditional solenoid valves, which can only be fully open or fully closed, are overcome, making precise control of inflation and deflation difficult. This effectively avoids over-inflation or over-deflation of the air spring.

[0072] In step S53, the pneumatic actuator adjusts the damping force of the variable damping shock absorber according to the pneumatic control signal.

[0073] It should be noted that the pneumatic actuator contains a pneumatic piston and an adjusting rod. A pneumatic control signal drives a proportional servo valve, controlling the air pressure entering the pneumatic actuator's air chamber. The air pressure in the chamber pushes the pneumatic piston, causing it to move, which in turn moves the adjusting rod. The adjusting rod is connected to a throttle valve inside the variable damping shock absorber; its displacement changes the opening of the throttle valve. The smaller the throttle valve opening, the greater the resistance to the oil flow inside the shock absorber, and the greater the damping force, i.e., the larger the damping coefficient. Conversely, the larger the throttle valve opening, the smaller the damping coefficient.

[0074] For example, during cornering, the vehicle body is prone to tilting due to centrifugal force. In this situation, the pneumatic control signal drives the proportional servo valve to increase the air pressure entering the pneumatic actuator chamber. The pneumatic piston then pushes the adjusting linkage to decrease the throttle valve opening, increasing the damper's damping coefficient to 0.8. This increased damping force effectively suppresses body roll, improving vehicle stability and tire grip during cornering. Conversely, under low-speed, stable driving conditions, the pneumatic control signal decreases the chamber pressure, increasing the throttle valve opening and reducing the damping coefficient to 0.3. This makes the suspension system softer, filtering out vibrations from minor road surface undulations.

[0075] Step S54: The pressure sensor monitors the pressure of the air tank and controls the air compressor to replenish air and purify the air.

[0076] Understandably, the air tank provides a stable air pressure source for the air spring's inflation action. A pressure sensor collects the actual air pressure value inside the air tank in real time. When the actual air pressure value falls below a preset threshold of 0.6 MPa, the air replenishment mechanism is triggered. The air compressor starts, drawing in outside air, compressing it, and delivering it to the air tank. Before entering the air tank, the air must pass through a filter and a dryer. The filter element inside the filter intercepts dust and impurity particles in the air, while the desiccant inside the dryer absorbs moisture from the air. The clean, dry air after filtration and drying enters the air tank for storage. This process ensures the air quality entering the high-speed switching solenoid valve, pneumatic actuator, and air spring, avoiding problems such as impurities wearing down the valve core, moisture condensation causing airway blockage, or component corrosion. This reduces wear on suspension system components, extends the service life of the entire intelligent control system, and ensures the long-term stable and reliable operation of the inflation / deflation system.

[0077] S6, the sensor collects the feedback parameters of the suspension system, compares the feedback parameters with the predicted parameters, and corrects the adjustment command through the intelligent control unit.

[0078] Step S6: The sensor collects feedback parameters from the suspension system, compares the feedback parameters with the predicted parameters, and corrects the adjustment command through the intelligent control unit. Specifically, this includes: Step S61: Real-time acquisition of feedback parameters of the suspension system through sensors. The feedback parameters include the actual working air pressure of the air spring, the actual damping coefficient of the shock absorber, the actual value of the vehicle body posture, and the adhesion coefficient between the tire and the ground.

[0079] Step S62: The feedback parameters are transmitted to the intelligent control unit and compared with the pre-acquired prediction parameters to calculate the deviation value.

[0080] In step S63, the intelligent control unit, based on the deviation value, uses a built-in fuzzy neural network controller combined with a proportional-integral-derivative adjustment algorithm to perform real-time correction of the adjustment command and adjust the action parameters of the actuator.

[0081] Step S64: The dynamic response of the suspension system is optimized using a stability optimization algorithm to suppress vehicle body vibration.

[0082] In one embodiment, the acquisition of feedback parameters in step S61 is completed by a feedback module, which consists of various high-precision sensors to acquire the actual state of the suspension system in real time after the actuators move. The actual working air pressure of the air spring is obtained through a pressure sensor, the actual damping coefficient of the shock absorber is indirectly calculated through measurements from displacement and force sensors, and the actual values ​​of the vehicle body attitude, including roll angle, pitch angle, and vehicle height, are directly read by attitude sensors. These feedback parameters can accurately reflect the current physical state of the suspension system, providing basic data support for subsequent closed-loop control.

[0083] For example, the process of calculating the deviation value in step S62 is a key node in closed-loop control. After receiving the feedback parameters, the intelligent control unit compares them one by one with the target parameters predicted based on the vehicle dynamics model.

[0084] For example, when the vehicle is traveling at high speed, the predicted target operating air pressure of the air spring is 0.35 MPa, while the actual operating air pressure collected by the air pressure sensor is 0.33 MPa. The intelligent control unit calculates the difference and obtains an air pressure deviation value of -0.02 MPa. This deviation value quantifies the gap between the current actual state and the ideal target state, and serves as the direct input for subsequent correction algorithms.

[0085] Specifically, the correction process in step S63, where the fuzzy neural network controller combines the proportional-integral-derivative (PID) control algorithm, is the core of addressing the nonlinear and time-varying characteristics of the suspension system. The fuzzy neural network controller integrates the advantages of fuzzy inference in handling uncertain information with the powerful self-learning ability of neural networks. It can dynamically adjust the proportional, integral, and derivative coefficients of the PID control algorithm based on the magnitude and rate of change of the deviation. When the deviation is large, the fuzzy neural network controller increases the proportional coefficient to accelerate the system's response and rapidly reduce the gap between the actual and target values. When the deviation is small and close to the target value, it increases the integral and derivative coefficients to eliminate steady-state errors and suppress overshoot. Through this dynamic adjustment, the intelligent control unit outputs the corrected control command, precisely controlling the pulse width modulation duty cycle and inflation / deflation time of the high-speed switching solenoid valve.

[0086] In one possible implementation, for the aforementioned air pressure deviation of -0.02 MPa, the fuzzy neural network controller analyzes and determines that a small amount of air replenishment is needed. At this point, the controller outputs a corrected adjustment command, controlling the high-speed switching solenoid valve to open with a 20% pulse width modulation duty cycle, setting the inflation time to 0.2 seconds. Through this high-precision fine-tuning, the actual working air pressure of the air spring is precisely increased to the target value of 0.35 MPa. If, under bumpy road conditions, the predicted target damping coefficient of the shock absorber is 0.3, while the actual damping coefficient is 0.4 (a deviation of +0.1), the controller will output a command to control the pneumatic actuator to reduce the damping force, causing the damping coefficient to quickly drop back to 0.3, thus ensuring the vehicle's smooth driving on bumpy roads. This real-time dynamic correction based on the deviation effectively avoids the overcharging and over-discharging phenomena of the air spring that are prone to occur in traditional open-loop control, significantly improving the adjustment accuracy of the vehicle's posture.

[0087] It should be noted that the stability optimization algorithm in step S64 is mainly used to handle the dynamic transition phase during the adjustment process. When the stiffness and damping of the suspension system change rapidly, the vehicle body is prone to additional vibration or attitude instability. The stability optimization algorithm monitors the rate of change of the actual values ​​of the vehicle body attitude in real time. When it detects that the rate of change of the roll angle or pitch angle exceeds the safety threshold, the algorithm will actively intervene to smooth the corrected adjustment command and limit the action rate of the actuator.

[0088] For example, during emergency braking, vehicles are prone to severe pitching, meaning the pitch angle increases dramatically. In this situation, the stability optimization algorithm forces the front suspension's air springs to inflate rapidly and increases the shock absorber damping, while simultaneously limiting the rear suspension's deflation rate to prevent excessively rapid changes in vehicle attitude that could cause oscillations. Through this smoothing and rate limiting, the stability optimization algorithm effectively suppresses vehicle vibration, avoids overshoot during the adjustment process, ensures the stability and reliability of the suspension system under complex road conditions and extreme operating conditions, and significantly improves the ride comfort of passengers.

[0089] S107. The sensors collect vehicle driving status parameters in real time, including vehicle speed, longitudinal and lateral acceleration, steering angle, road surface excitation, vehicle body posture, and load changes. The collected signals are preprocessed, including: S21, the vehicle speed sensor collects vehicle speed, the acceleration sensor collects longitudinal and lateral acceleration, the steering angle sensor collects steering angle, the road surface sensor collects road surface excitation, the posture sensor collects vehicle body posture, and the load sensor collects load changes; S22, the collected signals are filtered, denoised, normalized, and abnormal data is removed. The preprocessed signal is obtained by using a Kalman filter algorithm and transmitted to the intelligent control unit.

[0090] Step S21: The vehicle speed sensor collects the vehicle speed, the acceleration sensor collects the longitudinal and lateral acceleration, the steering angle sensor collects the steering angle, the road surface sensor collects the road surface excitation, the attitude sensor collects the vehicle body attitude, and the load sensor collects the load change.

[0091] Specifically, step S21 includes the following sub-steps: Step S211: The vehicle's current driving speed signal is obtained through the vehicle speed sensor, and the longitudinal acceleration signal and lateral acceleration signal of the vehicle during driving are obtained through the acceleration sensor. The real-time steering angle signal of the steering wheel is obtained through the steering angle sensor.

[0092] For example, during the driving of a passenger vehicle, the vehicle speed sensor monitors the wheel rotation speed in real time and converts it into driving speed; the longitudinal acceleration signal reflects the vehicle's acceleration or braking state; the lateral acceleration signal reflects the lateral force on the vehicle when driving in a curve; and the steering angle signal directly reflects the driver's steering intention. These kinematic parameters together constitute the basic data matrix of the vehicle's current motion state.

[0093] Step S212: Obtain the road excitation signal received by the wheels through the road surface sensor, obtain the roll angle, pitch angle and vehicle height signals of the vehicle body through the attitude sensor, and obtain the change signal of the vehicle's load mass through the load sensor.

[0094] In one possible implementation, the road excitation signal reflects the vertical impact intensity of road unevenness on the wheels. Attitude sensors, using multi-axis gyroscopes, calculate the vehicle's three-dimensional attitude changes in space in real time. Load sensors, installed at suspension support points, sense load fluctuations caused by changes in the number of passengers or the weight of cargo in the trunk. These multi-source sensors work synchronously, converting the collected continuous physical quantities into discrete digital electrical signals, forming a raw signal set containing multi-dimensional vehicle driving state parameters.

[0095] Step S22: The acquired signal is filtered, denoised, normalized, and abnormal data is removed. The Kalman filter algorithm is used to process the signal to obtain the preprocessed signal, which is then transmitted to the intelligent control unit.

[0096] Specifically, step S22 includes the following sub-steps: Step S221: The Kalman filter algorithm is called to recursively estimate and filter the various types of acquired signals in the original signal set to eliminate high-frequency random noise introduced by electromagnetic interference and mechanical vibration, and obtain smooth state estimates.

[0097] It should be noted that the Kalman filter algorithm is an algorithm that uses the state equations of a linear system to optimally estimate the system state using system input and output observation data. Since the observation data includes noise and interference from the system, the optimal estimation can also be viewed as a filtering process. For a vehicle suspension system, a discrete-time state-space model of the vehicle's driving state is first constructed, setting the state variables as actual physical quantities such as vehicle speed, acceleration, and attitude angle, and the observation variables as the actual signals collected by the sensors. In the time update phase, the smoothed state estimate and state transition matrix from the previous time step are used to predict the prior state estimate for the current time step, and the prior estimation error covariance matrix is ​​calculated. Then, in the measurement update phase, the Kalman gain matrix is ​​calculated, which reflects the trust weight distribution between the prediction model and the actual observed signals. Next, the difference between the current acquired signal and the predicted observation value is calculated to obtain the measurement residual. The Kalman gain matrix is ​​then used to correct the prior state estimate, ultimately obtaining the optimal smoothed state estimate for the current time step, and the posterior estimation error covariance matrix is ​​updated simultaneously, providing a basis for the recursive calculation in the next time step.

[0098] For example, when a passenger vehicle travels at 80 km / h on a gravel road, the road excitation signal collected by the road surface sensor will be mixed with a large amount of high-frequency mechanical vibration noise generated by gravel impacting the chassis. If the raw signal is used directly, it will cause frequent fluctuations in subsequent control commands. By introducing a Kalman filter algorithm, the system can adjust the Kalman gain in real time according to the dynamic changes in road excitation. When the variance of the observation noise increases, the Kalman gain decreases, and the algorithm tends to trust the prediction model more, thereby effectively filtering out the high-frequency spikes caused by gravel impacts and outputting a smooth state estimate that truly reflects the macroscopic undulations of the road surface, significantly improving the signal-to-noise ratio.

[0099] Step S222: Based on the smoothed state estimate, use statistical threshold judgment rules to identify and remove abnormal data, remove out-of-bounds data caused by instantaneous sensor failure or extreme interference, and obtain valid state signals.

[0100] In one possible implementation, the system sets reasonable physical boundary thresholds and rate of change thresholds for each state parameter.

[0101] For example, the physical boundary of the steering angle for passenger vehicles is set to -90 degrees to 90 degrees, and the maximum rate of change of the steering angle between adjacent sampling periods is set to a preset safety limit. When the smoothed state estimate exceeds the above physical boundary, or its first derivative exceeds the rate of change threshold, the system determines that the data point is abnormal data. For the identified abnormal data, the system removes it and uses the valid state signal from the previous moment combined with the current vehicle speed and acceleration for linear interpolation replacement, thereby ensuring the continuity and logical rationality of the data sequence and obtaining a valid state signal. This abnormal data removal mechanism can effectively prevent erroneous signals from misleading the intelligent control unit, avoid abrupt malfunctions in the suspension system, and ensure the safety of vehicle operation.

[0102] Step S223: Perform maximum and minimum value normalization processing on the valid state signal, and uniformly map the state parameters of different dimensions and orders of magnitude to the standard dimensionless interval from 0 to 1, generate a preprocessed signal and transmit it to the intelligent control unit.

[0103] Specifically, since vehicle speed is measured in kilometers per hour (km / h) and typically ranges from 0 to 120, while acceleration is measured in meters per second squared (m / s²) and ranges from -5 to 5, the magnitudes of these parameters differ significantly. Without standardized processing, larger parameters would mask the effects of smaller, crucial parameters in subsequent fuzzy neural network calculations, leading to an imbalance in network weight updates. Therefore, a maximum-minimum normalization method is employed. For each valid state signal, the theoretical minimum value of that parameter under preset conditions is subtracted, and then divided by the difference between the theoretical maximum and minimum values. This linear transformation compresses all parameters to the range of 0 to 1, eliminating the influence of dimensionality.

[0104] For example, assuming the currently acquired and filtered effective vehicle speed signal is 60 km / h, with a preset theoretical minimum of 0 and a maximum of 120 km / h, the normalized preprocessed vehicle speed signal is 0.5. Similarly, if the effective longitudinal acceleration signal is 2.5 m / s², with a preset minimum of -5 and a maximum of 5, the normalized preprocessed acceleration signal is 0.75. These standardized preprocessed signals are transmitted to the intelligent control unit in real time as the standard input matrix for the multi-mode control strategy and the fuzzy neural network controller. Through the above rigorous filtering, elimination, and normalization processes, not only is the purity and reliability of the input data significantly improved, but the convergence speed of subsequent intelligent algorithms is also significantly accelerated. This enables the intelligent control unit to more accurately and quickly identify vehicle operating conditions and output suspension adjustment commands, fundamentally solving the problems of delayed adjustment response and unstable vehicle posture caused by inaccurate parameter acquisition in traditional control methods.

[0105] S108. The intelligent control unit receives the preprocessed signal and combines it with a preset multi-mode control strategy to identify the vehicle's operating conditions, including: S31, extracting feature values ​​of each parameter in the preprocessed signal, including vehicle speed threshold, acceleration change rate, steering angle range, and road excitation intensity; S32, performing pattern matching between the feature values ​​and a preset operating condition feature library to determine the operating condition result, and dynamically adjusting the identification threshold according to load changes.

[0106] S31, extract the feature values ​​of each parameter in the preprocessed signal, including vehicle speed threshold, acceleration change rate, steering angle range, and road excitation intensity.

[0107] S311 acquires the real-time driving speed of the vehicle through a vehicle speed sensor, compares the speed signal with preset low-speed, medium-speed, and high-speed threshold ranges to determine the current speed level, which serves as a basic reference value for judging the vehicle's driving status.

[0108] S312 calculates the change in longitudinal and lateral acceleration per unit time, i.e., the rate of change of acceleration. This value reflects the degree of dynamic impact of the vehicle during starting, braking or turning, and is used to distinguish between smooth driving and rapid acceleration or deceleration conditions.

[0109] S313 monitors the real-time output of the steering angle sensor, dividing the absolute value of the steering angle into three intervals: small-angle steering, medium-angle steering, and large-angle steering. By analyzing the frequency of steering angle changes, it identifies whether the vehicle is in a continuous lane-changing or cornering state.

[0110] S314 uses the vertical acceleration signal collected by the vehicle body vibration sensor to convert the time domain signal into the frequency domain signal through fast Fourier transform, and extracts the road excitation intensity. This intensity value characterizes the energy input generated by the road surface unevenness to the suspension system. The larger the value, the more severe the road bumps.

[0111] S32, perform pattern matching between the feature values ​​and the preset working condition feature library to determine the working condition result, and dynamically adjust the recognition threshold according to the load change.

[0112] S321 constructs a multi-dimensional working condition feature library, combining vehicle speed, acceleration change rate, steering angle range, and road excitation intensity into feature vectors, pre-storing standard values ​​of feature vectors under different driving scenarios, and achieving rapid matching of working conditions by calculating the Euclidean distance between real-time feature vectors and standard vectors.

[0113] S322 introduces real-time load data collected by load sensors to establish a load correction coefficient model. When the load increases, the system automatically raises the trigger threshold of the road excitation intensity to avoid the suspension system becoming oversensitive due to the increase in vehicle weight, ensuring that it can still accurately identify bumpy road surfaces under heavy load conditions.

[0114] S323 dynamically adjusts the threshold for determining the rate of change of acceleration based on load changes. Under full load conditions, it appropriately relaxes the restriction on the amplitude of acceleration fluctuations to prevent false triggering caused by increased vehicle inertia and ensure the robustness of the operating condition identification results.

[0115] In one embodiment, when the vehicle is traveling on a highway, the vehicle speed feature value extracted in S31 remains within the high-speed threshold range, and S32 identifies it as a high-speed driving condition through pattern matching. At this time, based on the light-load signal fed back by the load sensor, the system maintains a low acceleration change rate threshold, enabling the suspension system to sensitively detect minute fluctuations in vehicle body posture. By adjusting the air spring pressure, the vehicle height is reduced, thereby reducing wind resistance and improving driving stability.

[0116] In another embodiment, when the vehicle is traveling on a rugged mountain road, S31 detects that the road surface excitation intensity frequently exceeds a preset threshold. During mode matching, S32, combined with the heavy load status detected by the load sensor, automatically raises the threshold for determining the road surface excitation intensity by 20%. This dynamic adjustment mechanism effectively filters out the natural vibrations of the suspension system caused by heavy loads, ensuring that the system only switches to the bumpy road driving mode when truly encountering severe bumps. This avoids the suspension system frequently switching adjustment modes under complex road conditions, significantly improving the vehicle's ride comfort and the lifespan of its components.

[0117] S109. The intelligent control unit calls the vehicle dynamics model and combines the preprocessed signal and the working condition result to predict the key parameters of the suspension system, including: S41. The vehicle dynamics model covers the dynamic characteristics of the vehicle body, wheels, air springs, and shock absorbers, considering nonlinear stiffness and damping characteristics and load effects; S42. The predicted parameters are obtained by using radial basis function neural network compensation model parameters to predict the target working air pressure shock absorber target damping coefficient, the trend of vehicle body attitude change, and subsequent changes in road surface excitation.

[0118] S41, the vehicle dynamics model covers the dynamic characteristics of the vehicle body, wheels, air springs, and shock absorbers, taking into account nonlinear stiffness and damping characteristics as well as load effects.

[0119] S411, obtain the vehicle body mass distribution parameters and wheel rotational inertia, and combine the effective area and initial volume of the air spring to establish the basic four-degree-of-freedom vehicle vertical vibration differential equation.

[0120] S412, based on the gas polytropic exponential variation law of the air spring under different inflation and deflation states, the stiffness of the air spring is defined as a nonlinear function that changes in real time with air pressure and displacement, and is substituted into the four-degree-of-freedom vehicle vertical vibration differential equation.

[0121] S413, extract the asymmetric piecewise linear characteristics of the damping force of the shock absorber during the recovery and compression strokes as a function of relative velocity, construct the damping characteristic function, and introduce the damping characteristic function as a damping term into the four-degree-of-freedom vehicle vertical vibration differential equation.

[0122] S414 receives the real-time load change data collected by the load sensor, and dynamically corrects the vehicle body mass distribution parameters and the initial volume of the air spring based on the real-time load change data to obtain a vehicle dynamics model that covers nonlinear stiffness, damping characteristics and load effects.

[0123] In one embodiment, step S41 establishes a vehicle dynamics model covering multiple characteristics, specifically including step S411, obtaining the vehicle body mass distribution parameters and the rotational inertia of the wheels, and combining the effective area and initial volume of the air spring to establish a basic four-degree-of-freedom vehicle vertical vibration differential equation.

[0124] Specifically, during vehicle operation, the mass distribution of the vehicle body and wheels directly determines the inertial basis of the vibration system. By using pre-calibrated vehicle body mass distribution parameters and wheel rotational inertia, combined with the physical structural parameters of the air spring, namely the effective area and initial volume, it is possible to construct the fundamental differential equations describing the vertical displacement of the vehicle body, the vehicle body pitch angle displacement, the vertical displacement of the front wheels, and the vertical displacement of the rear wheels.

[0125] Step S412: Based on the gas polytropic exponential variation law of the air spring under different inflation and deflation states, the stiffness of the air spring is defined as a nonlinear function that changes in real time with air pressure and displacement, and then substituted into the four-degree-of-freedom vehicle vertical vibration differential equation.

[0126] It should be noted that the stiffness of an air spring is not constant, but changes nonlinearly with the increase or decrease of internal air pressure and the compression or elongation of the spring itself. By introducing the polytropic variation law of the gas, this nonlinear change can be accurately described, allowing the differential equation to truly reflect the dynamic stiffness performance of the air spring in actual operation.

[0127] Step S413: Extract the asymmetric piecewise linear characteristics of the damping force of the shock absorber during the recovery and compression strokes as a function of relative velocity, construct the damping characteristic function, and introduce the damping characteristic function as a damping term into the four-degree-of-freedom vehicle vertical vibration differential equation.

[0128] For example, the opening state of the internal valve system of a shock absorber differs during tensile recovery and compression, resulting in an asymmetric damping force. Transforming this asymmetric piecewise linear characteristic into a damping characteristic function and introducing it into a differential equation allows for accurate simulation of the damping dissipation process of the shock absorber under different motion directions and velocities.

[0129] Step S414: Receive the real-time load change data collected by the load sensor, dynamically correct the vehicle body mass distribution parameters and the initial volume of the air spring based on the real-time load change data, and obtain a vehicle dynamics model that covers nonlinear stiffness, damping characteristics and load effects.

[0130] It is understandable that changes in passenger numbers or loading / unloading of cargo can cause sudden changes in vehicle load. Real-time acquisition of load changes and subsequent adjustments to mass distribution and initial spring volume can eliminate the impact of load fluctuations on model accuracy, ensuring that the vehicle dynamics model accurately describes the dynamic response of the suspension system under any loading condition.

[0131] S42, the predicted parameters are obtained by using radial basis function neural network compensation model parameters to predict the target working air pressure shock absorber damping coefficient, vehicle body attitude change trend, and subsequent road excitation changes.

[0132] S421, the vehicle speed, longitudinal acceleration, lateral acceleration, steering angle and the working condition result in the preprocessed signal are used as input vectors and input to the input layer of the radial basis function neural network.

[0133] S422, in the hidden layer of the radial basis function neural network, the Euclidean distance between the input vector and the preset hidden layer center vector is calculated, and the Euclidean distance is substituted into the Gaussian radial basis function to obtain the hidden layer output value.

[0134] S423, the hidden layer output value is linearly weighted and summed with the output layer weight matrix of the radial basis function neural network to obtain the model parameter compensation amount.

[0135] S424, the compensation amount of the model parameters is superimposed on the vehicle dynamics model, and the state extrapolation is performed using the compensated vehicle dynamics model to calculate the target working air pressure of the air spring, the target damping coefficient of the shock absorber, the trend of vehicle body attitude change, and the subsequent changes of road excitation, which are used as the prediction parameters.

[0136] In one embodiment, step S42 uses a radial basis function neural network for parameter compensation and prediction. Specifically, step S421 involves inputting the vehicle speed, longitudinal acceleration, lateral acceleration, steering angle, and the operating condition result from the preprocessed signal as input vectors into the input layer of the radial basis function neural network.

[0137] Specifically, the radial basis function neural network is a feedforward neural network with a single hidden layer, capable of approximating any continuous function with arbitrary precision. By combining vehicle speed, acceleration, steering angle, and identified operating conditions—reflecting the vehicle's current motion—into a multi-dimensional input vector, the neural network can provide a comprehensive representation of the vehicle's driving state, serving as the basis for subsequent parameter compensation calculations.

[0138] Step S422: In the hidden layer of the radial basis function neural network, calculate the Euclidean distance between the input vector and the preset hidden layer center vector, and substitute the Euclidean distance into the Gaussian radial basis function to obtain the hidden layer output value.

[0139] It should be noted that the hidden layer center vectors represent the cluster centers of different typical driving states in multidimensional space. By calculating the Euclidean distance between the input vector and these center vectors, the similarity between the current vehicle state and each typical state can be measured. The Gaussian radial basis function has local response characteristics; the closer to the center, the larger the output value, thus accurately capturing the local nonlinear features of the vehicle state and generating hidden layer output values ​​that reflect state similarity.

[0140] Step S423: The hidden layer output value is linearly weighted and summed with the output layer weight matrix of the radial basis function neural network to obtain the model parameter compensation amount.

[0141] For example, the output layer weight matrix stores empirical values ​​of parameter compensation corresponding to different state similarities. By linearly weighted summation, the nonlinear features extracted by the hidden layer can be transformed into specific numerical compensation quantities. This compensation quantity is specifically used to correct the calculation bias of the basic vehicle dynamics model under complex conditions caused by unmodeled dynamics or time-varying parameters, thereby significantly improving the accuracy of the model.

[0142] Step S424: The model parameter compensation amount is superimposed on the vehicle dynamics model. The compensated vehicle dynamics model is used to perform state extrapolation to calculate the target working air pressure of the air spring, the target damping coefficient of the shock absorber, the trend of vehicle body attitude change, and the subsequent changes of road excitation, which are used as the prediction parameters.

[0143] Understandably, the ability of a basic vehicle dynamics model to describe the dynamic response of a suspension system is greatly enhanced after the addition of compensation. By integrating and extrapolating the compensated model over time, the optimal state required by the suspension system at future moments can be calculated in advance.

[0144] For example, when identifying high-speed driving conditions with continuously increasing vehicle speed, the compensated model predicts the need for a lower target air spring operating pressure to reduce vehicle height, while simultaneously predicting a higher target damping coefficient for the shock absorbers to enhance driving stability. Furthermore, the model can extrapolate the subsequent changes in road surface excitation based on the current input, and predict the trend of vehicle attitude changes accordingly. This forward-looking prediction based on a high-precision model provides a reliable lead time for subsequent precise adjustments, effectively overcoming the response lag problem of traditional control methods and significantly improving the smoothness and stability of vehicle driving.

[0145] In one embodiment, for driving conditions on bumpy roads, the state extrapolation process in step S424 involves the radial basis function neural network outputting a large model parameter compensation amount to correct the nonlinear stiffness deviation of the suspension system when the input vector reflects drastic fluctuations in road excitation intensity. During state extrapolation, the compensated vehicle dynamics model calculates that subsequent changes in road excitation will continue to maintain a high-frequency, large-amplitude state. Based on this prediction, the model calculates that the target working air pressure of the air spring needs to be significantly increased to enhance its shock absorption capacity, while the target damping coefficient of the shock absorber needs to be correspondingly reduced to accelerate the response speed of the suspension system and prevent excessive roll or pitch in the vehicle's attitude change trend. Through this precise parameter prediction, the system can pre-plan the stiffness and damping matching scheme of the suspension system before the bumps are actually transmitted to the vehicle body, thereby maximizing the attenuation of vehicle vibration and improving the ride comfort of passengers.

[0146] S1010, the fuzzy neural network controller receives the preprocessed signal, the operating condition result, and the predicted parameters, and outputs adjustment instructions. The adjustment instructions include air spring inflation / deflation control instructions and shock absorber damping coefficient adjustment instructions, including: S51, processing the input through fuzzy inference in combination with preset fuzzy rules; S52, optimizing the fuzzy rules and network weights through neural network self-learning, and outputting the air spring inflation / deflation control instructions and shock absorber damping coefficient adjustment instructions according to the operating condition requirements.

[0147] S51, combine preset fuzzy rules to process the input through fuzzy inference.

[0148] Specifically, the fuzzy inference process includes three stages: fuzzification, rule base matching, and defuzzification. First, parameters such as real-time collected vehicle speed, acceleration, steering angle, road surface excitation, vehicle posture, and load changes are mapped into fuzzy sets through a preset membership function. The membership function adopts a Gaussian function, which can smoothly describe the transition state of the parameters in different intervals.

[0149] For example, vehicle speed is categorized into three fuzzy levels: low, medium, and high; road surface excitation is categorized into three fuzzy levels: flat, moderate, and bumpy. Then, inference is performed based on a pre-defined fuzzy rule base, which consists of a series of if-then logical statements. For instance, if the vehicle speed is high and the road surface excitation is flat, then the air spring stiffness requirement is low, and the shock absorber damping requirement is high. The fuzzy inference engine performs logical operations on the input parameters to obtain the fuzzy output quantities of suspension system stiffness and damping. Finally, defuzzification is performed using the center-of-gravity method to transform the fuzzy output quantities into specific control reference values, providing preliminary decision-making basis for subsequent neural network processing.

[0150] The S52, through neural network self-learning, optimizes fuzzy rules and network weights to output air spring inflation / deflation control commands and shock absorber damping coefficient adjustment commands based on operating conditions. The neural network employs a radial basis function network structure, with its hidden layer nodes corresponding to the trigger strength of the fuzzy rules. During self-learning, the neural network receives reference values ​​from fuzzy inference and predicted parameters from the vehicle dynamics model, adjusting the center position and width of the hidden layer nodes and the connection weights of the output layer in real time using a gradient descent algorithm. When vehicle driving conditions change, the neural network automatically corrects the membership function parameters of the fuzzy rules based on the actual feedback of suspension system response deviations, thereby achieving dynamic adaptation to different road conditions and load changes. In the output stage, the neural network converts the optimized control signals into pulse width modulation signals with varying duty cycles and frequencies, directly driving the opening and closing times of the high-speed switching solenoid valve to achieve precise control of the air spring inflation / deflation volume. Simultaneously, it outputs corresponding current signals to adjust the solenoid valve opening of the variable damping shock absorber, thereby precisely changing the damping force of the shock absorber.

[0151] In one embodiment, for high-speed driving scenarios, the fuzzy neural network controller maps the vehicle speed parameters to a high-speed fuzzy set through S51, and combines the self-learning mechanism in S52 to identify that the vehicle stability requirement is higher than the comfort requirement at this time.

[0152] Specifically, the neural network adjusts the weights to output a smaller air spring inflation duty cycle, thereby reducing the vehicle height and air resistance. At the same time, it increases the damping coefficient of the shock absorber to suppress minor vibrations during high-speed driving, ensuring the vehicle's driving stability under lane changes or crosswind interference.

[0153] In another embodiment, for bumpy road surface scenarios, the fuzzy neural network controller identifies the road excitation parameters as a high-intensity fuzzy set in S51. In S52, the neural network automatically adjusts the weights of the fuzzy rules based on real-time feedback of the suspension system's vibration acceleration, outputting a larger air spring inflation command to increase the suspension system's support stiffness and prevent the air springs from bottoming out. Simultaneously, the neural network outputs a command to reduce the damping coefficient of the shock absorbers, enabling the suspension system to absorb road impact energy more quickly, thereby significantly improving passenger comfort.

[0154] It should be noted that through the collaborative work of S51 and S52, the system can effectively overcome the shortcomings of fixed parameters and lag response in traditional control methods. Fuzzy inference provides an experience-based logical framework, while the neural network endows the system with the ability to handle nonlinear dynamic characteristics through self-learning. This combination enables the suspension system to achieve millisecond-level adaptive parameter adjustment when facing sudden load changes or complex road surface excitations, ensuring that the air springs and shock absorbers are always in optimal working condition, effectively reducing fatigue wear of suspension components and extending the overall service life of the vehicle suspension system.

[0155] S1011. The actuator receives the adjustment command and controls the air spring inflation / deflation and the damper damping coefficient through the high-speed switching solenoid valve and the pneumatic actuator, including: S61. The high-speed switching solenoid valve uses pulse width modulation technology to control the inflation / deflation rate and time according to the air spring inflation / deflation control command to adjust the air spring working pressure; S62. The pneumatic actuator adjusts the damping force of the variable damping shock absorber according to the damper damping coefficient adjustment command. When the air pressure sensor monitors that the air pressure in the air tank is lower than a preset threshold, the air compressor starts to replenish the air, which is then processed by the filter and dryer.

[0156] Step S61: The high-speed switching solenoid valve uses pulse width modulation technology to control the air filling and discharging rate and time according to the air spring filling and discharging control command, and adjusts the working air pressure of the air spring.

[0157] Step S611: Analyze the air spring inflation / deflation control command to obtain the target inflation / deflation rate and target inflation / deflation time.

[0158] Specifically, the air spring inflation / deflation control command output by the intelligent control unit includes the dynamic adjustment requirements of the suspension system under the current driving conditions. Upon receiving this command, the actuator first decodes the command signal, extracting the target inflation / deflation rate to characterize the speed of inflation / deflation and the target inflation / deflation time to limit the duration of the inflation / deflation action. The target inflation / deflation rate determines the mass of gas entering or exiting the air spring per unit time, while the target inflation / deflation time determines the time span of the entire adjustment process. These two parameters together determine the final change in air spring pressure.

[0159] Step S612: Calculate the duty cycle parameter of the pulse width modulation signal based on the target inflation / deflation rate.

[0160] In one embodiment, pulse width modulation (PWM) is an effective technique for controlling analog circuits by changing the ratio of the high-level duration of a digital signal to the total cycle time. The duty cycle parameter is the percentage of the high-level time in the total cycle. The calculation of the duty cycle parameter involves comparing the target charge / discharge rate with the maximum rated flow rate of the high-speed switching solenoid valve to obtain the required flow rate ratio. This is then combined with the solenoid valve's nonlinear flow characteristic curve, which reflects the variation in the solenoid valve's flow capacity under different inlet / outlet pressure differentials. The corresponding duty cycle parameter is obtained by polynomial interpolation on the curve. A larger duty cycle parameter results in a longer opening time for the solenoid valve within one cycle and a higher average charge / discharge rate.

[0161] Step S613: Generate a drive electrical signal according to the duty cycle parameter and the target inflation / deflation time to control the high-frequency opening and closing action of the high-speed switching solenoid valve.

[0162] For example, the drive circuit generates a pulse width modulation drive signal with a corresponding frequency and duty cycle based on the calculated duty cycle parameters, and continues this signal for the target inflation / deflation time. Under the action of this drive signal, the high-speed switching solenoid valve's internal electromagnetic coil is energized and de-energized at high frequency. The valve core of the solenoid valve reciprocates rapidly under the alternating action of electromagnetic force and return spring force, realizing high-frequency opening and closing of the air path. Through this high-frequency opening and closing action, the continuous high-pressure airflow is converted into discrete airflow pulses, thereby precisely controlling the actual gas flow rate entering or exiting the air spring, ultimately allowing the working air pressure of the air spring to smoothly transition to the target air pressure value.

[0163] In one embodiment, when the vehicle is traveling at high speed, the intelligent control unit predicts that the vehicle height needs to be lowered to improve driving stability, and thus outputs an air spring deflation command. The actuator analyzes this command to obtain the target deflation rate and the target deflation time of 0.5 seconds. Based on the target deflation rate, the duty cycle parameter of the pulse width modulation signal is calculated to be 30%. The drive circuit then generates a pulse signal with a 30% duty cycle, driving the high-speed switching solenoid valve to operate for 0.5 seconds. During this period, the solenoid valve core operates at high frequency, precisely discharging a fixed amount of gas, causing the air spring working pressure to drop smoothly to 0.35 MPa. This precise control based on pulse width modulation technology effectively avoids the problem of over- or under-inflation caused by the full opening and closing of traditional ordinary solenoid valves, significantly improving the accuracy and response speed of suspension system stiffness adjustment, avoiding overshoot and oscillation during vehicle posture adjustment, and significantly improving the ride comfort and safety of the vehicle at high speeds.

[0164] In step S62, the pneumatic actuator adjusts the damping force of the variable damping shock absorber according to the damper damping coefficient adjustment command. When the air pressure sensor detects that the air pressure in the air tank is lower than the preset threshold, the air compressor starts to replenish the air, which is then processed by the filter and dryer.

[0165] In step S621, the pneumatic actuator receives the damping coefficient adjustment command of the shock absorber and drives the throttle valve of the variable damping shock absorber to change the flow cross-sectional area to adjust the damping force.

[0166] Specifically, the damper damping coefficient adjustment command includes target damping coefficient information. Based on this information, the pneumatic actuator generates a corresponding mechanical displacement, actuating a throttle valve inside the variable damper. The displacement of the throttle valve changes the flow cross-sectional area of ​​the hydraulic oil or gas inside the damper. When the flow cross-sectional area decreases, fluid resistance increases, and the damper damping force increases accordingly. When the flow cross-sectional area increases, fluid resistance decreases, and the damper damping force decreases accordingly.

[0167] Step S622: The pressure sensor collects the current pressure value of the gas storage tank in real time and compares it with a preset threshold to determine whether the gas replenishment conditions are met.

[0168] In one embodiment, the air tank serves as the energy source for the entire air suspension system, and the stability of its air pressure directly affects the reliability of the air spring inflation. A pressure sensor continuously monitors the current air pressure inside the air tank and transmits this value to the control system. The control system compares the current air pressure value with a preset minimum operating air pressure threshold in real time. When the current air pressure value is greater than or equal to the preset threshold, the system is deemed to have sufficient air supply and no additional air is needed. When the current air pressure value is lower than the preset threshold, the system is deemed to meet the conditions for air replenishment.

[0169] Step S623: When the air replenishment conditions are met, the air compressor is triggered to start drawing in outside air, which is then purified and dehumidified by passing through a filter and a dryer before being pressed into the air storage tank.

[0170] For example, once the air replenishment conditions are determined to be met, the control system immediately sends a start electrical signal to the air compressor. The air compressor operates, generating negative pressure to draw ambient air into the system. The drawn-in air first passes through a filter, whose filter material effectively intercepts dust, particulate matter, and other solid impurities in the air, preventing them from entering the air path and damaging the solenoid valve core or abrading the pneumatic actuator. Subsequently, the pre-filtered air enters a dryer, which is filled with desiccant material to absorb moisture from the air and reduce humidity. The purified and dehumidified air is then pressurized by the compressor and delivered to the air tank for storage until the air pressure in the air tank returns to the normal operating range.

[0171] In one embodiment, when a vehicle travels on bumpy roads for extended periods, the suspension system frequently adjusts its air pressure by inflating and deflating, consuming a significant amount of high-pressure air from the air tank. A pressure sensor detects that the current air pressure in the tank has dropped to 0.6 MPa, below the preset threshold of 0.65 MPa. At this point, the system automatically triggers the air compressor. After being drawn in, outside air is filtered to remove dust from the road surface, and a dryer removes moisture from the air, preventing condensation in the air passages that could cause the solenoid valves to rust and become stuck, or to freeze and block the air passages in low temperatures. The clean, high-pressure air is continuously replenished to the air tank, ensuring that subsequent air spring inflation commands are executed quickly and in sufficient quantities. This closed-loop air replenishment and air handling mechanism not only ensures the continuous and stable operation of the suspension adjustment system but also significantly reduces the wear and failure rate of pneumatic components, extending the service life of the entire intelligent control system.

[0172] S1012, The sensor collects feedback parameters of the suspension system, compares the feedback parameters with the predicted parameters, and corrects the adjustment command through the intelligent control unit, including: S71, the air pressure sensor collects the actual working air pressure of the air spring, collects the actual damping coefficient of the shock absorber, the actual value of the vehicle body posture, and the tire adhesion coefficient to obtain the feedback parameters; S72, the deviation value between the feedback parameters and the predicted parameters is calculated, and the adjustment command is corrected by a fuzzy neural network and PID combined algorithm to adjust the action parameters of the actuator.

[0173] S71, the air pressure sensor collects the actual working air pressure of the air spring, collects the actual damping coefficient of the shock absorber, the actual value of the vehicle body posture, and the tire adhesion coefficient to obtain the feedback parameters. In one embodiment, step S71 includes step S711, which involves obtaining the actual working air pressure value inside the air spring in real time through an air pressure sensor installed in the air tank and the air spring circuit.

[0174] Step S712: The relative motion speed and displacement of the shock absorber piston are obtained by the displacement sensor, and the actual damping coefficient of the shock absorber is calculated by combining it with the pre-calibrated damping characteristic curve.

[0175] Step S713: Use attitude sensors installed at key nodes of the vehicle body to obtain the vehicle roll angle, pitch angle and vehicle height. At the same time, use data from wheel speed sensors and longitudinal acceleration sensors to estimate the tire slip ratio with respect to the ground, and then derive the tire adhesion coefficient. The above actual working air pressure value, actual damping coefficient, vehicle roll angle, pitch angle, vehicle height and tire adhesion coefficient are summarized to form the feedback parameters.

[0176] Specifically, the above parameters are collected continuously during vehicle operation, with a collection frequency set to 100 times per second, to ensure that the obtained feedback parameters can accurately reflect the physical state of the suspension components at the current instant.

[0177] S72, calculate the deviation between the feedback parameter and the predicted parameter, and correct the adjustment command to adjust the action parameters of the actuator by using a fuzzy neural network and a proportional-integral-derivative algorithm.

[0178] In one embodiment, step S72 includes step S721, performing subtraction operations on each value in the feedback parameters with the pre-calculated target working air pressure, target damping coefficient, and target vehicle body attitude values ​​to obtain a deviation matrix containing air pressure deviation value, damping deviation value, and attitude deviation value.

[0179] Step S722: Input the deviation matrix into the controller that combines fuzzy neural network and proportional-integral-derivative (PID) algorithm. Use fuzzy inference rules to fuzzify the deviation values ​​in the deviation matrix and determine the dynamic adjustment amounts of the proportional coefficient, integral coefficient, and derivative coefficient.

[0180] Step S723: The node weights of the fuzzy neural network are updated in real time using the error reverse propagation calculation process. The corrected air spring inflation / deflation duty cycle and shock absorber damping valve opening are calculated based on the updated weights and dynamic adjustment amount, and the corrected adjustment command is generated.

[0181] Step S724: According to the corrected adjustment command, a corresponding pulse width modulation signal is sent to the high-speed switching solenoid valve to adjust the charging and discharging time. At the same time, a displacement control signal is sent to the pneumatic actuator to change the flow area of ​​the shock absorber throttle orifice, thus completing the adjustment of the actuator's action parameters.

[0182] Specifically, the fuzzy neural network combined with proportional-integral-derivative (PID) control integrates the advantages of fuzzy logic in handling nonlinear problems, the self-learning optimization capability of neural networks, and the precision of PID control. In the closed-loop control of vehicle suspension components, due to the highly nonlinear changes in air spring stiffness and the slight delay in solenoid valve response, traditional single control methods are prone to overshoot or oscillation in vehicle height adjustment. By using the deviation matrix as input, the fuzzy neural network can autonomously match the most suitable PID parameters based on the current deviation magnitude and rate of change.

[0183] For example, when a vehicle encounters a crosswind while driving at high speed, causing the body roll angle deviation to increase instantaneously, the fuzzy neural network will quickly identify the large deviation and output a large adjustment amount for the proportional coefficient and the derivative coefficient.

[0184] In one embodiment, when the air pressure deviation value indicates that the actual working air pressure of the air spring is 0.05 MPa lower than the target working air pressure, the controller determines that an air replenishment operation is required. At this time, the fuzzy neural network infers the required inflation rate based on the current vehicle speed and load status, and adjusts the parameters of the proportional-integral-derivative (PID) controller. The corrected adjustment command instructs the high-speed switching solenoid valve to open at 60% duty cycle for 0.3 seconds. If, during inflation, the attitude deviation value indicates that the vehicle height rises too quickly, posing a risk of overshoot, the neural network will reduce the proportional coefficient and increase the derivative coefficient in real time, rapidly reducing the duty cycle to 20%, thereby smoothly approaching the target air pressure and avoiding unnecessary vertical oscillations of the vehicle body.

[0185] For example, under bumpy road conditions, the undulations in the road surface cause frequent positive and negative alternating damping deviations between the actual and target damping coefficients of the shock absorber. In this situation, the fuzzy neural network continuously iterates the weights of its internal nodes, recording the deviation patterns from previous cycles to predict the damping demand trend for the next moment. This allows for pre-emptive fine-tuning of the integral coefficient, mitigating integral saturation. The corrected adjustment command controls the pneumatic actuator to fine-tune the flow area of ​​the shock absorber's throttle orifice fifty times per second, enabling the shock absorber's damping force to dynamically match the high-frequency undulations of the road surface. This multi-algorithm fusion correction method not only significantly improves the adaptability of the suspension components to complex road conditions but also effectively suppresses vehicle vibration, significantly enhancing ride smoothness and comfort.

[0186] S1013. The vehicle dynamics model combines the preprocessed signal and the operating condition result to predict key parameters of the suspension system, including: S81. The vehicle dynamics model establishes a description of the dynamic response law of the suspension system based on classical mechanics methods; S82. By predicting and obtaining the dynamic demand of the suspension in advance, it provides the advance amount of adjustment command.

[0187] Step S81: The vehicle dynamics model is established based on classical mechanics methods to describe the dynamic response law of the suspension system.

[0188] Step S811: Obtain vehicle body mass parameters, wheel mass parameters, air spring nonlinear stiffness parameters, and variable damping shock absorber damping characteristic parameters.

[0189] Step S812: Construct a multi-degree-of-freedom vehicle vibration differential equation based on the vehicle body mass parameters, wheel mass parameters, air spring nonlinear stiffness parameters, and variable damping shock absorber damping characteristic parameters.

[0190] Step S813: Substitute the preprocessed signal and operating condition results into the multi-degree-of-freedom vehicle vibration differential equation to calculate the vehicle body vertical displacement, vehicle body pitch angle change rate, and vehicle body roll angle change rate, which are used as the dynamic response law of the suspension system.

[0191] Specifically, classical mechanics methods, primarily based on Newton's second law, simplify a vehicle into a multi-degree-of-freedom vibrating entity containing both sprung and unsprung mass. By constructing multi-degree-of-freedom vehicle vibration differential equations, the degree of disturbance of road surface unevenness to the vehicle's attitude can be accurately quantified, allowing for the calculation of the vehicle's vertical displacement, pitch angle change rate, and roll angle change rate over a short period.

[0192] Step S82 provides advance advance of adjustment commands by predicting and obtaining the dynamic demand of the suspension in advance.

[0193] Step S821: The radial basis function neural network is used to perform nonlinear mapping and error compensation on the vertical displacement of the vehicle body, the rate of change of the vehicle body pitch angle, and the rate of change of the vehicle body roll angle obtained from the multi-degree-of-freedom vehicle vibration differential equation, so as to obtain high-precision dynamic response prediction values.

[0194] Step S822: Based on the high-precision dynamic response prediction value, the target working air pressure of the air spring and the target damping coefficient of the shock absorber required to maintain vehicle stability are derived in reverse as the dynamic requirements of the suspension.

[0195] Step S823: The target working air pressure of the air spring and the target damping coefficient of the shock absorber are used as the advance amount of the adjustment command and transmitted to the fuzzy neural network controller for command generation.

[0196] In one embodiment, the radial basis function neural network is a feedforward artificial neural network with a single hidden layer. Its hidden layer nodes use radial basis functions as activation functions, enabling it to approximate any continuous function with arbitrary precision. Since multi-degree-of-freedom vehicle vibration differential equations typically neglect some higher-order nonlinear factors during their establishment, directly calculated parameters such as the vehicle's vertical displacement may contain errors. Introducing a radial basis function neural network allows for error compensation of the differential equation calculation results through its powerful nonlinear mapping capabilities, thereby outputting more accurate and high-precision dynamic response predictions.

[0197] For example, when a vehicle is traveling at 80 km / h and is about to enter a bumpy section, the multi-degree-of-freedom vehicle vibration differential equation initially calculates that the vertical displacement of the vehicle body will increase dramatically. A radial basis function neural network, combined with historical driving data, compensates for this increase in displacement, obtaining a high-precision dynamic response prediction. Subsequently, based on this prediction, standardized values ​​are derived requiring the air spring's target working pressure to be increased to 0.45 MPa and the shock absorber's target damping coefficient to be reduced to 0.3. These two values ​​represent the suspension's dynamic requirements. The standardized values ​​of 0.45 MPa and 0.3 are transmitted as adjustment command lead times to the fuzzy neural network controller, allowing the controller to generate inflation and damping reduction commands before the vehicle actually enters the bumpy section. This mechanism of predicting and providing lead times effectively overcomes the lag in command output in traditional adjustment methods, avoiding severe vibrations caused by insufficient suspension adjustment at the initial stage of entering a bumpy section. This significantly improves the vehicle's ride smoothness and comfort, while reducing wear on suspension components caused by sudden impacts.

[0198] It should be noted that when a vehicle experiences a sudden load change, such as a sudden increase of 500 kg in the rear load, the multi-degree-of-freedom vehicle vibration differential equation will calculate a significant increase in the rate of change of the vehicle's pitch angle, resulting in a decrease in the rear height of the vehicle. After the radial basis function neural network performs high-precision compensation for this rate of change, it inversely derives the target working air pressure of the rear suspension air springs, which needs to be increased to 0.50 MPa. After this advance measure is issued, the inflation / deflation actuator controls the high-speed switching solenoid valve through pulse width modulation, continuously inflating for 0.8 seconds with a 50% duty cycle. This completes the air pressure replenishment before the load fully presses down on the vehicle body, maintaining the horizontal stability of the vehicle's attitude and further highlighting the excellent beneficial effect of obtaining the suspension dynamic requirements in advance under complex conditions.

[0199] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A suspension adjustment method based on an air spring damping device in a vehicle intelligent control system, characterized in that, include: The sensor collects vehicle driving status parameters in real time, including vehicle speed, longitudinal and lateral acceleration, steering angle, road surface excitation, vehicle body posture and load changes. The collected signals are preprocessed to obtain preprocessed signals which are then transmitted to the intelligent control unit. The intelligent control unit receives the preprocessed signal and combines it with a preset multi-mode control strategy to identify the vehicle's operating conditions and obtain the operating condition results. The intelligent control unit calls the vehicle dynamics model and combines the preprocessed signal and the working condition result to predict the key parameters of the suspension system to obtain the predicted parameters; The fuzzy neural network controller receives the preprocessed signal, the operating condition result, and the predicted parameters, and outputs adjustment instructions, including air spring inflation / deflation control instructions and shock absorber damping coefficient adjustment instructions. The actuator receives the adjustment command and controls the inflation and deflation of the air spring and the damping coefficient of the shock absorber through the high-speed switching solenoid valve and the pneumatic actuator; The sensor collects feedback parameters from the suspension system, compares these feedback parameters with the predicted parameters, and then corrects the adjustment command through the intelligent control unit.

2. The suspension adjustment method based on the air spring damping device of a vehicle intelligent control system according to claim 1, characterized in that, The sensor collects vehicle driving status parameters in real time, including vehicle speed, longitudinal and lateral acceleration, steering angle, road surface excitation, vehicle posture, and load changes. The collected signals undergo preprocessing, including: The vehicle speed sensor collects vehicle speed, the acceleration sensor collects longitudinal and lateral acceleration, the steering angle sensor collects steering angle, the road surface sensor collects road surface excitation, the attitude sensor collects vehicle attitude, and the load sensor collects load changes. The preprocessed signal, obtained by filtering, denoising, normalizing, and removing abnormal data from the acquired signal using the Kalman filter algorithm, is transmitted to the intelligent control unit.

3. The suspension adjustment method based on the air spring damping device of a vehicle intelligent control system according to claim 1, characterized in that, The intelligent control unit receives the preprocessed signal and combines it with a preset multi-mode control strategy to identify the vehicle's operating conditions, including: Extract the feature values ​​of each parameter in the preprocessed signal, including vehicle speed threshold, rate of change of acceleration, steering angle range, and road excitation intensity; The feature values ​​are matched with a preset working condition feature library to determine the working condition result, and the identification threshold is dynamically adjusted according to the load change.

4. The suspension adjustment method based on the air spring damping device of a vehicle intelligent control system according to claim 1, characterized in that, The intelligent control unit calls upon the vehicle dynamics model, combining the preprocessed signals and the operating condition results, to predict key parameters of the suspension system, including: The vehicle dynamics model encompasses the dynamic characteristics of the vehicle body, wheels, air springs, and shock absorbers, taking into account nonlinear stiffness and damping characteristics as well as load effects. The predicted parameters are obtained by using a radial basis function neural network compensation model to predict the target working air pressure shock absorber damping coefficient, the trend of vehicle body attitude change, and subsequent changes in road surface excitation.

5. A suspension adjustment method based on an air spring damping device of a vehicle intelligent control system according to claim 1, characterized in that, The fuzzy neural network controller receives the preprocessed signal, the operating condition result, and the predicted parameters, and outputs adjustment commands. These adjustment commands include air spring inflation / deflation control commands and shock absorber damping coefficient adjustment commands, including: Input is processed through fuzzy inference by combining preset fuzzy rules; The neural network self-learns and optimizes the fuzzy rules and network weights to output the air spring inflation / deflation control command and the shock absorber damping coefficient adjustment command according to the working conditions.

6. A suspension adjustment method based on an air spring damping device of a vehicle intelligent control system according to claim 1, characterized in that, The actuator receives the adjustment command and controls the inflation and deflation of the air spring and the damping coefficient of the shock absorber through a high-speed switching solenoid valve and a pneumatic actuator, including: The high-speed switching solenoid valve uses pulse width modulation technology to control the inflation / deflation rate and time according to the air spring inflation / deflation control command, and adjusts the working air pressure of the air spring. The pneumatic actuator adjusts the damping force of the variable damping shock absorber according to the damping coefficient adjustment command. When the air pressure sensor detects that the air pressure in the air tank is lower than the preset threshold, the air compressor starts to replenish the air, which is then processed by the filter and dryer.

7. A suspension adjustment method based on an air spring damping device of a vehicle intelligent control system according to claim 1, characterized in that, The sensor collects feedback parameters from the suspension system, compares these feedback parameters with the predicted parameters, and corrects the adjustment command through the intelligent control unit, including: The air pressure sensor collects the actual working air pressure of the air spring, the actual damping coefficient of the shock absorber, the actual value of the vehicle body posture, and the tire adhesion coefficient to obtain the feedback parameters. The deviation between the feedback parameter and the predicted parameter is calculated, and the adjustment command is corrected by using a fuzzy neural network combined with a PID algorithm to adjust the actuator's action parameters.

8. A suspension adjustment method based on an air spring damping device of a vehicle intelligent control system according to claim 1, characterized in that, The vehicle dynamics model, combining the preprocessed signal and the operating condition results, predicts key parameters of the suspension system, including: The vehicle dynamics model is established based on classical mechanics methods to describe the dynamic response law of the suspension system. By predicting and obtaining the dynamic demand for suspension in advance, adjustment commands can be given a lead time.

Images