Electronic control air suspension online self-calibration method and system based on machine learning
By embedding a machine learning model into the electronically controlled air suspension system, the system uses vehicle sensors to identify changes in the stiffness and damping of the suspension system in real time and dynamically adjusts the control parameters, thus solving the problem of performance degradation caused by aging and achieving intelligent adaptive control and performance maintenance.
Patent Information
- Application Number
- CN202511620606.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-11-07
- Publication Date
- 2026-02-03
AI Technical Summary
Existing electronically controlled air suspension systems experience a gradual decline in performance throughout the vehicle's lifespan due to component aging. They are unable to achieve online monitoring and real-time compensation, which affects vehicle comfort, handling, and system reliability.
By employing machine learning models combined with the vehicle's inherent sensors, the equivalent stiffness and damping rate of the suspension system are identified in real time. Control parameters are dynamically adjusted through feature vector comparison to achieve online self-calibration and maintain superior system performance.
It achieves intelligent adaptability of the suspension system, maintains superior performance throughout its lifespan, improves the long-term quality of vehicle use, reduces hardware costs, and provides predictive maintenance prompts.
Smart Images
Figure CN121453433A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of automotive electronic control technology, and in particular to an online self-calibration method and system for electronically controlled air suspension based on machine learning. Background Technology
[0002] The suspension system is an elastic device in a car that connects the chassis and wheels. It consists of elastic elements, guiding mechanisms, shock absorbers, and other components, and is mainly used to buffer road impacts and improve ride comfort. The performance of an air suspension system is reflected in the stiffness characteristics of the air springs and the damping characteristics of the dampers.
[0003] However, during the entire lifespan of a vehicle, the rubber air springs will experience changes in effective volume and stiffness characteristics due to creep and aging; the dampers will suffer from damping characteristics degradation due to internal fluid deterioration and valve wear. This slow performance degradation is a gradual process, causing the ECAS system's control performance to gradually deviate from its design optimum, manifesting as follows: Decreased comfort: The ability to filter road impacts is reduced, and vehicle vibration is aggravated.
[0004] Reduced handling: Vehicle attitude control (such as suppressing positioning and pitch) responds more slowly, resulting in reduced stability.
[0005] Reduced system reliability: Control based on inaccurate models will increase the burden on other components (such as air pumps and solenoid valves), resulting in decreased comfort, weakened operability, and reduced system reliability.
[0006] Traditional ECAS system control strategies rely on factory-preset fixed parameter models. Currently, the industry lacks the capability for online monitoring and real-time compensation of the performance degradation of these components. Replacement is typically only performed when the user experiences significant discomfort or the component completely fails, making it impossible to maintain continuous system performance.
[0007] To address the problems in existing technologies, this invention provides a machine learning-based online self-calibration method and system for electronically controlled air suspension. Summary of the Invention
[0008] The purpose of this invention is to provide an online self-calibration method and system for electronically controlled air suspension based on machine learning, so as to solve the technical problem of gradual performance degradation caused by slow aging in the prior art.
[0009] The technical solution of this invention is: an online self-calibration method for electronically controlled air suspension based on machine learning, comprising: During vehicle operation, vibration signals of the vehicle body mass and the non-vehicle body mass of the wheels are collected and preprocessed. From the preprocessed vibration signals, feature vectors describing the dynamic characteristics of the suspension system are extracted. The feature vectors are input into a pre-trained machine learning model, and the machine learning model outputs the identification values of the equivalent stiffness and equivalent damping rate of the current suspension system. Based on the comparison results between the identified equivalent stiffness value and equivalent damping ratio and the initial calibration parameters of the system, the control parameters of the electronically controlled air suspension system are dynamically adjusted.
[0010] Preferably, the feature vector includes at least one of the following: Frequency domain features obtained based on Fast Fourier Transform include the dominant frequency and the amplitude of the dominant frequency; Statistical characteristics obtained from time-domain analysis include root mean square value and variance; Transitivity characteristics calculated based on the acceleration signals of sprung and unsprung mass include the peak transitivity, peak frequency, and half-power bandwidth at the natural frequency of the suspension system.
[0011] Preferably, the control parameters of the electronically controlled air suspension system are dynamically adjusted based on the comparison results between the identified equivalent stiffness value and the equivalent damping ratio and the initial calibration parameters of the system. The method includes: The stiffness calibration factor is calculated based on the equivalent stiffness value and the pre-stored initial stiffness calibration value, and the damping calibration factor is calculated based on the equivalent damping ratio and the pre-stored initial damping calibration value. Based on the stiffness calibration factor and / or the damping calibration factor, query the preset control parameter mapping table or calculate the control parameter adjustment function; The adjustment amount of at least one control parameter is obtained based on the control parameter adjustment function; The corresponding control parameters are updated based on the adjustment amount.
[0012] Preferably, the stiffness calibration factor is the ratio of the equivalent stiffness value to the initial stiffness calibration value; the damping calibration factor is the ratio of the equivalent damping ratio to the initial damping calibration value.
[0013] Preferably, querying a preset control parameter mapping table or calculating the quantization relationship of the control parameter adjustment function based on the damping calibration factor includes: Based on the numerical range of the damping calibration factor, determine the compensation coefficient of the control gain of the control parameters; The product of the initial control parameter gain and the compensation coefficient represents the updated control gain of the control parameter. Based on the updated control gain of the control parameter, the adjustment amount of the control parameter is determined to compensate for the damping attenuation.
[0014] Preferably, querying a preset control parameter mapping table or calculating the quantization relationship of the control parameter adjustment function based on the stiffness calibration factor includes: Substituting the stiffness calibration factor into a preset linear or piecewise linear function, the adjustment amount of the proportional gain and / or integral gain in the vehicle height control PID loop is calculated.
[0015] Preferably, the control parameters include at least one of the following: feedback gain of the overhead hook control algorithm, feedback gain of the ground hook control algorithm, parameters of the vehicle height control PID loop, intervention threshold and intensity of the pitch suppression strategy, and intervention threshold and intensity of the roll suppression strategy.
[0016] Preferably, the machine learning model is a model obtained through supervised training, and the training data comes from vibration signals and their corresponding feature vectors and label values collected in bench tests or simulation environments under different spring stiffness and damper damping rates. The machine learning model is a deep neural network, support vector machine regression, or random forest regression model.
[0017] An electronically controlled air suspension system, used to implement the aforementioned machine learning-based online self-calibration method for electronically controlled air suspension, includes: The sensor module, located on the vehicle body and wheels, includes: a vehicle body acceleration sensor, a wheel acceleration sensor, and a suspension height sensor, used to collect vibration signals and vehicle status information; The data processing and recognition module, integrated into the vehicle's electronic control unit, includes: The preprocessing unit is used to preprocess the vibration signals and vehicle status information collected by the sensor module, including filtering, noise removal, and standardization. The feature extraction unit is used to extract frequency domain, time domain, and transmission characteristic features from the preprocessed vibration signal; The virtual sensor unit has a pre-trained machine learning model built in, which is used to identify the current equivalent stiffness and equivalent damping rate based on the feature vector. The control parameter update module, based on the comparison results between the identified equivalent stiffness and equivalent damping rate and the initial calibration parameters of the system, queries the mapping table or runs an adaptive algorithm to calculate the control parameter adjustment function, and dynamically adjusts the control parameters of the electronically controlled air suspension system. The pre-stored unit internally stores: a parameter update mapping table and initial calibration parameters; The suspension actuators, including the solenoid valves of the air springs and the damping adjusters, receive commands from the electronic control unit and execute specific suspension adjustment actions.
[0018] A computer-readable storage medium storing at least one instruction or at least one program, wherein the at least one instruction or at least one program is loaded and executed by a processor to implement the aforementioned machine learning-based online self-calibration method for electronically controlled air suspension.
[0019] Compared with the prior art, the advantages of the present invention are: (1) Intelligence and Adaptability: The ECAS system has achieved a leap from "fixed parameters" to "adaptive parameters", enabling the suspension to maintain superior performance throughout its life.
[0020] (2) Zero additional hardware cost: Fully explore and utilize the data value of the existing sensors in the ECAS system, and improve performance through algorithm innovation. No new physical sensors need to be added, resulting in extremely high cost-effectiveness.
[0021] (3) Enhance product life cycle value: It alleviates the performance degradation caused by component aging and improves the long-term use quality of the vehicle and user satisfaction.
[0022] (4) Predictive maintenance: The stiffness / damping decay trend identified by the system can serve as an advanced early warning signal for predictive maintenance, prompting users that the damper or spring may be nearing the end of its life, making it easier to plan for replacement in advance.
[0023] This invention provides an online self-calibration method based on machine learning. This method embeds a machine learning model into an automotive-grade ECU controller. Based on signals collected by the vehicle's inherent sensors, the machine learning model acts as a "virtual sensor" to perceive and quantify mechanical component characteristics (stiffness, damping) that cannot be directly measured by physical sensors in real time. It also calibrates compensation and updates control parameters in real time, so that the suspension system maintains optimal performance close to that of a new car, thus maintaining comfort. Attached Figure Description
[0024] The present invention will be further described below with reference to the accompanying drawings and embodiments: Figure 1 This is a schematic diagram of the principle architecture of the electronically controlled air suspension system described in this invention; Figure 2 This is a flowchart illustrating the overall process of the online self-calibration method described in this invention. Figure 3 This is a schematic diagram illustrating the control parameter update principle described in this invention; Figure 4 This is a schematic diagram of the structure of the electronically controlled air suspension system described in this invention; Figure 5 This is an example diagram illustrating the gain compensation mapping of the top hook in an embodiment of the present invention.
[0025] The components include: 1. Sensor module; 2. Data processing and recognition module; 3. Control parameter update module; 4. Suspension actuator. 21. Preprocessing unit; 22. Feature extraction unit; 23. Virtual sensor unit. Detailed Implementation
[0026] The present invention will be further described in detail below with reference to specific embodiments: This invention provides a machine learning-based online self-calibration method for electronically controlled air suspension, as shown in the attached figure. Figure 1 The provided system principle block diagram shows that this method treats the entire vehicle suspension system as a "black box" or "gray box" model. It utilizes the inherent sensor signals, takes the signals collected by the inherent sensors as input, and uses a machine learning model as a virtual sensor to reverse identify and output the characteristic values of its internal key components (springs, dampers). Based on the characteristic changes, the parameters are adjusted. The method combines online machine learning to identify component parameter attenuation with real-time compensation of control parameters to maintain the superior performance of the control components.
[0027] like Figure 2 As shown, the method specifically includes the following steps: Step 1: Data Acquisition and Preprocessing.
[0028] During normal vehicle operation, the ECAS controller continuously collects vibration signals from the body acceleration sensor (located on the sprung mass) and the wheel acceleration sensor (or the differential signal from the height sensor, located on the unsprung mass) at a fixed frequency. Simultaneously, the signals are filtered (e.g., low-pass filtering to remove high-frequency noise), mean-reduced, and standardized.
[0029] Step 2: Feature Engineering and Feature Extraction.
[0030] From the preprocessed time-domain signal, key feature vectors describing the dynamic characteristics of the suspension system are extracted. These features include, but are not limited to: Frequency domain characteristics: The dominant frequency, amplitude, and power spectral density (PSD) of the vibration signal are calculated using Fast Fourier Transform (FFT).
[0031] Time-domain statistical characteristics: Calculate the root mean square (RMS), variance, kurtosis, and margin factor of the signal.
[0032] Transmission characteristics: The transmission rate between the acceleration of the sprung mass and the unsprung mass is calculated. This transmission rate curve directly reflects the damping and stiffness characteristics of the system.
[0033] The peak value, peak frequency, and half-power bandwidth at the system's natural frequency directly reflect the system's damping and stiffness characteristics; damping attenuation is usually manifested as an increase in the peak transmissibility and a narrowing of the bandwidth.
[0034] Step 3: Model reasoning and parameter identification.
[0035] The feature vectors extracted in step two are input into a pre-trained machine learning model. The model outputs the identification values of the equivalent stiffness (denoted as K) and equivalent damping ratio (denoted as C) of the current system.
[0036] Machine learning models can be selected from supervised learning algorithms such as deep neural networks (DNN), support vector machines (SVR), or random forests.
[0037] Taking a deep neural network (DNN) as an example, the number of nodes in its input layer is consistent with the dimension of the feature vector, the hidden layer can be designed to have 3-5 layers, and the output layer has 2 nodes, which correspond to the equivalent stiffness and equivalent damping rate of the current system, respectively.
[0038] During the model development phase, a series of spring stiffnesses with different parameter levels and dampers with different attenuation degrees are artificially introduced into the bench test or simulation environment. A large amount of corresponding vibration data and features are collected for supervised training to establish a mapping relationship between "feature vector" and "stiffness / damping value".
[0039] Step 4: Adaptive update of control parameters.
[0040] See attached document Figure 3 As shown, the ECAS controller compares the identified current equivalent stiffness and equivalent damping ratio with the initial calibration parameters of the system. Based on the deviation value, it updates the mapping table or adaptive control algorithm according to the preset control parameters, dynamically adjusting the core control parameters of ECAS. For example: Update the feedback gain of control algorithms such as Skyhook and Groundhook.
[0041] Update the integral and derivative parameters of the vehicle height control PID loop to match the changed system response speed.
[0042] Update the intervention threshold and intensity of the roll / pitch suppression strategy.
[0043] Specifically, the content of adaptive parameter adjustment and updating based on parameter comparison results includes: The calibration factor is calculated as follows: The ratio of the identified current equivalent stiffness to the initial stiffness value of the system is used as the stiffness calibration factor (denoted as α); a stiffness calibration factor of 1 indicates normal stiffness characteristics, while a stiffness calibration factor less than 1 indicates stiffness decay. The ratio of the identified current equivalent damping ratio to the initial damping ratio of the system is used as the damping calibration factor (denoted as β); a damping calibration factor of 1 indicates normal damping characteristics, while a damping calibration factor less than 1 indicates damping decay.
[0044] Establishing a mapping relationship: The system pre-stores a control parameter update mapping table or embeds a control parameter adjustment function. This mapping table or function defines the quantitative relationship between stiffness calibration factors and damping calibration factors and specific control parameter adjustment amounts. This quantitative relationship is derived during the vehicle development phase through extensive simulation and bench testing, aiming to optimally compensate for performance losses caused by component degradation. This quantitative relationship can be linear, piecewise linear, or based on a specific transfer function.
[0045] Perform quantitative calibration: Based on the calculated calibration factor, accurately calculate the required adjustment values for the control parameters by querying a mapping table or using a calculation function, and then complete the update. The following provides a detailed description of specific embodiments for the implementation of quantitative benchmarking and dynamic updating of control parameters.
[0046] Example 1: Damping control for top hooks.
[0047] Taking damper attenuation as an example, the specific implementation process is as follows: Data Acquisition: The vehicle travels at 60 km / h on a Class B road surface, and the ECAS controller collects the vertical acceleration signal of the vehicle body for 10 seconds.
[0048] Feature extraction: Fast Fourier Transform (FFT) analysis of the signal revealed that the peak value of the transmissivity curve at the system's natural frequency (e.g., 1.2 Hz) was significantly higher than that of the new vehicle calibration data (indicating a weakened damping effect).
[0049] Model recognition: Input the feature (peak frequency, peak amplitude, curve bandwidth, etc.) into the trained deep neural network model (DNN).
[0050] Parameter update: The ECU controller increases the feedback gain according to the built-in strategy and the "hook damping control algorithm" to compensate for the attenuation of physical damping.
[0051] In detail, the content of the hook damping control algorithm includes: The initial damping ratio is based on the calibration of a brand new damper.
[0052] If the output damping calibration factor is β=0.75, then the damping attenuation is 25%. The corresponding gain compensation coefficient γ is obtained by querying the preset gain compensation mapping table.
[0053] See attached document Figure 5Table 1 in the table provides an example diagram of the gain compensation mapping for the antenna. Referring to Table 1, when β=0.75, the gain compensation coefficient γ=1.0 / 0.75≈1.33, so the updated antenna gain is the initial damping rate × 1.33. In other words, if a 25% decrease in damping rate is detected, the feedback gain of the antenna control algorithm will be increased by approximately 30% to compensate for the attenuation of physical damping.
[0054] Therefore, by quantifying the identified component characteristic attenuation into one or more "calibration factors" and based on a preset quantization mapping relationship, precise compensation of control parameters can be achieved.
[0055] Example 2: Adjustment of PID parameters for height control.
[0056] The PID parameters (Kp, Ki, Kd) of the vehicle height control loop are closely related to the overall stiffness of the suspension.
[0057] When the ratio α of the equivalent stiffness K to the initial stiffness value is identified as 0.8, it indicates that the stiffness decay is 20%, and the parameters are updated according to the preset PID parameter adjustment function.
[0058] The new Kp value of the PID is equal to the product of the original Kp value and the first update coefficient, which is (1+K1×(1-α)); the new Ki value of the PID is equal to the product of the original Ki value and the second update coefficient, which is (1+K2×(1-α)).
[0059] Among them, K 1、 K2 is an empirical coefficient obtained through system identification, for example, K1 = 0.5, K2 = 0.3; The PID parameter adjustment function is used to quantify and adjust the controller's response to match the changed controlled parameters.
[0060] By introducing calibration factors and mapping adjustment functions, the current equivalent parameter values are compared with the initial parameter values and quantized, thus completing the accurate correction of the system control parameters.
[0061] The calculated new control parameters (including overhead hook damping and suspension stiffness) are written into the active memory of the ECAS controller, overwriting the old parameters. The system can then monitor indicators such as the RMS value of vehicle acceleration within a short period of time to verify whether the performance has improved, thus forming a closed loop.
[0062] Test results showed that when the acceleration signal was collected again and its RMS value was calculated, the vehicle body vibration was significantly reduced and the comfort was restored.
[0063] Step 5: Continuous learning and model optimization.
[0064] Alternatively, the system can be designed as a closed-loop learning system, using the system's performance after updating control parameters (such as whether the vehicle's acceleration decreases) as feedback to fine-tune the machine learning model's recognition results, thereby achieving lifelong iterative optimization.
[0065] See attached document Figure 4 As shown, the present invention provides an electronically controlled air suspension system for implementing the above-mentioned online self-calibration method for electronically controlled air suspension based on machine learning. The system includes a sensor module 1 arranged on the vehicle body and wheels, a data processing and recognition module 2 integrated in the vehicle's electronic control unit, a control parameter update module 3, and a suspension actuator 4.
[0066] Sensor module 1 includes: a vehicle body acceleration sensor, a wheel acceleration sensor, and a suspension height sensor, used to collect vibration signals and vehicle status information; The data processing and recognition module 2 includes: a preprocessing unit 21, a feature extraction unit 22, and a virtual sensor unit 23.
[0067] The preprocessing unit 21 is used to preprocess the vibration signals and vehicle status information collected by the sensor module, including filtering, noise removal and standardization; the feature extraction unit 22 is used to extract frequency domain, time domain and transmission characteristic features from the preprocessed vibration signals; the virtual sensor unit 23 has a pre-trained machine learning model built in, which is used to identify the current equivalent stiffness and equivalent damping rate based on the feature vector.
[0068] The control parameter update module 3 has a built-in pre-storage unit, which is a non-volatile memory. The pre-storage unit stores a parameter update mapping table and initial calibration parameters (initial system stiffness value and initial system damping ratio). Based on the comparison results between the identified equivalent stiffness and equivalent damping ratio and the initial system calibration parameters, the control parameter update module 3 queries the mapping table or runs an adaptive algorithm to calculate the control parameter adjustment function, dynamically adjusting the control parameters of the electronically controlled air suspension system.
[0069] The suspension actuator 4 includes the solenoid valve of the air spring, the damping adjuster, etc., and receives instructions from the electronic control unit (e.g., ECU controller) to perform specific suspension adjustment actions.
[0070] The machine learning model is embedded in the automotive-grade ECU controller. Based on the signals collected by the vehicle's inherent sensors, this invention uses the machine learning model as a "virtual sensor" to perceive and quantify the characteristics (stiffness, damping) of mechanical components that cannot be directly measured by physical sensors in real time. Based on this, the control parameters are adaptively updated in a data-driven manner to perform "component aging compensation" and achieve real-time control of characteristic parameters in the vehicle's vibration environment to maintain comfort.
[0071] This invention also provides a computer-readable storage medium, which can be disposed in an electronic device to store at least one instruction, at least one program, code set, or instruction set related to implementing the online self-calibration method for electronically controlled air suspension in the method embodiments. The at least one instruction, at least one program, code set, or instruction set can be loaded and executed by the processor of the electronic device to implement the machine learning-based online self-calibration method for electronically controlled air suspension provided in the above method embodiments.
[0072] Optionally, in this embodiment, the storage medium may include, but is not limited to, various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.
[0073] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments, while other embodiments fall within the scope of the appended claims. In some cases, the actions or steps described in the claims can be performed in a different order than those shown in the embodiments and still achieve the desired results. Additionally, the processes depicted in the drawings do not necessarily require a specific or sequential order to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.
[0074] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the apparatus embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.
[0075] Those skilled in the art will understand that all or part of the steps of the above embodiments can be implemented by hardware or by a program instructing related hardware. The program can be stored in a computer-readable storage medium, such as a read-only memory, a disk, or an optical disk.
[0076] The above description is merely a preferred embodiment of the present invention and should not be construed as limiting the scope of the invention. Therefore, any equivalent variations made in accordance with the claims of the present invention are still within the scope of the present invention.
Claims
1. A machine learning-based online self-calibration method for electronically controlled air suspension, characterized in that, include: During vehicle operation, vibration signals of the vehicle body mass and the non-vehicle body mass of the wheels are collected, and the vibration signals are preprocessed. From the preprocessed vibration signal, extract the feature vector to describe the dynamic characteristics of the suspension system; The feature vector is input into a pre-trained machine learning model, which outputs the identification values of the equivalent stiffness and equivalent damping rate of the current suspension system. Based on the comparison results between the identified equivalent stiffness value and equivalent damping rate and the initial calibration parameters of the system, the control parameters of the electronically controlled air suspension system are dynamically adjusted.
2. The online self-calibration method for electronically controlled air suspension based on machine learning according to claim 1, characterized in that, The feature vector includes at least one of the following: Frequency domain features obtained based on Fast Fourier Transform, including the dominant frequency and its amplitude; Statistical characteristics obtained from time-domain analysis include root mean square value and variance; Transitivity characteristics calculated based on the acceleration signals of sprung and unsprung mass include the peak transitivity, peak frequency, and half-power bandwidth at the natural frequency of the suspension system.
3. The online self-calibration method for electronically controlled air suspension based on machine learning according to claim 1, characterized in that, The method for dynamically adjusting the control parameters of the electronically controlled air suspension system based on the comparison results between the identified equivalent stiffness value and the equivalent damping ratio and the initial calibration parameters of the system includes: The stiffness calibration factor is calculated based on the equivalent stiffness value and the pre-stored initial stiffness calibration value, and the damping calibration factor is calculated based on the equivalent damping ratio and the pre-stored initial damping calibration value. Based on the stiffness calibration factor and / or the damping calibration factor, query the preset control parameter mapping table or calculate the control parameter adjustment function; The adjustment amount of at least one control parameter is obtained based on the control parameter adjustment function; The corresponding control parameters are updated based on the adjustment amount.
4. The online self-calibration method for electronically controlled air suspension based on machine learning according to claim 3, characterized in that, The stiffness calibration factor is the ratio of the equivalent stiffness value to the initial stiffness calibration value; the damping calibration factor is the ratio of the equivalent damping ratio to the initial damping calibration value.
5. The online self-calibration method for electronically controlled air suspension based on machine learning according to claim 3, characterized in that, Based on the damping calibration factor, query the preset control parameter mapping table or calculate the quantization relationship of the control parameter adjustment function, including: Based on the numerical range of the damping calibration factor, determine the compensation coefficient of the control gain of the control parameters; The product of the initial control parameter gain and the compensation coefficient represents the updated control gain of the control parameter. Based on the updated control gain of the control parameter, the adjustment amount of the control parameter is determined to compensate for the damping attenuation.
6. The online self-calibration method for electronically controlled air suspension based on machine learning according to claim 3, characterized in that, Based on the stiffness calibration factor, query the preset control parameter mapping table or calculate the quantization relationship of the control parameter adjustment function, including: Substituting the stiffness calibration factor into a preset linear or piecewise linear function, the adjustment amount of the proportional gain and / or integral gain in the vehicle height control PID loop is calculated.
7. The online self-calibration method for electronically controlled air suspension based on machine learning according to claim 1, characterized in that, The control parameters include at least one of the following: feedback gain of the overhead hook control algorithm, feedback gain of the ground hook control algorithm, parameters of the vehicle height control PID loop, intervention threshold and intensity of the pitch suppression strategy, and intervention threshold and intensity of the roll suppression strategy.
8. The online self-calibration method for electronically controlled air suspension based on machine learning according to claim 1, characterized in that, The machine learning model is a model obtained through supervised training. The training data comes from vibration signals and their corresponding feature vectors and label values collected in bench tests or simulation environments under different spring stiffness and damper damping rates. The machine learning model is a deep neural network, support vector machine regression, or random forest regression model.
9. An electronically controlled air suspension system for implementing the online self-calibration method for electronically controlled air suspension based on machine learning as described in any one of claims 1-8, characterized in that, include: The sensor module, located on the vehicle body and wheels, includes: a vehicle body acceleration sensor, a wheel acceleration sensor, and a suspension height sensor, used to collect vibration signals and vehicle status information; The data processing and recognition module, integrated into the vehicle's electronic control unit, includes: The preprocessing unit is used to preprocess the vibration signals and vehicle status information collected by the sensor module, including filtering, noise removal, and standardization. The feature extraction unit is used to extract frequency domain, time domain, and transmission characteristic features from the preprocessed vibration signal; The virtual sensor unit has a pre-trained machine learning model built in, which is used to identify the current equivalent stiffness and equivalent damping rate based on the feature vector. The control parameter update module, based on the comparison results between the identified equivalent stiffness and equivalent damping rate and the initial calibration parameters of the system, queries the mapping table or runs an adaptive algorithm to calculate the control parameter adjustment function, and dynamically adjusts the control parameters of the electronically controlled air suspension system. The pre-stored unit internally stores: a parameter update mapping table and initial calibration parameters; The suspension actuators, including the solenoid valves of the air springs and the damping adjusters, receive commands from the electronic control unit and execute specific suspension adjustment actions.
10. A computer-readable storage medium, characterized in that, The storage medium stores at least one instruction or at least one program segment, which is loaded and executed by a processor to implement a machine learning-based online self-calibration method for electronically controlled air suspension as described in any one of claims 1-8.