Electric vehicle semi-active suspension vibration reduction method and system based on magnetorheological fluid damper

Abstract

The invention relates to the technical field of automobile vibration reduction, in particular to an electric automobile semi-active suspension vibration reduction method and system based on a magnetorheological fluid damper, and the method comprises the steps: testing a target suspension through a vehicle road simulation test bench according to a preset scheme, and obtaining an initial vibration reduction performance data set; performing deviation analysis in combination with the structural characteristics of the whole vehicle, and outputting an influence coefficient set to adjust the initial data to obtain an optimized data set; predicting performance attenuation according to a plurality of preset mileage intervals according to the road environment characteristics of the common driving area, outputting an attenuation coefficient set, correcting optimized data, and generating a multi-stage prediction data set; the self-adaptive semi-active suspension vibration reduction control result is generated by integrating the mileage intervals and the prediction data set, accurate mapping from laboratory static testing to real vehicle full-life-cycle dynamic application is achieved, the problems that a traditional method neglects the whole vehicle coupling effect and the long-term service performance declines are effectively solved, and the method is suitable for large-scale popularization and application. And the self-adaptability and the long-term reliability of vibration reduction control are obviously improved.

Description

Technical Field

[0001] This invention relates to the field of automotive vibration reduction technology, and in particular to a semi-active suspension vibration reduction method and system for electric vehicles based on a magnetorheological fluid damper. Background Technology

[0002] With the rapid development of electric vehicle technology, vehicle ride smoothness and passenger comfort have become important indicators for evaluating overall vehicle performance. Semi-active suspension systems, due to their combination of excellent damping performance and low energy consumption, are widely used in high-end electric vehicles. Among them, semi-active suspensions based on magnetorheological fluid dampers have become a hot topic in current intelligent suspension system research and application due to their advantages such as fast response speed, continuously adjustable damping, and high control precision. However, in actual use, the damping performance of existing semi-active suspensions is not only affected by their own structural parameters but also closely related to the vehicle's mass distribution, driving conditions, and long-term service environment. This makes it difficult for laboratory test data to accurately reflect the vehicle's actual performance throughout its entire lifecycle.

[0003] Currently, most vibration reduction performance evaluation methods rely on bench tests or simulation analysis under standard operating conditions, lacking a comprehensive consideration of the coupled effects of multiple factors during actual vehicle operation. In particular, they neglect the matching influence of different vehicle model structural characteristics on suspension performance and the time-varying performance degradation caused by complex road environments. Furthermore, magnetorheological fluid dampers are susceptible to temperature changes, road surface excitation, and seal aging during long-term use, leading to a decrease in damping force output and thus weakening the dynamic control capability of the suspension system. Therefore, traditional static testing methods are no longer sufficient to meet the requirements of high-precision, long-cycle, and adaptive vibration reduction control.

[0004] The above content is only used to help understand the technical solution of the present invention and does not represent an admission that the above content is prior art. Summary of the Invention

[0005] The main objective of this invention is to provide a semi-active suspension vibration reduction method and system for electric vehicles based on magnetorheological fluid dampers. This aims to solve the technical problem that existing semi-active suspension vibration reduction evaluation methods neglect the matching of vehicle structural characteristics, the influence of complex road environments, and long-term service performance degradation, resulting in laboratory test data that cannot accurately reflect the actual vibration reduction performance throughout the vehicle's entire life cycle.

[0006] To achieve the above objectives, the present invention provides a semi-active suspension vibration reduction method for electric vehicles based on a magnetorheological fluid damper, the method comprising:

[0007] According to the preset test plan, the target semi-active suspension was tested for vibration reduction performance using a vehicle road simulation test bench to obtain an initial vibration reduction performance dataset.

[0008] By combining the overall vehicle structural characteristics of the electric vehicle where the target semi-active suspension is located, a vibration reduction performance deviation analysis is performed, and a set of vibration reduction performance influence coefficients is output. The initial vibration reduction performance dataset is then adjusted to obtain an optimized vibration reduction performance dataset.

[0009] Combining the road environment characteristics of the target semi-active suspension's commonly used driving areas, the damping performance attenuation is predicted according to multiple preset driving mileage intervals, and multiple predicted damping performance attenuation coefficient sets are output. The optimized damping performance dataset is then corrected to obtain multiple predicted damping performance datasets.

[0010] Semi-active suspension damping control results are generated based on the multiple preset driving mileage ranges and the multiple predicted damping performance datasets.

[0011] Optionally, the step of conducting vibration reduction performance tests on the target semi-active suspension using a vehicle road simulation test bench according to a preset test plan to obtain an initial vibration reduction performance dataset includes:

[0012] Obtain a preset test plan, wherein the preset test plan includes test indicators, test procedures and parameter configurations, and the test indicators include at least vehicle body acceleration, suspension dynamic deflection, tire dynamic displacement, damping adjustment response speed and vibration reduction energy consumption rate;

[0013] Based on the aforementioned test indicators, and following the test procedures and parameter configurations, a multi-condition road loading test was conducted on a semi-active suspension equipped with a magnetorheological fluid damper using a vehicle road simulation test bench, and an initial vibration reduction performance dataset was output.

[0014] Optionally, the vibration damping performance deviation analysis, which combines the overall vehicle structural characteristics of the electric vehicle where the target semi-active suspension is located, outputs a set of vibration damping performance influence coefficients, including:

[0015] Obtain the overall vehicle structural characteristics of the electric vehicle in which the target semi-active suspension is located, wherein the overall vehicle structural characteristics include at least vehicle type, curb weight, chassis structure type and axle load distribution;

[0016] Obtain the structural attribute information of the target semi-active suspension, and expand the structural attribute information according to a preset feature tolerance range to obtain a structural attribute range. The structural attribute information includes at least the damper specifications, suspension configuration, spring stiffness, and lower control arm mechanical parameters.

[0017] Guided by the target semi-active suspension, using the structural attribute range as the retrieval and comparison conditions, and constrained by a preset time range, big data technology is used to retrieve sample data, obtain the sample whole vehicle structural feature set and multiple sample vibration reduction performance influence coefficient sets;

[0018] Using the sample whole vehicle structural feature set as input and the multiple sample vibration reduction performance influence coefficient sets as supervision, a BP neural network is trained until convergence to obtain the vibration reduction performance influence analysis model.

[0019] Using the vibration reduction performance influence analysis model, vibration reduction performance deviation analysis is performed based on the vehicle structural characteristics, and a set of vibration reduction performance influence coefficients is output, wherein the vibration reduction performance influence coefficients correspond one-to-one with the test indicators.

[0020] Optionally, the step of using big data technology to retrieve sample data and obtain a set of structural features of the whole vehicle and a set of vibration reduction performance influence coefficients for multiple samples includes:

[0021] Using big data technology, sample data is retrieved to obtain multiple sample vehicle structural features that satisfy the structural attribute interval and the preset time range, and a sample vehicle structural feature set is constructed.

[0022] Multiple historical vibration reduction data of different sample vehicle structural features under actual driving load are obtained. Based on the test index, the mapping deviation of the multiple historical vibration reduction data and multiple vibration reduction test data under the same test road load is compared to obtain multiple index deviation sets. The mean set of index deviations is calculated as the sample vibration reduction performance influence coefficient set, and multiple sample vibration reduction performance influence coefficient sets are obtained. Among them, the index deviation is the ratio of the difference between the historical vibration reduction data and the vibration reduction test data to the vibration reduction test data.

[0023] Optionally, the method combines the road environment characteristics of the target semi-active suspension's commonly used driving areas, predicts the damping performance degradation according to multiple preset driving mileage intervals, and outputs multiple sets of predicted damping performance degradation coefficients, including:

[0024] Multiple preset driving mileage intervals are configured according to preset mileage intervals;

[0025] Obtain the road environment characteristics of the target semi-active suspension's commonly used driving areas. These road environment characteristics include road surface smoothness distribution, average annual driving temperature difference, frequency of road water accumulation, and proportion of bumpy road sections.

[0026] A vibration reduction performance degradation prediction model was obtained based on BP neural network training.

[0027] Using the vibration reduction performance attenuation prediction model, based on the road environment characteristics, vibration reduction performance attenuation is predicted according to multiple preset driving mileage intervals, and multiple sets of predicted vibration reduction performance attenuation coefficients are output.

[0028] Optionally, the vibration reduction performance degradation prediction model obtained based on BP neural network training includes:

[0029] Based on the historical operation and maintenance monitoring records of similar magnetorheological fluid semi-active suspensions, a sample mileage set and a sample road environment feature set were collected. The performance degradation ratio of multiple test indicators under different sample mileage and sample environment features was calculated and set as the sample performance degradation coefficient, thus obtaining the sample performance degradation coefficient set.

[0030] Using the sample driving mileage set and sample road environment feature set as inputs, and the sample performance attenuation coefficient set as supervision, a BP neural network is trained until convergence to obtain a vibration reduction performance attenuation prediction model.

[0031] Optionally, generating semi-active suspension damping control results based on the plurality of preset driving mileage ranges and the plurality of predicted damping performance datasets includes:

[0032] The multiple preset driving mileage intervals and the multiple predicted vibration reduction performance datasets are mapped and combined to generate a semi-active suspension vibration reduction control parameter table adapted to different usage stages, which serves as the semi-active suspension vibration reduction control result.

[0033] Furthermore, to achieve the above objectives, the present invention also provides a semi-active suspension damping system for electric vehicles based on a magnetorheological fluid damper, the system comprising:

[0034] The bench test module is used to test the damping performance of the target semi-active suspension using a vehicle road simulation test bench according to a preset test plan, and to obtain the initial damping performance dataset.

[0035] The structural correction module is used to perform vibration reduction performance deviation analysis based on the overall vehicle structural characteristics of the electric vehicle where the target semi-active suspension is located, output a set of vibration reduction performance influence coefficients, and adjust the initial vibration reduction performance dataset to obtain an optimized vibration reduction performance dataset.

[0036] The attenuation prediction module is used to combine the road environment characteristics of the target semi-active suspension in the commonly used driving area, predict the attenuation of vibration reduction performance according to multiple preset driving mileage intervals, output multiple sets of predicted vibration reduction performance attenuation coefficients, and correct the optimized vibration reduction performance datasets respectively to obtain multiple predicted vibration reduction performance datasets.

[0037] The control generation module is used to generate semi-active suspension damping control results based on the multiple preset driving mileage intervals and the multiple predicted damping performance datasets.

[0038] Furthermore, to achieve the above objectives, the present invention also provides a semi-active suspension damping device for electric vehicles based on a magnetorheological fluid damper. The device includes: a memory, a processor, and a semi-active suspension damping program for electric vehicles based on a magnetorheological fluid damper stored in the memory and executable on the processor. The semi-active suspension damping program for electric vehicles based on a magnetorheological fluid damper is configured to implement the steps of the semi-active suspension damping method for electric vehicles based on a magnetorheological fluid damper as described above.

[0039] Furthermore, to achieve the above objectives, the present invention also provides a computer-readable storage medium storing a semi-active suspension damping program for an electric vehicle based on a magnetorheological fluid damper. When the semi-active suspension damping program for an electric vehicle based on a magnetorheological fluid damper is executed by a processor, it implements the steps of the semi-active suspension damping method for an electric vehicle based on a magnetorheological fluid damper as described above.

[0040] This invention provides a semi-active suspension vibration reduction method for electric vehicles based on magnetorheological fluid dampers. This method incorporates vehicle structural characteristics for deviation analysis and outputs influence coefficients, overcoming the limitations of traditional bench tests that only test single components. It corrects initial data from ideal laboratory environments to optimized data that closely matches the actual installation state of a specific vehicle model, eliminating performance evaluation errors caused by vehicle coupling effects and making the test results more accurately reflect the actual performance of the suspension on the target vehicle. It innovatively combines road environment characteristics from commonly used driving areas and predicts performance degradation according to mileage intervals. This not only considers the current vibration reduction capability but also proactively predicts the performance degradation trend of the magnetorheological fluid damper due to aging, wear, and environmental corrosion at different stages of use. This phased correction mechanism allows the vibration reduction control strategy to cover the entire lifecycle of the vehicle, from new to old, avoiding control failure or decreased comfort due to component performance degradation. The vibration reduction control results generated based on multi-stage prediction datasets are no longer static, fixed parameter tables but adaptive strategies that can dynamically adjust with increasing mileage and environmental changes. This allows the semi-active suspension system to maintain optimal damping force output throughout different stages of the vehicle's service life, thus ensuring both ride comfort and handling stability, significantly extending the effective service life of the suspension system. It also utilizes big data technology and neural network models to replace numerous repetitive and costly long-term real-vehicle road tests. By simulating performance evolution under different structures and environments in a virtual environment, potential problems can be identified and control algorithms optimized early in the design process, significantly shortening the development cycle and improving R&D efficiency. Attached Figure Description

[0041] Figure 1This is a flowchart illustrating an embodiment of the semi-active suspension vibration reduction method for electric vehicles based on magnetorheological fluid dampers according to the present invention.

[0042] Figure 2 This is a structural block diagram of an embodiment of the semi-active suspension damping system for electric vehicles based on a magnetorheological fluid damper according to the present invention.

[0043] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0044] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0045] Reference Figure 1 , Figure 1 This is a flowchart illustrating an embodiment of the semi-active suspension vibration reduction method for electric vehicles based on a magnetorheological fluid damper according to the present invention.

[0046] In one embodiment, the semi-active suspension vibration reduction method for electric vehicles based on magnetorheological fluid dampers includes:

[0047] Step S100: According to the preset test plan, the target semi-active suspension is tested for vibration reduction performance using a vehicle road simulation test bench to obtain an initial vibration reduction performance dataset.

[0048] The target semi-active suspension can be a semi-active suspension system installed on a specific electric vehicle, with a magnetorheological fluid damper as the core actuator. It can be used to dynamically adjust the damping force according to control commands to achieve real-time suppression of vehicle body vibration. In this embodiment, the target semi-active suspension can collect vehicle body state signals through onboard sensors and combine them with a control algorithm to drive the magnetorheological fluid damper to adjust the internal magnetic field strength, thereby changing the damping characteristics. Furthermore, the target semi-active suspension can couple with the overall vehicle structural characteristics to affect actual vibration reduction performance. Initial performance testing is conducted using a vehicle road simulation test bench, and its performance degradation is affected by road environment characteristics and driving mileage range. The magnetorheological fluid damper is a hydraulic vibration reduction device that utilizes the principle of reversible change in the rheological properties of magnetorheological fluid under the action of a magnetic field to achieve continuously adjustable damping force. It can be used as the core actuator of a semi-active suspension, providing fast response and high-precision controllable damping force output. For example, the magnetorheological fluid damper can generate a controllable magnetic field by energizing a coil, changing the yield stress of the magnetorheological fluid, thereby adjusting the piston movement resistance.

[0049] Electric vehicles are motor vehicles powered primarily by electricity and can be used as mounting platforms for target semi-active suspensions. Their mass distribution, stiffness characteristics, and powertrain layout form the basis of the overall vehicle structural features. A vehicle road simulation test bench can be a bench testing device that reproduces the road excitation spectrum in a laboratory environment. It can be used to perform repeatable vibration damping performance tests on the target semi-active suspension under controlled conditions. In one embodiment, the vehicle road simulation test bench can simulate road surface unevenness input using hydraulic or electric actuators, simultaneously acquiring suspension response data. The preset test scheme can be a standardized test procedure that specifies test conditions, excitation types, sampling frequencies, and evaluation indicators, used to ensure the comparability and repeatability of the initial vibration damping performance dataset. The initial vibration damping performance dataset can be an uncorrected set of suspension dynamic response data obtained on the vehicle road simulation test bench according to the preset test scheme, which can be used as a benchmark input for subsequent deviation analysis and performance correction.

[0050] According to the pre-set test plan, the vibration reduction performance of the target semi-active suspension is tested using a vehicle road simulation test bench. This can be done in a laboratory environment by applying simulated road surface excitation according to standardized test procedures, while simultaneously acquiring the suspension's dynamic response signals. Furthermore, this operation can be achieved through broadband response testing using time-domain frequency sweep excitation, or through time-history reproduction testing based on measured road spectra, thereby obtaining repeatable and controlled initial vibration reduction performance benchmark data. The initial vibration reduction performance dataset can be obtained by exporting pre-processed suspension response data from the vehicle road simulation test bench's data acquisition system, thus forming the raw input for subsequent deviation correction and attenuation prediction.

[0051] Step S200: Combine the structural characteristics of the electric vehicle where the target semi-active suspension is located to perform vibration reduction performance deviation analysis, output the vibration reduction performance influence coefficient set, adjust the initial vibration reduction performance dataset, and obtain the optimized vibration reduction performance dataset.

[0052] The vehicle structural features can be a set of parameters reflecting the overall mechanical characteristics of an electric vehicle, including but not limited to sprung mass, unsprung mass, center of gravity position, body stiffness distribution, and suspension geometry. These features can be used to quantify the impact of vehicle coupling effects on the actual suspension performance. For example, vehicle structural features may include mass distribution characteristics, stiffness coupling characteristics, and inertial coupling characteristics. Vibration damping performance deviation analysis can be a quantitative assessment process comparing the performance differences between ideal bench test results and the actual vehicle installation state. This can be used to identify and quantify performance prediction errors caused by neglecting vehicle coupling effects. In a specific embodiment, this operation can involve establishing a dynamic model that includes the vehicle's mass-stiffness-damping coupling relationship and comparing the differences between bench test and vehicle simulation results. Furthermore, this operation can be achieved through coupling effect simulation analysis based on a multibody dynamics simulation platform, or by back-deriving structural coupling parameters and calculating deviations through actual vehicle modal testing, thereby quantifying the actual impact of the vehicle installation state on suspension performance. The set of vibration damping performance influence coefficients can be a multi-dimensional parameter set characterizing the proportion of correction of the initial vibration damping performance by the vehicle's structural features. It can be used to convert the initial vibration damping performance dataset into optimized data that closely matches the actual installation conditions. The optimized vibration damping performance dataset can be a set of vibration damping performance data corrected for deviations in the vehicle's structural features. It can be used to more accurately reflect the true dynamic performance of the target semi-active suspension on a specific vehicle model.

[0053] The output set of vibration reduction performance influence coefficients can be used to convert deviation analysis results into correction ratios or offsets for the initial performance data, thereby generating structural coupling compensation parameters that can be used for data correction. Adjusting the initial vibration reduction performance dataset to obtain an optimized vibration reduction performance dataset can be achieved by applying the set of vibration reduction performance influence coefficients to the initial data, performing point-by-point or segmented corrections, thus ensuring that the vibration reduction performance data conforms to the actual installation boundary conditions of a specific vehicle model.

[0054] Step S300: Combining the road environment characteristics of the target semi-active suspension's commonly used driving areas, predict the damping performance degradation according to multiple preset driving mileage intervals, output multiple predicted damping performance degradation coefficient sets, and correct the optimized damping performance datasets respectively to obtain multiple predicted damping performance datasets.

[0055] The frequently used driving area can be a set of geographical areas where the target vehicle operates frequently throughout its life cycle. This can be used to define the data source range for road environment characteristics, improving the regional specificity of degradation prediction. Road environment characteristics can be a set of statistical parameters describing typical road surface conditions within the frequently used driving area, including road surface grade, excitation spectrum, temperature fluctuation range, and humidity level. These can be used as key external input variables for performance degradation prediction. In an exemplary embodiment, road environment characteristics may include road surface excitation characteristics, climate environment characteristics, and corrosive medium characteristics. The preset mileage interval can be several discrete stages dividing the entire life cycle of the vehicle according to the cumulative mileage. This can be used to provide a time-dimensional segmentation basis for performance degradation prediction. For example, the preset mileage interval may include the break-in period interval, the stable service period interval, and the performance degradation period interval.

[0056] Vibration damping performance degradation prediction can be a modeling process based on road environment characteristics and driving mileage intervals to estimate the performance degradation of magnetorheological fluid dampers over time. This can be used to proactively quantify the decreasing trend of damping force output due to aging, wear, and environmental erosion during long-term service. In one specific embodiment, this operation can involve constructing a life prediction model that integrates environmental stress and usage intensity to estimate the degree of performance degradation in stages. Furthermore, this operation can be achieved by training a neural network degradation prediction model based on accelerated aging test data, or by establishing a performance degradation differential equation using a physics-data hybrid driving method, thereby enabling proactive modeling of the long-term performance evolution of magnetorheological fluid dampers. The predicted vibration damping performance degradation coefficient set can be a set of performance degradation correction factors corresponding to each preset driving mileage interval, which can be used to dynamically correct the optimized vibration damping performance dataset in stages. Multiple predicted vibration damping performance datasets can be sets of vibration damping performance data generated separately for different driving mileage intervals, which can be used to support the formulation of adaptive control strategies throughout the entire life cycle.

[0057] Outputting multiple sets of predicted vibration reduction performance attenuation coefficients can be achieved by calculating the corresponding performance attenuation correction factor for each preset driving mileage interval, thus providing a dynamic attenuation compensation basis for phased correction. The optimized vibration reduction performance dataset is then corrected separately to obtain multiple predicted vibration reduction performance datasets. This can be achieved by applying the attenuation coefficient sets for each mileage interval to the optimized dataset, generating predicted performance data for the corresponding stage, thereby forming a multi-stage performance evolution dataset covering the entire lifecycle.

[0058] Step S400: Generate semi-active suspension damping control results based on multiple preset driving mileage ranges and multiple predicted damping performance datasets.

[0059] Among them, the semi-active suspension damping control result can be a set of damping control strategies that are dynamically adjusted with driving mileage and environment based on a multi-stage prediction dataset. It can be used to maintain the optimal damping force output at different service stages of the vehicle, taking into account both comfort and handling stability.

[0060] The semi-active suspension damping control results are generated based on multiple preset driving mileage ranges and multiple predicted damping performance datasets. This can be achieved by mapping the predicted performance data at each stage to corresponding damping control parameter tables or control laws. Furthermore, this operation can be implemented by constructing a lookup-based segmented control strategy library and automatically switching according to the current mileage, or by training a deep reinforcement learning agent and generating real-time control commands with predicted performance data as state input, thereby generating a dynamic control strategy with mileage and environment adaptability.

[0061] Taking the suspension tuning of a high-end electric sedan throughout its entire lifecycle as an example, the semi-active suspension vibration reduction method for electric vehicles based on magnetorheological fluid dampers in this embodiment can be as follows: During the development phase of a high-end electric sedan, the magnetorheological semi-active suspension is first tested on a vehicle road simulation test bench according to the ISO standard spectrum to obtain an initial vibration reduction performance dataset; then, combined with the low center of gravity battery layout and lightweight body structure characteristics unique to this model, deviation analysis is performed through a multibody dynamics model to output an influence coefficient set and generate an optimized dataset; furthermore, based on the main sales areas of this model (such as urban roads in East China and mountainous roads in Southwest China)... Typical road surface spectra and annual average temperature and humidity data were used to divide the driving mileage into three ranges: 0–50,000 km, 50,000–100,000 km, and 100,000–150,000 km. A neural network model was used to predict the damping force attenuation trend caused by the aging of magnetorheological fluid seals and changes in fluid viscosity at each stage, and three sets of attenuation coefficients were output. Finally, the three sets of corrected predicted vibration reduction performance datasets were converted into Skyhook control parameters for the corresponding mileage ranges and embedded into the vehicle controller. This allows the vehicle to provide a soft and comfortable experience in the early stages of new vehicle delivery and maintain sufficient roll suppression capability after 100,000 km, avoiding a decrease in cornering stability due to damping attenuation.

[0062] This embodiment overcomes the evaluation distortion caused by neglecting the vehicle coupling effect in traditional methods by introducing the structural characteristics of the entire vehicle to correct the deviation of the initial bench test data. This makes the optimized vibration damping performance dataset more consistent with the actual installation state of a specific vehicle model. Furthermore, it combines the road environment characteristics of commonly used driving areas and performs phased vibration damping performance degradation prediction according to preset driving mileage intervals. The optimized data is dynamically corrected using the predicted vibration damping performance degradation coefficient set, thereby proactively compensating for the performance degradation of the magnetorheological fluid damper caused by long-term service factors such as temperature changes, seal aging, and road surface excitation. Finally, the vibration damping control result generated based on the multi-stage prediction dataset has full life cycle adaptive capability. It is no longer a static parameter table, but a control strategy that dynamically adjusts with driving mileage and environmental evolution, ensuring that the vehicle can maintain optimal damping force output at different stages of use, improving ride comfort while ensuring handling stability. In addition, this method relies on big data and neural network models to efficiently simulate the performance evolution under structure-environment coupling in a virtual environment, replacing a large number of high-cost and long-cycle real vehicle road tests, significantly shortening the development cycle and improving R&D efficiency. It achieves the technical effect of a technological leap from component-level static testing to system-level dynamic prediction.

[0063] In one embodiment, according to a preset test plan, the target semi-active suspension is subjected to vibration reduction performance testing using a vehicle road simulation test bench to obtain an initial vibration reduction performance dataset, including:

[0064] Obtain a preset test plan, wherein the preset test plan includes test indicators, test procedures and parameter configurations, and the test indicators include at least vehicle body acceleration, suspension dynamic deflection, tire dynamic displacement, damping adjustment response speed and vibration reduction energy consumption rate.

[0065] The pre-set test plan can be a standardized procedure specifying the index system, execution process, and equipment parameter configuration for vibration reduction performance testing. This ensures the initial vibration reduction performance dataset possesses multidimensionality, repeatability, and engineering representativeness. Test indices can be a set of key physical quantities used to quantify the vibration reduction performance of semi-active suspensions. These indices can characterize the system's response characteristics and control effectiveness under dynamic excitation from multiple dimensions. The test process can specify the test execution sequence, operating condition switching logic, and data acquisition node operation sequence. This standardizes the implementation order of multi-condition loading, ensuring the systematic nature and integrity of the testing process. Parameter configuration can be a set of set values ​​for the excitation amplitude, frequency range, sampling rate, and control mode used by the vehicle road simulation test bench in the test. This ensures that the test conditions match the statistical characteristics of the target road environment.

[0066] Vehicle acceleration can be the acceleration response of the vehicle's center of gravity or key locations in the vertical, lateral, or longitudinal directions. It directly reflects the level of ride comfort and is a core evaluation criterion for vibration damping performance. Suspension dynamic deflection can be the change in compression and extension travel of the suspension system under dynamic loads. It characterizes suspension travel utilization and impact limiting risk, affecting ride comfort and structural safety. Tire dynamic displacement can be the dynamic vertical displacement of the tire relative to the wheel hub or vehicle body. It reflects wheel contact performance and directly affects driving safety and handling stability. Damping adjustment response speed can be the time required for a magnetorheological fluid damper to output the target damping force from receiving a control command. It is a key dynamic indicator for measuring the real-time adjustment capability of a semi-active system. Vibration damping energy consumption rate can be the ratio of electrical or thermal energy consumed by the suspension system per unit time to the vibration damping effect. It can be used to evaluate the energy efficiency level of semi-active suspensions and support the optimization design of electric vehicle range.

[0067] Obtaining a preset test plan can be achieved by retrieving a complete test procedure containing test metrics, test procedures, and parameter configurations from a test database or standard document. Furthermore, obtaining a preset test plan can be accomplished through structured calls, thus providing a structured execution basis for subsequent multi-condition testing.

[0068] Based on the test indicators, and following the test procedures and parameter configurations, a multi-condition road loading test was conducted on a semi-active suspension equipped with a magnetorheological fluid damper using a vehicle road simulation test bench, and an initial vibration reduction performance dataset was output.

[0069] The semi-active suspension equipped with a magnetorheological fluid damper can be a suspension system with real-time adjustable damping characteristics, using the magnetorheological fluid damper as the actuating unit. It can be used as a test object, and its dynamic response constitutes the source of the initial vibration reduction performance dataset. In an exemplary embodiment, the semi-active suspension equipped with a magnetorheological fluid damper can undergo multi-condition road loading tests on a vehicle road simulation test bench, and its response data is quantified by a test index system. Multi-condition road loading testing can be a composite test method that sequentially or in combination applies multiple typical road excitation spectra in a single test process, which can be used to improve the coverage of the complexity of real roads in laboratory tests. In a specific embodiment, multi-condition road loading testing can include random-sinusoidal composite loading, urban-highway-off-road spectrum sequence loading, temperature-excitation coupled loading, etc.

[0070] Based on test indicators and following the test procedure and parameter configuration, a semi-active suspension equipped with a magnetorheological fluid damper undergoes multi-condition road loading tests using a vehicle road simulation test bench. This can be achieved by sequentially executing different road spectrum excitations on the test bench according to a pre-set test plan, simultaneously collecting dynamic response data for various test indicators. Furthermore, this operation can be implemented by using time-domain road spectrum reproduction technology to drive the actuator based on measured road elevation data, or by synthesizing random excitations based on power spectral density (PSD) and combining them with frequency-sweeping sine waves to simulate resonance conditions. This allows for the reproduction of complex road loads in a controlled environment, obtaining high-fidelity, multi-dimensional initial performance data. The output initial damping performance dataset can be generated by organizing and storing data such as vehicle acceleration and suspension dynamic deflection collected during multi-condition tests according to the operating conditions and time series. Furthermore, the output initial damping performance dataset can be implemented using a structured data format, thus forming a structured, multi-dimensional initial performance benchmark to support subsequent deviation correction and attenuation prediction.

[0071] Taking suspension bench verification during the development phase of a new vehicle model as an example, the semi-active suspension vibration reduction method for electric vehicles based on magnetorheological fluid dampers in this embodiment can be as follows: During the suspension selection phase, an electric SUV development team first obtains a pre-defined test plan containing five indicators: vehicle vertical acceleration, suspension dynamic deflection, tire dynamic displacement, damping response speed, and energy consumption rate. This plan specifies that a Class B random road surface spectrum test is first conducted, followed by a 3Hz sinusoidal sweep frequency to induce vehicle resonance, and finally, repeated testing in a high-temperature environment to evaluate thermal decay. Based on this, the team performs multi-condition loading on a semi-active suspension equipped with magnetorheological fluid dampers on a vehicle road simulation test bench, simultaneously collecting data from various sensors, and finally outputting an initial vibration reduction performance dataset covering comfort, safety, responsiveness, and energy efficiency. This dataset is then used for vehicle structural deviation correction, ensuring that subsequent control strategy development is based on a more realistic performance baseline.

[0072] This embodiment, by constructing a standardized test index system, a standardized test process, and parameter configurations that match real road characteristics, enables multi-condition excitation to fully stimulate the dynamic response characteristics of the suspension system. This achieves the technical effect of efficiently acquiring high-fidelity, multi-dimensional, and highly representative initial damping performance datasets in a laboratory environment. This dataset not only improves the input quality for subsequent deviation correction and attenuation prediction but also enhances the comparability of performance data between different suspension configurations. It provides structured, high-quality samples for big data modeling, neural network training, and virtual simulation to replace real-vehicle road testing, thereby shortening the development cycle and improving the robustness of the control algorithm.

[0073] In one embodiment, a vibration damping performance deviation analysis is performed based on the overall vehicle structural characteristics of the electric vehicle where the target semi-active suspension is located, outputting a set of vibration damping performance influence coefficients, including:

[0074] Obtain the overall structural characteristics of the electric vehicle in which the target semi-active suspension is located. The overall structural characteristics include at least the vehicle type, curb weight, chassis structure, and axle load distribution.

[0075] Among these, vehicle type can be a classification identifier for electric vehicles based on purpose, size, or drive type, serving as a fundamental dimension of the vehicle's structural characteristics and influencing mass distribution and suspension layout design. Curb weight can be the total mass of an electric vehicle in its standard equipped state, unloaded and loaded with cargo, directly affecting the sprung mass and serving as a key input parameter for matching suspension stiffness and damping. Chassis structure type can describe the structural type of the vehicle chassis's load-bearing and connection methods, determining the stiffness of the suspension mounting points and the body-suspension coupling characteristics. For example, chassis structure types can include monocoque chassis, body-on-frame, and semi-monocoque platforms. Axle load distribution can be the proportion or absolute value of the static load borne by the front and rear axles, influencing the suspension preload state and dynamic load transfer characteristics, thereby altering the damping response.

[0076] Obtaining the overall structural characteristics of the electric vehicle with the target semi-active suspension can be achieved by extracting parameters such as vehicle type, curb weight, chassis structure, and axle load distribution from a vehicle database or CAD / BOM system. Furthermore, this operation can be automated by integrating with the enterprise's product data management system to extract structured parameters, thereby providing vehicle-side input for deviation analysis.

[0077] Obtain the structural attribute information of the target semi-active suspension, and expand the structural attribute information according to the preset feature tolerance range to obtain the structural attribute range. The structural attribute information includes at least the damper specifications, suspension configuration, spring stiffness and lower control arm mechanical parameters.

[0078] The structural attribute information can be a set of physical and geometric parameters of the target semi-active suspension itself, which can be used to characterize the suspension's inherent characteristics and for matching analysis with the vehicle structure. The damper specifications can be the model, size, maximum damping force, and current-damping force mapping relationship of the magnetorheological fluid damper, used to determine the adjustable damping range and control precision of the suspension. The suspension architecture can be the mechanical topology of the suspension, which can influence kinematic characteristics and force transmission paths. The spring stiffness can be the restoring force coefficient generated by the suspension spring under unit deformation, which can be used in conjunction with the damper to determine the system's natural frequency and vibration decay rate. The lower control arm mechanical parameters can be the lower control arm's length, installation angle, material elastic modulus, and moment of inertia, etc., which can influence the suspension's lateral stiffness and wheel bounce trajectory, indirectly modulating tire contact performance. The preset feature tolerance range can be a reasonable deviation range allowed for the structural attribute information, which can be used to expand search conditions to cover similar suspension configurations with engineering equivalence. Furthermore, the preset feature tolerance interval can include a narrow tolerance interval for high-sensitivity parameters, a wide tolerance interval for low-sensitivity parameters, and a discrete parameter enumeration set. The structural attribute interval can be a multi-dimensional parameter range formed by expanding the structural attribute information through the preset feature tolerance interval. It can be used as a comparison condition for large-scale data sample retrieval, improving the generalization ability of historical data matching.

[0079] Obtaining the structural attribute information of the target semi-active suspension can be achieved by reading damper specifications, suspension configuration, spring stiffness, and lower control arm mechanical parameters from suspension design documents or product specifications. Furthermore, this operation can automatically extract key parameters by parsing 3D CAD models or BOM lists, thereby constructing a feature description of the suspension body for subsequent generalization and retrieval. Expanding the structural attribute information according to a preset feature tolerance interval yields a structural attribute interval. This can be achieved by setting upper and lower limits for each structural attribute parameter based on its engineering tolerance or sensitivity, forming a multi-dimensional hyperrectangular retrieval space. Further, this operation can be implemented by setting a relative tolerance interval of ±10% for continuous parameters (such as spring stiffness) and using all subtypes within the same category as the expansion set for discrete parameters (such as suspension configuration), thereby improving the inclusiveness of sample retrieval and covering approximately functionally equivalent configurations.

[0080] Guided by the target semi-active suspension, using structural attribute intervals as retrieval and comparison conditions, and constrained by a preset time range, big data technology is used to retrieve sample data, obtain the sample whole vehicle structural feature set and multiple sample vibration reduction performance influence coefficient sets;

[0081] The preset time range can be a time window that limits the collection or generation of sample data, ensuring that the samples used reflect the current technological level and avoiding data mismatch caused by material and process iterations. Big data technology can be a system for storing, processing, and analyzing massive amounts of historical suspension test and real-vehicle operation data, supporting efficient retrieval and modeling of high-dimensional structure-performance related samples. Sample data retrieval can be a process of filtering and matching historical cases from a database based on structural attribute intervals and time constraints, obtaining training samples with physical similarity to ensure model generalization ability. The sample vehicle structural feature set can be the set of vehicle structural features corresponding to the retrieved sample vehicles, serving as input variables for neural network models. The set of multiple sample vibration damping performance influence coefficients can be a set of historical vibration damping performance correction coefficients corresponding to the sample vehicle structural features, serving as supervision labels for neural network models to establish structure-deviation mapping relationships.

[0082] Guided by a target semi-active suspension, using structural attribute intervals as retrieval and comparison criteria, and constrained by a preset time range, this method leverages big data technology to retrieve sample data. This can involve filtering records from a historical suspension test database that meet the structural attribute intervals and whose timestamps fall within the preset range. Furthermore, this operation can be accelerated by employing vector similarity search (such as FAISS) to speed up high-dimensional parameter matching, or by combining a graph database to establish a suspension-vehicle topological relationship index to improve retrieval accuracy. This achieves the technical effect of efficiently obtaining training samples with strong physical comparability. Obtaining the sample vehicle structural feature set and multiple sample vibration damping performance influence coefficient sets can be achieved by extracting the corresponding vehicle structural features for each sample and their influence coefficients obtained through bench-to-real-vehicle comparison from the retrieval results. Furthermore, this operation can batch export matching records through a structured data interface, thereby achieving the technical effect of constructing the input-output pairs required for supervised learning.

[0083] Using the sample whole vehicle structural feature set as input and multiple sample vibration reduction performance influence coefficient sets as supervision, a BP neural network is trained until convergence to obtain the vibration reduction performance influence analysis model.

[0084] The BP neural network can be a multi-layer feedforward artificial neural network trained using the backpropagation algorithm, which can be used to learn the nonlinear mapping relationship between the vehicle structural features and the vibration reduction performance influence coefficients. In an exemplary embodiment, the BP neural network can minimize the loss function between the predicted coefficients and the true coefficients using the gradient descent method, adjusting the weight parameters layer by layer. The vibration reduction performance influence analysis model can be a trained and converged BP neural network used to predict the vibration reduction performance deviation under a specific vehicle structure, and can be used to achieve a fast mapping from vehicle structural features to the influence coefficients of various test indicators. Furthermore, the vibration reduction performance influence analysis model can receive vehicle structural features as input and output a set of vibration reduction performance influence coefficients that correspond one-to-one with the test indicators.

[0085] Using a sample set of whole-vehicle structural features as input and multiple sample sets of vibration reduction performance influence coefficients as supervision, a backpropagation neural network is trained until convergence. This can be achieved by dividing the sample data into training and validation sets, and iteratively optimizing the network weights through backpropagation until the loss function stabilizes. Furthermore, this operation can be achieved by using an Adam optimizer with an early stopping strategy to prevent overfitting, or by introducing an attention mechanism to enhance the weight learning of highly sensitive structural parameters. This achieves the technical effect of establishing a nonlinear prediction capability from the whole-vehicle structure to multi-dimensional performance deviations. The resulting vibration reduction performance influence analysis model can be obtained by saving the trained BP neural network weights and structural configuration. Furthermore, this operation can be serialized and stored in a deployable format, thereby achieving the technical effect of forming a deployable deviation prediction tool.

[0086] Using the vibration reduction performance influence analysis model, the vibration reduction performance deviation is analyzed based on the structural characteristics of the whole vehicle, and the vibration reduction performance influence coefficient set is output. Among them, the vibration reduction performance influence coefficient corresponds one-to-one with the test index.

[0087] The vibration damping performance influence coefficient set can be a performance correction ratio calculated for each test index (such as vehicle body acceleration, suspension dynamic deflection, etc.). It can be used to correct the initial vibration damping performance dataset index by index, eliminating evaluation bias caused by vehicle coupling effects. Furthermore, the vibration damping performance influence coefficient set can be generated by a vibration damping performance influence analysis model, maintaining a one-to-one correspondence with the test indexes. Using the vibration damping performance influence analysis model, vibration damping performance deviation analysis can be performed based on the vehicle's structural characteristics. This can involve inputting the target vehicle's structural characteristics into the model and inferring the performance influence coefficients corresponding to each test index. Furthermore, this operation can be executed in real-time via API calls or an embedded inference engine, achieving a fast, data-driven quantification of vehicle coupling effects. Outputting the vibration damping performance influence coefficient set can be achieved by organizing the model inference results into a structured coefficient set according to the test index dimensions. Furthermore, this operation can be output in standardized JSON or CSV format, providing precise correction factors for the correction of the initial vibration damping performance dataset.

[0088] Taking the development of a new platform electric sports car suspension as an example, the semi-active suspension damping method for electric vehicles based on magnetorheological fluid dampers in this embodiment can be as follows: A car manufacturer develops a rear-wheel-drive electric sports car with a curb weight of 2150kg, an axle load distribution of 48:52, and a multi-link rear suspension. The team first obtains the overall structural characteristics of the vehicle and extracts structural attribute information such as the spring stiffness (28N / mm) and damper specifications (maximum damping force 5000N) of the target magnetorheological suspension; then, the spring stiffness tolerance range is set to ±8%, the suspension type is extended to all multi-link variants, and the time range is limited to data from the past three years. 37 matching samples are retrieved through a big data platform, including vehicle models with different masses but similar axle load distributions and their corresponding influence coefficients for indicators such as body acceleration and tire dynamic displacement. A BP neural network is then trained to obtain a vibration reduction performance influence analysis model. After inputting the overall structural features of the coupe, the model outputs the influence coefficients of five test indicators (such as the body acceleration correction coefficient of 1.12), which are used to correct the bench test data and make the subsequent control strategy more in line with the actual vehicle performance.

[0089] This embodiment can transform the initial data obtained from traditional isolated component testing into corrected data adapted to the installation environment of a specific vehicle model, fundamentally solving the technical problem of performance evaluation distortion caused by ignoring the matching of the whole vehicle structure. At the same time, the data-driven modeling method based on neural networks avoids the high computational cost of complex multibody dynamics simulation, and uses historical measured or simulated big data to achieve fast and high-precision deviation prediction, providing a reliable foundation for subsequent full life cycle decay correction and adaptive control strategies, significantly improving the control effectiveness and comfort performance of the suspension system on a real vehicle platform.

[0090] In one embodiment, big data technology is used to retrieve sample data to obtain a sample vehicle structural feature set and multiple sample vibration reduction performance influence coefficient sets, including:

[0091] Using big data technology, sample data retrieval is performed to obtain multiple sample vehicle structural features that meet the structural attribute interval and preset time range, and a sample vehicle structural feature set is constructed.

[0092] The multiple sample vehicle structural features can be sets of electric vehicle structural parameters retrieved from historical databases that satisfy structural attribute intervals and preset time ranges. These sets can serve as the basic units for constructing the sample set and establishing a structure-performance mapping relationship. In an exemplary embodiment, the multiple sample vehicle structural features can be obtained by filtering and matching records in the historical vehicle database based on multidimensional parameter ranges of structural attribute intervals and timestamp constraints.

[0093] Utilizing big data technology for sample data retrieval allows for the acquisition of multiple sample vehicle structural features that satisfy structural attribute intervals and preset time ranges. This can be achieved by filtering and matching records in a historical vehicle database based on multidimensional parameter ranges of structural attribute intervals and timestamp constraints. Furthermore, this operation can be implemented by setting the range of structural parameters such as suspension configuration and spring stiffness of the target vehicle and a time window of approximately 24 months, ensuring that the selected samples are comparable to the target vehicle in terms of physical configuration and technological generation. Constructing a sample vehicle structural feature set can be achieved by organizing the retrieved sample vehicle structural features into a structured dataset using a unified format. This operation can be further implemented through standardized field naming, missing value imputation, and unit normalization, thereby forming the set of input variables required for neural network training.

[0094] Multiple historical vibration reduction data of different sample vehicle structural features under actual driving load are obtained. Based on the test index, the mapping deviation of multiple historical vibration reduction data and multiple vibration reduction test data under the same test road load is compared to obtain multiple index deviation sets. The mean set of index deviations is calculated as the sample vibration reduction performance influence coefficient set, and multiple sample vibration reduction performance influence coefficient sets are obtained. Among them, the index deviation is the ratio of the difference between the historical vibration reduction data and the vibration reduction test data to the vibration reduction test data.

[0095] The actual driving load can be a set of dynamic excitations experienced by the vehicle during operation in a real road environment, including complex conditions such as road surface unevenness, acceleration and deceleration, and steering. This can be used to drive the suspension system to generate a realistic response, forming the input conditions for historical damping data. In one specific embodiment, the actual driving load can be one or more of the following: random road surface excitation load, maneuvering transient load, and thermo-mechanical coupling load. Multiple historical damping data can be time-series data of test indicators such as body acceleration and suspension dynamic deflection collected by onboard sensors from different sample vehicles under actual driving loads. This can be used to reflect the actual performance of the suspension in a real-world usage environment. For example, multiple historical damping data can be obtained from sensor records of corresponding sample vehicles on typical routes extracted from a vehicle networking platform or a real-vehicle test database. The same test road surface load can be a standard excitation spectrum reproduced on a vehicle road simulation test bench, consistent with the statistical characteristics of the road experienced during the historical damping data collection period. This can be used to provide benchmark input conditions for laboratory testing that are comparable to real-vehicle data. Multiple vibration reduction test data can be response values ​​of various test indicators corresponding to historical vibration reduction data obtained through bench testing under the same test pavement load. These data can be used as performance benchmarks under ideal boundary conditions and for deviation calculations with historical data.

[0096] Mapping deviation comparison can be a process of comparing historical vibration reduction data with vibration reduction test data item by item under the same test indicators and operating conditions. It can be used to quantify the performance differences between ideal laboratory conditions and real-world usage environments. In an exemplary embodiment, mapping deviation comparison can use the RMS value of the time-domain signal for overall deviation calculation, or perform segmented comparison of the response amplitude in key frequency bands (e.g., 1–5Hz) in the frequency domain. Multiple indicator deviation sets can be sets of relative deviations between historical data and test data calculated separately for each sample vehicle according to the test indicators. They can be used to characterize the multidimensional performance deviations of a single sample caused by the vehicle-wide coupling effect. Indicator deviation can be the relative error obtained by dividing the difference between historical vibration reduction data and vibration reduction test data by the vibration reduction test data. It can be used to standardize the expression of the degree of performance deviation, eliminate the influence of dimensions, and facilitate cross-indicator comparisons. The indicator deviation mean set can be a set of correction coefficients formed by statistically averaging the deviations of the same test indicator for multiple samples. It can be used to eliminate individual noise and extract the common patterns of the impact of structural configuration on performance. Furthermore, the indicator deviation mean set can be the numerical source of the sample vibration reduction performance influence coefficient set, corresponding one-to-one with the test indicators.

[0097] Obtaining multiple historical vibration reduction data points for different sample vehicle structural features under actual driving loads can be achieved by extracting sensor records of the corresponding sample vehicles on typical routes from a vehicle networking platform or real-vehicle test database. Furthermore, this operation can be implemented by matching the vehicle's VIN code with structural feature records, thus obtaining first-hand data reflecting suspension performance under real-world usage conditions. Based on test indicators, mapping deviations can be compared between multiple historical vibration reduction data points and multiple vibration reduction test data points under the same test road load. This can be done by comparing the historical values ​​of the real vehicle and the bench reproduction values ​​for each test indicator (such as vehicle acceleration) point-by-point or statistically under the same excitation spectrum. Further, this operation can be achieved by using the RMS value of the time-domain signal for overall deviation calculation, or by segmenting and comparing the response amplitude of key frequency bands (such as 1–5Hz) in the frequency domain, thereby establishing a quantitative deviation relationship between ideal testing and real-world performance.

[0098] Multiple sets of indicator deviations are obtained, which can be achieved by organizing and calculating the indicator deviations for each sample vehicle according to the test indicator dimension, forming a multi-dimensional deviation vector. Further, this operation can be implemented by constructing a deviation matrix with test indicators as columns and samples as rows, thus preserving the individual performance deviation characteristics of the samples and supporting subsequent statistical analysis. The mean set of indicator deviations is calculated as the sample vibration reduction performance influence coefficient set. This can be achieved by taking the arithmetic mean of the deviations of all samples on the same test indicator, generating a representative correction coefficient for that indicator. Further, this operation can be achieved by using a weighted average (the weights are determined by the sample data quality or mileage coverage) or by calculating the mean after removing outliers to improve robustness, thereby extracting the systematic performance deviation patterns dominated by the vehicle's structural characteristics. Multiple sets of sample vibration reduction performance influence coefficients are obtained by binding and storing the mean set of indicator deviations corresponding to each sample with its vehicle structural characteristics. Further, this operation can be achieved by establishing a key-value pair database of structural features and influence coefficients, thus forming input-output pairs that can be used for supervised learning, supporting the training of the vibration reduction performance influence analysis model.

[0099] Taking suspension performance correction based on historical fleet data as an example, the semi-active suspension vibration reduction method for electric vehicles based on magnetorheological fluid dampers in this embodiment can be as follows: A car manufacturer uses the operating data of 5,000 electric SUVs equipped with semi-active suspension that it has sold. First, it sets the structural attribute range according to the suspension configuration (multi-link) and spring stiffness (26±2N / mm) of the target new model, and limits the data time to the past 24 months. 127 matching vehicles are retrieved through a big data platform, and their curb weight, axle load distribution, and other vehicle structural characteristics are obtained to form a sample vehicle structural feature set. Simultaneously, historical data of the vehicle's vertical acceleration when these vehicles are driving on urban expressways (Class B roads) are extracted, and the loading is reproduced on a test bench using the same PSD spectrum to obtain the corresponding bench vibration reduction test data. The deviation of the vehicle's acceleration index ((actual vehicle RMS - bench RMS) / bench RMS) is calculated for each vehicle, resulting in 127 deviation values. The average value is 0.18, meaning that the vehicle's acceleration is overestimated by an average of 18% under this structural configuration. Similarly, the mean deviations of indicators such as suspension dynamic deflection and tire dynamic displacement are obtained to form a sample set of vibration reduction performance influence coefficients, which are used to train the neural network model so that the control strategy of the new model can compensate for such systematic deviations in advance.

[0100] In one embodiment, considering the road environment characteristics of the target semi-active suspension's commonly used driving areas, vibration damping performance attenuation is predicted according to multiple preset driving mileage intervals, outputting multiple sets of predicted vibration damping performance attenuation coefficients, including:

[0101] Multiple preset driving mileage intervals are configured according to preset mileage intervals.

[0102] The preset mileage interval can be a fixed mileage step size used to divide the entire vehicle lifecycle. This provides a unified time-use scale benchmark for configuring preset mileage intervals, ensuring consistency and comparability in the degradation prediction stage division. In this embodiment, configuring multiple preset mileage intervals according to the preset mileage interval can divide the entire vehicle lifecycle into several continuous and non-overlapping intervals using a fixed mileage step size. Furthermore, this operation can be achieved by using an equidistant division method or by non-uniform division based on the vehicle warranty period and typical user vehicle replacement cycle, thereby establishing a standardized service stage division framework, facilitating staged modeling and control strategy mapping.

[0103] The road environment characteristics of the target semi-active suspension's commonly used driving areas are obtained. These characteristics include road surface smoothness distribution, average annual driving temperature difference, frequency of road water accumulation, and percentage of bumpy road sections.

[0104] The road surface smoothness distribution can be a set of parameters describing the proportion of different grades of road surfaces and their statistical characteristics according to the International Roughness Index within a commonly used driving area. It can reflect the intensity and frequency of high-frequency and low-frequency excitations experienced by the vehicle, serving as a major driving factor for mechanical wear and fatigue aging of the magnetorheological fluid damper. For example, the road surface smoothness distribution can include, but is not limited to, the smoothness characteristics of high-grade highways, urban arterial roads, and unpaved road surfaces. The average annual driving temperature difference can be the statistical average of the difference between the highest and lowest ambient temperatures experienced by the target vehicle within a year in its commonly used driving area. It can be used to characterize the influence of temperature cycles on the viscosity-temperature properties of the magnetorheological fluid and the elastic modulus of the sealing material, and is a key environmental variable for predicting fluid performance degradation and seal failure. In an exemplary embodiment, the average annual driving temperature difference can include annual temperature difference characteristics of cold temperate zones, subtropical zones, and plateau / mountainous areas. The frequency of road flooding can be the statistical probability of encountering flooded or waterlogged sections per unit of driving distance. It can be used to assess the impact of moisture intrusion risk on the corrosion rate of the damper sealing system and internal metal components. Furthermore, the frequency of road waterlogging can include the frequency of waterlogging in rainy urban areas, the frequency of waterlogging in seasonally flooded areas, and the frequency of waterlogging in poorly drained road sections. The proportion of bumpy road sections can be the percentage of high-impact road sections (such as continuous speed bumps, potholes, or construction sections) in the total mileage of commonly used driving areas. This can be used to quantify the cumulative impact damage effect of extreme transient loads on the damper piston rod, valve system structure, and magnetic circuit components. In a specific embodiment, the proportion of bumpy road sections can include the proportion of periodic speed bump sections, the proportion of random pothole sections, and the proportion of sections disturbed by temporary construction.

[0105] Obtaining the road environment characteristics of commonly used driving areas for the target semi-active suspension can be achieved by extracting and statistically analyzing the four types of parameters mentioned above from geographic information systems, publicly available data from transportation departments, or historical trajectories from onboard OBD systems. Furthermore, this operation can be performed by extracting environmental features through the fusion analysis of high-precision maps and meteorological databases, or by identifying typical driving areas and statistically analyzing environmental parameters through crowdsourced fleet data clustering. This provides regionally representative environmental input variables for the attenuation prediction model.

[0106] A vibration reduction performance degradation prediction model was obtained based on BP neural network training.

[0107] The BP neural network can be a multi-layer feedforward artificial neural network trained using the backpropagation algorithm. It can be used to construct a high-precision prediction model by learning the nonlinear mapping relationship between road environment characteristics and damper performance degradation in historical or simulated data. In an exemplary embodiment, the BP neural network can use road environment characteristics as the input layer and the degradation coefficients for each mileage interval as the output layer, iteratively optimizing the network weights using gradient descent. The vibration reduction performance degradation prediction model can be a data-driven model based on the BP neural network, used to predict the performance degradation degree of magnetorheological fluid dampers in different driving mileage intervals. It can be used to convert multi-dimensional road environment characteristics into staged performance degradation coefficients, supporting the dynamic correction of the entire life cycle control strategy. Furthermore, the vibration reduction performance degradation prediction model can receive road environment characteristics as input and output the corresponding predicted vibration reduction performance degradation coefficient set according to a preset driving mileage interval.

[0108] The vibration reduction performance degradation prediction model trained based on a BP neural network can be obtained by using a sample set containing road environment features and corresponding measured or simulated degradation data to supervise the training of the BP neural network until convergence. For example, this operation can be achieved by using the Levenberg-Marquardt algorithm to accelerate network convergence, or by introducing Dropout or L2 regularization to prevent overfitting. This allows the establishment of a nonlinear mapping relationship between environmental stress and performance degradation, achieving degradation prediction with high generalization ability.

[0109] Using a vibration reduction performance attenuation prediction model, based on road environment characteristics, vibration reduction performance attenuation is predicted for multiple preset driving mileage intervals, and multiple sets of predicted vibration reduction performance attenuation coefficients are output.

[0110] By utilizing a vibration damping performance degradation prediction model, and based on road environment characteristics, vibration damping performance degradation can be predicted for multiple preset driving mileage intervals. This can be achieved by inputting road environment characteristics into a trained model and calculating the performance degradation degree corresponding to each preset driving mileage interval. Furthermore, this operation can be implemented through parallel inference of the degradation coefficients for each mileage interval, or by using a recursive approach with the state of the previous interval as the initial condition for the next interval for sequential prediction. This generates refined degradation prediction results that match the usage stage and regional environment. Outputting multiple sets of predicted vibration damping performance degradation coefficients can be achieved by organizing the model prediction results into a set of degradation correction coefficients corresponding one-to-one with each preset driving mileage interval. This provides direct input for subsequent phased correction of the optimized vibration damping performance dataset.

[0111] Taking the prediction of suspension life of electric SUVs for rainy cities in the south as an example, the semi-active suspension vibration reduction method of electric vehicles based on magnetorheological fluid dampers in this embodiment can be as follows: A certain electric SUV is mainly sold in South China, and its commonly used driving areas are characterized by high humidity, frequent short-term heavy rainfall, smooth urban roads but many potholes in the suburbs. The development team first set three preset mileage intervals at 50,000-kilometer intervals. Then, they obtained IRI data for the region over the past five years from the provincial traffic maintenance platform. The statistics showed that high-grade highways accounted for 60%, urban roads 30%, and rural roads 10%, with an average annual driving temperature difference of 28°C, a road water accumulation frequency of 1.2 times per 100 kilometers, and bumpy road sections (including areas with dense speed bumps) accounting for 18%. Using the above four-dimensional road environment characteristics as input, a BP neural network model trained based on historical accelerated aging test data was used to predict the damping force retention rate in three stages: 0–50,000 kilometers, 50,000–100,000 kilometers, and 100,000–150,000 kilometers, respectively, and output three sets of predicted vibration reduction performance degradation coefficients (e.g., [0.98, 0.92, 0.85]). These coefficients were then used to correct and optimize the vibration reduction performance dataset, so that the on-board controller automatically increases the current command when the vehicle travels to 80,000 kilometers to compensate for the decrease in damping force caused by seal micro-leakage and oil oxidation, thus maintaining the ride comfort.

[0112] This embodiment achieves the technical effect of accurate modeling and dynamic compensation of the performance degradation of magnetorheological fluid dampers throughout their entire life cycle by establishing a standardized service stage division framework, introducing regionally representative multidimensional environmental input variables, constructing a nonlinear data-driven prediction model, and generating staged refined attenuation coefficients. It effectively captures the coupled influence of complex environmental factors and usage intensity on damper aging, overcomes the bias of traditional life assessment that ignores the effects of time-varying environments, supports adaptive adjustment of control strategies to maintain optimal vibration reduction performance, and significantly reduces the cost and development cycle of real vehicle road tests, thereby improving the intelligence and forward-looking level of suspension system development.

[0113] In one embodiment, a vibration reduction performance degradation prediction model is obtained based on BP neural network training, including:

[0114] Based on the historical operation and maintenance monitoring records of similar magnetorheological fluid semi-active suspensions, a sample mileage set and a sample road environment feature set were collected. The performance degradation ratio of multiple test indicators under different sample mileage and sample environment features was calculated and set as the sample performance degradation coefficient, thus obtaining the sample performance degradation coefficient set.

[0115] Using the sample driving mileage set and sample road environment feature set as inputs, and the sample performance degradation coefficient set as supervision, a BP neural network is trained until convergence to obtain a vibration reduction performance degradation prediction model.

[0116] Among them, the similar magnetorheological fluid semi-active suspension can be a mass-produced magnetorheological suspension system that is highly similar to the target semi-active suspension in terms of structural form, damper model, control logic, or application vehicle level. It can serve as a physical carrier for historical data, and its service records can reflect performance degradation patterns under real-world conditions. Furthermore, the historical maintenance and monitoring records of the similar magnetorheological fluid semi-active suspension can be used to construct training samples to support supervised learning of the vibration damping performance degradation prediction model. The historical maintenance and monitoring records can be a long-term operational data set collected from vehicles with similar magnetorheological fluid semi-active suspensions already on the market, including timestamps, mileage, environmental parameters, and suspension performance indicators. This data can be used to provide observational evidence of performance evolution under the coupling of multiple factors in the real world, and to construct high-fidelity training samples. In an exemplary embodiment, the historical maintenance and monitoring records can be periodically uploaded to a cloud database through an in-vehicle telematics system or an after-sales diagnostic interface.

[0117] The sample mileage set can be a collection of cumulative mileage values ​​extracted from historical operation and maintenance monitoring records, corresponding to each performance observation point. It can be used as the time-usage intensity dimension of the model input to characterize the service life of the component. The sample road environment feature set can be a multi-dimensional feature set describing the environmental conditions of the geographical area where the vehicle is located, collected synchronously with the sample mileage. It can be used as the environmental stress dimension of the model input to quantify the impact of external factors on performance degradation. For example, the sample road environment feature set and the sample mileage set can together constitute the input vector of a BP neural network. The test index can be a key performance parameter used to quantify the vibration reduction capability of the magnetorheological fluid damper. It can be used as a benchmark for calculating the performance degradation ratio, reflecting the functional state of the damper. In a specific embodiment, the test index can include, but is not limited to, one or more of the following: peak damping force, phase lag angle, and response delay time. The performance degradation ratio can be the rate of decrease of a specific test index in its current service state relative to its initial factory value. It can be used to objectively measure the degree of functional degradation of the magnetorheological fluid damper under a specific mileage and environmental combination. Furthermore, the performance degradation ratio can be calculated by comparing the measured values ​​in historical monitoring with the initial values ​​of bench calibration. The sample performance degradation coefficient can be a normalized representation value formed by combining the performance degradation ratios of multiple test indicators for a specific observed sample. It can be used as a supervisory signal to guide the BP neural network in learning the mapping relationship between input features and performance degradation. The sample performance degradation coefficient set can be a collection of performance degradation coefficients corresponding to all historical samples, corresponding one-to-one with the input features. It can be used to form the label set for supervised learning tasks to drive the BP neural network training convergence.

[0118] Based on historical operation and maintenance monitoring records of similar magnetorheological fluid semi-active suspensions, sample mileage sets and sample road environment feature sets are collected. This can be achieved by filtering the operation logs of similar suspension vehicles that meet the criteria from a cloud database and extracting mileage and environmental parameters within the corresponding time window. Furthermore, this operation can be enhanced by grouping samples by region to increase environmental diversity, or by using a sliding window method to slice and sample continuous operation data to increase sample density, thereby constructing an input feature dataset representative of real operating conditions. The performance degradation ratios of multiple test indicators under different sample mileages and sample environmental characteristics are statistically analyzed and set as sample performance degradation coefficients. This can be achieved by calculating the degradation rate of the current test indicator value relative to the initial value for each sample point, and generating a single degradation coefficient through weighted fusion or multi-indicator principal component analysis. For example, this operation can be achieved by linearly combining different test indicators based on expert weights, or by using unsupervised dimensionality reduction methods (such as PCA) to extract the dominant degradation pattern as a coefficient, thereby compressing multidimensional performance degradation information into a scalar label that can be learned under supervision.

[0119] Obtaining the sample performance degradation coefficient set can be achieved by organizing the performance degradation coefficients of all samples into label vectors aligned with the input features, thus forming a complete supervised learning dataset to support model training. Using the sample mileage set and sample road environment feature set as input, and the sample performance degradation coefficient set as supervision, a BP neural network is trained until convergence. This can be achieved by feeding the input feature vector and corresponding labels into the BP neural network, iteratively updating the weights through backpropagation until the loss function stabilizes. In a specific embodiment, this operation can be implemented by using early stopping to prevent overfitting, using the minimum validation set loss as the convergence criterion, or by introducing an adaptive learning rate optimizer (such as Adam) to accelerate the convergence process, thereby enabling the network to learn the nonlinear mapping law from service conditions to performance degradation. Obtaining the vibration reduction performance degradation prediction model can be achieved by saving the trained BP neural network weights and structural parameters, encapsulating them into a callable prediction function, thus obtaining a performance degradation prediction tool that can be deployed in subsequent control processes.

[0120] Taking intelligent suspension life modeling based on fleet big data as an example, the semi-active suspension vibration reduction method for electric vehicles based on magnetorheological fluid dampers in this embodiment can be implemented by a car manufacturer using remote monitoring data from 50,000 electric cars equipped with magnetorheological suspensions that it has sold, and selecting 30,000 vehicles that have been operating for a long time in East China, North China, and Southwest China as similar samples. From their historical operation and maintenance records, the cumulative mileage at every 10,000-kilometer interval, the monthly average temperature difference of the corresponding region, the IRI value, the number of water crossings, and the frequency of bump events are extracted to form a sample mileage set and a sample road environment feature set. Simultaneously, the peak damping force is retrieved by the on-board current-displacement sensor and compared with the factory calibration value to calculate the performance degradation ratio of each node. The sample performance degradation coefficient is generated by weighted fusion. Finally, a dataset containing 120,000 samples is formed to train a three-layer BP neural network. After training, the model can accurately predict the degree of damping force degradation of the new model at 80,000 km and 120,000 km in Chengdu (high humidity, many potholes) or Harbin (large temperature difference, icy and snowy road surface) environments, respectively, so as to inject compensation gain into the controller in advance and avoid sudden changes in comfort.

[0121] This embodiment constructs high-fidelity input-label pairs from real service data and drives the neural network to learn the nonlinear mapping relationship between environment, usage intensity, and performance degradation. This enables the obtained model to accurately reflect the damper performance evolution trend under the coupled effects of complex factors such as temperature changes, bump excitation, and seal aging. It supports the generation of adaptive vibration reduction strategies throughout the entire life cycle and significantly reduces R&D costs and cycles, while improving the technical effect of suspension system adaptability and robustness to actual working conditions.

[0122] In one embodiment, a semi-active suspension damping control result is generated based on multiple preset driving mileage ranges and multiple predicted damping performance datasets, including:

[0123] Multiple preset driving mileage ranges and multiple predicted vibration reduction performance datasets are mapped and combined to generate a semi-active suspension vibration reduction control parameter table adapted to different usage stages, which serves as the semi-active suspension vibration reduction control result.

[0124] The semi-active suspension damping control parameter table can be a structured set of control parameters generated by associating different driving mileage intervals with corresponding predicted damping performance datasets. This set guides the real-time adjustment of the magnetorheological fluid damper and can serve as the basis for damping control invoked by the vehicle controller at different vehicle usage stages, achieving adaptive damping throughout the entire lifecycle. In this embodiment, the semi-active suspension damping control parameter table can be constructed by mapping and combining multiple preset driving mileage intervals with multiple predicted damping performance datasets. Each mileage interval corresponds to a set of control parameters corrected by the vehicle structure and environmental attenuation. The mapping and combination process can be a data integration process that establishes a one-to-one correspondence between discrete driving mileage intervals and their corresponding predicted damping performance datasets. This can be used to form a phased, structured control strategy input basis, supporting the generation of the parameter table. In an exemplary embodiment, the mapping and combination can associate each preset driving mileage interval identifier with its own predicted damping performance dataset through index matching or rule binding.

[0125] Mapping and combining multiple preset mileage intervals and multiple predicted damping performance datasets can establish a relationship between mileage interval identifiers and corresponding predicted damping performance datasets, forming structured mapping pairs. Further, this operation can be achieved by storing each mileage interval ID and its corresponding performance dataset using a key-value pair data structure, or by defining the logical binding relationship between intervals and datasets through database views or configuration files. This provides a data organization foundation for generating a control parameter table adapted to different usage stages. Generating a semi-active suspension damping control parameter table adapted to different usage stages, as the semi-active suspension damping control result, can be based on the mapping and combination results. This involves converting the predicted damping performance data of each stage into parameter forms usable for real-time control (such as current-damping force mapping curves, control gain matrices, etc.) and integrating them into a unified parameter table. In a specific embodiment, this operation can be achieved by fitting the performance data of each stage into a piecewise continuous control function and embedding it into the ECU firmware, or by constructing a multidimensional lookup table (LUT) and automatically switching the subset of control parameters using the current mileage as an index. This results in a dynamic control strategy carrier with full lifecycle adaptive capabilities, replacing traditional static control parameters.

[0126] Taking the adaptive suspension calibration of an urban commuter electric SUV as an example, the semi-active suspension damping method for electric vehicles based on magnetorheological fluid dampers in this embodiment can be as follows: A certain electric SUV is mainly aimed at users in first-tier cities, and the commonly used driving areas are well-paved but frequently stop-and-go urban roads. During the development process, based on the three preset mileage ranges of 0-40,000 km, 40,000-80,000 km, and 80,000-120,000 km obtained in the early stage, and the corresponding three sets of predicted damping performance datasets (all of which have been integrated with the low sprung mass ratio and high body stiffness characteristics of the vehicle model, and the aging effect of the high temperature and humidity environment in East China on the seals), the interval-dataset correspondence is established through mapping combination operations; then, each set of data is converted into the damping gain coefficient and magnetorheological current mapping curve in the Skyhook algorithm to generate a semi-active suspension damping control parameter table containing three sub-tables. After the vehicle leaves the factory, the on-board system automatically activates the corresponding sub-tables according to the accumulated mileage: the new car stage focuses on comfort, the mid-term balances comfort and support, and the later stage increases damping compensation to offset the response lag caused by the decrease in oil viscosity, ensuring consistent ride quality within 120,000 km.

[0127] This embodiment provides a semi-active suspension damping method for electric vehicles based on magnetorheological fluid dampers. By mapping and combining multiple preset driving mileage intervals and multiple predicted damping performance datasets, a semi-active suspension damping control parameter table adapted to different usage stages is generated as the semi-active suspension damping control result. By establishing a structured correspondence between driving stages and performance data and converting predicted performance into executable control parameters, the technical effect of transforming the control strategy from a static, fixed mode to a full life-cycle adaptive mode can be achieved. This method can accurately match the actual performance state of the magnetorheological fluid damper at different service stages, effectively compensating for damping force attenuation caused by factors such as temperature changes, seal aging, and road excitation accumulation. Simultaneously, since each driving mileage interval corresponds to performance data corrected by vehicle structural characteristics and road environment attenuation prediction, the control parameters possess a high degree of vehicle customization and regional environmental adaptability, thereby continuously maintaining optimal damping output throughout the vehicle's entire service life, improving ride comfort while ensuring handling stability. In addition, this mapping combination mechanism relies on virtual simulation prediction based on big data and neural networks in the early stage, avoiding the traditional approach of relying on a large number of real vehicle long-term road tests to obtain control parameters, which significantly improves the efficiency of control strategy development and the feasibility of engineering implementation.

[0128] In addition, refer to Figure 2 To achieve the above objectives, the present invention also provides a semi-active suspension damping system for electric vehicles based on a magnetorheological fluid damper, the system comprising:

[0129] The bench test module 10 is used to test the damping performance of the target semi-active suspension using a vehicle road simulation test bench according to a preset test plan, and to obtain an initial damping performance dataset.

[0130] The structural correction module 20 is used to perform vibration reduction performance deviation analysis based on the overall vehicle structural characteristics of the electric vehicle where the target semi-active suspension is located, output a set of vibration reduction performance influence coefficients, and adjust the initial vibration reduction performance dataset to obtain an optimized vibration reduction performance dataset.

[0131] The attenuation prediction module 30 is used to combine the road environment characteristics of the target semi-active suspension in the commonly used driving area, predict the attenuation of vibration reduction performance according to multiple preset driving mileage intervals, output multiple sets of predicted vibration reduction performance attenuation coefficients, and correct the optimized vibration reduction performance datasets respectively to obtain multiple predicted vibration reduction performance datasets.

[0132] The control generation module 40 is used to generate semi-active suspension damping control results based on the multiple preset driving mileage intervals and the multiple predicted damping performance datasets.

[0133] Other embodiments or specific implementations of the electric vehicle semi-active suspension damping system based on magnetorheological fluid damper described in this invention can be referred to the above-mentioned method embodiments, and will not be repeated here.

[0134] Furthermore, to achieve the above objectives, the present invention also provides a semi-active suspension damping device for electric vehicles based on a magnetorheological fluid damper. The device includes: a memory, a processor, and a semi-active suspension damping program for electric vehicles based on a magnetorheological fluid damper stored in the memory and executable on the processor. The semi-active suspension damping program for electric vehicles based on a magnetorheological fluid damper is configured to implement the steps of the semi-active suspension damping method for electric vehicles based on a magnetorheological fluid damper as described above.

[0135] Furthermore, to achieve the above objectives, the present invention also provides a computer-readable storage medium storing a semi-active suspension damping program for an electric vehicle based on a magnetorheological fluid damper. When the semi-active suspension damping program for an electric vehicle based on a magnetorheological fluid damper is executed by a processor, it implements the steps of the semi-active suspension damping method for an electric vehicle based on a magnetorheological fluid damper as described above.

[0136] The above are merely preferred embodiments of the present invention and do not limit the scope of the patent. Any equivalent structural or procedural transformations made based on the description and drawings of the present invention, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of the present invention.

Claims

1. A method for semi-active suspension vibration reduction in electric vehicles based on magnetorheological fluid dampers, characterized in that, The method includes: According to the preset test plan, the target semi-active suspension was tested for vibration reduction performance using a vehicle road simulation test bench to obtain an initial vibration reduction performance dataset. By combining the overall vehicle structural characteristics of the electric vehicle where the target semi-active suspension is located, a vibration reduction performance deviation analysis is performed, and a set of vibration reduction performance influence coefficients is output. The initial vibration reduction performance dataset is then adjusted to obtain an optimized vibration reduction performance dataset. Combining the road environment characteristics of the target semi-active suspension's commonly used driving areas, the damping performance attenuation is predicted according to multiple preset driving mileage intervals, and multiple predicted damping performance attenuation coefficient sets are output. The optimized damping performance dataset is then corrected to obtain multiple predicted damping performance datasets. Semi-active suspension damping control results are generated based on the multiple preset driving mileage ranges and the multiple predicted damping performance datasets.

2. The semi-active suspension vibration reduction method for electric vehicles based on magnetorheological fluid dampers as described in claim 1, characterized in that, The test involves conducting vibration damping performance tests on the target semi-active suspension using a vehicle road simulation test bench according to a preset test plan, and obtaining an initial vibration damping performance dataset, including: Obtain a preset test plan, wherein the preset test plan includes test indicators, test procedures and parameter configurations, and the test indicators include at least vehicle body acceleration, suspension dynamic deflection, tire dynamic displacement, damping adjustment response speed and vibration reduction energy consumption rate; Based on the aforementioned test indicators, and following the test procedures and parameter configurations, a multi-condition road loading test was conducted on a semi-active suspension equipped with a magnetorheological fluid damper using a vehicle road simulation test bench, and an initial vibration reduction performance dataset was output.

3. The semi-active suspension vibration reduction method for electric vehicles based on magnetorheological fluid dampers as described in claim 2, characterized in that, The analysis of vibration reduction performance deviation is performed by combining the overall vehicle structural characteristics of the electric vehicle in which the target semi-active suspension is located, and the output set of vibration reduction performance influence coefficients includes: Obtain the overall vehicle structural characteristics of the electric vehicle in which the target semi-active suspension is located, wherein the overall vehicle structural characteristics include at least vehicle type, curb weight, chassis structure type and axle load distribution; Obtain the structural attribute information of the target semi-active suspension, and expand the structural attribute information according to a preset feature tolerance range to obtain a structural attribute range. The structural attribute information includes at least the damper specifications, suspension configuration, spring stiffness, and lower control arm mechanical parameters. Guided by the target semi-active suspension, using the structural attribute range as the retrieval and comparison conditions, and constrained by a preset time range, big data technology is used to retrieve sample data, obtain the sample whole vehicle structural feature set and multiple sample vibration reduction performance influence coefficient sets; Using the sample whole vehicle structural feature set as input and the multiple sample vibration reduction performance influence coefficient sets as supervision, a BP neural network is trained until convergence to obtain the vibration reduction performance influence analysis model. Using the vibration reduction performance influence analysis model, vibration reduction performance deviation analysis is performed based on the vehicle structural characteristics, and a set of vibration reduction performance influence coefficients is output, wherein the vibration reduction performance influence coefficients correspond one-to-one with the test indicators.

4. The semi-active suspension vibration reduction method for electric vehicles based on magnetorheological fluid dampers as described in claim 3, characterized in that, The method of using big data technology to retrieve sample data, obtaining a sample vehicle structural feature set and multiple sample vibration reduction performance influence coefficient sets, includes: Using big data technology, sample data is retrieved to obtain multiple sample vehicle structural features that satisfy the structural attribute interval and the preset time range, and a sample vehicle structural feature set is constructed. Multiple historical vibration reduction data of different sample vehicle structural features under actual driving load are obtained. Based on the test index, the mapping deviation of the multiple historical vibration reduction data and multiple vibration reduction test data under the same test road load is compared to obtain multiple index deviation sets. The mean set of index deviations is calculated as the sample vibration reduction performance influence coefficient set, and multiple sample vibration reduction performance influence coefficient sets are obtained. Among them, the index deviation is the ratio of the difference between the historical vibration reduction data and the vibration reduction test data to the vibration reduction test data.

5. The semi-active suspension vibration reduction method for electric vehicles based on magnetorheological fluid dampers as described in claim 1, characterized in that, The method combines the road environment characteristics of the target semi-active suspension's commonly used driving areas, predicts the damping performance degradation according to multiple preset driving mileage intervals, and outputs multiple sets of predicted damping performance degradation coefficients, including: Multiple preset driving mileage intervals are configured according to preset mileage intervals; Obtain the road environment characteristics of the target semi-active suspension's commonly used driving areas. These road environment characteristics include road surface smoothness distribution, average annual driving temperature difference, frequency of road water accumulation, and proportion of bumpy road sections. A vibration reduction performance degradation prediction model was obtained based on BP neural network training. Using the vibration reduction performance attenuation prediction model, based on the road environment characteristics, vibration reduction performance attenuation is predicted according to multiple preset driving mileage intervals, and multiple sets of predicted vibration reduction performance attenuation coefficients are output.

6. The semi-active suspension vibration reduction method for electric vehicles based on magnetorheological fluid dampers as described in claim 5, characterized in that, The vibration reduction performance degradation prediction model obtained based on BP neural network training includes: Based on the historical operation and maintenance monitoring records of similar magnetorheological fluid semi-active suspensions, a sample mileage set and a sample road environment feature set were collected. The performance degradation ratio of multiple test indicators under different sample mileage and sample environment features was calculated and set as the sample performance degradation coefficient, thus obtaining the sample performance degradation coefficient set. Using the sample driving mileage set and sample road environment feature set as inputs, and the sample performance attenuation coefficient set as supervision, a BP neural network is trained until convergence to obtain a vibration reduction performance attenuation prediction model.

7. The semi-active suspension vibration reduction method for electric vehicles based on magnetorheological fluid dampers as described in claim 1, characterized in that, The process of generating semi-active suspension damping control results based on the multiple preset driving mileage ranges and the multiple predicted damping performance datasets includes: The multiple preset driving mileage intervals and the multiple predicted vibration reduction performance datasets are mapped and combined to generate a semi-active suspension vibration reduction control parameter table adapted to different usage stages, which serves as the semi-active suspension vibration reduction control result.

8. A semi-active suspension damping system for electric vehicles based on a magnetorheological fluid damper, characterized in that, The system includes: The bench test module is used to test the damping performance of the target semi-active suspension using a vehicle road simulation test bench according to a preset test plan, and to obtain the initial damping performance dataset. The structural correction module is used to perform vibration reduction performance deviation analysis based on the overall vehicle structural characteristics of the electric vehicle where the target semi-active suspension is located, output a set of vibration reduction performance influence coefficients, and adjust the initial vibration reduction performance dataset to obtain an optimized vibration reduction performance dataset. The attenuation prediction module is used to combine the road environment characteristics of the target semi-active suspension in the commonly used driving area, predict the attenuation of vibration reduction performance according to multiple preset driving mileage intervals, output multiple sets of predicted vibration reduction performance attenuation coefficients, and correct the optimized vibration reduction performance datasets respectively to obtain multiple predicted vibration reduction performance datasets. The control generation module is used to generate semi-active suspension damping control results based on the multiple preset driving mileage intervals and the multiple predicted damping performance datasets.

9. A semi-active suspension damping device for electric vehicles based on a magnetorheological fluid damper, characterized in that, The device includes: a memory, a processor, and a semi-active electric vehicle suspension damping program based on a magnetorheological fluid damper stored in the memory and executable on the processor, the semi-active electric vehicle suspension damping program based on a magnetorheological fluid damper configured to implement the steps of the semi-active electric vehicle suspension damping method based on a magnetorheological fluid damper as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a semi-active suspension damping program for electric vehicles based on a magnetorheological fluid damper. When the semi-active suspension damping program for electric vehicles based on a magnetorheological fluid damper is executed by a processor, it implements the steps of the semi-active suspension damping method for electric vehicles based on a magnetorheological fluid damper as described in any one of claims 1 to 7.

Images