A Simulation and Optimization Method for Automobile Chassis Structure Based on Digital Twin

Abstract

The application relates to the technical field of automobile chassis simulation, and discloses a simulation optimization method for an automobile chassis mechanism based on digital twinning. The method first collects a real-time running data set of the automobile chassis mechanism, which contains multi-axis acceleration time sequence information, suspension displacement monitoring sequences and the like, and synchronously transmits the real-time running data set to a cloud digital twinning platform; in the platform, a hierarchical digital twinning body model containing a spatial constraint relationship of a suspension subsystem node, a steering mechanism node and connecting components is constructed according to a chassis topological structure; multi-physical field feature extraction is performed on the real-time running data set to generate a chassis multi-physical field data set; multi-scale joint simulation calculation is performed based on the above model and data set, and a chassis dynamic response prediction result containing key connecting piece fatigue damage distribution and the like is output; finally, an optimization parameter adjustment instruction set is generated according to the result and is fed back to a physical chassis executing mechanism. The method realizes real-time and precision of chassis simulation optimization, and improves chassis performance.

Description

Technical Field

[0001] This invention relates to the field of automotive chassis simulation technology, specifically to a simulation optimization method for automotive chassis mechanisms based on digital twins. Background Technology

[0002] As the core load-bearing structure for vehicle performance, the chassis's dynamic characteristics and stability directly affect the vehicle's handling safety, ride comfort, and service life. With the automotive industry moving towards intelligent and lightweight designs, traditional chassis design and optimization methods are gradually revealing significant limitations.

[0003] Currently, chassis simulation analysis largely relies on offline modeling and static parameter input, making it difficult to reflect the vehicle's dynamic response under complex road conditions in real time. For example, in the performance evaluation of suspension systems, traditional methods typically use pre-set road excitation models for simulation, but the randomness and variability of road conditions during actual driving can lead to significant deviations between simulation results and real-world conditions.

[0004] The various subsystems of an automotive chassis are interconnected by complex coupling relationships, such as the dynamic interaction between the steering mechanism and the suspension system, and the vibration transmission between the powertrain and the chassis structure. Existing simulation methods often analyze individual subsystems independently, lacking the ability to coordinate and optimize the entire chassis system. This can easily lead to the problem of optimizing local performance while degrading overall performance.

[0005] In the optimization of chassis structures, traditional methods require extensive physical experiments to obtain data, which is not only time-consuming and costly but also difficult to cover all possible operating conditions. With the development of vehicle electrification and intelligence, the structure of chassis systems is becoming increasingly complex, and the requirements for simulation accuracy and optimization efficiency are constantly increasing. Traditional methods can no longer meet the needs of modern automotive R&D. Summary of the Invention

[0006] The purpose of this invention is to provide a simulation optimization method for automobile chassis mechanisms based on digital twins, so as to solve the problems mentioned in the background art.

[0007] To achieve the above objectives, this invention provides a simulation optimization method for automotive chassis mechanisms based on digital twins, the method comprising:

[0008] A real-time operating data set of the vehicle chassis mechanism is collected, which includes multi-axis acceleration time sequence information, suspension displacement monitoring sequence, steering mechanism angle change amount and temperature distribution map, and the real-time operating data set is synchronously transmitted to the cloud digital twin platform.

[0009] In the cloud-based digital twin platform, a hierarchical digital twin model is constructed based on the chassis topology. The hierarchical digital twin model includes the spatial constraint relationships of suspension subsystem nodes, steering mechanism nodes, and connecting components.

[0010] Multiphysics feature extraction processing is performed on the real-time running data set to generate a chassis multiphysics dataset, which includes vibration mode feature vectors, structural stress distribution matrix and thermo-coupling parameter set;

[0011] Based on the hierarchical digital twin model and the chassis multiphysics dataset, a multi-scale joint simulation calculation process is executed to output the chassis dynamic response prediction results, which include the fatigue damage distribution of key connecting parts and the deformation offset of the mechanism.

[0012] Based on the chassis dynamic response prediction results, an optimized parameter adjustment instruction set is generated and fed back to the physical chassis actuator.

[0013] Preferably, the set of real-time operating data of the vehicle chassis mechanism includes:

[0014] A triaxial vibration sensor is deployed at the hinge point of the chassis suspension control arm to obtain spatial acceleration timing information at a fixed sampling frequency;

[0015] The extension and retraction of the piston rod of each shock absorber is collected by a linear displacement sensor to form a suspension displacement monitoring sequence.

[0016] An angle encoder is used to record the change in the swing angle of the steering tie rod in real time, and the change in the steering mechanism angle is generated by combining the current fluctuation data of the steering motor.

[0017] An infrared thermal imager was used to scan the metal connection area of ​​the chassis to generate a temperature distribution map.

[0018] The spatial acceleration timing information, suspension displacement monitoring sequence, steering mechanism angle change and temperature distribution map are aligned by timestamp and stored as a real-time running data set.

[0019] Preferably, the multiphysics feature extraction process for the real-time running data set includes:

[0020] Perform short-time Fourier transform processing on the spatial acceleration time-series information to extract the proportion of vibration energy within a specific frequency band as the basic element of the vibration mode feature vector;

[0021] The displacement change rate of adjacent sampling points is calculated based on the suspension displacement monitoring sequence, and the initial value of the structural stress distribution matrix is ​​derived by combining the material elastic modulus parameter.

[0022] Analyze the geometric center coordinates of the high-temperature region in the temperature distribution map, and associate the corresponding vibration mode eigenvectors to establish a mapping relationship of the thermo-mechanical coupling parameter set;

[0023] By fusing vibration mode feature vectors, structural stress distribution matrices, and thermo-coupling parameter sets using convolutional neural networks, a multi-physics dataset for chassis containing spatiotemporal correlation features is generated.

[0024] Preferably, the multi-scale co-simulation calculation process includes:

[0025] The vibration mode feature vector is input into the finite element sub-model to calculate the dynamic load transfer function of the suspension subsystem nodes;

[0026] The stress distribution matrix of the structure is input into the multibody dynamics sub-model to solve for the motion trajectory deviation of the steering mechanism nodes;

[0027] The set of thermo-coupling parameters is input into the fluid dynamics sub-model to simulate the air convection heat dissipation efficiency of the chassis connection area;

[0028] Establish a coupling interface between the finite element sub-model, the multibody dynamics sub-model, and the fluid dynamics sub-model, so that the dynamic load transfer function of the suspension subsystem node is used as the boundary condition input to the multibody dynamics sub-model, and the motion trajectory deviation of the steering mechanism node is fed back to the finite element sub-model.

[0029] The calculation is iterated until the output error rate of each sub-model is lower than the preset threshold, generating a chassis dynamic response prediction result that includes the coordinates of the stress concentration area and the deformation trend vector.

[0030] Preferably, the step of generating the set of optimization parameter adjustment instructions based on the chassis dynamic response prediction results includes:

[0031] Identify high fatigue damage areas in the chassis dynamic response prediction results and extract the material stress amplitude spectrum at the corresponding locations;

[0032] Based on the peak frequency characteristics of the stress amplitude spectrum, the adjustment amount of suspension bushing stiffness and the correction value of shock absorber damping coefficient are generated.

[0033] Analyze the spatial distribution of deformation offset of the analysis mechanism, and calculate the ball joint clearance compensation of the steering mechanism and the fine adjustment of the control arm installation angle.

[0034] Based on the thermal gradient variation trend of the temperature distribution map, the optimal parameters for the angle of the heat dissipation guide plate are derived.

[0035] The system integrates the suspension bushing stiffness adjustment, shock absorber damping coefficient correction, steering mechanism ball joint clearance compensation, control arm mounting angle fine-tuning, and heat dissipation deflector angle optimization parameters to form a set of optimized parameter adjustment instructions.

[0036] Preferably, the step of constructing a hierarchical digital twin model based on the chassis topology includes:

[0037] A hierarchical topology map of the chassis mechanism components is established, which includes a main load-bearing component layer, a kinematic pair connection layer, and an auxiliary actuator layer.

[0038] A six-degree-of-freedom motion constraint equation is defined for each suspension subsystem node. The six-degree-of-freedom motion constraint equation includes displacement boundary conditions and torque transfer functions.

[0039] A gear meshing parameterized model is established for the steering mechanism nodes, and the gear meshing parameterized model covers the backlash nonlinearity characteristics and friction coefficient variables;

[0040] The viscoelastic constitutive relationship of setting a rubber bushing in the kinematic pair connection layer and the ball joint clearance tolerance threshold;

[0041] The hierarchical topology map is matched and calibrated with the sensor spatial coordinates in the real-time running data set to complete the initialization of the hierarchical digital twin model.

[0042] Preferably, the method further includes a virtual stimulus loading step:

[0043] Based on the suspension bushing stiffness adjustment amount in the set of optimized parameter adjustment instructions, modify the material property parameters of the corresponding nodes in the hierarchical digital twin model;

[0044] Based on the corrected value of the damper damping coefficient, update the damping matrix elements in the finite element sub-model;

[0045] The backlash compensation amount of the steering mechanism ball joint is injected into the backlash tolerance variable of the gear meshing parameterization model;

[0046] A virtual road surface excitation spectrum is applied to the updated hierarchical digital twin model using the modal superposition method. The virtual road surface excitation spectrum includes random amplitude waveforms and step impact components.

[0047] Collect model response data under virtual stimulation to verify the effectiveness of the set of optimization parameter adjustment instructions.

[0048] Preferably, the method further includes a dynamic twin update step:

[0049] Real-time feedback data from the physical chassis actuators is received, including the actual shock absorber displacement curve and the steering motor torque output.

[0050] Calculate the residual index between the feedback data and the predicted results of the chassis dynamic response, wherein the residual index includes phase lag and amplitude attenuation rate;

[0051] When the residual index exceeds the preset tolerance threshold, the parameter correction process of the hierarchical digital twin model is initiated.

[0052] The mass distribution parameters of the finite element sub-model are adjusted according to the phase lag, and the energy dissipation coefficient of the multibody dynamics sub-model is corrected according to the amplitude decay rate.

[0053] The updated hierarchical digital twin model will be used as the computational basis for the next simulation cycle.

[0054] Preferably, the method further includes a model iterative optimization step:

[0055] Accumulate measured fatigue damage data of the physical chassis actuators within a preset time period;

[0056] Extract the simulated fatigue damage prediction values ​​for the corresponding period of the hierarchical digital twin model;

[0057] Construct the error distribution matrix between measured data and simulated predicted values;

[0058] The material degradation rate parameters and thermal conductivity coefficient in the multi-scale co-simulation calculation process are adjusted by using the backpropagation algorithm.

[0059] A validated multi-scale co-simulation calculation process is generated for subsequent cycles of chassis dynamic response prediction.

[0060] Preferably, the method further includes a digital twin visualization step:

[0061] Map the coordinates of stress concentration areas in the chassis dynamic response prediction results to the 3D visualization engine;

[0062] Render and display a dynamic change cloud map of the deformation offset of the mechanism in the 3D model;

[0063] Animated simulation of the execution effect of the optimized parameter adjustment instruction set is displayed overlay;

[0064] A remaining life warning and identification system for critical connectors is established, wherein the remaining life warning and identification system is driven by color gradient changes based on fatigue damage distribution data.

[0065] Compared with the prior art, the beneficial effects of the present invention are:

[0066] By collecting real-time operational data of the vehicle chassis and synchronously transmitting it to a cloud-based digital twin platform, real-time perception and data feedback of the chassis's dynamic operating status are achieved, breaking the limitations of traditional offline simulation that is disconnected from actual working conditions. The introduction of real-time data enables simulation analysis to closely match the vehicle's actual driving state, providing a realistic foundation for subsequent model building and simulation calculations.

[0067] The hierarchical digital twin model built on the cloud-based digital twin platform encompasses the spatial constraints of suspension subsystem nodes, steering mechanism nodes, and connecting components, clearly presenting the structural relationships and interactions of various parts of the chassis. This hierarchical model construction method can accurately reflect the overall topology of the chassis system, providing a structural foundation for collaborative simulation between subsystems, avoiding the limitations of analyzing a single subsystem, and helping to grasp chassis performance from a holistic perspective.

[0068] Multiphysics feature extraction is performed on real-time operational data to generate a chassis multiphysics dataset containing vibration mode feature vectors, structural stress distribution matrices, and a set of thermo-coupling parameters, enriching the dimensions of simulation analysis. The comprehensive consideration of multiphysics features can fully reflect the chassis's performance in mechanics, thermodynamics, and other aspects, making the simulation results more comprehensive and in-depth, no longer limited to the analysis of a single physics field.

[0069] Based on a multi-scale co-simulation computational process executed using a hierarchical digital twin model and a chassis multiphysics dataset, the output chassis dynamic response prediction results include the fatigue damage distribution of key connecting components and the deformation offset of the mechanism, achieving accurate prediction of the chassis dynamic response. Multi-scale co-simulation can analyze chassis performance at different scales, considering both the macroscopic overall response and the microscopic local details, thus improving the accuracy and breadth of the simulation.

[0070] Based on the chassis dynamic response prediction results, optimization parameter adjustment commands are generated and fed back to the physical chassis actuators, forming a complete closed loop from data acquisition and simulation analysis to optimization execution. This closed-loop optimization mechanism can promptly transform the results of simulation analysis into actual optimization actions, enabling the chassis mechanism's performance to be dynamically adjusted according to real-time operating conditions, adapting to performance requirements under different driving conditions. Attached Figure Description

[0071] Figure 1 This is a timing diagram of the digital twin-based automotive chassis mechanism simulation optimization method described in this invention.

[0072] Figure 2 A flowchart for real-time data collection;

[0073] Figure 3 A flowchart for multiphysics feature extraction processing;

[0074] Figure 4 The flowchart generated to optimize parameter adjustment instructions;

[0075] Figure 5 The flowchart for verifying virtual stimulus loading. Detailed Implementation

[0076] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0077] Please see Figure 1 This invention provides a simulation optimization method for automotive chassis mechanisms based on digital twins, the method comprising:

[0078] First, a multi-sensor array, including triaxial vibration sensors, linear displacement sensors, angle encoders, and infrared thermal imagers, is deployed on the physical chassis to form a comprehensive monitoring network covering mechanical motion and thermodynamic state. The sensors capture spatial acceleration, suspension displacement, steering angle, and temperature field data at millisecond-level sampling frequencies, which are then uploaded to a cloud-based digital twin platform after time synchronization. The platform establishes a hierarchical digital twin model based on the chassis topology, including the suspension subsystem, steering mechanism, and connecting components. This model describes the motion constraints and mechanical transmission relationships of each node through parameterized equations. A multiphysics feature extraction module performs frequency domain transformation, stress derivation, and thermo-coupling analysis on the raw data, generating a chassis multiphysics dataset that integrates vibration, stress, and temperature characteristics. A multi-scale co-simulation engine uses finite element method, multibody dynamics, and fluid dynamics sub-models for collaborative calculation, outputting dynamic response results including fatigue damage and deformation predictions. Finally, an optimization algorithm generates instruction sets for suspension stiffness adjustment and shock absorber damping correction, driving the physical actuators to complete closed-loop control.

[0079] Example 1: See Figure 2 This technology enables high-precision acquisition and synchronous transmission of real-time operational data from automotive chassis mechanisms. A miniature triaxial MEMS vibration sensor array is used at the hinge point of the chassis suspension control arm to measure spatial acceleration. This sensor utilizes the semiconductor piezoresistive effect and integrates an internal temperature compensation circuit to effectively suppress the impact of ambient temperature changes on measurement accuracy. The sensor is packaged in an aerospace-grade aluminum alloy shell with an IP67 protection rating, enabling it to withstand the harsh working environment of the chassis. Each sensor node is connected to the data acquisition unit via a CAN bus interface, with a sampling frequency set to 2048Hz to meet the requirements for capturing high-frequency vibration characteristics. The raw acceleration signal first undergoes hardware-level anti-aliasing filtering, with the cutoff frequency set to 40% of the sampling frequency, and then undergoes analog-to-digital conversion via a 24-bit Σ-Δ ADC. A differential signal transmission protocol is used during digital signal transmission to effectively reduce the impact of electromagnetic interference on signal quality.

[0080] The linear displacement measurement system uses a magnetostrictive displacement sensor to monitor the movement of the shock absorber piston rod. The sensor's measuring rod is installed parallel to the piston rod, employing a non-contact measurement method to avoid mechanical wear. The sensor internally incorporates a permanent magnet and waveguide wire structure, determining the displacement by measuring the strain pulse propagation time, achieving a linearity of 0.05% of full scale. A sampling interval of 5ms accurately records the dynamic response of the suspension system on bumpy roads. Measurement data is transmitted via an RS485 interface, encapsulated using the Modbus-RTU protocol, and verified using CRC-16. The steering angle monitoring system consists of a photoelectric encoder and a current acquisition module. The encoder uses an incremental measurement principle, outputting 5000 pulses per revolution, and a quadruple frequency circuit improves the angle resolution to 0.05°. The steering motor current signal is acquired by a Hall effect sensor with a sampling rate of 10kHz, and a digital filtering algorithm eliminates high-frequency noise from PWM control. The current-angle conversion algorithm considers parameters such as the motor torque constant and the transmission system reduction ratio to establish a precise correspondence between current fluctuations and steering angle changes.

[0081] Temperature field monitoring utilizes an infrared thermal imager array for non-contact measurement. The thermal imagers operate in the 8-14μm wavelength range, with a temperature measurement range of -20℃ to 300℃ and an optical resolution of 640×512 pixels. Each thermal imager is equipped with a programmable pan-tilt unit, capable of scanning key connection areas of the chassis along a preset trajectory. The thermal image acquisition interval is set to 30 seconds, automatically triggering a higher frequency acquisition mode in high-temperature areas. Image data is transmitted after H.264 encoding compression, while simultaneously recording ambient temperature and humidity compensation parameters. The time synchronization system is based on the IEEE 1588 precision time protocol, with a GPS-disciplined atomic clock as the master clock, achieving microsecond-level time synchronization accuracy. Each sensor node is equipped with hardware timestamp marking, ensuring precise recording of data acquisition moments and embedding them in the data frame header. The data transmission network employs a 5G industrial module, establishing a dedicated network slice to guarantee real-time and reliable data transmission. Data encapsulation uses a lightweight JSON format, with each data packet containing fields such as device ID, timestamp, data value, and checksum. A data integrity check module is deployed at the cloud-based receiving end to ensure complete data reception through verification sequence detection and retransmission mechanisms.

[0082] Data preprocessing is completed at the edge computing node, including signal filtering, outlier removal, and unit unification. Acceleration signals undergo band-limiting processing using a digital Butterworth filter to remove high-frequency noise and DC drift. Displacement data is smoothed using a moving average filter, with the window width dynamically adjusted based on vehicle speed. Steering angle data uses median filtering to eliminate impulse interference, and current signals undergo RMS conversion. Thermal image data undergoes non-uniformity correction and radiometric calibration to eliminate inherent response differences in the optical system. All preprocessed data is categorized and stored in a time-series database, establishing a timestamp-based index structure. The data association module establishes mapping relationships between different sensor data based on spatial location, for example, correlating control arm vibration data with temperature data from neighboring areas. The data quality control module continuously monitors the operating status of each sensor, identifying sensor faults or performance degradation through self-diagnostic algorithms. When an anomaly is detected, the system automatically switches to redundant sensors or triggers an early warning mechanism.

[0083] The real-time running dataset is stored using a distributed architecture. Raw data is stored in the Hadoop Distributed File System, while processed feature data is stored in a time-series database. The data access layer provides a unified API interface, supporting multi-dimensional queries by time range, device type, and spatial location. The data compression algorithm uses the Snappy compression library, reducing storage space usage while ensuring query efficiency. The data backup strategy employs a multi-site active-active architecture, using blockchain technology to ensure data immutability. For security, data transmission is encrypted using the TLS 1.3 protocol, stored data is encrypted using AES-256, and access control is based on the RBAC model for fine-grained permission management. The monitoring system displays the online status and data quality indicators of each sensor in real time, automatically triggering alarms when data transmission delays or anomalies are detected. The system maintenance module supports remote firmware upgrades and parameter configuration, and most maintenance operations can be completed through a web interface. To cope with network interruptions and other anomalies, the local storage module can cache at least 72 hours of running data, automatically resuming interrupted transmissions once the network is restored.

[0084] The sensor calibration system establishes a regular self-checking mechanism, including zero-point calibration, sensitivity calibration, and linearity testing. Accelerometers are calibrated using a centrifuge to generate a standard acceleration field, displacement sensors use a laser interferometer as a reference, and angle encoders verify measurement accuracy through an optical indexing head. Thermal imager calibration uses a blackbody radiation source as a standard temperature reference, and the radiation response curve is updated regularly. Calibration data is stored separately and used for compensation and correction of subsequent measurement data. An environmental adaptation mechanism automatically adjusts sensor operating parameters based on monitored temperature and humidity changes; for example, it automatically activates heating to prevent frost formation in low-temperature environments. The power supply system employs a dual-redundant design, with the main power source being the vehicle's electrical system and the backup power source being a supercapacitor bank, capable of maintaining continuous operation for at least 10 minutes in the event of a main power outage. Electromagnetic compatibility design includes power filtering, signal shielding, and grounding optimization measures to ensure stable sensor operation in complex electromagnetic environments. For mechanical protection, all exposed sensors are equipped with shock-absorbing brackets and protective covers, and critical connections utilize waterproof connectors and anti-loosening designs.

[0085] Example 2: See Figure 3 This document describes the technical implementation process of multi-physics feature extraction and multi-scale joint simulation. In the data processing stage, the system employs a sliding window mechanism to segment the acquired raw signals, with the window width dynamically adjusted based on signal characteristics. The vibration signal analysis module performs windowed Fourier transform on the triaxial acceleration data, using the Hanning window, which has excellent spectral characteristics. The transform result undergoes frequency band energy integration to extract the vibration energy distribution characteristics of specific frequency bands. Frequency band division references the inherent frequency characteristics of the chassis structure, focusing on potentially resonant frequency bands. Energy proportion calculation uses a relative value representation to eliminate the dimensional influence of absolute amplitude. Displacement signal processing employs a digital differential algorithm, obtaining a smooth velocity change curve through polynomial fitting and differentiation. Combined with elastic modulus data from the material parameter library, the structural stress distribution is derived based on fundamental mechanical principles. The material parameter library contains performance parameters under different temperature conditions, automatically selecting matching parameter values ​​based on real-time temperature data.

[0086] The temperature field analysis module performs region segmentation on infrared thermal images. The algorithm employs an improved watershed method, which can accurately identify the boundaries of temperature anomaly regions. Thermal region feature extraction includes operations such as geometric center coordinate calculation, temperature gradient analysis, and area change rate statistics. Thermo-mechanical coupling analysis establishes a correlation model between the temperature field and the vibration field, considering the nonlinear characteristics of the material's thermal expansion coefficient with temperature. Spatiotemporal alignment of vibration signals and temperature data is achieved through an interpolation algorithm, ensuring that data from different sampling rates can be correlated and analyzed under the same time reference. The feature fusion network adopts a deep convolutional structure. The first convolutional kernel is designed to operate in the time dimension, extracting the time-varying features of the vibration signal; the second convolutional kernel operates in the spatial dimension, capturing the spatial correlation of stress distribution; and the third convolutional kernel realizes feature interaction across physical fields. The network training process adopts an unsupervised learning approach, learning the low-dimensional representation of features through an autoencoder structure. The attention mechanism module dynamically adjusts the weight coefficients of different physical field features, highlighting feature changes under abnormal operating conditions.

[0087] The multi-scale simulation system employs a hierarchical solution strategy, with the finite element analysis module handling structural-scale problems. Model discretization uses high-order tetrahedral elements, and the mesh size is automatically adjusted based on the stress gradient. Material constitutive relations consider the nonlinear hardening effect under cyclic loading, and boundary conditions are dynamically updated based on measured vibration data. The solver uses an implicit time integration algorithm, with the iteration step size adaptively adjusted based on convergence. The multibody dynamics module establishes a multibody system model including flexible components. Joint constraints are represented parametrically, and the contact algorithm considers the influence of surface roughness. Kinematic solutions employ a constraint stabilization method, and dynamic solutions use the generalized-alpha algorithm. The fluid analysis module uses the finite volume method to discretize the governing equations, and a modified version of the turbulence model suitable for low-speed flows is selected. Mesh generation uses polyhedral elements, with boundary layer refinement in the near-wall region. Thermal boundary conditions are set based on infrared thermography data, and the convective heat transfer coefficient considers the influence of vehicle speed variations.

[0088] The coupling interface employs a loose coupling strategy, with each sub-model solved independently and then co-simulated through data exchange. The data mapping algorithm uses a conserved interpolation method to ensure the accuracy of energy and momentum transfer at the interface. Coupling between the finite element model and the multibody model is achieved through interface force transfer, with contact force distribution applied using equivalent nodal forces. The fluid-structure interaction interface utilizes bidirectional data exchange, and structural deformation is reflected in the fluid domain through dynamic meshing. The iterative convergence criterion is based on the relative error norms of multiple physical quantities; calculation terminates when the errors of all parameters are below a set threshold. Convergence acceleration techniques employ a quasi-Newton method to update the Jacobian matrix, reducing the number of iterations. Computational resource allocation uses a dynamic load balancing strategy, automatically adjusting processor core allocation based on the computational complexity of each sub-model.

[0089] The post-processing module performs in-depth analysis of the simulation output. Stress results utilize an isosurface extraction algorithm to identify high-stress regions, while the displacement field is visualized using vector fields to display deformation patterns. Fatigue analysis is based on stress-time histories and employs a cyclic counting method to statistically analyze the load spectrum. Thermal analysis results generate temperature contour maps and simultaneously calculate heat flux density distribution. A data fusion algorithm unifies the analysis results from different physical fields into a single spatiotemporal coordinate system, establishing a complete performance evaluation map. The anomaly detection module uses pattern recognition algorithms to identify potential failure risks, and an early warning mechanism triggers different response strategies based on the risk level. The visualization interface supports multi-view collaborative display and allows interactive querying of the physical quantity change history at any location. Data export formats support standard engineering analysis software interfaces, facilitating data exchange with other systems. The calculation process monitoring displays the solution status and resource usage of each sub-model in real time, and an automatic recovery mechanism is triggered in case of anomalies.

[0090] Example 3: See Figure 4 This paper describes the analysis and optimization parameter generation process of chassis dynamic response prediction results, as well as the construction method of a hierarchical digital twin model. In the dynamic response result processing stage, the system uses a region growth algorithm to identify high fatigue risk areas from the stress cloud map. The seed point selection is based on the statistical distribution characteristics of stress amplitude, and the growth criterion considers the stress gradient changes of adjacent nodes. Stress amplitude spectrum analysis uses an improved rainflow counting method to iteratively extract and statistically analyze the load time history, establishing a relationship matrix between stress range and the number of iterations. The calculation of suspension bushing stiffness adjustment incorporates the material stress-strain hysteresis characteristics, considering the nonlinear response of rubber materials under dynamic loads. The damper damping coefficient correction is based on velocity-force characteristic curve reconstruction, using an iterative search method to determine the optimal damping parameter that minimizes the phase difference. Deformation offset analysis uses principal component decomposition technology to extract the main deformation modes of the spatial displacement field, and the ball joint clearance compensation calculation considers the cumulative wear effect and the influence of fit tolerances. Fine-tuning of the control arm installation angle is achieved through inverse kinematics, with the solution process incorporating compensation requirements for manufacturing and assembly errors.

[0091] The derivation of thermal management optimization parameters is based on thermal flow field analysis. The objective function for optimizing the angle of the heat dissipation guide plate is defined as follows:

[0092]

[0093] in, Indicates the integral index of the thermal gradient. Represents the surface area of ​​the deflector. For temperature field distribution, The normal temperature gradient is represented. Minimization of this index is achieved using the conjugate gradient method, updating the guide vane tilt angle parameter in each iteration. A parameter correlation matrix is ​​established during the multi-objective optimization integration phase, with matrix elements reflecting the coupling strength between different adjustment amounts. The genetic algorithm uses real-number encoding, and the fitness function comprehensively considers three objectives: fatigue life extension, vibration suppression effect, and temperature control. The non-dominated solution selection uses a crowding distance sorting strategy to maintain the diversity and distribution of the solution set. The final optimization instruction set is generated in JSON-LD format, containing metadata such as parameter adjustment amounts, execution priorities, and effective conditions.

[0094] The hierarchical digital twin modeling process begins with chassis topology decomposition. The main load-bearing component layer is discretized using beam-shell hybrid elements, with node definitions including mass attributes and connection relationships. The modeling accuracy of the kinematic pair connection layer reaches the microscopic geometric feature level, and the gear meshing model includes tooth surface modification parameters and lubrication state variables. The viscoelastic behavior of the rubber bushing is described using a generalized Maxwell model, with parameter identification based on dynamic mechanical analysis test data. The ball joint clearance model considers collision effects in a multibody system, setting tolerance thresholds for different wear states. Parametric modeling of steering mechanism nodes employs a feature-dimensional driven method, with key dimensions linked to a design tolerance database. During model initialization, coordinate system unification is performed, rigidly aligning the CAD coordinate system with the sensor measurement coordinate system. The matching and calibration process uses a feature point registration algorithm, selecting reference points with clear geometric features as registration markers. Model validation is achieved through white noise excitation testing, comparing the correlation index between the simulated frequency response function and the measured data.

[0095] The execution strategy for parameter adjustment commands employs a tiered activation mechanism, with safety-critical parameters such as steering mechanism adjustments requiring multiple confirmations. Suspension bushing stiffness updates are achieved by modifying the hyperelastic material parameters in the finite element model, using the Ogden constitutive model to describe the nonlinear characteristics of the rubber. Shock absorber damping coefficient corrections are reflected in the force element definitions of the multibody model, constructing a velocity-damping force lookup table. Ball joint clearance compensation is achieved by adjusting the contact detection distance parameter, considering the influence of preload on the effective clearance. Control arm installation angle modifications are reflected in the initial assembly position of the multibody model, while simultaneously updating relevant constraint equations. The optimized parameters for the heat dissipation guide vane angle are converted into dynamic mesh boundary conditions in the CFD model, setting mesh deformation constraints to prevent element distortion. The command verification phase employs a virtual execution mode using a digital twin to simulate system response changes after parameter adjustments and check for violations of design constraints.

[0096] Model parameter management employs a version control mechanism, generating a new model branch version with each adjustment. Change logs include information such as modification content, execution time, and operator, supporting regression to any historical state. Parameter sensitivity analysis uses the Morris screening method to identify key parameters significantly impacting system response. Uncertainty quantification is achieved through Monte Carlo simulation, assessing the confidence interval of prediction results based on parameter fluctuations. Interface protocol definitions adopt the ASAM standard format, ensuring data compatibility across different software platforms. Real-time data exchange is achieved through a shared memory mechanism, reducing latency caused by I / O operations. Computational resource scheduling considers task priority and timeliness requirements, allocating dedicated computing nodes to critical analysis tasks. Anomaly handling mechanisms include parameter out-of-bounds detection, model convergence monitoring, and computation interruption recovery, ensuring continuous and stable system operation. A visual monitoring interface displays the comparison effects before and after parameter adjustments, supporting 3D interactive viewing of local detail changes.

[0097] Example 4: See Figure 5 This document outlines the complete technical process for virtual excitation loading and dynamic twin updates. In the virtual excitation loading stage, the system automatically modifies the attribute parameters of the digital twin model through a parametric script engine. Taking the front suspension system of a certain SUV as an example, when a suspension bushing stiffness adjustment command is received, the system calls the ANSYS APDL script interface to modify the hyperelastic parameters in the rubber material definition. Specific operations include updating the C10 and C01 coefficients in the Mooney-Rivlin model, with the adjustment values ​​calculated based on the percentage change in the optimization command. Shock absorber damping parameter modification is achieved through the damper element attribute editor in the ADAMS / Car module, importing the corrected velocity-force characteristic curve in the form of a two-dimensional lookup table. Steering mechanism backlash parameter adjustment involves multiple parameters defined in the gear pair, including basic tooth profile deviation, pitch error, and tooth profile modification. These parameters are updated in a linked manner in the parametric model through a variable association mechanism.

[0098] The virtual pavement excitation is constructed by combining standard spectrum and measured data. The system uses the pavement roughness classification spectrum defined by ISO8608 as a basis, while also incorporating pavement profile data collected from a specific test site. The excitation signal generator produces a composite waveform containing random amplitude and discrete impacts, with the time series sampling interval set to 0.001 seconds. Table 1 below shows the excitation parameter configuration used in a certain virtual test:

[0099] Table 1: Excitation parameter configuration table used in a certain virtual experiment.

[0100]

[0101] The dynamic twin update process uses real vehicle feedback data as input. The CAN bus acquisition module captures shock absorber displacement sensor and steering motor torque signals at a frequency of 100Hz. The data packet contains information such as timestamp, channel ID, and engineering values. The signal preprocessing stage performs sensor characteristic compensation, such as software correction for temperature drift of the magnetostrictive displacement sensor. The data alignment module uses a dynamic time warping algorithm to solve the time base difference problem between simulation and measured signals. Residual calculation uses specific indicators for different types of signals: vibration signals are mainly analyzed for differences in frequency domain energy distribution, displacement signals focus on phase lag, and torque signals are compared for peak hold characteristics.

[0102] When the amplitude attenuation rate exceeds a threshold, the system initiates a model parameter correction process. The mass distribution update of the finite element model is based on the modal confidence criterion, iteratively adjusting density parameters to improve the correlation coefficient between simulated and measured modal shapes. The energy dissipation coefficient correction for the multibody model involves multiple components, including the damping coefficient of the bushings, the friction parameters of the kinematic pairs, and the hysteresis characteristics of the tires. The update algorithm uses gradient descent to find the optimal parameter combination, and the model's effectiveness is verified by rerunning under standard operating conditions after each modification. Taking the steering system model update as an example, when a deviation of more than 20% in the steering torque feedback is detected, the system sequentially adjusts the gear transmission efficiency parameter, steering shaft stiffness value, and steering gear internal friction coefficient until the error rate between the simulation results and measured data decreases to within 5%.

[0103] Model version management employs a branch-merge strategy, creating a new model branch for each major parameter update. The version control system records a complete modification history, including metadata such as a list of parameter changes, modification times, and verification results. A rollback mechanism allows for rapid restoration to a stable version in case of anomalies, while preserving all intermediate states needed for problem analysis. A parameter sensitivity analysis module runs periodically to identify the key parameters that have the greatest impact on system response. The analysis results are stored in the form of a parameter impact factor matrix, guiding the prioritization of subsequent update operations.

[0104] The anomaly handling system establishes a multi-level monitoring mechanism, including parameter boundary checks, model convergence diagnosis, and hardware resource monitoring. When parameter out-of-bounds conditions are detected, the system automatically freezes modification commands and triggers an expert review process. Convergence issues during model solving are mitigated by automatically adjusting numerical parameters such as time step and iteration tolerance. Computational resource management employs a dynamic allocation strategy, allowing critical simulation tasks to exclusively occupy computing nodes to ensure timeliness.

[0105] The visual monitoring interface provides a comparison display of parameters before and after adjustment, supporting simultaneous operation of multiple views. The 3D model view uses shading and deformation animations to intuitively display the modification effect, while the curve comparison view can overlay the differences between simulation and measured data. The parameter influence path diagram shows the transmission relationship of adjusted parameters, helping to understand the cascading effects of the system. The user interaction function allows for fine-tuning of parameter values ​​through drag-and-drop, and real-time observation of the changing trends of the system response. The data export module supports the generation of diagnostic reports compliant with ASAMODX standards, containing detailed parameter modification records and verification data.

[0106] Example 5: Specific Implementation Process of Model Iterative Optimization and Digital Twin Visualization Technology. During the model iterative optimization phase, the system establishes a periodic data acquisition mechanism, periodically acquiring actual operating status data of key components of the physical chassis through the on-board diagnostic interface. The fatigue damage monitoring module reads the cumulative values ​​of preset counters from the vehicle control unit; these counters record the number of cycles and amplitude information under specific load conditions. The data acquisition cycle is dynamically adjusted according to component characteristics. A daily acquisition strategy is adopted for key components such as the suspension system, while a weekly or monthly acquisition scheme is used for auxiliary components. A strict data verification process is executed during the acquisition process, including range checks, rationality verification, and integrity confirmation steps, to eliminate abnormal data caused by sensor malfunctions or communication interference.

[0107] The simulation prediction value extraction module queries the calculation results for the corresponding time period from the digital twin database. These results are stored in a standard format during each simulation run. The data matching process considers the differences between actual operating conditions and simulation conditions, and normalizes the data using operating condition conversion factors. Error analysis employs a multi-dimensional evaluation method, comparing not only the differences in overall damage values ​​but also analyzing the morphological characteristics and trends of the damage accumulation curve. Weighting factors are introduced during the construction of the error distribution matrix to reflect the differences in the importance of components at different locations. The backpropagation algorithm uses an adaptive learning rate strategy when adjusting material parameters, dynamically adjusting the update step size based on parameter sensitivity. The thermal conductivity optimization considers the influence of temperature gradients, employing different correction strategies for high-temperature and normal-temperature regions. Safety boundary constraints are set during the parameter update process to prevent the optimization process from causing physical property parameters to exceed reasonable ranges.

[0108] The visualization system is implemented based on modern graphics rendering technology, establishing a complete processing chain from simulation data to 3D display. Visualization of stress concentration areas employs a physically based rendering method, with surface shading considering the anisotropic reflectivity of metallic materials. The color mapping scheme is ergonomically optimized, using a gradient color spectrum that aligns with engineering intuition, transitioning from cool tones in low-stress areas to warm tones in high-stress areas. Dynamic cloud map generation of deformation offsets utilizes a combination of particle systems and isosurface rendering, effectively representing both overall deformation trends and local detail changes. The animation refresh rate is synchronized with the simulation time step, supporting slow-motion playback and keyframe marking functions.

[0109] The visualization of parameter adjustment commands employs a difference comparison display method, simultaneously showing the states before and after adjustment through semi-transparent overlay technology. The execution effect simulation utilizes a physically based animation engine to accurately represent the mechanical system response caused by parameter changes. The remaining life warning system for critical connectors provides multi-level visual cues, including color coding, icon markings, and text annotations. Warning thresholds are dynamically set according to component design standards, and color gradient changes are achieved smoothly using the HSL color space interpolation algorithm. Interactive functionality allows users to click to query detailed data at any location, with pop-up windows displaying stress history curves and predicted remaining life values.

[0110] The 3D scene construction employs hierarchical detailing technology, dynamically adjusting model complexity based on viewing distance. Viewpoint control provides standard engineering views and a free-browsing mode, supporting analysis tools such as section planes and transparency. The lighting model considers the specific needs of engineering analysis, setting a shadowless, uniform lighting mode to accurately represent surface details. The rendering pipeline is optimized for smooth interaction in complex scenes on ordinary workstations. Data compression technology reduces network latency, and a progressive loading strategy enhances the user experience.

[0111] The model version comparison function visualizes differences across time and space, supporting side-by-side display of simulation results from different optimization stages. Historical data backtracking can reproduce the system state at specific points in time, aiding in the analysis of the long-term effects of parameter adjustments. The collaborative review tool allows multiple users to simultaneously annotate and discuss visualization results, with annotation information stored in association with the 3D model. The report generation module automatically organizes the visualization analysis results, outputting 2D drawings and 3D screenshots conforming to engineering standards. The decision support function guides engineers to focus on the areas most in need of optimization by highlighting key problem areas.

[0112] In terms of system integration, data interoperability with the PLM platform is achieved, allowing direct import of component design information into the visualization scene. The real-time data interface supports linkage with the test bench system, enabling synchronized display of the digital twin and the actual component. The access control system manages the operational permissions of different users, ensuring secure access to core parameters. The logging function records all visualization operations and parameter modifications in detail, meeting the traceability requirements for engineering changes. Cross-platform compatibility is achieved through WebGL technology, supporting access to visualization results on various terminal devices.

[0113] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0114] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A digital-twin-based simulation optimization method for a vehicle chassis mechanism, characterized by, The method comprises the following steps: Collecting a real-time running data set of a vehicle chassis mechanism, the real-time running data set including multi-axis acceleration time series information, suspension displacement monitoring sequences, steering mechanism angle change amounts, and temperature distribution maps, and synchronously transmitting the real-time running data set to a cloud digital twin platform; In the cloud digital twin platform, a hierarchical digital twin model is constructed according to a chassis topology structure, the hierarchical digital twin model including suspension subsystem nodes, steering mechanism nodes, and spatial constraint relationships of connecting components; Multi-physical field feature extraction processing is performed on the real-time running data set to generate a chassis multi-physical field data set, the chassis multi-physical field data set covering vibration modal feature vectors, structural stress distribution matrices, and thermal coupling parameter sets; Based on the hierarchical digital twin model and the chassis multi-physical field data set, a multi-scale joint simulation calculation process is performed to output chassis dynamic response prediction results, the chassis dynamic response prediction results including fatigue damage distributions of key connecting components and mechanism deformation offset amounts; Optimization parameter adjustment instruction sets are generated according to the chassis dynamic response prediction results, and the optimization parameter adjustment instruction sets are fed back to physical chassis execution mechanisms.

2. The method of claim 1, wherein, The collecting of the real-time running data set of the vehicle chassis mechanism comprises: A three-axis vibration sensor is arranged at a hinge point of a chassis suspension control arm to obtain multi-axis acceleration time series information at a fixed sampling frequency; A linear displacement sensor is used to collect piston rod extension and retraction amounts of each shock absorber to form suspension displacement monitoring sequences; An angle encoder is used to record steering tie rod swing angle change amounts in real time, and steering motor current fluctuation data are combined to generate steering mechanism angle change amounts; An infrared thermal imager is used to scan chassis metal connecting areas to generate temperature distribution maps; The multi-axis acceleration time series information, the suspension displacement monitoring sequences, the steering mechanism angle change amounts, and the temperature distribution maps are stored as the real-time running data set after being aligned according to time stamps.

3. The method of claim 2, wherein, The multi-physical field feature extraction processing on the real-time running data set comprises: Short-time Fourier transform processing is performed on the multi-axis acceleration time series information to extract vibration energy proportions in specific frequency bands as basic elements of vibration modal feature vectors; Based on the suspension displacement monitoring sequences, displacement change rates of adjacent sampling points are calculated, and initial values of structural stress distribution matrices are derived in combination with material elastic modulus parameters; Geometric center coordinates of high-temperature areas in the temperature distribution maps are analyzed, vibration modal feature vectors of corresponding positions are associated, and a mapping relationship of thermal coupling parameter sets is established; Convolutional neural networks are used to fuse the vibration modal feature vectors, the structural stress distribution matrices, and the thermal coupling parameter sets to generate a chassis multi-physical field data set including space-time correlation features.

4. The method of claim 3, wherein, The performing of the multi-scale joint simulation calculation process comprises: The vibration modal feature vectors are input into a finite element submodel to calculate dynamic load transfer functions of suspension subsystem nodes; The structural stress distribution matrices are input into a multi-body dynamics submodel to solve motion trajectory deviation amounts of steering mechanism nodes; inputting the thermal coupling parameter set into a fluid mechanics submodel to simulate air convection heat dissipation efficiency of the chassis connecting area; establishing a coupling interface of the finite element submodel, the multi-body dynamics submodel and the fluid mechanics submodel, so that a dynamic load transfer function of a suspension subsystem node is inputted into the multi-body dynamics submodel as a boundary condition, and a motion trajectory deviation of a steering mechanism node is fed back to the finite element submodel; iteratively calculating until an output error rate of each submodel is lower than a preset threshold, and generating a chassis dynamic response prediction result containing a stress concentration area coordinate and a deformation trend vector.

5. The method of claim 4, wherein, The generating of the optimization parameter adjustment instruction set according to the chassis dynamic response prediction result comprises: identifying a high fatigue damage area in the chassis dynamic response prediction result, and extracting a material stress amplitude spectrum of a corresponding position; generating a suspension bushing stiffness adjustment amount and a shock absorber damping coefficient correction value according to a peak frequency characteristic of the stress amplitude spectrum; analyzing a spatial distribution law of a mechanism deformation offset, and calculating a steering mechanism ball joint gap compensation amount and a control arm installation angle fine adjustment amount; deducing a heat dissipation guide plate angle optimization parameter based on a thermal gradient change trend of a temperature distribution map; integrating the suspension bushing stiffness adjustment amount, the shock absorber damping coefficient correction value, the steering mechanism ball joint gap compensation amount, the control arm installation angle fine adjustment amount and the heat dissipation guide plate angle optimization parameter to form the optimization parameter adjustment instruction set.

6. The method of claim 5, wherein, The constructing of the hierarchical digital twin model according to the chassis topology structure comprises: establishing a hierarchical topology map of a chassis mechanism assembly, the hierarchical topology map containing a main load-bearing member layer, a kinematic pair connection layer and an auxiliary actuator layer; defining a six-degree-of-freedom motion constraint equation for each suspension subsystem node, the six-degree-of-freedom motion constraint equation containing a displacement boundary condition and a torque transfer function; establishing a gear meshing parameterized model for a steering mechanism node, the gear meshing parameterized model covering a gear backlash nonlinear characteristic and a friction coefficient variable; setting a viscoelastic constitutive relation of a rubber bushing and a ball joint gap tolerance threshold in the kinematic pair connection layer; matching and calibrating the hierarchical topology map with sensor spatial coordinates in a real-time running data set to complete initialization of the hierarchical digital twin model.

7. The method of claim 6, wherein, It also comprises a virtual excitation loading step: modifying material attribute parameters of a corresponding node in the hierarchical digital twin model according to a suspension bushing stiffness adjustment amount in the optimization parameter adjustment instruction set; updating a damping matrix element in the finite element submodel based on a shock absorber damping coefficient correction value; injecting a steering mechanism ball joint gap compensation amount into a gear backlash tolerance variable of the gear meshing parameterized model; applying a virtual road surface excitation spectrum to the updated hierarchical digital twin model by using a modal superposition method, the virtual road surface excitation spectrum containing a random amplitude waveform and a step impact component; collecting model response data under virtual excitation to verify effectiveness of the optimization parameter adjustment instruction set.

8. The method of claim 7, wherein, It also comprises a dynamic twin updating step: real-time receiving feedback data of a physical chassis execution mechanism, the feedback data containing an actual shock absorber displacement change curve and a steering motor torque output amount; calculating a residual index of the feedback data and the chassis dynamic response prediction result, the residual index including a phase lag amount and an amplitude attenuation rate; when the residual index exceeds a preset tolerance threshold, starting a parameter correction process of the hierarchical digital twin model; adjusting a mass distribution parameter of the finite element sub-model according to the phase lag amount, and correcting an energy dissipation coefficient of the multi-body dynamics sub-model according to the amplitude attenuation rate; using the updated hierarchical digital twin model as a calculation basis for a next simulation cycle.

9. The method of claim 8, wherein, Further comprising a model iteration optimization step: accumulating fatigue damage measured data of the physical chassis actuator within a preset time period; extracting a simulation fatigue damage prediction value of the corresponding cycle of the hierarchical digital twin model; constructing an error distribution matrix of the measured data and the simulation fatigue damage prediction value; adjusting a material degradation rate parameter and a heat conduction coefficient in the multi-scale joint simulation calculation process through a back propagation algorithm; generating a verified multi-scale joint simulation calculation process for chassis dynamic response prediction in a subsequent cycle.

10. The method of claim 9, wherein, Further comprising a digital twin visualization step: mapping stress concentration region coordinates in the chassis dynamic response prediction result to a three-dimensional visualization engine; rendering and displaying a dynamic change cloud chart of the actuator deformation offset in the three-dimensional model; superimposedly displaying an execution effect simulation animation of the optimization parameter adjustment instruction set; establishing a remaining life warning identification system of the key connecting piece, the remaining life warning identification system driving color gradient changes based on fatigue damage distribution data.

Images