Simulation model parameter calibration method, electronic device, and computer program product

Abstract

The application discloses a simulation model parameter calibration method, an electronic device and a computer program product, relates to the technical field of simulation, and comprises the following steps: acquiring each working condition information of a simulation model, and determining to-be-calibrated parameters of the simulation model based on the each working condition information, wherein the to-be-calibrated parameters comprise global parameters and local parameters; the calibration parameters are divided into global parameters and local parameters, different types of parameters can be distinguished, the consistency of cross-working condition physical parameters is ensured through the global parameters, the disturbance deviation special for each working condition is adapted through the local parameters, error of the local parameters is prevented from interfering with optimization of other parameters, and on this basis, the global parameters and the local parameters are calibrated in layers, so that mutual interference of multi-working condition errors is avoided. Through parameter classification and layered calibration, the application not only ensures physical consistency and universality of cross-working condition operation of the simulation model, but also greatly improves calibration precision and adaptation capability of parameters of the multi-working condition simulation model.

Description

Technical Field

[0001] This application relates to the field of simulation technology, and in particular to simulation model parameter calibration methods, electronic devices, and computer program products. Background Technology

[0002] Digital simulation technology is widely used in industrial fields such as electromechanical, vehicle, fluid, thermal management and control systems. As a core tool for equipment performance analysis and optimization design, the calibration accuracy of the simulation model directly determines the credibility of the simulation results and is a key foundation for realistically reproducing the operating response of equipment.

[0003] Currently, simulation model calibration mainly employs two methods: independent calibration under a single operating condition and black-box optimization. The independent calibration method can only meet the accuracy requirements of a single operating condition. However, actual engineering equipment typically encompasses multiple operating conditions such as speed, load, temperature, and pressure. Adjusting parameters individually for each condition leads to conflicting parameters and inconsistent physical meanings, resulting in poor cross-condition adaptability and insufficient versatility. Black-box optimization, on the other hand, typically integrates errors from multiple operating conditions into a single objective function for iterative optimization. Errors caused by various parameters can interfere with each other, leading to parameter distortion and loss of physical interpretability. In summary, existing simulation model calibration methods suffer from poor parameter consistency and parameter distortion across operating conditions, making it difficult to meet the high-precision calibration requirements of simulation models under multiple operating conditions.

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

[0005] The main purpose of this application is to provide a method for calibrating simulation model parameters, electronic equipment, and computer program products, aiming to solve the technical problems of poor parameter consistency and parameter distortion in existing simulation model calibration methods, which make it difficult to meet the high-precision calibration requirements of simulation models under multiple operating conditions.

[0006] To achieve the above objectives, this application proposes a method for calibrating simulation model parameters, which includes: Obtain the operating condition information of the simulation model, and determine the calibration parameters of the simulation model based on the operating condition information, wherein the calibration parameters include global parameters and local parameters; Candidate global parameter values ​​are generated for the global parameters according to a preset initial generation strategy; Based on the candidate global parameter values ​​and the information on each working condition, the model for each experimental working condition is determined; Based on the experimental working condition models, working condition simulation experiments were conducted to calibrate the local parameter values ​​of the local parameters and obtain the experimental results for each working condition. Based on the experimental results of each working condition, new candidate global parameter values ​​are generated; Based on the new candidate global parameter values, the steps of determining the experimental working condition model based on the candidate global parameter values ​​and the working condition information are returned until the preset calibration conditions are reached. The new candidate global parameter values ​​and local parameter values ​​obtained when the preset calibration conditions are reached are respectively used as the target parameter values ​​of the global parameters and the local parameters.

[0007] In one embodiment, the global parameters include constant parameters and function parameters, and the step of determining the calibration parameters of the simulation model based on the various operating condition information includes: Based on the information of each working condition and the simulation model, determine each candidate working condition model of the simulation model; Obtain the operating condition constant parameters, operating condition mapping parameters, and operating condition exclusive parameters of each candidate operating condition model. For each candidate operating condition model, the operating condition constant parameters are constant parameters that can be shared with other candidate operating condition models. The operating condition mapping parameters are determined by a functional relationship shared with other candidate operating condition models. The operating condition exclusive parameters are parameters unique to the candidate operating condition model. The operating condition constant parameter is used as the constant parameter of the global parameter; Obtain the function expression of the working condition mapping parameters, and use the coefficients of the function expression as the function parameters of the global parameters.

[0008] The operating condition exclusive parameters are used as the local parameters.

[0009] In one embodiment, the candidate global parameter values ​​include candidate constant parameter values ​​for the constant parameters and candidate function parameter values ​​for the function parameters; The step of determining the experimental condition model based on the candidate global parameter values ​​and the information on each operating condition includes: The experimental parameter values ​​are determined by traversing each candidate working condition model. The experimental parameter values ​​include working condition constant experimental parameter values ​​and working condition mapping experimental parameter values. For any candidate working condition model of the simulation model, if the working condition constant parameter exists in the candidate working condition model, the working condition constant experimental parameter value of the working condition constant parameter is determined from the candidate constant parameter values. If the working condition mapping parameter exists in the candidate working condition model, the working condition mapping experimental parameter value is obtained by substituting the candidate function parameter value into the function expression of the working condition mapping parameter. The experimental condition models are obtained by substituting the experimental parameter values ​​into each candidate operating condition model.

[0010] In one embodiment, the step of conducting working condition simulation experiments based on each experimental working condition model, calibrating the local parameter values ​​of the local parameters, and obtaining the experimental results for each working condition includes: Using the aforementioned experimental working condition models, working condition simulation experiments were performed, and simulation results for each working condition were obtained. For each experimental working condition model containing a working condition-specific parameter, a local error objective function is constructed based on the working condition simulation results of the target working condition model and the target signal in the corresponding target working condition information of the target working condition model. Under preset local constraints, the local error objective function is optimized using the preset inner layer optimizer of the target working condition model to obtain the working condition-specific experimental parameter values ​​of the working condition-specific parameters in the target working condition model. Based on the exclusive experimental parameter values ​​under the aforementioned operating conditions, the local parameter values ​​are obtained. Based on the simulation results of each working condition and the local parameter values, the experimental results of each working condition are obtained.

[0011] In one embodiment, the experimental results for each working condition include the experimental results for a first working condition and the experimental results for a second working condition. The step of obtaining the experimental results for each working condition based on the simulation results for each working condition and the local parameter values ​​includes: The simulation results of each working condition are processed by a preset result processing method to obtain the experimental results of each working condition. Specifically, for a first experimental working condition model containing post-processing-type working condition exclusive parameters, the simulation results of the first working condition model are processed by a preset post-processing method to obtain the first working condition experimental results. The post-processing-type working condition exclusive parameters do not change the state trajectory of the simulation model. The preset post-processing method includes at least one of time shifting, signal biasing, and signal scaling. For a second experimental working condition model containing re-simulation-type working condition exclusive parameters, the parameter values ​​of the re-simulation-type working condition exclusive parameters are substituted into the second experimental working condition model for re-simulation to obtain the second working condition experimental results. The re-simulation-type working condition exclusive parameters change the state trajectory of the simulation model. The re-simulation-type working condition exclusive parameters include at least one of initial state offset parameters, input scaling parameters, and input biasing parameters.

[0012] In one embodiment, the preset local constraint conditions include at least one of regularization constraints and amplitude boundary constraints.

[0013] In one embodiment, the step of generating new candidate global parameter values ​​based on the experimental results of each operating condition includes: Calculate the error between the experimental results of each working condition and the target signal in the information of each working condition to obtain the working condition error of each working condition experimental result, and aggregate the working condition errors to obtain the aggregated error; The candidate global parameter values ​​are optimized based on the aggregation error to obtain new candidate global parameter values.

[0014] In one embodiment, the step of aggregating the errors of each operating condition to obtain the aggregated error includes: Calculate the 90th percentile error and the maximum error under each operating condition; The average value of the errors under each working condition is taken as the average error; The average error, the 90th percentile error, and the maximum operating error are each assigned a corresponding weight, and then weighted and summed to obtain the aggregated error.

[0015] In addition, to achieve the above objectives, this application also proposes an electronic device, the device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the simulation model parameter calibration method as described above.

[0016] In addition, to achieve the above objectives, this application also provides a computer program product, which includes a computer program that, when executed by a processor, implements the steps of the simulation model parameter calibration method described above.

[0017] One or more technical solutions proposed in this application have at least the following technical effects: In this embodiment, the simulation model's operating condition information is first obtained, and based on this information, two categories of parameters to be calibrated are identified: global parameters and local parameters. This classification of simulation model parameters facilitates subsequent fine-grained calibration. Next, candidate global parameter values ​​are generated for the global parameters. An experimental operating condition model is constructed using these fixed candidate global parameter values, and operating condition simulation experiments are conducted. This determines the local parameter values ​​for the local parameters and yields the experimental results for each operating condition. Then, new candidate global parameter values ​​are generated based on the experimental results, thereby optimizing the values ​​of the global parameters. By performing hierarchical calibration of global and local parameters, interference between different types of parameters is avoided. The above steps are iterated until the preset calibration conditions are met, thus achieving multi-operating condition joint simulation calibration.

[0018] This application classifies calibration parameters into global and local parameters, distinguishing between different types of parameters. Global parameters ensure consistency of physical parameters across operating conditions, mitigating parameter conflicts, unifying the physical meaning of parameters, and thus improving the model's adaptability and versatility. Local parameters adapt to the unique disturbances and deviations of each operating condition, preventing errors in local parameters from interfering with the optimization of other parameters, thereby avoiding parameter distortion. Furthermore, hierarchical calibration of global and local parameters avoids mutual interference between errors from multiple operating conditions, accurately meeting the demands for high-precision calibration across complex and changing operating conditions. Through parameter classification and hierarchical calibration, this application not only ensures the physical consistency and versatility of the simulation model across operating conditions but also significantly improves the calibration accuracy and adaptability of multi-condition simulation model parameters. Attached Figure Description

[0019] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0020] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a flowchart illustrating an embodiment of the simulation model parameter calibration method of this application. Figure 2 This is a schematic diagram illustrating the logic for classifying exclusive parameters under operating conditions in this application. Figure 3 This is a simplified flowchart illustrating the simulation model parameter calibration method of this application; Figure 4 This is a schematic diagram of the simulation curves after the tire physical simulation model was calibrated, as shown in the example. Figure 5 This is a schematic diagram of the hardware operating environment involved in the simulation model parameter calibration method in this application embodiment.

[0022] The purpose, features, and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0023] It should be understood that the specific embodiments described herein are merely illustrative of the technical solutions of this application and are not intended to limit this application.

[0024] To better understand the technical solution of this application, a detailed description will be provided below in conjunction with the accompanying drawings and specific implementation methods.

[0025] In engineering fields such as electromechanical engineering, automotive, fluid dynamics, thermal management, and control systems, physical simulation models typically consist of a large number of parameters with physical meaning, such as stiffness, damping, inertia, efficiency, friction, control gain, valve coefficient, pump characteristics, or tire characteristics. To ensure that the simulation model accurately reflects the response of the real equipment or experimental system, the model parameters need to be calibrated based on experimental data.

[0026] Traditional calibration methods often involve repeatedly adjusting parameters based on a single experimental condition. When a real engineering model needs to simultaneously meet multiple conditions, such as the vertical load, lateral load, and lateral deflection input conditions for a vehicle tire model, or different speed and load conditions for a motor model, calibrating independently for each condition will result in inconsistent parameters, causing the model to lose its unified physical meaning when used across different conditions.

[0027] Existing automatic optimization methods can apply black-box optimization algorithms to model parameter search, but they typically compress errors from multiple operating conditions into a single objective function without distinguishing the sources of error. Local measurement biases, time alignment errors, input excitation deviations, and initial state differences can easily affect global physical parameters, causing them to lose their physical interpretation. Furthermore, the continuous variation of parameters with operating conditions such as load, speed, temperature, and pressure is difficult to accurately express using fixed constants or independent parameters for each operating condition. Therefore, a calibration method is needed that can maintain physical consistency across multiple operating conditions while also considering condition-specific disturbances and expressing the variation of parameters with operating condition characteristics.

[0028] The main objective of this invention is to resolve the contradiction between "physical consistency across operating conditions" and "absorption of operating condition-specific disturbances" in multi-operating condition calibration. If all errors are borne by global physical parameters, local disturbances will contaminate the global parameters; if each operating condition is calibrated independently, the parameters of each operating condition are fragmented, making it difficult for the model to generalize.

[0029] The main solution of this application embodiment is as follows: Obtain the operating condition information of the simulation model, and determine the parameters to be calibrated of the simulation model based on the operating condition information, wherein the parameters to be calibrated include global parameters and local parameters; generate candidate global parameter values ​​for the global parameters according to a preset initial generation strategy; determine each experimental operating condition model based on the candidate global parameter values ​​and the operating condition information; conduct operating condition simulation experiments based on each experimental operating condition model, calibrate the local parameter values ​​of the local parameters, and obtain the experimental results for each operating condition; generate new candidate global parameter values ​​based on the experimental results for each operating condition; return to the step of determining each experimental operating condition model based on the candidate global parameter values ​​and the operating condition information based on the new candidate global parameter values, until a preset calibration condition is reached, and use the new candidate global parameter values ​​and local parameter values ​​obtained when the preset calibration condition is reached as the target parameter values ​​for the global parameters and the local parameters, respectively.

[0030] In this embodiment, by dividing calibration parameters into global and local parameters, different types of parameters can be distinguished. Global parameters ensure consistency of physical parameters across operating conditions, thereby mitigating parameter conflicts, unifying the physical meaning of parameters, and improving the model's adaptability and versatility. Local parameters adapt to the specific disturbances and deviations of each operating condition, preventing errors in local parameters from interfering with the optimization of other parameters, thus avoiding parameter distortion. Based on this, hierarchical calibration of global and local parameters avoids mutual interference between errors from multiple operating conditions, accurately adapting to the complex and ever-changing high-precision calibration requirements of multiple operating conditions. This application, through parameter classification and hierarchical calibration, ensures both the physical consistency and versatility of the simulation model across operating conditions and significantly improves the calibration accuracy and adaptability of multi-operating condition simulation model parameters. In summary, the main purpose of this application is to resolve the contradiction between "physical consistency across operating conditions" and "the influence of operating condition-specific disturbances" in multi-operating condition calibration. If all errors are borne by global physical parameters, local disturbances will pollute the global parameters; if each operating condition is calibrated independently, the parameters of each operating condition are fragmented, making the model difficult to generalize.

[0031] Based on this, the embodiments of this application provide a method for calibrating simulation model parameters, referring to... Figure 1 , Figure 1 This is a flowchart illustrating the first embodiment of the simulation model parameter calibration method of this application.

[0032] In this embodiment, the simulation model parameter calibration method includes steps S1 to S6: Step S1: Obtain the operating condition information of the simulation model, and determine the calibration parameters of the simulation model based on the operating condition information, wherein the calibration parameters include global parameters and local parameters; It should be noted that the simulation model is a physical simulation model corresponding to industrial equipment, which can realistically represent the physical operating mechanism of equipment such as electromechanical, vehicle, fluid, thermal management, and control systems; the operating condition information is a set of information corresponding to multiple independent operating conditions included in the calibration task, specifically including experimental data, target signals to be fitted, simulation configuration parameters, operating condition characteristic parameters, and optional local parameter references for each operating condition; the parameters to be calibrated are unknown parameters of the simulation model that are strongly correlated with the operating state of multiple operating conditions and need to be determined through iterative optimization using experimental data.

[0033] Additionally, it should be noted that, based on the functional characteristics of the parameters and the scope of operation, this embodiment divides the parameters to be calibrated into global parameters and local parameters. Global parameters are shared across all operating conditions and are used to characterize the inherent physical characteristics of the equipment. Local parameters are exclusive parameters that only take effect under a single operating condition and are used to absorb the deviations specific to that operating condition. This embodiment accurately selects and divides all the parameters to be calibrated to adapt to the multi-operating-condition calibration task by using the experimental characteristics, simulation configuration, and parameter scope corresponding to each operating condition.

[0034] Understandably, by distinguishing between global and local layers of the calibration parameters, it is possible to decouple the calibration of inherent deviations of the operating conditions from the global physical parameters, effectively avoid local errors such as single-condition measurement disturbances and input disturbances from polluting the global physical parameters, significantly improve the physical interpretability of the calibration parameters, and at the same time ensure the consistency of parameters when the simulation model runs across different operating conditions.

[0035] In one feasible implementation, the global parameters include constant parameters and function parameters. Step S1: Obtaining the operating condition information of the simulation model and determining the calibration parameters of the simulation model based on the operating condition information may include steps S11~S15: Step S11: Based on the information of each working condition and the simulation model, determine each candidate working condition model of the simulation model; It should be noted that the candidate working condition model refers to the simulation model mapped to the corresponding basic model state under each independent working condition scenario. Specific parameter values ​​are not loaded. It is used to configure the corresponding parameters, generate the final computable experimental working condition model, and complete the corresponding working condition simulation experiment.

[0036] Step S12: Obtain the operating condition constant parameters, operating condition mapping parameters, and operating condition exclusive parameters of each candidate operating condition model. For each candidate operating condition model, the operating condition constant parameters are constant parameters that can be shared with other candidate operating condition models. The operating condition mapping parameters are determined by a functional relationship shared with other candidate operating condition models. The operating condition exclusive parameters are parameters exclusive to the candidate operating condition model. It should be noted that operating condition constant parameters are inherent physical properties of the equipment that do not change with operating conditions. These parameters have fixed values ​​used throughout the entire operating process and under all operating conditions. Examples include basic physical parameters such as stiffness, damping, and tire shape factor. Operating condition-specific parameters are correction parameters introduced during calibration, effective only within the corresponding single operating condition. They are used to absorb measurement deviations, experimental errors, and operational disturbances unique to that condition. Operating condition mapping parameters are parameters with physical significance but dynamically change according to the characteristics of the equipment's operating conditions. These parameters can be mapped based on load, speed, temperature, pressure, etc. The parameter values ​​are adaptively adjusted according to the characteristics of the working conditions. For example, a load-related parameter function relationship can be constructed: KZ_eff=a0+a1·Fz, where a0 and a1 are working condition mapping parameters. It can be understood that by constructing a functional parameter sharing structure of p=f(x;a), the traditional fixed constant parameter configuration method is replaced, so that the key physical parameters of the model can dynamically change with the characteristics of continuous working conditions. This effectively makes up for the defect that fixed parameters cannot adapt to the dynamic changes of working conditions, and significantly improves the interpolation calculation ability and working condition generalization ability of the simulation model under unseen working conditions.

[0037] Step S13: Use the operating condition constant parameter as the constant parameter of the global parameter; Understandably, this step involves uniformly collecting the common operating condition constant parameters of all candidate operating condition models and using them as constant parameters of the global parameters of the simulation model. This ensures that all operating conditions share the same set of basic physical parameters, thereby guaranteeing the uniformity of the physical mechanisms of multiple operating conditions from the root.

[0038] Step S14: Obtain the function expression of the working condition mapping parameters, and use the coefficients of the function expression as the function parameters of the global parameters.

[0039] It is understandable that the working condition mapping parameters are used to accurately describe the continuous variation of the model's physical parameters with the working condition characteristic variables. This embodiment can adapt to various mapping function forms, including linear functions, polynomial functions, piecewise linear functions, lookup table functions, and two-dimensional graph functions. For example, a linear mapping can be expressed as p=a0+a1x, a second-order polynomial mapping can be expressed as p=a0+a1x+a2x², and a two-dimensional graph mapping can be expressed as p=f(x1,x2;a), where a is the corresponding interpolation coefficient or grid node value. The breakpoints of the piecewise linear mapping can be automatically generated based on the user-defined working condition characteristic range or the actual experimental working condition characteristic values. The two-dimensional graph mapping can be implemented using regular grid bilinear interpolation or irregular sampling point radial basis function interpolation. This embodiment replaces the traditional method of independent parameter tuning for each working condition by optimizing and unifying the mapping coefficients, enabling all working conditions to share a unified parameter variation law and physical structure, thus completely solving the problems of parameter fragmentation and inconsistent physical laws caused by parameter tuning for each working condition.

[0040] Step S15: Use the operating condition exclusive parameters as the local parameters.

[0041] Understandably, this step collects the unique operating condition parameters of each candidate operating condition model and uses them as local parameters of the simulation model to achieve independent correction of the disturbance errors specific to each operating condition without interfering with the calibration results of the global parameters.

[0042] As a specific implementation example, the calibration task includes three tire calibration conditions: vertical_load, lateral_load, and lateral_slip. Each condition is configured with independent experimental data, target signals, simulation operation settings, and a nominal vertical load characteristic Fz (Vertical Force). Specifically, the vertical_load condition corresponds to a nominal load Fz=2kN, and the target signal is the vertical force summary_fW_z; the lateral_load condition corresponds to a nominal load Fz=4kN, and the target signal is the lateral force summary_fW_y; the lateral_slip condition corresponds to a nominal load Fz=6kN, and the target signal is the lateral force summary_fW_y under the condition of lateral slip input. The three operating conditions share globally unique physical parameters such as CZ (Vertical Stiffness Coefficient), PCY1 (Pacejka Lateral Shape Coefficient 1), PDY1 (Pacejka Lateral Peak Coefficient 1), and PKY1 (Pacejka Lateral Stiffness Coefficient 1). Furthermore, this embodiment abandons the traditional calibration method of fixing the KZ (Vertical Stiffness Scaling Coefficient) parameter for each operating condition. Instead, it constructs a polynomial mapping relationship KZ_eff (Effective Vertical Stiffness Scaling Coefficient under the current operating condition) = a0 + a1·Fz + a2·Fz², which continuously varies with the vertical load. The mapping coefficients a0, a1, and a2 are uniformly solved through an outer optimizer, achieving adaptive dynamic updating of parameters according to the characteristics of the operating conditions.

[0043] In this embodiment, a multi-condition joint calibration architecture consisting of "constant parameters + function parameters + local parameters" is constructed by performing three-layer refined classification and hierarchical calibration of simulation model parameters. On the one hand, constant parameters solidify the inherent physical characteristics of the equipment, ensuring physical consistency across all operating conditions; function parameters establish a continuous mapping relationship between parameters and operating condition characteristics, improving the model's dynamic adaptability and generalization ability for unknown operating conditions. On the other hand, by independently absorbing the perturbation deviations specific to a single operating condition through local parameters, decoupling calibration between the global physical mechanism and local operating condition errors is achieved, effectively avoiding local errors from contaminating global parameters. This solves the technical defects of traditional calibration methods, such as parameter conflicts, poor physical interpretability, and weak cross-operating condition adaptability, significantly improving the accuracy, stability, and versatility of multi-condition simulation model parameter calibration.

[0044] Step S2: Generate candidate global parameter values ​​for the global parameters according to the preset initial generation strategy; It should be noted that the candidate global parameter values ​​are a complete set of candidate parameters to be evaluated output by the outer optimizer in a single evaluation process. These candidate points include at least constant parameters and function parameters, forming a complete parameter combination for subsequent multi-condition joint simulation evaluation and error solving. During the outer iterative optimization process, the optimizer can generate a single set of candidate points sequentially or in parallel according to iteration requirements. Each set of candidate global parameter values ​​is an independent evaluation object. Each set of candidate points must independently complete the simulation evaluation process for all conditions and output its corresponding error, which is then used by the inner and outer optimizers to perform parameter selection and iterative updates.

[0045] Additionally, it should be noted that the preset initial generation strategy is a parameter generation mechanism pre-configured based on the preset parameter value range, initial parameter values, and optimization algorithm rules. In this embodiment, the outer optimizer can generate a single set of candidate global parameter values ​​for serial calibration and evaluation according to the preset initial generation strategy. Alternatively, it can generate multiple sets of different candidate global parameter values ​​at once, and then perform subsequent simulation experiments under various working conditions in parallel using multi-threading and multi-task execution to achieve synchronous evaluation of multiple parameter schemes, effectively improving overall calibration efficiency.

[0046] It is worth noting that before generating candidate global parameter values ​​in step S2, this method also includes a model and data pre-check step. Specifically, this involves a comprehensive verification of the simulation model's validity, the completeness of the experimental data under operating conditions, the availability of the target signal, the boundaries of the parameters to be calibrated, the characteristics of each operating condition, the configuration relationships of the mapping function, and the dependencies and correlations of local parameters. The pre-check checks whether the task can be executed stably before formal optimization, avoiding the discovery of problems with data, model, or parameter configuration only halfway through the optimization process. Pre-testing may include: checking whether the physical simulation model can be loaded, instantiated, or compiled; checking whether the experimental data files for each working condition exist and whether the time series and target signal series are complete; checking whether the target signal can be read in the model simulation results; checking whether the parameters to be calibrated exist in the model and whether the parameter boundaries are reasonable, such as the lower limit being less than the upper limit; checking whether the working condition characteristics are complete, such as whether Fz is configured for each working condition that requires Fz; checking whether the mapping function can calculate effective parameters and does not produce obvious illegal values; checking whether the local parameter type is consistent with its position of action, such as time translation being only used for error calculation and input scaling triggering a re-simulation; optionally, a small-scale trial run is performed to confirm that the simulation backend, results directory, and log records are available; if the pre-test passes, the outer optimization is performed; if the pre-test fails, specific configuration problems are indicated and the formal optimization is not performed. Understandably, the pre-inspection mechanism can screen for problems such as abnormal model loading, missing data, out-of-bounds parameters, incorrect function configuration, and disordered parameter dependencies in advance, avoiding simulation errors, parameter failures, and abnormal results during subsequent parameter iteration and evaluation. This effectively ensures the stability and reliability of subsequent candidate parameter generation, simulation operation, and error evaluation processes, and improves the success rate and accuracy of the overall calibration process.

[0047] Step S3: Based on the candidate global parameter values ​​and the information on each working condition, determine the model for each experimental working condition; It should be noted that the experimental operating condition model is the final effective operating condition model that can be directly used for simulation calculations. It is an executable simulation model instance formed by loading the candidate global parameter values ​​of the current iteration on the basis of each candidate operating condition model, and is used for subsequent simulation experiments and error evaluation for each operating condition. This embodiment combines the candidate global parameter values ​​obtained from the outer layer optimization with the independent operating condition information and characteristics of each operating condition, assigns parameter values ​​and configures the model for each candidate operating condition model one by one, and then generates multiple sets of experimental operating condition models adapted to the operating characteristics of each operating condition, thus completing the model instantiation construction before simulation.

[0048] In one feasible implementation, the candidate global parameter values ​​include candidate constant parameter values ​​of the constant parameters and candidate function parameter values ​​of the function parameters. Step S3: Determining the experimental condition model based on the candidate global parameter values ​​and the various operating condition information may include steps S31-S32: Step S31: Traverse each candidate working condition model to determine the experimental parameter values, wherein the experimental parameter values ​​include working condition constant experimental parameter values ​​and working condition mapping experimental parameter values. For any candidate working condition model of the simulation model, when the working condition constant parameter exists in the candidate working condition model, determine the working condition constant experimental parameter value of the working condition constant parameter from the candidate constant parameter values. When the working condition mapping parameter exists in the candidate working condition model, substitute the candidate function parameter value into the function expression of the working condition mapping parameter to obtain the working condition mapping experimental parameter value of the working condition mapping parameter. It should be noted that the experimental parameter values ​​are the effective parameter values ​​actually written into the simulation model or working condition configuration and directly involved in the model state evolution and simulation calculation during the simulation experiments of each working condition. Specifically, these include the experimental parameter values ​​of the working condition constants and the experimental parameter values ​​of the working condition mapping. In this embodiment, during each parameter iteration evaluation, the system automatically traverses all working conditions included in the joint calibration object, reads the working condition feature information corresponding to each working condition, and uniformly adopts the globally shared candidate constant parameter values ​​as the experimental parameter values ​​for each working condition for the working condition constant parameters. For the working condition mapping parameters, the working condition features of the current working condition are substituted into the mapping function, and combined with the candidate function parameter values ​​output by the outer optimization layer, the real-time values ​​of the mapping parameters adapted to the specific working condition are obtained through analytical calculation. For example, for the load mapping parameter KZ_eff=a0+a1·Fz+a2·Fz², by substituting the nominal vertical load Fz of each working condition and the coefficients a0, a1, and a2 of the current iteration into the formula, the true effective parameter values ​​of KZ_eff under different load conditions can be obtained. Together with the global physical constant parameters, these parameters constitute the complete set of experimental parameters for the current working condition. It can be understood that the working condition mapping parameter is used to solve how to calibrate the parameters by sharing functional relationships rather than setting values ​​independently for each working condition when the parameters change with the characteristics of continuous working conditions such as load, speed, and temperature.

[0049] Step S32: Substitute the experimental parameter values ​​into each candidate working condition model to obtain each experimental working condition model.

[0050] Understandably, after traversing all working conditions and solving for the experimental parameter values ​​corresponding to each working condition, the experimental parameter values ​​of the working condition constants and the experimental parameter values ​​of the working condition mapping corresponding to each group of working conditions are uniformly written into the corresponding candidate working condition model to complete the model parameter configuration and instantiation, and finally obtain the experimental working condition models that can be directly used for simulation calculation.

[0051] In this embodiment, a hierarchical parameter assignment mechanism is used to assign uniform values ​​to the operating condition constant parameters across all operating conditions. This ensures that the inherent physical properties of the equipment remain consistent across all operating conditions, guaranteeing the uniformity and interpretability of the overall physical mechanism of the model. The operating condition mapping parameters are analyzed in real-time using a "operating condition feature + mapping function + function parameter value" approach. This allows different operating conditions to automatically generate specific parameter values ​​adapted to their current operating conditions based on their own operational characteristics, achieving continuous, smooth, and adaptive changes in model parameters with operating condition characteristics. Compared to traditional calibration methods that use fixed parameters for all operating conditions or random, independent parameters for each operating condition, this scheme retains a unified physical parameter structure shared by multiple operating conditions while also considering the parameter differences between different operating conditions. This effectively reflects the physical laws governing the continuous changes in real equipment parameters with operating condition loads, speeds, temperatures, and other conditions, significantly improving the realism and fit of multi-operating condition modeling. This provides a precise and reliable model foundation for subsequent high-precision multi-operating condition joint calibration and simulation evaluation.

[0052] Step S4: Conduct working condition simulation experiments based on the experimental working condition models, calibrate the local parameter values ​​of the local parameters, and obtain the experimental results for each working condition; It should be noted that the operating condition simulation experiment is a basic simulation experiment, which is the initial simulation operation performed under the premise that the experimental operating condition model is only configured with candidate global parameter values ​​and no operating condition-specific parameters are loaded. By performing basic simulations for each operating condition one by one, the output results of each operating condition simulation experiment can be obtained and the basic error can be calculated, providing unified benchmark data for subsequent local parameter calibration, error correction and operating condition result evaluation.

[0053] It is worth noting that during the execution of the operating condition simulation experiment, if abnormal situations occur such as model loading failure, parameter writing anomalies, simulation errors, simulation timeouts, missing result files, unreadable target signals, simulation results containing NaN (Not a Number) or infinity, or simulation duration not meeting the preset requirements, resulting in the inability to effectively calculate the operating condition error, then the current operating condition simulation experiment is determined to have failed. In this case, the aggregated error corresponding to the current candidate global parameter value is assigned a preset penalty value, and the current candidate global parameter value, failure condition, fault stage, and backend error information are recorded simultaneously. The penalty value is then sent back to the outer optimizer to achieve rapid identification and elimination of invalid parameter candidate points, avoiding unnecessary iterations that consume computing power.

[0054] Specifically, if a candidate point causes simulation failure, the system does not treat it as a feasible solution, but instead assigns it a very poor target value, i.e., a preset penalty value. The penalty value can be set to a value significantly larger than the normal error range; for example, when the normal study target value is usually between 0 and 10, the penalty value can be set to 1e6. Alternatively, it can be set to several times the current worst historical target value. The failure log should include: candidate point number and parameter value; the operating condition in which the failure occurred; the stage of the failure, such as model loading failure, parameter writing failure, simulation non-convergence, result file non-existent, or target signal reading failure; error messages returned by the simulation backend; the failure time; and the number of retries.

[0055] This processing method will cause the outer optimizer to consider the candidate point to be very poor, thus reducing the probability of continuing to search for similar unsimulatable regions. At the same time, failure logs are retained for later analysis to determine if parameter boundaries need to be tightened.

[0056] In a feasible implementation, step S4: conducting working condition simulation experiments based on each experimental working condition model, calibrating the local parameter values ​​of the local parameters, and obtaining the experimental results for each working condition may further include steps S41-S44: Step S41: Using the experimental working condition models, perform working condition simulation experiments to obtain simulation results for each working condition; It should be noted that the simulation results for each operating condition mainly include the basic simulation curve and basic error corresponding to each operating condition. The basic error is the bias error obtained by directly comparing the basic simulation curve with the experimentally measured curve. The operating condition simulation results can intuitively reflect the original fitting performance of the current candidate global parameter values ​​under each operating condition. On the one hand, it can provide reusable benchmark simulation data for error correction of post-processing local parameters, avoiding repeated simulation calculations; on the other hand, it can be used to quickly determine the simulability of the current candidate global parameter combination. If the operating condition simulation experiment fails, a penalty value is directly applied to the current candidate global parameter value and returned to the outer optimizer, terminating the subsequent calculation of the current branch. Simultaneously, the basic error can serve as the initial evaluation benchmark for the inner-layer optimization of local parameters. In operating conditions without configured local parameters, it is directly used as the final operating condition error. In operating conditions with configured local parameters, it can be used to compare the error improvement effect before and after inner-layer optimization, quantifying the optimization benefits of local correction.

[0057] Step S42: For each experimental working condition model containing a working condition-exclusive parameter, a local error objective function is constructed based on the working condition simulation results of the target working condition model and the target signal in the corresponding target working condition information of the target working condition model; under preset local constraints, the local error objective function is optimized using the preset inner layer optimizer of the target working condition model to obtain the working condition-exclusive experimental parameter values ​​of the working condition-exclusive parameters in the target working condition model.

[0058] It should be noted that the local error objective function refers to the objective function used in the local optimization process of the target working condition model to measure the deviation between the simulation result of the target working condition model under the current parameter values ​​and the expected local physical state or target signal of the working condition, i.e., the target signal in the target working condition information. Its core function is to serve as the optimization benchmark of the preset inner-layer optimizer of the target working condition model. By minimizing the local error through mathematical methods, the working condition-specific parameters can be precisely adjusted, thereby driving the target working condition model to reproduce the local response characteristics under the specific working condition as accurately as possible.

[0059] Additionally, it should be noted that the preset inner-layer optimizer is a dedicated optimization strategy for single-condition local deviation correction. It optimizes the condition-specific parameters and can employ lightweight optimization methods such as analytical estimation, coordinate search, low-budget local search, or black-box optimization with boundary constraints. Inner-layer optimization can use a fixed number of iterations, a maximum number of local evaluations, or an error improvement percentage threshold as termination conditions to ensure a balance between local optimization efficiency and accuracy. Under the premise of successful condition simulation experiments, this step keeps the current outer-layer global parameter values ​​unchanged and only performs local optimization on the condition-specific parameters to obtain the optimal condition-specific experimental parameter values. If the current candidate global parameter value causes simulation anomalies, a penalty value is directly assigned and a fault log is recorded to prevent the outer-layer optimization from converging to an invalid parameter range that cannot be simulated.

[0060] Additionally, it should be noted that the preset local constraints refer to the restrictive rules imposed on the solution range or iterative update path of the operating condition exclusive parameters when performing local error objective function optimization.

[0061] In one feasible implementation, the preset local constraint conditions include at least one of regularization constraints and amplitude boundary constraints.

[0062] It should be noted that regularization constraints add a penalty term to the objective function for calculating the error, penalizing local parameters for deviations from their initial values. The greater the deviation of the local parameters from their default values, the larger the penalty in the objective function, unless it can significantly improve the error. An example is a weight decay constraint strategy used to suppress overfitting of local parameters, which can be expressed by the constraint formula L_inner=L_signal+λ·Σ_j((θ_local,j-θ0,j) / s_j)^2, where θ0,j are the experimental parameter values ​​specific to the operating condition, s_j is the parameter normalization scale, λ is the regularization weight, and the amplitude boundary constraint is the parameter range... Interval constraint strategies are used to limit the fluctuation range of local parameters, such as limiting the maximum absolute value of time_shift, the maximum amplitude of signal_bias, and the deviation range of signal_scale relative to 1, to avoid unreasonable extreme values ​​of local parameters. For example, amplitude boundary constraints limit the maximum deviation range of local parameters, such as limiting time_shift to [-0.05s, 0.05s], signal_scale to [0.95, 1.05], and signal_bias to a certain proportion of the measurement range.

[0063] Understandably, preset constraints are used to prevent local parameters from excessively "swallowing up" global errors. Without constraints, local parameters might be finely tuned for each individual operating condition, but global parameters would lose their physical meaning.

[0064] In this embodiment, by applying regularization constraints and amplitude boundary constraints to the condition-specific parameters, the correction amplitude of local parameters can be effectively limited, avoiding overcompensation and excessive absorption of global physical errors and model mechanism errors by local parameters. These constraints mainly apply to condition-specific local parameters such as time translation, signal bias, signal scaling, input scaling, input bias, and initial state offset. The aim is to ensure that local parameters only absorb reasonable condition-specific disturbances, rather than replacing global physical parameters to complete the fitting. This embodiment retains the ability of local parameters to correct for single-condition measurement deviations and experimental disturbances, while preventing unconstrained fitting of local parameters from causing distortion of global physical parameters and degradation of the model's physical meaning. It effectively balances the fitting accuracy of single-condition fitting with the global physical consistency of multi-condition fitting, improving the stability and robustness of the overall calibration results.

[0065] As an example, the local error objective function uses the target signal of the target operating condition as the unit. It compares the simulated signal from the simulation experiment with the corresponding target signal to obtain the weighted root mean square error of each target signal, and aggregates them according to the target signal weights. Based on this, a normalized squared deviation penalty term for the local parameters relative to their initial values ​​is added. The candidate values ​​of the local parameters are simultaneously constrained by their preset lower and upper limits to avoid overcompensation of local parameters that could mask global physical parameter errors. It is worth noting that for simulation models containing post-processing-type operating condition-specific parameters, the simulation signal after time alignment and optional signal correction is compared with the corresponding target signal. Time alignment involves adjusting the sampling time of the simulation output signal according to the time-shifted local parameter Δt, and interpolating the adjusted simulation output signal based on the sampling time of the experimental data corresponding to the target signal to obtain a simulation signal sequence with the same time coordinate as the experimental data, so that the error can be calculated point-by-point. The current implementation uses one-dimensional linear interpolation; optional signal correction refers to: when the corresponding operating condition is configured with signal scaling parameters and / or signal bias parameters, correction is performed according to y_k correction(t) = s_k. y_k(t)+b_k performs amplitude scaling and / or bias correction on the k-th simulation output signal; when no corresponding parameters are configured, s_k=1 and b_k=0 are used, without changing the original simulation signal. This processing does not change the state trajectory of the simulation model, so existing simulation results can be reused.

[0066] The initial values ​​are nominal or default correction values ​​pre-configured in the local parameter template, and are also the center values ​​of the regularization penalty. They are not the condition-specific experimental parameter values ​​obtained after inner-layer optimization. For example, the default values ​​for time shift, signal bias, initial state offset, and input bias are usually 0, while the default values ​​for signal scaling and input scaling are usually 1. The starting point for inner-layer optimization can be estimated based on simulation signals and experimental data, but this estimated starting point does not change the preset initial values ​​used for regularization.

[0067] As an example, the local error objective function takes the target signal of the target operating condition as the unit. The target signal includes multiple signals. The simulated signal obtained from the simulation experiment is compared with the corresponding experimental data. For the k-th target signal in the target operating condition, the simulated signal corresponding to the target signal is compared, and the root mean square error of the k-th target signal is calculated.

[0068] In the above formula, This is the i-th simulated output value of the k-th target signal after time alignment and optional signal correction; This represents the i-th experimental data value corresponding to the k-th target signal; The total number of data sampling points (or data length) for k target signals; Let be the root mean square error of the k-th target signal. This formula is used to calculate the root mean square error of the k-th target signal in the target operating condition. Specifically, the simulated signal output from the simulation experiment is compared point by point with the corresponding experimental data of the target operating condition. The mean of the sum of squares of the deviations between the two is calculated and the square root is taken to obtain the overall time-series error of the target signal.

[0069] After obtaining the root mean square error of each target signal, the signals are weighted and aggregated according to their preset weights to obtain the signal error of the target operating condition:

[0070] In the above formula, The weighted signal error for the entire target operating condition; K is the total number of target signals included in the target operating condition; The preset weight for the k-th target signal is used to adjust the proportion of influence of each target signal in the aggregation of operating condition signal errors. This formula is used to perform weighted aggregation of the errors of all K target signals in the target operating condition. Since different target signals have different degrees of influence on the accuracy of the operating condition simulation, a preset weight Wk is set according to the importance of each target signal, and the root mean square error Ek of each signal is weighted and averaged to obtain the weighted signal error characterizing the overall signal deviation of the target operating condition. .

[0071] The objective function for local error is defined as:

[0072] and satisfy In the above formula, Let J be the objective function for the final local error, including regularization penalties; J is the total number of condition-specific parameters in the target working condition model. Let j be the exclusive parameter of the j-th working condition to be optimized in the target working condition model; This is the preset initial value (default value) corresponding to the exclusive parameter of the j-th working condition; The normalization scale is used for the j-th working condition exclusive parameter to eliminate the difference in magnitude of parameters of different magnitudes in the penalty term; This is the global regularization weight, used to adjust the overall strength of the regularization penalty term in the local error objective function; The regularization weight is the exclusive parameter for the j-th working condition, used for differentiated regularization strength control for different parameters; This is the lower limit constraint value for the allowable fluctuation of the exclusive parameter for the j-th operating condition; This represents the upper limit constraint value for the allowable fluctuation of the exclusive parameter under the j-th operating condition. This formula is the final local error objective function, derived from the aforementioned overall signal error term. Together with the regularization penalty term, it constitutes the system. The regularization term belongs to the weight decay constraint strategy, and its function is to add a penalty for local parameters deviating from their initial default values ​​when calculating local errors. If the local parameters... Deviation from initial value The larger the value, the greater the penalty for this term, unless the deviation actually significantly improves the signal error. Otherwise, the optimizer will tend to maintain the stability of local parameters.

[0073] It is worth noting that the inner local optimization process is only triggered when the working condition simulation experiment is successful and the corresponding experimental working condition model is configured with working condition exclusive parameters, so as to ensure that the hierarchical iteration logic of global parameters and local parameters is independent of each other and does not interfere with each other.

[0074] Step S43: Obtain the local parameter values ​​of the local parameters based on the exclusive experimental parameter values ​​of the operating conditions; As an example, the values ​​of the operating condition-specific experimental parameters are directly assigned to the corresponding local parameters to complete the configuration of the local parameter values.

[0075] It should be noted that the operating condition exclusive experimental parameter value is a unique parameter obtained through experimental calibration for a specific operating condition. Each operating condition exclusive experimental parameter value corresponds one-to-one with a local parameter. After filling the operating condition exclusive experimental parameter value into the corresponding local parameter, the local parameter value corresponding to that local parameter can be obtained directly.

[0076] Step S44: Based on the simulation results of each working condition and the local parameter values, obtain the experimental results of each working condition.

[0077] In a feasible implementation, step S44: obtaining the experimental results of each working condition based on the simulation results of each working condition and the local parameter values ​​may further include step S441: Step S441: Process the simulation results of each working condition using a preset result processing method to obtain the experimental results of each working condition. Specifically, for the first experimental working condition model containing post-processing-type working condition exclusive parameters, process the first working condition simulation results of the first experimental working condition model using a preset post-processing method to obtain the first working condition experimental results. The post-processing-type working condition exclusive parameters do not change the state trajectory of the simulation model. The preset post-processing method includes at least one of time shifting, signal biasing, and signal scaling. For the second experimental working condition model containing re-simulation-type working condition exclusive parameters, substitute the parameter values ​​of the re-simulation-type working condition exclusive parameters into the second experimental working condition model for re-simulation to obtain the second working condition experimental results. The re-simulation-type working condition exclusive parameters change the state trajectory of the simulation model. The re-simulation-type working condition exclusive parameters include at least one of initial state offset parameters, input scaling parameters, and input bias parameters.

[0078] It should be noted that the classification principle for local parameters can be based on whether they alter the state trajectory of the simulation model. This classification can be performed through preset types of local parameters: the time shift parameter `time_shift`, the signal bias parameter `signal_bias`, and the signal scaling parameter `signal_scale` are identified as post-processing condition-specific parameters, reusing the basic simulation results; the initial state offset parameter `initial_state_offset`, the input scaling parameter `input_scale`, and the input bias parameter `input_bias` are identified as re-simulation condition-specific parameters, writing the local parameter values ​​to the corresponding initial state or model input before re-executing the simulation. This type-driven classification method corresponds to the judgment principle that post-processing parameters do not change the state trajectory, while re-simulation parameters do.

[0079] Specifically, post-processing-type case-specific parameters only affect time-series data alignment, signal correction, and error calculation, without altering the model equations, input excitations, initial states, or system state evolution trajectories. These include the time alignment parameter `time_shift`, the signal bias parameter `signal_bias`, and the signal scaling parameter `signal_scale`. The corresponding preset post-processing methods are time shifting, signal superposition bias, and signal amplitude scaling, etc., requiring no rerun of the simulation. Re-simulation-type case-specific parameters directly affect the model's input excitations, initial states, internal parameters, and state evolution. These include the input scaling parameter `input_scale`, the input bias parameter `input_bias`, and the initial state offset parameter `initial_state_offset`. Optimized local parameter values ​​need to be written into the model configuration and the simulation re-executed to obtain accurate case output results and error values.

[0080] It is understandable that post-processing local parameters do not change the model's state trajectory, but only alter the way simulation results and experimental data are compared, thus allowing the reuse of basic simulation results. Common practices include: time shift (time_shift): without resimulating, simply shifting the simulation or experimental curve along the time axis, for example, comparing y_sim(t+Δt) with y_exp(t); signal bias (signal_bias): without resimulating, simply adding a constant to the simulation signal before error calculation, for example, comparing y_sim(t)+b with y_exp(t); signal scaling (signal_scale): without resimulating, simply multiplying the simulation signal by a scale before error calculation, for example, comparing s... The parameters y_sim(t) and y_exp(t) only affect how the curves are compared, and do not affect how the internal state of the model evolves. Therefore, the inner optimization only needs to repeatedly calculate the error on the existing basic simulation curve, which is relatively fast.

[0081] It's important to note that the preceding simulation experiments were run after only the outer candidate global parameters were written; at this point, the specific parameters that would alter the state trajectory had not yet been written. The re-run simulation occurs when, during inner-level optimization, a specific parameter is discovered that alters the model input, initial state, model equations, or state trajectory. This specific parameter is then written into the model or input configuration and run again. For example, `signal_bias` only measures the bias and does not change the model's motion, so a re-run is unnecessary. However, `input_scale` changes the input stimulus, and `initial_state_offset` changes the initial state; both will cause different model state trajectories, so a re-run simulation is necessary to calculate the accurate error.

[0082] As a specific example, refer to Figure 2 The physical simulation model is represented as F(x(t), dx / dt, u(t), theta) = 0, the model output is y_sim(t) = G(x(t), theta), and the error calculation function is J(y_sim, y_exp; theta_local). If the local parameter θ_local only participates in the calculation of the error function J and does not affect the model equation F, the output equation G, the input excitation u(t), the initial state x(t), and the physical parameter theta that affects the state evolution, then the following condition is met. x(t) / If theta_local=0, it is determined to be a post-processing parameter, and the basic simulation results can be directly reused; if the local parameter theta_local participates in model iteration and state evolution, causing the state trajectory to change with the parameter, it is determined to be a re-simulation parameter, triggering a re-simulation operation.

[0083] It is understandable that the operating condition-specific parameters are classified to address how to avoid unnecessary repetitive simulations when the operating condition-specific parameters only affect error calculation and not the state trajectory; and how to ensure accurate error assessment when the operating condition-specific parameters change the input, initial state, or dynamic trajectory of the model.

[0084] In this embodiment, based on the influence characteristics of condition-specific parameters on the state trajectory, a quantifiable automatic branch evaluation logic is constructed to achieve differentiated processing of post-processing parameters and re-simulation parameters. For post-processing local parameters that do not affect the model's state evolution, the completed basic simulation results are directly reused, and errors are quickly corrected through data post-processing, saving the overhead of repeated simulations, significantly reducing the computational power consumption and iteration time of multi-condition joint calibration, and improving local correction efficiency. For re-simulation local parameters that can change the model's state trajectory, the parameters are forcibly written and re-simulated to ensure that the model's state evolution and output results completely match the local parameter correction effect, avoiding evaluation bias introduced by ignoring the influence of parameters on the model's operation. This scheme balances calibration accuracy and computational efficiency, solving the dual defects of computational redundancy in the traditional unified re-simulation mode and insufficient accuracy in the unified post-processing mode, and significantly improving the efficiency and accuracy of multi-condition inner-layer local optimization.

[0085] It is worth noting that the local error objective function uses the target signal of the target operating condition as the unit. For the simulation model of post-processed local parameters, the simulation signal after time alignment and optional signal correction is compared with the corresponding target signal to obtain the weighted root mean square error of each target signal, and then aggregated according to the target signal weight. On this basis, a normalized squared deviation penalty term of the local parameters relative to the initial value is added. The candidate values ​​of the local parameters are simultaneously constrained by their preset lower and upper limits to avoid overcompensation of local parameters that would mask the global physical parameter error.

[0086] Step S5: Based on the experimental results of each working condition, generate new candidate global parameter values; It should be noted that the new candidate global parameter values ​​are a completely new combination of parameters obtained by the outer optimizer after optimizing and updating the original candidate global parameter values ​​before the iteration based on the fitting error of the experimental results of each working condition. These new parameters are used for simulation evaluation in the next iteration cycle.

[0087] In a feasible implementation, step S5, based on the experimental results of each operating condition, generates new candidate global parameter values, including steps S51-S52: Step S51: Obtain the error of each working condition from the experimental results of each working condition, and aggregate the errors of each working condition to obtain the aggregated error; It should be noted that the operating condition error is the comprehensive error value of the fitting deviation of all target signals under a single operating condition, used to quantify the overall fit between the simulation results and experimental measured data under a single operating condition. The preset aggregation algorithm adopts a multi-layer error fusion strategy of hierarchical aggregation, specifically divided into three layers of aggregation logic: signal level, operating condition level, and joint calibration task level. The aggregation error is the comprehensive error result obtained after multi-layer weighted statistical fusion, including single signal error, single operating condition aggregation error, and the total Study target value characterizing the overall fitting performance of the current candidate point.

[0088] The specific aggregation process is as follows: First, signal-level error aggregation is performed. For each target signal under each working condition, the single signal fitting error is calculated using either weighted root mean square error (weighted_rmse) or windowed weighted_rmse weighted for key time windows. Second, working condition-level error aggregation is performed. The signal errors corresponding to all target signals under the same working condition are fused and summarized to obtain the independent working condition error corresponding to that working condition. Finally, joint calibration task-level aggregation is performed. The working condition errors of all working conditions are further aggregated to obtain the total Study target value corresponding to the current candidate global parameter value, which is the final aggregated error.

[0089] Understandably, by constructing a three-tiered progressive error aggregation system at the signal, operating condition, and task levels, the traditional single overall error calculation method can be replaced, enabling refined, hierarchical, and comprehensive quantitative evaluation of fitting errors across multiple operating conditions. Through signal-level weighted error calculation, the fitting weights of key signals and key time windows can be significantly strengthened, ensuring the calibration accuracy of core performance indicators. Through operating condition-level error aggregation, the fitting quality of each operating condition can be independently quantified, accurately identifying abnormal and weak operating conditions.

[0090] In one feasible implementation, step S51: obtaining the operating condition errors of the experimental results for each operating condition, and aggregating the operating condition errors to obtain the aggregated error includes steps S511-S513: Step S511: Calculate the 90th percentile error and the maximum error of each operating condition; Specifically, the errors of each working condition are arranged in ascending order of error value, the 90th percentile error is calculated based on the arranged errors of each working condition, and the maximum value among the errors of each working condition is determined as the maximum working condition error. It should be noted that the 90th percentile error term is a value obtained by performing deterministic quantile statistics on all operating condition errors, indicating that approximately 90% of the operating condition errors are not greater than this value.

[0091] Step S512: Take the average value of the errors under each working condition as the average error; Specifically, based on the preset working condition weights corresponding to each working condition, the errors of each working condition are weighted and averaged to obtain the average error. Step S513: Assign average weight and high error weight to the average error and the high error respectively, and add them together to obtain the aggregated error.

[0092] Specifically, the joint calibration task-level aggregation can adopt the mean-plus-tail weighted aggregation method, specifically defined as L_study = L_mean + β·L_p90 + γ·L_max; where L_mean is the weighted average of all operating condition errors, L_p90 is the 90th percentile of all operating condition errors, L_max is the maximum value among all operating condition errors, and β and γ are configurable non-negative weighting coefficients. By simultaneously considering the average fit level, the 90th percentile error, and the worst-case error, the dilution of key high-error operating conditions when only the average value is used can be avoided.

[0093] In this embodiment, the study-level mean_plus_tail weighted aggregation mechanism is used to simultaneously consider the average fitting accuracy of all working conditions, the overall performance of the high-error working condition at the 90th percentile, and the fitting defects of the worst working condition. This effectively solves the problem that traditional mean aggregation is easily offset by the average of a large number of normal working conditions, thus masking the errors of key working conditions and local distortions.

[0094] Step S52: Optimize the candidate global parameter values ​​based on the aggregation error to obtain new candidate global parameter values.

[0095] It should be noted that this step feeds back the calculated aggregate error, i.e. the total objective value of Study, to the outer optimizer. The outer optimizer then iteratively optimizes and updates the current candidate global parameter values ​​based on the gradient of the overall error, generating a new round of candidate parameter combinations that are suitable for fitting multiple working conditions.

[0096] In this embodiment, an error aggregation system is used to replace the traditional single overall error calculation method, so as to achieve a refined, hierarchical, and comprehensive quantitative evaluation of the fitting error under multiple working conditions.

[0097] Step S6: Based on the new candidate global parameter values, return to the step of determining the experimental working condition model based on the candidate global parameter values ​​and the working condition information, until the preset calibration conditions are reached, and use the new candidate global parameter values ​​and local parameter values ​​obtained when the preset calibration conditions are reached as the target parameter values ​​of the global parameters and the local parameters, respectively.

[0098] It should be noted that this step is the closed-loop iterative control step of the entire calibration method. It is used to receive the new candidate global parameter values ​​output from step S5, forming a complete inner and outer nested iterative optimization closed loop. After each round of parameter update, the new candidate global parameter values ​​and new candidate function parameter values ​​are fed back to the front-end working condition modeling process. The entire process of experimental working condition model construction, working condition simulation experiment, local parameter calibration, multi-layer error aggregation, and global parameter update is executed cyclically, continuously iterating and optimizing the parameter combination performance. During the iteration process, it is continuously checked whether the preset calibration termination condition is met. If the termination condition is not met, iterative optimization continues in a loop. After the preset calibration condition is met and the iteration is determined to be converged, the iteration process is terminated, and the currently converged optimal new candidate global parameter values, new candidate function parameter values, and local parameter values ​​matched for each working condition are uniformly solidified as the target parameter values ​​of the final global parameters, function parameters, and local parameters of the simulation model, completing the entire multi-working condition layered parameter calibration.

[0099] It should be noted that the preset calibration conditions in this application include reaching the maximum number of iterations for evaluation, the overall target value converging to meet the preset accuracy threshold, and the error improvement amount for M consecutive iterations being lower than the preset accuracy ε. During the iteration process, the evaluation history, simulation failure records, and corresponding overall target values ​​for each set of candidate parameter points are retained throughout the entire process to ensure that the iteration process is traceable and the problem can be located. After the iteration convergence is completed, a complete calibration analysis report is generated only for the final optimal parameter solution, which simplifies the output content while retaining the data records of the entire process, thereby improving the readability and engineering applicability of the results.

[0100] It is understandable that steps S5 and S6 together form a complete multi-layered nested iterative optimization mechanism, which brings the following beneficial effects: This application, through the three-layer refined error aggregation and mean_plus_tail tail attention evaluation mechanism constructed in step S5, breaks through the limitations of traditional single mean error evaluation. It can simultaneously take into account the average fitting accuracy of conventional working conditions and the fitting defects of severe critical working conditions, providing a comprehensive, balanced, and engineering-practical optimization guide for outer parameter iteration, avoiding one-sided and distorted optimization direction. Through the closed-loop iterative mechanism of step S6, the global physical parameters, working condition mapping function parameters, and local correction parameters of each working condition are continuously updated, achieving a synergistic optimal solution for the global physical mechanism, the continuous change characteristics of working conditions, and the unique disturbance deviations of single working conditions. Simultaneously, relying on standardized iteration termination conditions, it ensures sufficient convergence of the overall fitting accuracy of multiple working conditions while effectively avoiding invalid iterations and saving computational power. The resulting hierarchical parameter system has clear physical meaning, strong consistency across operating conditions, and excellent generalization ability for unknown operating conditions. It completely solves the technical defects of traditional calibration methods, such as parameter conflicts across operating conditions, parameter distortion, poor physical interpretability, and weak generalization ability, and greatly improves the accuracy, stability, and engineering applicability of multi-operating condition calibration of industrial physical simulation models.

[0101] In one feasible implementation, after the calibration of global and local parameters is completed, a calibration report may also be output. The report includes global physical parameter values ​​that meet the convergence conditions, mapping parameter coefficients, local parameter values ​​for each working condition, effective parameter values ​​after analysis for each working condition, signal-level error, condition-level error, study-level error, worst-case or high-error tail condition, optimization history, failure statistics, comparison curves, structured report, and optimized physical model workpiece.

[0102] In one feasible implementation, the simulation model parameter calibration method also includes engineering steps such as dependency library pre-checking, pre-loading, model loading order stabilization, simulation settings retention, backend lock protection, and multi-instance isolation and parallelism to improve the stability of automated calibration.

[0103] Understandably, the above steps serve as auxiliary technical means in specific implementation, used to support the automated closed-loop execution of physical simulation models. In multi-instance isolated parallel mode, lightweight isolation strategies can be adopted based on the simulation backend file structure. For example, shared, symbolic links, or copy-on-write can be used for read-only library files and public model files, isolating only the working directory, cache directory, result directory, and session state, thereby reducing the I / O and memory overhead caused by large model copying. The effect of stabilizing stateful simulation backend execution is also significant. Through auxiliary steps such as pre-checking, loading order stabilization, simulation setting retention, and isolated execution, the repeatability of the automated calibration process and the stability of engineering delivery can be improved. Furthermore, industrial physical simulation backends are typically stateful. Loading models, opening libraries, writing parameters, running simulations, and reading results depend on the current session state, working directory, cache directory, and result directory. Simply executing multiple simulation evaluations in parallel during automatic calibration can easily lead to model state crosstalk, parameter writing failures, or result file contamination.

[0104] For example, to help understand the implementation process of the simulation model parameter calibration method, please refer to... Figure 3 , Figure 3 A simplified flowchart of a simulation model parameter calibration method is provided, specifically: First, the simulation model's operating condition information is obtained through a joint calibration object (a study containing multiple operating conditions). Next, the parameters to be calibrated are determined and categorized into global and local parameters. Then, the outer optimizer searches for candidate global parameter values ​​for the global parameters based on a preset initial generation strategy. Next, each operating condition is traversed to determine the experimental operating condition model, and operating condition simulation experiments are conducted to obtain the simulation results. The parameter values ​​of the operating condition-specific parameters are calibrated based on the simulation results. Then, the category of the operating condition-specific parameters in each experimental operating condition model is determined. If it is a post-processing type operating condition-specific parameter, the post-processing branch is entered, reusing the basic simulation results (i.e., the operating condition simulation results), and the basic simulation results are processed. Post-processing yields the simulation results for the operating condition. If the parameters are unique to the re-simulation type, the re-simulation branch is entered, and the unique parameters (local parameters) are written into the experimental operating condition model for re-simulation to obtain the simulation results for the operating condition. The operating condition errors in the simulation results are aggregated, including the study error of the entire study, the condition error of each condition in the study, and the error of the target signal in each condition, to obtain the aggregated error. Based on the aggregated error, it is determined whether the preset calibration conditions are met. If they are met, a report is output; otherwise, candidate global parameters are regenerated based on the aggregated error, and the above steps are iterated until the preset calibration conditions are met.

[0105] Furthermore, to aid in understanding the implementation process of the simulation model parameter calibration method, a specific embodiment is provided as follows: It should be noted that this embodiment uses a tire physical simulation model as an example to illustrate the complete operation flow of the present invention from input to output. This embodiment is only for illustrating specific implementation methods of this application and does not constitute a limitation on the application field, model type, or parameter type.

[0106] The calibration objects in this embodiment include three tire operating conditions: vertical_load, lateral_load, and lateral_slip. Each operating condition has its own experimental data, target signal, simulation settings, and nominal vertical load characteristic Fz. Specifically, the vertical_load condition corresponds to Fz=2kN, and the target signal is the vertical force summary_fW_z; the lateral_load condition corresponds to Fz=4kN, and the target signal is the lateral force summary_fW_y; the lateral_slip condition corresponds to Fz=6kN, and the target signal is the lateral force summary_fW_y with lateral slip input. The three working conditions share the same set of constant parameters of global parameters, such as CZ, PCY1, PDY1, and PKY1. In order to reflect the role of the working condition mapping parameters and the function parameters of global parameters, this embodiment does not treat KZ as an independent constant for each working condition, but establishes a function relationship that changes continuously with the load: KZ_eff=a0+a1·Fz+a2·Fz², where a0, a1, and a2 are the function coefficients uniformly searched by the outer optimizer, Fz is the nominal vertical load characteristic value of the current working condition, and Fz is the working condition mapping parameter value obtained according to the above function relationship.

[0107] Before formal optimization, the system performs a pre-check to verify the completeness and executableness of the model files or model configurations, experimental data files, target signal sequences, parameter boundaries, operating condition characteristics Fz, KZ mapping functions, local parameter definitions, and operating condition-level parameter path coverage relationships for the three operating conditions. If the pre-check fails, the system outputs a specific configuration error and does not proceed to formal optimization.

[0108] After the pre-detection is successful, the outer optimizer generates a set of candidate points and candidate global parameter values, including constant parameter values ​​and function parameter values, such as {CZ,PCY1,PDY1,PKY1,a0,a1,a2}. The system iterates through the three operating conditions, substitutes each Fz into the KZ_eff mapping function, obtains the effective KZ_eff for each of the three operating conditions, and writes it along with the shared global parameters into the corresponding candidate operating condition model or operating condition configuration to obtain the experimental operating condition model.

[0109] Subsequently, basic simulations, i.e., simulation experiments, are performed for each load case to obtain simulated curves of vertical or lateral forces, and the basic error is calculated by comparing them with the respective experimental curves. For the vertical_load load case without configured local parameters, the basic error can be directly used as the error for that load case. For the lateral_load load case, time_shift or signal_bias can be configured as post-processing load case-specific parameters. The system reuses the basic simulation results for this load case while keeping the outer candidate points unchanged, recalculating the error through time shifting or signal biasing, without rerunning the simulation. For the lateral_slip load case, input_scale or initial_state_offset can be configured as re-simulation load case-specific parameters. Since these load case-specific parameters change the model input, initial state, or state evolution trajectory, the system writes the local parameter values ​​into the model input or initial state during inner-layer optimization, reruns the simulation, and then calculates the error for that load case.

[0110] The errors from the three operating conditions are first aggregated at the signal level into condition-level errors within their respective operating conditions. Then, at the study level, they are aggregated using methods such as weighted_mean, max, or mean_plus_tail to obtain the overall study objective value for the candidate point, resulting in the aggregated error. The outer optimizer continues to generate the next set of candidate points based on this overall objective value, repeating the process of "candidate point generation, operating condition feature analysis, basic simulation, branching by local parameter type, and multi-level aggregation" until the termination condition is met. See [link to relevant documentation]. Figure 4 , Figure 4 The diagram shows the simulation curves after calibration of the tire physical simulation model in this embodiment. As can be seen, the simulation curves calibrated by the simulation model parameter calibration method of this application basically match the expected curves, while the simulation curves before calibration differ significantly from the expected curves.

[0111] After optimization, the system outputs the final optimal global physical parameters, optimal KZ mapping coefficients, effective KZ_eff for each of the three working conditions, local parameter values ​​for each working condition, error matrix, worst working condition, comparison curve, optimization history, failure statistics, structured report, and optimized model file.

[0112] It should be noted that the above examples are only for understanding this application and do not constitute a limitation on the simulation model parameter calibration method of this application. Any simple transformations based on this technical concept are within the protection scope of this application.

[0113] This application provides an electronic device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the simulation model parameter calibration method in Embodiment 1 above.

[0114] The following is for reference. Figure 5 The diagram illustrates a structural schematic of an electronic device suitable for implementing embodiments of this application. The electronic devices in these embodiments may include, but are not limited to, mobile terminals such as mobile phones, laptops, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Descriptions), PMPs (Portable Media Players), in-vehicle terminals (e.g., in-vehicle navigation terminals), and fixed terminals such as digital TVs and desktop computers. Figure 5 The electronic device shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of this application.

[0115] like Figure 5 As shown, the electronic device may include a processing unit 1001 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory 1002 or a program loaded from a storage device 1003 into a random access memory 1004. The random access memory 1004 also stores various programs and data required for the operation of the electronic device. The processing unit 1001, the read-only memory 1002, and the random access memory 1004 are interconnected via a bus 1005. An input / output interface 1006 is also connected to the bus. Typically, the following systems can be connected to the input / output interface 1006: input devices 1007 including, for example, touchscreens, touchpads, keyboards, mice, image sensors, microphones, accelerometers, gyroscopes, etc.; output devices 1008 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 1003 including, for example, magnetic tapes, hard disks, etc.; and communication devices 1009. The communication device 1009 allows the electronic device to communicate wirelessly or wiredly with other devices to exchange data. Although the diagrams show electronic devices with various systems, it should be understood that it is not required to implement or have all of the systems shown. More or fewer systems may be implemented alternatively.

[0116] Specifically, according to the embodiments disclosed in this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments disclosed in this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device, or installed from storage device 1003, or installed from read-only memory 1002. When the computer program is executed by processing device 1001, it performs the functions defined in the methods of the embodiments disclosed in this application.

[0117] The electronic device provided in this application, employing the simulation model parameter calibration method described in the above embodiments, can solve the technical problems of poor parameter consistency and parameter distortion in existing simulation model calibration methods, making it difficult to meet the high-precision calibration requirements of simulation models under multiple operating conditions. Compared with the prior art, the beneficial effects of the electronic device provided in this application are the same as those of the simulation model parameter calibration method provided in the above embodiments, and other technical features of this electronic device are the same as those disclosed in the method of the previous embodiment, and will not be repeated here.

[0118] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.

[0119] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0120] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the simulation model parameter calibration method described above.

[0121] The computer program product provided in this application can solve the technical problems of poor parameter consistency and parameter distortion in existing simulation model calibration methods, which make it difficult to meet the high-precision calibration requirements of simulation models under multiple operating conditions. Compared with the prior art, the beneficial effects of the computer program product provided in this application are the same as those of the simulation model parameter calibration method provided in the above embodiments, and will not be repeated here.

[0122] The above description is only a part of the embodiments of this application and does not limit the patent scope of this application. All equivalent structural transformations made under the technical concept of this application and using the contents of the specification and drawings of this application, or direct / indirect applications in other related technical fields, are included in the patent protection scope of this application.

Claims

1. A method for calibrating simulation model parameters, characterized in that, The simulation model parameter calibration method includes: Obtain the operating condition information of the simulation model, and determine the calibration parameters of the simulation model based on the operating condition information, wherein the calibration parameters include global parameters and local parameters; Candidate global parameter values ​​are generated for the global parameters according to a preset initial generation strategy; Based on the candidate global parameter values ​​and the information on each working condition, the model for each experimental working condition is determined; Based on the experimental working condition models, working condition simulation experiments were conducted to calibrate the local parameter values ​​of the local parameters and obtain the experimental results for each working condition. Based on the experimental results of each working condition, new candidate global parameter values ​​are generated; Based on the new candidate global parameter values, the steps of determining the experimental working condition model based on the candidate global parameter values ​​and the working condition information are returned until the preset calibration conditions are reached. The new candidate global parameter values ​​and local parameter values ​​obtained when the preset calibration conditions are reached are respectively used as the target parameter values ​​of the global parameters and the local parameters.

2. The simulation model parameter calibration method as described in claim 1, characterized in that, The global parameters include constant parameters and function parameters. The step of determining the calibration parameters of the simulation model based on the information of each working condition includes: Based on the information of each working condition and the simulation model, determine each candidate working condition model of the simulation model; Obtain the operating condition constant parameters, operating condition mapping parameters, and operating condition exclusive parameters of each candidate operating condition model. For each candidate operating condition model, the operating condition constant parameters are constant parameters that can be shared with other candidate operating condition models. The operating condition mapping parameters are determined by a functional relationship shared with other candidate operating condition models. The operating condition exclusive parameters are parameters unique to the candidate operating condition model. The operating condition constant parameter is used as the constant parameter of the global parameter; Obtain the function expression of the working condition mapping parameters, and use the coefficients of the function expression as the function parameters of the global parameters; The operating condition exclusive parameters are used as the local parameters.

3. The simulation model parameter calibration method as described in claim 2, characterized in that, The candidate global parameter values ​​include candidate constant parameter values ​​of the constant parameters and candidate function parameter values ​​of the function parameters; The step of determining the experimental condition model based on the candidate global parameter values ​​and the information on each operating condition includes: The experimental parameter values ​​are determined by traversing each candidate working condition model. The experimental parameter values ​​include working condition constant experimental parameter values ​​and working condition mapping experimental parameter values. For any candidate working condition model of the simulation model, if the working condition constant parameter exists in the candidate working condition model, the working condition constant experimental parameter value of the working condition constant parameter is determined from the candidate constant parameter values. If the working condition mapping parameter exists in the candidate working condition model, the working condition mapping experimental parameter value is obtained by substituting the candidate function parameter value into the function expression of the working condition mapping parameter. The experimental parameter values ​​are substituted into each candidate working condition model to obtain each experimental working condition model.

4. The simulation model parameter calibration method as described in claim 1, characterized in that, The steps of conducting simulation experiments based on each experimental working condition model, calibrating the local parameter values ​​of the local parameters, and obtaining the experimental results for each working condition include: Using the aforementioned experimental working condition models, working condition simulation experiments were performed, and simulation results for each working condition were obtained. For each experimental working condition model containing a working condition-specific parameter, a local error objective function is constructed based on the working condition simulation results of the target working condition model and the target signal in the corresponding target working condition information of the target working condition model. Under preset local constraints, the local error objective function is optimized using the preset inner layer optimizer of the target working condition model to obtain the working condition-specific experimental parameter values ​​of the working condition-specific parameters in the target working condition model. Based on the exclusive experimental parameter values ​​under the aforementioned operating conditions, the local parameter values ​​are obtained. Based on the simulation results of each working condition and the local parameter values, the experimental results of each working condition are obtained.

5. The simulation model parameter calibration method as described in claim 4, characterized in that, The experimental results for each working condition include the experimental results for the first working condition and the experimental results for the second working condition. The step of obtaining the experimental results for each working condition based on the simulation results for each working condition and the local parameter values ​​includes: The simulation results of each working condition are processed by a preset result processing method to obtain the experimental results of each working condition. Specifically, for a first experimental working condition model containing post-processing-type working condition exclusive parameters, the simulation results of the first working condition model are processed by a preset post-processing method to obtain the first working condition experimental results. The post-processing-type working condition exclusive parameters do not change the state trajectory of the simulation model. The preset post-processing method includes at least one of time shifting, signal biasing, and signal scaling. For a second experimental working condition model containing re-simulation-type working condition exclusive parameters, the parameter values ​​of the re-simulation-type working condition exclusive parameters are substituted into the second experimental working condition model for re-simulation to obtain the second working condition experimental results. The re-simulation-type working condition exclusive parameters change the state trajectory of the simulation model. The re-simulation-type working condition exclusive parameters include at least one of initial state offset parameters, input scaling parameters, and input biasing parameters.

6. The simulation model parameter calibration method as described in claim 4, characterized in that, The preset local constraints include at least one of regularization constraints and amplitude boundary constraints.

7. The simulation model parameter calibration method as described in claim 1, characterized in that, The step of generating new candidate global parameter values ​​based on the experimental results of each working condition includes: Calculate the error between the experimental results of each working condition and the target signal in the information of each working condition to obtain the working condition error of each working condition experimental result, and aggregate the working condition errors to obtain the aggregated error; The candidate global parameter values ​​are optimized based on the aggregation error to obtain new candidate global parameter values.

8. The simulation model parameter calibration method as described in claim 7, characterized in that, The step of aggregating the errors of each operating condition to obtain the aggregated error includes: Calculate the 90th percentile error and the maximum error under each operating condition; The average value of the errors under each working condition is taken as the average error; The average error, the 90th percentile error, and the maximum operating error are each assigned a corresponding weight, and then weighted and summed to obtain the aggregated error.

9. An electronic device, characterized in that, The device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the simulation model parameter calibration method as described in any one of claims 1 to 8.

10. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the steps of the simulation model parameter calibration method as described in any one of claims 1 to 8.

Images