Integrated Test Method and System for Axis Current and Radiated Emission in Electric Drive Systems

Abstract

The application discloses a shaft current and radiation emission integrated test method and system for an electric drive system, and belongs to the technical field of testing. Through real road working conditions, the method and system realize the synchronous collection of multiple physical quantities of the electric drive system, and combine a high-fidelity digital twin model to perform dynamic correction and simulation analysis, thereby solving the problems of misalignment of data, missing correlation between internal shaft current and external radiation field, and incapability of quantifying the contribution of the shaft current in the traditional test. Therefore, the specific influence of the shaft current in the total radiation emission of the electric drive system can be accurately identified, a quantitative basis is provided for the diagnosis and optimization of electromagnetic compatibility problems, and the efficiency and accuracy of the EMC test and rectification of the electric drive system of the new energy vehicle are improved.

Description

Technical Field

[0001] This application relates to the field of testing technology, and in particular to an integrated testing method and system for shaft current and radiated emission of electric drive systems. Background Technology

[0002] In electromagnetic compatibility (EMC) testing of electric drive systems for new energy vehicles, shaft current and radiated emissions are two key indicators, and existing standard testing methods have significant shortcomings. Test items are independent of each other; shaft current measurement and radiated emission testing are typically conducted on separate platforms and under separate operating conditions, resulting in data that is not aligned in time and under different conditions. This fragmentation makes it impossible to directly establish a quantitative correlation between the internal shaft current and the external radiated field under real-world operating conditions, leading to discrepancies between test results and the actual electromagnetic state of the vehicle. Furthermore, the testing process cannot quantify the contribution of interference sources; when system radiation exceeds limits, existing methods lack effective means to calculate the specific proportion of shaft current involved. Engineers are forced to rely on experience for trial and error rectification, which is inefficient and costly.

[0003] Therefore, the relevant technologies are insufficient for integrated testing and quantitative analysis of shaft current and radiated emissions of electric drive systems under realistic dynamic operating conditions. Summary of the Invention

[0004] This application provides a method and system for integrated testing of shaft current and radiated emission in an electric drive system, the technical solution of which is as follows: On the one hand, an integrated testing method for shaft current and radiated emission of an electric drive system is provided, the method comprising: A dynamic load excitation based on real road conditions is applied to the electric drive system under test, and multi-physical quantity synchronous acquisition is performed to obtain a multi-physical field synchronous data stream. The multi-physical field synchronous data stream includes time-aligned shaft current data, measured near-field radiation data, measured far-field radiation data, vibration data, and temperature data. Based on the shaft current data, vibration data, and temperature data in the multiphysics synchronous data stream, the multiphysics digital twin model of the electric drive system is driven to run, generating predicted near-field radiation data and predicted far-field radiation data synchronized with the physical test. The multiphysics digital twin model of the electric drive system includes a complete coupling link from shaft current conduction to space radiation. A time-frequency domain comparative analysis is performed on the predicted near-field radiation data, the predicted far-field radiation data, and the corresponding measured near-field radiation data and measured far-field radiation data in the multi-physics synchronous data stream to obtain a deviation spectrum. Based on the deviation spectrum, the critical path impedance parameters in the multi-physics digital twin model of the electric drive system are dynamically corrected to obtain a high-fidelity digital twin model consistent with the current test state. Based on the high-fidelity digital twin model, benchmark simulation results and comparative simulation results are obtained by performing benchmark simulations containing shaft current paths and comparative simulations with shielded shaft current paths in the model. Frequency domain difference calculations are performed on the baseline simulation results and the comparison simulation results to obtain the spectrum of the contribution of the shaft current to the total radiated emission. Attached Figure Description

[0005] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0006] Figure 1 This is a schematic diagram of the implementation environment of an integrated test method for shaft current and radiated emission of an electric drive system provided in an embodiment of this application; Figure 2 This is a flowchart of an integrated test method for shaft current and radiated emission of an electric drive system provided in an embodiment of this application; Figure 3 This is a flowchart of another integrated test method for shaft current and radiated emission of an electric drive system provided in an embodiment of this application; Figure 4 This is a flowchart of another integrated test method for shaft current and radiated emission of an electric drive system provided in the embodiments of this application; Figure 5 This is a flowchart of another integrated test method for shaft current and radiated emission of an electric drive system provided in the embodiments of this application; Figure 6 This is a flowchart of another integrated test method for shaft current and radiated emission of an electric drive system provided in the embodiments of this application; Figure 7 This is a flowchart of another integrated test method for shaft current and radiated emission of an electric drive system provided in the embodiments of this application; Figure 8 This is a flowchart of another integrated test method for shaft current and radiated emission of an electric drive system provided in the embodiments of this application; Figure 9 This is a schematic diagram of the structure of an integrated shaft current and radiated emission testing system for an electric drive system provided in an embodiment of this application; Figure 10 This is a schematic diagram of the structure of a system provided in an embodiment of this application. Detailed Implementation

[0007] To make the objectives, technical solutions, and advantages of this application clearer, the embodiments of this application will be described in further detail below with reference to the accompanying drawings.

[0008] In this application, the terms "first," "second," etc., are used to distinguish identical or similar items with essentially the same function. It should be understood that there is no logical or temporal dependency between "first," "second," and "nth," nor are there any restrictions on quantity or execution order.

[0009] It should be noted that the information (including but not limited to user device information, user personal information, etc.), data (including but not limited to data used for analysis, data stored, data displayed, etc.) and signals involved in this application are all authorized by the user or fully authorized by all parties, and the collection, use and processing of related data must comply with the relevant laws, regulations and standards of the relevant countries and regions.

[0010] Figure 1 This is a schematic diagram illustrating the implementation environment of an integrated testing method for shaft current and radiated emission in an electric drive system, as provided in an embodiment of this application. (See attached diagram.) Figure 1 The implementation environment may include node 110 and system 140.

[0011] Node 110 is connected to system 140 via a wireless or wired network. Optionally, node 110 can be a smartphone, tablet, laptop, desktop computer, etc., but is not limited to these. Node 110 has an application installed and running that supports integrated testing of shaft current and radiated emissions for electric drive systems.

[0012] System 140 is a standalone physical server, a server cluster or distributed system consisting of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, Content Delivery Network (CDN), and big data and artificial intelligence platforms. System 140 can provide background services for applications running on node 110.

[0013] In electromagnetic compatibility (EMC) testing of electric drive systems for new energy vehicles, shaft current and radiated emissions are key indicators. However, existing testing methods have limitations such as independent testing items and data that cannot be aligned in time and state. This makes it difficult to establish a quantitative correlation between internal shaft current and external radiated field under real operating conditions, and it is also impossible to effectively quantify the specific contribution of shaft current to total radiated emissions, resulting in low efficiency in fault diagnosis and rectification.

[0014] To address this issue, this application proposes an integrated testing method for shaft current and radiated emission of electric drive systems. This method applies a dynamic load excitation based on real road conditions to the electric drive system under test and performs synchronous acquisition of multiple physical quantities to obtain a time-aligned multi-physics synchronous data stream. Based on this data stream, a multi-physics digital twin model of the electric drive system is run to generate predicted radiation data. By comparing the predicted data with the measured data, the model parameters are dynamically adjusted to obtain a high-fidelity digital twin model. Based on this high-fidelity model, a benchmark simulation including the shaft current path and a comparative simulation of the shielded shaft current path are performed, and frequency domain differential calculations are conducted to obtain the spectrum of the contribution of shaft current to total radiated emission.

[0015] For ease of understanding, the following explains some key terms in this embodiment: Electric drive system: usually refers to the powertrain that drives the vehicle in new energy vehicles. Its core components include motors, inverters, reducers, etc., and it is the main source of shaft current and radiated emissions.

[0016] Shaft current data: refers to the high-frequency current flowing through the motor shaft, bearings, and housing grounding circuit during the operation of the electric drive system. This current is one of the important sources of electromagnetic interference.

[0017] Radiated emission data includes near-field and far-field radiation data. Near-field radiation data refers to the electromagnetic field distribution measured in the space near the electric drive system (typically within the wavelength range), reflecting the local characteristics of the interference source. Far-field radiation data refers to the electromagnetic field strength measured at a standard test distance (typically several meters), used to assess the system's electromagnetic interference level to the external environment.

[0018] Multiphysics synchronous data stream: refers to the synchronous acquisition of various physical quantities of an electric drive system under dynamic operating conditions, such as shaft current, near-field radiation, far-field radiation, vibration, and temperature, under a unified time reference. These data are strictly aligned in time and can reflect the multiphysics coupling state of the system under specific operating conditions.

[0019] Multiphysics Digital Twin Model of Electric Drive System: This refers to a virtual model that accurately reflects the multiphysics coupling behavior of a physical electric drive system in electromagnetic, mechanical, and thermal fields. The model includes a complete coupling link from shaft current conduction to spatial radiation and can predict the system's multiphysics response in real time based on input excitation and boundary conditions.

[0020] Deviation spectrum: This refers to the error distribution spectrum obtained by comparing radiation data predicted by a digital twin model with radiation data measured by physical tests in the time and frequency domain. This spectrum is used to quantify the differences between the model and the actual system and to guide the correction of model parameters.

[0021] A high-fidelity digital twin model refers to a digital twin model whose predicted results, after dynamic correction, are highly consistent with the actual physical test results in the time and frequency domain. This model can accurately reflect the multi-physics coupling characteristics of the electric drive system under the current test conditions.

[0022] Benchmark simulation results: These refer to the simulation output obtained by simulating the operation of an electric drive system in a digital twin model under a complete physical mechanism that includes shaft current paths. They are usually expressed as radiation spectra.

[0023] Comparison of simulation results: refers to the simulation output obtained in the digital twin model by shielding or removing the shaft current path to simulate the operation of the electric drive system without the influence of shaft current, which is usually also expressed as a radiation spectrum.

[0024] Contribution spectrum: refers to the distribution of the quantitative contribution of shaft current to the total radiated emission of electric drive system at different frequencies, obtained by frequency domain difference calculation of the baseline simulation results and the comparison simulation results.

[0025] This embodiment provides an integrated testing method for shaft current and radiated emission in electric drive systems. (See also...) Figure 2 This includes the following steps.

[0026] 201. Apply dynamic load excitation based on real road conditions to the tested electric drive system and perform synchronous acquisition of multiple physical quantities to obtain a synchronous multi-physics data stream. This data stream includes time-aligned shaft current data, measured near-field radiation data, measured far-field radiation data, vibration data, and temperature data.

[0027] In one implementation, a fixed load or a simple step load can be applied to the electric drive system, and independent measuring devices can be used to collect shaft current, radiation, vibration, and temperature data, followed by attempts to time-align them through post-processing. However, this approach struggles to simulate the complex dynamic changes of real road conditions, and the independently collected data may have persistent temporal discrepancies. As another approach, a preset cyclic operating condition can be applied to the electric drive system, and multiple independent acquisition systems can be used for measurement, followed by coarse alignment using timestamps. While this method can simulate some dynamics, it may still suffer from insufficient synchronization accuracy due to differences in time references between different acquisition systems.

[0028] 202. Based on shaft current data, vibration data, and temperature data in the multiphysics synchronous data stream, the multiphysics digital twin model of the electric drive system is driven to run, generating predicted near-field radiation data and predicted far-field radiation data synchronized with physical testing. The multiphysics digital twin model of the electric drive system contains a complete coupling link from shaft current conduction to space radiation.

[0029] In one implementation, a simplified digital model can be constructed, considering only the single conduction path from shaft current to radiation, and driven using static boundary conditions. This model may fail to accurately reflect the complex electromagnetic, mechanical, and thermal coupling effects within the electric drive system, resulting in limited prediction accuracy. As an alternative, multiple independent single-physics models, such as electromagnetic, vibrational, and thermal models, can be constructed separately, and their outputs can be combined to generate predicted radiation data. However, the coupling relationships between these independent models are difficult to establish precisely, and real-time interactions between multiple physics fields cannot be achieved.

[0030] 203. Time-frequency domain comparative analysis was performed on the predicted near-field radiation data, predicted far-field radiation data and the corresponding measured near-field radiation data and measured far-field radiation data in the multi-physics synchronous data stream to obtain the deviation spectrum.

[0031] 204. Based on the deviation spectrum, the critical path impedance parameters in the multiphysics digital twin model of the electric drive system are dynamically corrected to obtain a high-fidelity digital twin model consistent with the current test state.

[0032] One implementation involves a simple time-domain waveform comparison between predicted and measured data, with model parameters manually adjusted based on the engineer's experience. This method relies on manual judgment, resulting in low correction efficiency and difficulty in achieving high accuracy. Another approach involves a direct comparison of the frequency domain amplitude spectra of predicted and measured data, with a one-time correction of the fixed impedance parameters in the model based on the amplitude differences. This method may fail to capture instantaneous changes in model parameters under dynamic conditions, and the correction process lacks consideration for multi-physics coupling effects.

[0033] 205. Based on a high-fidelity digital twin model, benchmark simulation results and comparative simulation results are obtained by performing a benchmark simulation containing the shaft current path and a comparative simulation with the shielded shaft current path in the model.

[0034] In one implementation, a simulation encompassing all physical effects can be performed only once in the model without explicitly masking the shaft current path. This approach fails to effectively isolate the contribution of the shaft current and is difficult to quantify. Another implementation involves manually removing the shaft current source from the model and then performing a simulation, using the results for comparison. However, this simple removal may not fully simulate the true physical state after the shaft current path is "masked" and may overlook other indirect coupling effects.

[0035] 206. Perform frequency domain difference calculations on the baseline simulation results and the comparison simulation results to obtain the spectrum of the contribution of the shaft current to the total radiated emission.

[0036] In one implementation, the benchmark simulation result and the comparison simulation result can be simply subtracted in the time domain to obtain a time domain difference signal. This method cannot directly provide frequency domain contribution information and may be affected by time domain noise and phase differences. As another implementation, the frequency domain amplitudes of the benchmark simulation result and the comparison simulation result can be directly subtracted without further calibration or ratio calculation. This method may not eliminate baseline errors inherent in the model simulation itself, leading to insufficient accuracy in the contribution quantification results.

[0037] This method achieves simultaneous acquisition of multiple physical quantities of the electric drive system under real road conditions, and combines this with a high-fidelity digital twin model for dynamic correction and simulation analysis. This solves the problems of data misalignment, lack of correlation between internal shaft current and external radiation field, and inability to quantify the contribution of shaft current in traditional testing. Therefore, it can accurately identify the specific impact of shaft current on the total radiated emissions of the electric drive system, providing a quantitative basis for the diagnosis and optimization of electromagnetic compatibility issues, thereby improving the efficiency and accuracy of EMC testing and rectification of new energy vehicle electric drive systems.

[0038] In some of the solutions mentioned above in this application, a dynamic load excitation based on real road conditions is applied to the tested electric drive system and multi-physical quantity synchronous acquisition is performed to obtain a multi-physical field synchronous data stream, which is used to obtain synchronous multi-physical field data under real operating conditions as the basis for subsequent digital twin models. However, in this process, due to the lack of specific implementation details, such as how to generate dynamic load excitation and how to ensure that multi-physical quantities are accurately aligned in time, data acquisition may be asynchronous or inaccurate, thereby affecting the fidelity of subsequent models and the accuracy of analysis results, and failing to effectively solve the problem of data fragmentation in the prior art.

[0039] To address this, this application further proposes applying dynamic load excitation based on real road conditions to the tested electric drive system and performing synchronous acquisition of multiple physical quantities to obtain a synchronous multi-physics data stream. See [link to relevant documentation]. Figure 3 Specifically, it includes: 301. On the dynamometer test bench, a dynamic load excitation is generated based on the comprehensive road cycle working condition file, and the dynamometer test bench is controlled to apply a dynamic torque load including acceleration, deceleration and load change to the tested electric drive system.

[0040] 302. While applying the dynamic torque load, the shaft current data is collected by a high-frequency current probe configured in the shaft grounding circuit of the electric drive system under test, the measured near-field radiation data is collected by a magnetic near-field probe array arranged at key positions in the housing of the electric drive system under test, the measured far-field radiation data is collected by a broadband antenna set at a standard test distance, and the vibration data and temperature data are collected by a vibration acceleration sensor and a temperature sensor installed on the bearing housing and housing of the electric drive system under test, respectively.

[0041] 303. Connect the output signals of the high-frequency current probe, the magnetic near-field probe array, the broadband antenna, the vibration acceleration sensor and the temperature sensor to the same multi-channel synchronous data acquisition system, and perform synchronous sampling and recording under the control of a unified time base clock to generate the multi-physics synchronous data stream.

[0042] In this process, a dynamic load excitation is generated on a dynamometer bench based on a comprehensive road cycle driving condition file. The dynamometer bench is then controlled to apply dynamic torque loads, including acceleration, deceleration, and sudden load changes, to the tested electric drive system. This aims to simulate various dynamic conditions that the electric drive system might encounter in actual vehicle operation, ensuring the authenticity and representativeness of the test data. A dynamometer bench is an experimental device capable of simulating vehicle driving resistance and applying controllable loads to the tested power system. The comprehensive road cycle driving condition file contains information such as vehicle speed, acceleration, and gradient under specific driving cycles (e.g., WLTP, NEDC), forming the basis for generating the dynamic load excitation. By controlling the dynamometer bench, these dynamic torque loads can be accurately reproduced, allowing the tested electric drive system to operate close to its actual operating condition. Specifically, this can be achieved in one of the following ways: One method involves using the real-time simulation function of the dynamometer bench, taking the vehicle operating parameters from the comprehensive road cycle driving condition file as input, and having the vehicle model inside the dynamometer bench calculate and apply the corresponding dynamic torque loads in real time. Another approach is to pre-process the comprehensive road cycle operating condition file offline to generate the torque-time curve that the electric drive system needs to withstand throughout the entire operating cycle. Then, this curve is used as the control target and input to the controller of the dynamometer test bench to accurately reproduce the dynamic torque load.

[0043] While applying the dynamic torque load, the shaft current data is collected using a high-frequency current probe configured in the shaft grounding circuit of the tested electric drive system. Shaft current is the current generated between the shaft and bearing within the electric drive system due to common-mode voltage and other factors, and is a significant factor leading to bearing damage and radiated emissions. A high-frequency current probe is a non-contact or contact sensor capable of accurately measuring high-frequency current signals. By configuring it in the shaft grounding circuit, the shaft current flowing through the circuit can be captured, thereby obtaining the shaft current characteristics of the electric drive system under dynamic operating conditions. Specifically, this can be achieved in either of the following ways: One method is to use a non-contact high-frequency current probe, such as a split-type current clamp or a Rogowski coil, wrapped around the grounding wire of the electric drive system shaft, and collect the shaft current signal through electromagnetic induction. Another method is to connect a low-impedance current shunt in series in the shaft grounding circuit and measure the voltage drop across the shunt using a high-bandwidth voltage probe, thereby indirectly calculating the shaft current.

[0044] The measured near-field radiation data is acquired by using a magnetic near-field probe array positioned at key locations on the housing of the tested electric drive system. Near-field radiation refers to the electromagnetic field in the region near the electromagnetic source, and its characteristics differ from far-field radiation, allowing it to more directly reflect the local electromagnetic activity of the radiation source. The magnetic near-field probe array is a sensor system composed of multiple small magnetic field probes, capable of measuring the magnetic field strength and direction at different spatial locations. Positioning it at key locations on the electric drive system housing allows for the acquisition of near-field radiation distribution across different areas of the system surface, aiding in the location of the radiation source. Specifically, this can be achieved using either of the following methods: One method involves using an array of multiple ring or dipole magnetic field probes, precisely placed at predetermined key points on the electric drive system housing surface using a robotic arm or fixed support, and simultaneously acquiring the output signals of each probe. Another method is to use a scanning near-field probe system, automatically controlling the probes to perform a gridded scan on the electric drive system housing surface, acquiring near-field magnetic field data point by point, and then synthesizing the near-field radiation distribution map using software.

[0045] The measured far-field radiation data is acquired by using a broadband antenna positioned at a standard test distance. Far-field radiation refers to the electromagnetic field at a sufficiently far distance from the electromagnetic wave source, where the electric and magnetic fields are perpendicular to each other and perpendicular to the direction of propagation. It is a key indicator for measuring the electromagnetic compatibility (EMC) of equipment. A broadband antenna is an antenna capable of effectively receiving electromagnetic waves over a wide frequency range. Setting it at a standard test distance (such as 3 meters, 10 meters, etc.) can simulate the EMC test environment and acquire far-field electromagnetic wave data radiated by the electric drive system. Specifically, this can be achieved in one of the following ways: One method is to use a broadband antenna such as a biconical antenna, log-periodic antenna, or horn antenna, placed in a semi-anechoic chamber or fully anechoic chamber that meets EMC test standards, maintaining the specified test distance and height from the electric drive system under test, and connected to a spectrum analyzer for data acquisition. Another method is to conduct the test in an open field, but it is necessary to ensure that the test environment meets the relevant EMC standards for background noise and reflection requirements, and to use a broadband antenna to acquire far-field radiation data.

[0046] Vibration and temperature data are collected by vibration acceleration sensors and temperature sensors installed on the bearing housing and casing of the tested electric drive system, respectively. Vibration and temperature are important physical quantities affecting the performance and radiation characteristics of the electric drive system. Vibration acceleration sensors measure the vibration acceleration of the mechanical structure, reflecting the mechanical operating state and potential resonance problems of the electric drive system. Temperature sensors measure local temperature, reflecting changes in the thermal state and material properties of the electric drive system. Installing them on key components such as the bearing housing and casing allows for the acquisition of mechanical and thermal environment data closely related to shaft current and radiative emission. Specifically, this can be achieved in either of the following ways: One method is to use piezoelectric acceleration sensors installed on the surface of the bearing housing and casing, acquiring vibration acceleration data by measuring their output voltage signal. Another method is to use temperature sensors such as thermocouples or thermistors, attaching or embedding them to the surface of the bearing housing and casing to monitor temperature changes in key components of the electric drive system in real time.

[0047] The output signals of the high-frequency current probe, the magnetic near-field probe array, the broadband antenna, the vibration accelerometer, and the temperature sensor are connected to the same multi-channel synchronous data acquisition system. Under the control of a unified time base clock, they are synchronously sampled and recorded to generate a multi-physics synchronous data stream, aiming to ensure precise temporal alignment of all physical quantity data. A multi-channel synchronous data acquisition system is a device capable of simultaneously receiving and processing signals from multiple sensors. It typically includes a high-precision analog-to-digital converter (ADC) and a unified time base clock. By connecting the output signals of all sensors to this system and synchronously sampling them under the strict control of a unified time base clock, time drift and delay between different sensors or acquisition devices can be eliminated, thereby generating a time-aligned multi-physics synchronous data stream. Specifically, this can be achieved using any of the following methods: One method is to employ a modular data acquisition system based on a PXI or PCIe bus architecture, which includes multiple synchronous acquisition modules and provides a unified clock signal through a main control module to ensure synchronous sampling of all channels. Another approach is to use a distributed data acquisition system, where each acquisition unit has a built-in high-precision clock and synchronizes its clock via Network Time Protocol (NTP) or GPS time synchronization to ensure that the data collected by each unit is consistent in global time.

[0048] Through the aforementioned technical solution, this application simulates the complex operating state of an electric drive system under real road conditions by generating and applying dynamic torque loads, including acceleration, deceleration, and load mutations, based on a comprehensive road cycle working condition file on a dynamometer rig. This makes the collected data more meaningful. Simultaneously, during the application of the dynamic torque load, multiple sensors, including high-frequency current probes, magnetic near-field probe arrays, broadband antennas, vibration acceleration sensors, and temperature sensors, are used to synchronously collect multiple physical quantities such as shaft current, near-field radiation, far-field radiation, vibration, and temperature. Furthermore, all sensor output signals are connected to the same multi-channel synchronous data acquisition system for synchronous sampling and recording under strict control of a unified time base clock. This ensures precise alignment of all physical field data in the time dimension. This integrated and synchronized data acquisition mechanism effectively solves the data fragmentation problem caused by independent test items and misalignment of data time and state in existing technologies. It provides high-fidelity and high-reliability basic data input for the subsequent construction and correction of multi-physics digital twin models of electric drive systems and the analysis of shaft current contribution, improving the accuracy of test results and the diagnostic capability for electromagnetic compatibility issues of electric drive systems.

[0049] In some of the embodiments described above in this application, a dynamic load excitation based on a comprehensive road cycle working condition file is proposed to control the dynamometer rig to apply dynamic torque load. However, in this process, due to the lack of accurate modeling of the actual vehicle operating behavior, the directly generated dynamic load excitation may not be able to truly reproduce the acceleration, deceleration and load change characteristics in the actual road working condition, resulting in a deviation between the test data and the actual vehicle dynamic response, affecting the accuracy of subsequent synchronous acquisition of multiple physical quantities and the reliability of integrated testing.

[0050] To address this, this application further proposes a method for generating dynamic load excitation based on a comprehensive road cycle operating condition file, and controlling a dynamometer rig to apply a dynamic torque load, including acceleration, deceleration, and load abrupt changes, to the tested electric drive system. The specific steps include: parsing the comprehensive road cycle operating condition file to extract the vehicle speed curve, road gradient curve, and load demand curve; performing inversion calculations using a vehicle dynamics model based on the vehicle speed curve, road gradient curve, and load demand curve to obtain the demand torque time-series curve driving the tested electric drive system; and generating a control command sequence for the dynamometer rig based on the demand torque time-series curve to drive the dynamometer rig to reproduce the dynamic torque load including acceleration, deceleration, and load abrupt changes.

[0051] The process involves parsing the comprehensive road cycle test data file to extract vehicle speed curves, road gradient curves, and load demand curves. The aim is to obtain key kinematic and dynamic parameters of the vehicle under specific driving scenarios from standardized or customized test datasets. For example, specialized data parsing software, such as scripts written in Python or MATLAB, can read and process test data files in formats like CSV, Excel, or ASAM MDF to extract the vehicle's speed changes throughout the cycle, the road gradient information, and the vehicle's load demands on the drive system. Another approach is to utilize the built-in parsing module of professional test bench control software to directly import and identify the data structure in the test data file, automatically separating and extracting the required vehicle speed curves, road gradient curves, and load demand curves, providing accurate input data for subsequent torque calculations.

[0052] Based on the vehicle speed curve, road gradient curve, and load demand curve, an inverse calculation is performed using a vehicle dynamics model to obtain the time-series curve of the required torque driving the tested electric drive system. The vehicle dynamics model is a mathematical representation describing the relationship between the vehicle's motion state and driving force and resistance, comprehensively considering factors such as vehicle mass, aerodynamic characteristics, tire rolling resistance, transmission efficiency, and road gradient. Inverse calculation refers to using known vehicle motion states (such as speed and gradient) and external load demands to deduce the torque required by the drive system at each moment. Specifically, a physical model based on Newton's second law can be constructed, using the vehicle speed curve, road gradient curve, and load demand curve as input. By solving the vehicle's motion equations, the driving torque required to overcome various resistances and achieve the target motion state can be calculated. Alternatively, data-driven or hybrid modeling methods can be used to learn and train on a large amount of real-world driving data to establish an intelligent model capable of predicting the required torque under specific operating conditions, thereby generating a high-precision time-series curve of the required torque.

[0053] Based on the demand torque time-series curve, a control command sequence is generated for the dynamometer test bench to drive it to reproduce the dynamic torque load, which includes acceleration, deceleration, and sudden load changes. The demand torque time-series curve represents the ideal value of the torque required to drive the tested electric drive system under real-world operating conditions. To enable the dynamometer test bench to accurately simulate this ideal value, it needs to be converted into a control command sequence that the dynamometer controller can recognize and execute. This control command sequence can be an analog signal (such as voltage or current) or a digital signal (such as data packets transmitted via CAN bus or EtherCAT protocol). For example, the demand torque time-series curve can be imported into the host computer software of the dynamometer control system. The software will automatically generate a series of time-synchronized torque setpoints or speed setpoints based on the dynamometer's dynamic response characteristics and control algorithms (such as PID control, model predictive control, etc.), and convert them into drive signals for the dynamometer actuators (such as motors or hydraulic loaders). Another approach is to pre-store the demand torque timing curve in the memory of the dynamometer controller. The controller reads and executes these instructions in real time during the test, and through closed-loop control, ensures that the torque output by the dynamometer can follow the demand torque timing curve with high precision, thereby accurately reproducing dynamic torque loads such as acceleration, deceleration, and sudden load changes.

[0054] Through the above technical solution, this application can accurately simulate the dynamic behavior of real vehicles to generate dynamic load excitation, ensuring that the testing process can realistically reproduce the complex changes in road conditions, thereby improving the accuracy of integrated testing. Specifically, by analyzing the comprehensive road cycle condition file and extracting the vehicle speed curve, road slope curve, and load demand curve, basic parameters are provided for subsequent accurate calculations, avoiding the problem of incomplete excitation generation due to missing or inaccurate data. On this basis, inversion calculations are performed through a vehicle dynamics model, fully considering the mechanical interactions in actual vehicle operation, such as inertia, air resistance, rolling resistance, and slope resistance, ensuring that the obtained demand torque time series curve can accurately reflect real dynamic characteristics such as acceleration, deceleration, and sudden load changes, effectively preventing errors that may be introduced by simple linear processing. Based on the high-precision demand torque time series curve, a control command sequence for the dynamometer rig is generated, enabling the dynamometer rig to reproduce these dynamic torque loads with high fidelity, thereby simulating an excitation highly consistent with real road conditions in a laboratory environment. This provides a reliable and authentic excitation source for subsequent synchronous acquisition of multiple physical quantities, ensuring that the acquired shaft current data, measured near-field radiation data, measured far-field radiation data, vibration data, and temperature data can truly reflect the response of the tested electric drive system under dynamic operating conditions. This provides high-quality input for the construction and correction of the multi-physics digital twin model of the electric drive system, greatly improving the authenticity and data consistency of the entire shaft current and radiated emission integrated testing method.

[0055] In some embodiments described above in this application, a multi-physics digital twin model of a drive electric system is proposed to generate predicted radiation data. However, ensuring that the model accurately represents the multi-physics coupling relationship and effectively inputs synchronous data to generate high-precision predicted data consistent with physical tests in real time presents a challenge. Specifically, under dynamic operating conditions, if the model cannot accurately capture the interaction between shaft current conduction, mechanical vibration, thermal distribution, and spatial radiation field, or if the input data is not synchronously driven by the model in a joint manner, the time reference of the predicted data and the measured data will be inconsistent, affecting the accuracy of subsequent comparative analysis and correction, thus failing to reliably support the quantification of shaft current contribution.

[0056] In this regard, this application further proposes to drive the multi-physics digital twin model of the electric drive system based on shaft current data, vibration data, and temperature data in the multi-physics synchronous data stream, generating predicted near-field radiation data and predicted far-field radiation data synchronized with physical testing. See [link to relevant documentation]. Figure 4 ,include: 401. Construct a multi-physics digital twin model of the electric drive system. The multi-physics digital twin model of the electric drive system is a parameterized multi-physics coupling model that can be used to represent the coupling relationship between shaft current conduction, mechanical vibration, heat distribution and spatial radiation field.

[0057] 402. The shaft current data, vibration data and temperature data in the multi-physics synchronous data stream are used as a set of joint excitation sources and dynamic boundary conditions that are synchronized in the time domain, and are input into the multi-physics digital twin model of the electric drive system.

[0058] 403. Synchronously drive the operation of the multi-physics digital twin model of the electric drive system, so that the multi-physics digital twin model of the electric drive system generates in real time the predicted near-field radiation data and predicted far-field radiation data with the same time reference as the multi-physics synchronous data stream under the drive of the joint excitation source and dynamic boundary conditions.

[0059] Specifically, in constructing the multiphysics digital twin model of this electric drive system, the model performs a high-fidelity mapping of the physical electric drive system in digital space, simulating its behavior and interactions under different physical fields (such as electromagnetic, mechanical, and thermal fields). This parameterized multiphysics coupling model means that it can not only simulate these physical fields but also reflect the changes in the physical system under different operating conditions or states by adjusting its internal parameters, and can capture the mutual influence between different physical fields. For example, the thermal effect generated by shaft current conduction affects the electromagnetic properties of materials, and mechanical vibration may change contact impedance; these are all manifestations of coupling relationships. In one implementation, finite element analysis (FEA) software, such as ANSYS or COMSOL, can be used to construct a detailed geometric model and define material properties, boundary conditions, and physical field equations. The multiphysics module can then be used to achieve the coupled solution of electromagnetic, thermal, and structural vibration physical fields. Key geometric dimensions, material parameters, and boundary conditions in the model can be parameterized for subsequent adjustments. In another implementation, a method combining an equivalent circuit model (ECM) and a lumped parameter model can also be used. For example, the shaft current conduction path is abstracted into an RC equivalent circuit, mechanical vibration is simplified into a vibration response function through modal analysis, heat distribution is represented by a thermal network model, and multi-physics coupling is achieved by establishing mathematical relationships between these models (such as lookup tables, empirical formulas, or neural networks), and these equivalent parameters are parameterized.

[0060] When the shaft current data, vibration data, and temperature data from the multiphysics synchronous data stream are input into the multiphysics digital twin model of the electric drive system as a set of time-synchronized joint excitation sources and dynamic boundary conditions, this step aims to inject real-world operational data collected during physical testing into the digital twin model in a unified and time-aligned manner, serving as external inputs to drive the model's operation. Here, "joint excitation source" refers to using the shaft current data as the excitation source for the model's internal electromagnetic field, while "dynamic boundary conditions" refer to using the vibration and temperature data as real-time inputs of changes in the model's structure and thermal boundaries. Emphasizing "time-domain synchronization" ensures that the model's internal state changes remain consistent with the actual changes in the physical system on the time axis during simulation, thereby improving the realism and accuracy of the simulation. In one implementation, a data interface layer can be designed to receive data streams from a multi-channel synchronous data acquisition system. Based on preset mapping rules, this layer converts shaft current data into instantaneous amplitudes of common-mode voltage or current sources within the model, vibration data into instantaneous displacement or acceleration boundary conditions of structural components, and temperature data into instantaneous temperature values ​​of the model's thermal boundaries. These converted data are packaged into data frames based on timestamps and transmitted to the digital twin model in real time via API or shared memory mechanisms. In another implementation, a message queue or data bus-based approach can be used. The physical testing system publishes the acquired shaft current, vibration, and temperature data, along with their timestamps, to a specific message queue. The digital twin model subscribes to these message queues, synchronizes and reassembles data based on timestamps, and then uses this data as real-time excitation and dynamic boundary conditions for the model through its internal input ports or variable assignment mechanisms.

[0061] When synchronously driving the multiphysics digital twin model of the electric drive system, enabling it to generate predicted near-field and far-field radiation data with the same time reference as the multiphysics synchronous data stream under the drive of the joint excitation source and the dynamic boundary conditions, the core of this step lies in achieving "real-time synchronization" and "time reference consistency" between the digital twin model and the physical test. "Synchronous driving" means that the simulation step size of the model and the sampling interval of the physical data acquisition are consistent or proportional, ensuring that the model can respond to the latest input data at every time point. "Real-time generation" emphasizes the timeliness of the model output, that is, after receiving input data, it can quickly complete the multiphysics coupling calculation and output the prediction results. "Same time reference" ensures that the predicted data and the measured data are completely aligned on the time axis, laying the foundation for subsequent comparative analysis. In one implementation, an event-driven simulation scheduling mechanism can be used. When a new joint excitation source and dynamic boundary condition data frame arrive, the digital twin model is triggered to execute a simulation step size calculation. The simulation scheduler manages the timing of data input, model computation, and result output, ensuring that the start and end times of each simulation step strictly correspond to the timestamps of the physical data. In another implementation, high-performance computing platforms and parallel computing techniques can be utilized. The digital twin model is decomposed into multiple sub-models (such as electromagnetic, structural, and thermal sub-models), and independent computing resources are allocated to each sub-model. Through a co-simulation platform, the computations of each sub-model are executed in parallel under the control of a unified clock signal, and coupling information is transmitted in real time through a data exchange interface. At each time step, the outputs of each sub-model are aggregated to generate predicted radiation data.

[0062] Through the above technical solutions, this application effectively solves the problems that, under dynamic operating conditions, the multiphysics digital twin model of an electric drive system cannot accurately represent the multiphysics coupling relationship, and the input data cannot effectively synchronize with the model to generate high-precision prediction data consistent with physical tests in real time. Specifically, by constructing a parameterized multiphysics coupling model that can represent the coupling relationship between shaft current conduction, mechanical vibration, heat distribution, and spatial radiation field, this application ensures that the model can accurately capture key physical interactions and avoids the distortion of coupling relationships caused by unclear model structure. At the same time, by inputting shaft current data, vibration data, and temperature data from the multiphysics synchronous data stream as a set of time-domain synchronized joint excitation sources and dynamic boundary conditions into the model, this solves the problem of inconsistent model response caused by scattered data input, making the model run more accurately under real dynamic conditions. On this basis, the synchronous driving of the model operation enables it to generate predicted near-field radiation data and predicted far-field radiation data with the same time base as the multiphysics synchronous data stream in real time under the drive of the joint excitation source and dynamic boundary conditions. This ensures the time alignment of the predicted data and the measured data, solves the problem of prediction asynchrony, and improves the synchronicity and accuracy of the prediction. Overall, this application effectively generates high-precision synchronous prediction data under dynamic operating conditions through this series of steps. This provides a reliable and accurate foundation for subsequent time-frequency domain comparison analysis of the prediction data and measured data, and for dynamically correcting the critical path impedance parameters in the multi-physics digital twin model of the electric drive system based on the deviation spectrum. This results in a high-fidelity digital twin model consistent with the current test state, and further enables more accurate quantification of the contribution of shaft current to total radiated emission.

[0063] In some of the embodiments described above in this application, a multi-physics digital twin model of an electric drive system is proposed to generate predicted radiation data. However, in this process, the model architecture may lack a complete coupling path from shaft current conduction to spatial radiation, and the multi-physics coupling relationship between electromagnetic response, mechanical vibration response and temperature distribution is not clearly defined, resulting in insufficient model fidelity and inability to accurately simulate the dynamic behavior of the real system.

[0064] To address this, this application further proposes a method for constructing a multiphysics digital twin model of the electric drive system. The method includes: constructing a model architecture containing a complete coupling path from shaft current conduction to spatial radiation. This complete coupling path includes at least a common-mode voltage source inside the motor, a distributed parameter network between the windings and the housing, a bearing conductive loop, and the motor housing structure. Within this model architecture, the multiphysics coupling relationship between the electromagnetic response of the complete coupling path, the mechanical vibration response of the motor housing structure, and the temperature distribution of the electric drive system is established. The complete coupling path and the multiphysics coupling relationship are then parameterized to form the parameterized multiphysics coupling model.

[0065] To address the aforementioned issues, this application proposes a method for constructing a multiphysics digital twin model of an electric drive system. A model architecture is constructed that includes a complete coupling path from shaft current conduction to spatial radiation. This model architecture aims to comprehensively cover the entire physical process from the generation of shaft current within the electric drive system to its radiation of electromagnetic waves into external space. Specifically, this complete coupling path includes at least the common-mode voltage source inside the motor, the distributed parameter network between the windings and the housing, the bearing conductive loop, and the motor housing structure. For example, the common-mode voltage source inside the motor, the distributed parameter network between the windings and the housing, and the bearing conductive loop can be represented using lumped parameter or distributed parameter circuit models. The motor housing structure and its external space can then be represented using a three-dimensional electromagnetic field simulation model, coupled through ports or boundary conditions, thereby achieving accurate modeling of the shaft current conduction path. Alternatively, a unified modeling can be performed using full three-dimensional multiphysics simulation software. The geometric model of the electric drive system can be directly established, and material properties and excitation sources can be defined. The software can then automatically solve for the electromagnetic field distribution, simulating the generation, conduction, and radiation of shaft current. By constructing such a complete coupling path, it is possible to ensure that the digital twin model fully captures every key link of the shaft current from its source to its radiation, avoiding deviations between the model's predictions and the actual situation due to missing physical paths, thereby improving the model's accuracy.

[0066] In this model architecture, the multi-physics coupling relationship is established between the electromagnetic response of the complete coupled path, the mechanical vibration response of the motor housing structure, and the temperature distribution of the electric drive system. This means that the model not only independently simulates the electromagnetic field, mechanical vibration field, and temperature field, but also establishes the mechanism of their mutual influence and interaction. For example, this can be achieved through a simulation strategy of bidirectional or unidirectional coupling. An iterative solver is set in the simulation software. In each time step, the electromagnetic field is calculated first, and then the electromagnetic loss is input as a heat source into the thermal field model to calculate the temperature distribution. Simultaneously, the electromagnetic force is input as an excitation into the structural mechanics model to calculate the vibration response. Then, the material properties or geometric changes caused by temperature and vibration are fed back to the electromagnetic field model for the next iteration. Alternatively, it can be described by establishing a set of coupled field equations, combining the electromagnetic field equations, structural dynamics equations, and heat conduction equations, and introducing cross terms to represent the interaction between different physical fields, and then solving them numerically. Given that electromagnetic, mechanical, and thermal effects are highly coupled in real electric drive systems under dynamic operating conditions, establishing these multi-physics coupling relationships enables digital twin models to accurately reflect the dynamic behavior of real systems, thereby significantly improving the model's fidelity and making it closer to actual operating conditions.

[0067] Furthermore, the complete coupling path and the multiphysics coupling relationship are parameterized to form a parameterized multiphysics coupling model. This means defining the key physical quantities describing the complete coupling path and the key coefficients in the multiphysics coupling relationship as adjustable parameters or variables. For example, in modeling software, these physical quantities and coefficients are set as variables instead of fixed values; for instance, the equivalent resistance and capacitance of the bearing oil film can be defined as parameters. In the electromagnetic field model, the conductivity of the material can be defined as a function of temperature. In the structural model, the stiffness of the contact interface can be defined as a function of vibration amplitude. Alternatively, this can be achieved by establishing a parameterized model template or script, using the script to control the simulation software, defining all adjustable parameters in the script, and allowing users or optimization algorithms to modify these parameters externally, then automatically updating the model and executing the simulation. Parameterized characterization is the foundation for realizing dynamic model correction and optimization iteration. By exposing key parameters as adjustable variables, the model can be calibrated and updated based on measured data, thereby improving its adaptability and accuracy under different operating conditions, forming a high-fidelity digital twin model, and providing a reliable foundation for subsequent simulation and analysis.

[0068] Through the above technical solutions, the multi-physics digital twin model of the electric drive system constructed in this application not only possesses a complete physical link from shaft current conduction to spatial radiation, ensuring comprehensive coverage of the shaft current radiation mechanism, but also, by establishing multi-physics coupling relationships between electromagnetic response, mechanical vibration response, and temperature distribution, enables the model to accurately capture the interactions between various physical fields under real operating conditions, greatly improving the model's dynamic simulation capability and fidelity. Furthermore, the parameterized characterization of the complete coupling path and multi-physics coupling relationships provides flexibility and operability for subsequent dynamic correction and optimization of the model based on measured data, thereby forming a high-fidelity digital twin model consistent with the current test state. This allows the model to more accurately predict the radiated emission behavior of the electric drive system, providing a solid foundation for the integrated testing and quantitative analysis of shaft current and radiated emission, effectively solving the problems of insufficient model fidelity and inability to accurately simulate the dynamic behavior of real systems in existing technologies.

[0069] In some of the embodiments described above in this application, a multi-physics coupling relationship is proposed to accurately simulate the interaction between electromagnetic response, mechanical vibration response and temperature distribution in the digital twin model of electric drive system. However, in its implementation, the specific dynamic coupling mechanism between electromagnetic response and mechanical vibration and temperature distribution is ignored, especially the dynamic influence of vibration deformation on structural parameters and electromagnetic parameters and the dynamic modulation effect of temperature change on the electromagnetic properties of materials. This results in the model being unable to accurately reflect the mutual modulation behavior of multi-physics under real working conditions.

[0070] To address this, this application further proposes establishing a multiphysics coupling relationship between the electromagnetic response of the complete coupling path, the mechanical vibration response of the motor housing structure, and the temperature distribution of the electric drive system within the model architecture. Specifically, this includes: establishing a first coupling relationship between the electromagnetic response and the mechanical vibration response, which represents the dynamic changes in structural and electromagnetic parameters within the complete coupling path caused by the vibration deformation of the motor housing structure; establishing a second coupling relationship between the electromagnetic response and the temperature distribution, which represents the dynamic changes in the electromagnetic properties of the material within the complete coupling path caused by changes in the temperature distribution; and constructing coupled field equations to describe the mutual modulation effects among the electromagnetic response, mechanical vibration response, and temperature distribution, thereby completing the establishment of the multiphysics coupling relationship.

[0071] The first coupling relationship aims to capture the dynamic correlation between the structural and electromagnetic parameters of the motor housing structure during dynamic operation, caused by geometric deformation or contact state changes due to vibration, and thus affecting the electromagnetic conduction path (such as bearing oil film thickness and grounding contact resistance). Specifically, this can be achieved by establishing a mapping function between the vibration displacement field and electromagnetic parameters. For example, finite element analysis (FEA) can be used to calculate the deformation of the motor housing under different vibration modes, and then these deformation data can be correlated with structural parameters such as bearing oil film thickness and grounding contact pressure. Furthermore, functional relationships between these structural parameters and electromagnetic parameters such as equivalent capacitance and equivalent resistance can be established using experimental data or theoretical models. Another approach is to use a data-driven method, simultaneously measuring the vibration response of the motor housing and parameters of key electromagnetic paths (such as bearing impedance) under different vibration states. Then, a predictive model can be trained using machine learning algorithms (such as neural networks or regression models), which can dynamically adjust the electromagnetic parameters based on real-time vibration data.

[0072] This second coupling relationship aims to describe how changes in temperature distribution within an electric drive system dynamically affect the electromagnetic properties of materials along the electromagnetic conduction path, such as resistivity, dielectric constant, and permeability, thereby altering the electromagnetic response. Specifically, this can be achieved by establishing a physical model relating the temperature field to the electromagnetic properties of the materials. For example, for conductive materials, resistivity typically increases with temperature, and this relationship can be described using linear or nonlinear temperature coefficient models. For dielectric materials, the dielectric constant may also change with temperature, and corresponding functional relationships can be established based on the material properties. These relationships can be directly embedded into the electromagnetic simulation model. Alternatively, this can be achieved through experimental measurements and lookup tables. Accurate measurements of the electromagnetic properties of key materials can be performed under different temperature conditions, and a temperature-electromagnetic property lookup table can be constructed. During the operation of the digital twin model, the corresponding material electromagnetic properties can be obtained through interpolation or table lookup based on the real-time temperature distribution.

[0073] The coupled field equations are a set of mathematical expressions describing the interaction and dynamic modulation mechanisms among electromagnetic response, mechanical vibration response, and temperature distribution. They integrate the interdependencies between different physical fields into a unified mathematical framework, enabling the model to simultaneously solve for and reflect the dynamic changes of these fields. Specifically, they can be constructed using a system of partial differential equations. For example, Maxwell's equations (describing the electromagnetic field), structural dynamics equations (describing mechanical vibration), and heat conduction equations (describing temperature distribution) can be coupled. In the coupling terms, the dynamic changes described by the first and second coupling relationships are introduced. For instance, in Maxwell's equations, material parameters are no longer constants but functions of vibration and temperature. In the structural dynamics equations, electromagnetic force can serve as the excitation source. In the heat conduction equations, Joule heat loss serves as the heat source. Another approach is to use the collaborative simulation function of multiphysics simulation software. By integrating solvers of different physical fields (such as electromagnetic field solvers, structural mechanics solvers, and thermal solvers) and defining their interfaces and data exchange rules, each solver iterates and calculates within each time step, and exchanges boundary conditions or material properties until convergence is achieved, thereby realizing the mutual modulation effect among the three.

[0074] Through the above technical solution, this application effectively solves the problem of neglecting the dynamic coupling mechanism between electromagnetic response and mechanical vibration and temperature distribution in the digital twin model of electric drive system. Specifically, by establishing a first coupling relationship, the model can capture the real-time influence of the vibration deformation of the motor housing structure on the structural and electromagnetic parameters in the electromagnetic conduction path, thereby more accurately simulating the modulation effect of vibration on shaft current conduction and radiation emission. Simultaneously, by establishing a second coupling relationship, the model can reflect the dynamic influence of temperature distribution changes on the electromagnetic properties of materials, enabling the model to accurately consider the modulation effect of temperature on electromagnetic performance. Based on this, coupled field equations are constructed, integrating the mutual modulation effects among electromagnetic response, mechanical vibration response, and temperature distribution into a unified mathematical framework, realizing a deeper level of dynamic interaction simulation between multiple physics fields. This allows the digital twin model to more realistically and precisely reflect the complex physical phenomena in the complete coupled link of shaft current conduction to spatial radiation under dynamic operating conditions in the electric drive system, improving the model's prediction accuracy and simulation capability for real physical behavior, and providing a more solid and accurate foundation for subsequent model correction and shaft current contribution analysis.

[0075] In some of the embodiments described above in this application, a parameterized characterization of the complete coupling path and the multiphysics coupling relationship is proposed to construct a parameterized multiphysics coupling model. However, in its implementation, the parameterized characterization is not specific enough, and the key transmission components and coupling coefficients are not clearly identified and set as adjustable variables, resulting in inaccurate adjustment of the model under dynamic conditions, which affects the accuracy of subsequent analysis.

[0076] To address this, this application further proposes a parameterized characterization of the complete coupling path and the multiphysics coupling relationship to form a parameterized multiphysics coupling model. Specifically, this includes: determining the equivalent circuit parameters of key conductive components in the complete coupling path, which at least include the bearing oil film and the housing grounding path; determining key coupling coefficients in the multiphysics coupling relationship, which at least include a first-type coefficient representing the modulation effect of vibration on contact impedance and a second-type coefficient representing the modulation effect of temperature on the electromagnetic properties of the material; and setting the equivalent circuit parameters and the key coupling coefficients as adjustable variables in the parameterized multiphysics coupling model to complete the construction of the parameterized multiphysics coupling model.

[0077] The process involves determining the equivalent circuit parameters of key conductive components in the complete coupling path. These key components, including at least the bearing oil film and the housing grounding path, are physical components that influence the current flow characteristics in the shaft current conduction path of the electric drive system. Their electrical characteristics are characterized using an equivalent circuit model. Determining these parameters is crucial for accurately simulating the shaft current conduction behavior in a digital twin model. The bearing oil film forms an insulating or semi-insulating medium between the inner and outer rings of the bearing. Its thickness and dielectric constant affect the equivalent capacitance and resistance of the bearing, thus influencing the shaft current conduction. The housing grounding path is the channel through which the shaft current flows from the motor housing to the ground or vehicle chassis. Its impedance characteristics directly determine the discharge efficiency and radiation characteristics of the shaft current. In one implementation, impedance spectrum analysis and dielectric constant measurements of an actual electric drive system can be performed to obtain electrical response data of the bearing oil film and housing grounding path at different frequencies. Then, circuit theory and optimization algorithms can be used to fit the corresponding equivalent resistance, capacitance, inductance, and other parameters. In another implementation, the equivalent circuit parameters of these components can be calculated using electromagnetic simulation tools such as finite element analysis (FEA) or boundary element method (BEM) based on physical parameters such as bearing geometry, oil film material properties, housing material and connection method. For example, the capacitance and leakage resistance of the bearing oil film, as well as the contact resistance and distributed inductance of the grounding path can be calculated.

[0078] The key coupling coefficients in this multiphysics coupling relationship are determined. These key coupling coefficients include at least a first-type coefficient representing the modulation effect of vibration on contact impedance and a second-type coefficient representing the modulation effect of temperature on the electromagnetic properties of the material. These coefficients quantify the degree of mutual influence between different physical fields. In electric drive systems, mechanical vibration and temperature changes dynamically alter the characteristics of the electromagnetic conduction path. These coefficients are used to capture these dynamic coupling effects in a digital twin model. The first-type coefficients describe the quantitative relationship between vibration amplitude, frequency, and contact impedance changes, as vibration causes dynamic changes in the pressure, contact area, and microstructure of the contact interface, thereby altering contact resistance and contact capacitance. The second-type coefficients describe the quantitative relationship between temperature changes and changes in the electromagnetic properties of the material, as increased temperature changes electromagnetic properties such as the resistivity of conductors, the dielectric constant of dielectric materials, and the permeability of magnetic materials. In one implementation, these coupling coefficients can be obtained by designing controlled multiphysics experiments, such as measuring contact impedance changes at different vibration intensities or measuring the resistivity and dielectric constant of materials at different temperatures, and then performing data regression analysis or curve fitting. In another approach, theoretical models or empirical formulas describing the effects of vibration on contact impedance and temperature on the electromagnetic properties of materials can be established based on materials science, contact mechanics, and electromagnetics theories, and the corresponding coupling coefficients can be extracted or derived from them.

[0079] Setting the equivalent circuit parameters and the key coupling coefficient as adjustable variables in the parameterized multiphysics coupling model to complete its construction means defining these parameters as externally modifiable variables, rather than fixed values, allowing them to be dynamically modified based on actual data during model optimization, calibration, or inversion. This endows the digital twin model with flexibility and adaptability, enabling it to better match the dynamic behavior of the actual system. In one implementation, when building the digital twin model, these parameters are defined as externally modifiable variables through software interfaces or configuration files. For example, in simulation software, these parameters are set as user-editable input fields or variables invoked via APIs. In another implementation, these parameters are used as input variables for the optimization algorithm. During model calibration or parameter inversion, the optimization algorithm automatically adjusts the values ​​of these variables based on the deviation between the model output and the measured data to minimize the deviation, thereby achieving dynamic model correction and high-fidelity model construction.

[0080] Through the above technical solutions, this application solves the problem of inaccurate identification and adjustment of key parameters in model construction by specifying the parameterization characterization process. By determining the equivalent circuit parameters of key conductive components in the complete coupling path, such as the bearing oil film and the housing grounding path, it ensures that the model can capture the core elements in the conductive path, as these components play a dominant role in shaft current conduction, avoiding model distortion due to parameter omission. By determining the key coupling coefficients in the multiphysics coupling relationship, including the first type coefficient representing the modulation effect of vibration on contact impedance and the second type coefficient representing the modulation effect of temperature on the electromagnetic properties of materials, the model can effectively reflect the interactive influence of vibration and temperature on electromagnetic response under dynamic operating conditions, solving the problem of unquantified coupling effects. Setting the equivalent circuit parameters and key coupling coefficients as adjustable variables gives the model flexibility in subsequent optimization, allowing for dynamic correction of parameters based on measured data, thereby constructing a high-fidelity parameterized multiphysics coupling model. This enables the multiphysics digital twin model of the electric drive system to more accurately reflect the dynamic behavior of the physical system, providing a solid foundation for subsequent prediction and correction, and improving the prediction accuracy and reliability of the model under real road conditions.

[0081] In some embodiments described above in this application, shaft current data, vibration data, and temperature data are proposed as a joint excitation source and dynamic boundary conditions input to a multiphysics digital twin model of an electric drive system to drive the model to generate predicted radiation data. However, in its implementation, due to the inaccurate synchronization or mapping of the data stream, the input excitation source and boundary conditions may be mismatched in the time domain, thereby affecting the real-time performance and accuracy of the model prediction results and increasing the deviation between the predicted radiation data and the measured data.

[0082] To address this, this application further proposes to input the shaft current data, vibration data, and temperature data from the multiphysics synchronous data stream as a set of time-domain synchronized joint excitation sources and dynamic boundary conditions into the multiphysics digital twin model of the electric drive system. This step includes: parsing the multiphysics synchronous data stream, extracting shaft current data segments, vibration data segments, and temperature data segments with the same timestamp, and combining them into a time-aligned joint data frame; mapping the shaft current data segment in the joint data frame to the time-varying amplitude value of a high-frequency common-mode voltage source in the multiphysics digital twin model of the electric drive system; mapping the vibration data segment and temperature data segment in the joint data frame to the instantaneous vibration state and local temperature state of the corresponding structural component in the multiphysics digital twin model of the electric drive system, respectively; and synchronously applying the mapped time-varying amplitude value, instantaneous vibration state, and local temperature state to the corresponding interface of the multiphysics digital twin model of the electric drive system to constitute the joint excitation source and dynamic boundary conditions.

[0083] Specifically, the analysis of this multiphysics synchronous data stream aims to identify and separate valid data of different physical quantities from the raw acquired data, ensuring strict consistency of these data in the time dimension. For example, by comparing the timestamps of the data stream and employing algorithms such as linear interpolation or nearest neighbor matching, the shaft current data, vibration data, and temperature data acquired by different sensors can be aligned on a unified time base, thereby extracting data segments occurring at the same moment or within the same time window. Furthermore, by using a pre-defined data frame structure and synchronization protocol, the synchronous encapsulation of multiphysics data can be achieved at the data acquisition end, allowing the subsequent analysis process to directly obtain time-aligned data segments. This approach effectively avoids time misalignment caused by data acquisition or transmission delays, providing a high-precision time-synchronized data foundation for subsequent model input.

[0084] After obtaining the time-aligned joint data frame, the shaft current data segment within the joint data frame is mapped to the time-varying amplitude of the high-frequency common-mode voltage source in the multiphysics digital twin model of the electric drive system. This mapping process transforms the shaft current signal measured in the physical world into an electromagnetic excitation source recognizable by the digital twin model. For example, based on a pre-established equivalent circuit model of the electric drive system, the measured shaft current data segment can be directly converted into the instantaneous amplitude of the high-frequency common-mode voltage source in the model by calculating the transfer function or impedance relationship between the shaft current and the common-mode voltage source. Another approach is to establish a nonlinear mapping relationship between the shaft current and the common-mode voltage source amplitude through experimental calibration or simulation analysis, and store it as a lookup table or fitting function. During runtime, the corresponding voltage source amplitude can be dynamically queried or calculated based on the shaft current data segment. This mapping ensures that the electromagnetic excitation source in the model accurately reflects the dynamic characteristics of the shaft current of the tested electric drive system under real operating conditions.

[0085] Simultaneously, the vibration data segment and temperature data segment in the joint data frame are mapped to the instantaneous vibration state and local temperature state of the corresponding structural component in the multiphysics digital twin model of the electric drive system, respectively. This step aims to transform the mechanical vibration and thermal distribution information collected in physical testing into dynamic boundary conditions for the digital twin model. For example, vibration data segments (such as acceleration, displacement, or velocity) can be directly applied to the mechanical boundaries of the corresponding bearing housing or housing component in the digital twin model as instantaneous displacement or force excitation to simulate the actual dynamic response of the structure. For temperature data segments, they can be used as thermal boundary conditions and applied to the surface or internal nodes of the corresponding component in the model to dynamically adjust the thermal properties of the material (such as electrical conductivity and magnetic permeability) or simulate thermal expansion effects. In addition, vibration data can be mapped to the excitation amplitude of specific modes in the model through modal analysis, or local temperature data can be extended to a broader temperature field distribution through a heat conduction model, thereby more comprehensively reflecting the dynamic behavior of the physical system.

[0086] The mapped time-varying amplitude, instantaneous vibration state, and local temperature state are synchronously applied to the corresponding interface of the multiphysics digital twin model of the electric drive system to form the joint excitation source and dynamic boundary conditions. This synchronous application ensures that the digital twin model receives excitations and boundary conditions that are completely synchronized and consistent with the physical test at each simulation time step. For example, a unified simulation scheduler can be designed that extracts the mapped values ​​at the current time step from the joint data frame at each time step and loads these values ​​into the electromagnetic field solver, structural mechanics solver, and thermal field solver simultaneously through the API interface provided by the model. Another approach is to use a co-simulation platform to integrate electromagnetic, mechanical, and thermal simulation modules and ensure that all excitation sources and boundary conditions are synchronously updated and applied at the same physical moment through shared memory or message passing mechanisms under unified clock control.

[0087] Through the above technical solution, this application effectively solves the problem of input excitation source and boundary condition mismatch caused by inaccurate data stream synchronization or mapping in the multi-physics digital twin model of a drive electric drive system. By accurately parsing and time-aligning the multi-physics synchronized data stream, the strict consistency of shaft current, vibration, and temperature data in the time dimension is ensured, thereby constructing a high-fidelity joint data frame. Furthermore, these time-aligned data segments are accurately mapped to the time-varying amplitude of the high-frequency common-mode voltage source, instantaneous vibration state, and local temperature state in the model, enabling the digital twin model to dynamically and realistically reflect the multi-physics coupling behavior of the tested electric drive system under complex operating conditions. By synchronously applying these mapped excitation sources and dynamic boundary conditions, the input delay or asynchronous problem is eliminated, greatly improving the real-time performance and accuracy of the digital twin model's prediction results. This ensures that the predicted near-field radiation data and predicted far-field radiation data generated by the model are highly consistent with the measured data, laying a solid foundation for subsequent model correction and shaft current contribution analysis.

[0088] In some of the embodiments described above in this application, a multi-physics digital twin model of an electric drive system is proposed to generate predicted near-field radiation data and predicted far-field radiation data based on a multi-physics synchronous data stream. However, in its implementation, without a strict timing synchronization control mechanism, the model calculation may not be precisely aligned with the time reference of the physical test data, resulting in the predicted data drifting or misaligning in the time domain, which in turn affects the accuracy and reliability of subsequent comparative analysis, model correction, and shaft current contribution calculation.

[0089] To address this, this application further proposes a method to synchronously drive the multiphysics digital twin model of the electric drive system, enabling the model to generate, in real time, predicted near-field radiation data and predicted far-field radiation data with the same time reference as the multiphysics synchronous data stream, under the drive of the joint excitation source and the dynamic boundary conditions. Specifically, the method includes: establishing a simulation timing control loop, the step of which is synchronized with the timestamp increment of the multiphysics synchronous data stream; at each step of the simulation timing control loop, obtaining the corresponding excitation and boundary condition values ​​from the joint excitation source and the dynamic boundary conditions; loading the corresponding excitation and boundary condition values ​​into the multiphysics digital twin model of the electric drive system, and performing a single-step multiphysics coupling calculation of the model at that moment; and outputting the predicted near-field magnetic field and far-field electric field values ​​obtained from the single-step multiphysics coupling calculation, matching the corresponding moment, as the predicted near-field radiation data and predicted far-field radiation data for that moment, respectively.

[0090] A simulation timing control loop is established, whose step size is synchronized with the timestamp increment of the multiphysics synchronous data stream. This aims to ensure that the digital twin model is strictly aligned with the physical test data in the time dimension, avoiding data drift or accumulated errors caused by inconsistent time steps. This can be achieved in several ways. For example, it can be implemented through software programming, setting a master clock in the simulation platform. The clock's tick is consistent with the timestamp increment of the physical data acquisition system (e.g., 1 millisecond, 1 microsecond), and the simulation system executes a calculation step each time the master clock ticks. Alternatively, it can be implemented through an event-driven mechanism, where the simulation model is triggered to execute a calculation step when a new timestamp of the multiphysics synchronous data stream is received, ensuring that the duration of this step matches the time interval between two consecutive timestamps.

[0091] At each step of the simulation timing control loop, the corresponding excitation and boundary condition values ​​are obtained from the joint excitation source and the dynamic boundary conditions. This step ensures that the digital twin model receives inputs that precisely correspond to the current state of the physical system at each calculation step. The joint excitation source and dynamic boundary conditions are key inputs driving the operation of the digital twin model, including shaft current data, vibration data, and temperature data. Real-time acquisition of these values ​​is crucial for synchronizing the model with the physical test. This can be achieved through a data interface or shared memory mechanism. At each step, the simulation program reads values ​​matching the current simulation time's timestamp from the stored or real-time transmitted joint excitation source and dynamic boundary condition data structure via a preset API or data channel. Alternatively, this can be achieved through a message queue-based or publish / subscribe communication mechanism. The physical data acquisition system acts as a publisher, publishing timestamped excitation and boundary condition values ​​to the message queue. The digital twin model, as a subscriber, consumes the corresponding data from the queue at each simulation step.

[0092] The excitation and boundary condition values ​​at the corresponding time point are loaded into the multiphysics digital twin model of the electric drive system. This performs a single-step multiphysics coupling calculation of the digital twin model at that time point; this is the core calculation step of the digital twin model. Using the real-time acquired excitation and boundary condition values ​​as input, the model performs multiphysics coupling calculations to simulate the response of the electric drive system under the current physical state. Single-step calculation means that the model completes a full physics solution within each time step, such as the interaction of electromagnetic fields, mechanical vibration fields, and temperature fields. This can be achieved in simulation software (such as COMSOL Multiphysics, ANSYS Maxwell / Mechanical / Fluent, etc.) through scripts or programming interfaces, dynamically assigning the acquired values ​​to the model's input parameters or boundary conditions, and then calling the solver to perform a transient or temporal coupling simulation calculation for one time step. Alternatively, it can be achieved through a customized simulation engine that integrates a multiphysics solution algorithm. At each time step, the engine receives external input excitations and boundary conditions, and then iteratively calculates the instantaneous state of each physical field according to the preset coupling equations and solution strategies until convergence.

[0093] The model outputs the predicted near-field magnetic field and far-field electric field values ​​obtained from the single-step multiphysics coupling calculation, matching the corresponding time moment. These values ​​serve as the predicted near-field radiation data and predicted far-field radiation data for that moment, respectively. This step is crucial for the digital twin model to output its prediction results. After completing the single-step calculation, the model generates the electromagnetic field distribution for the current moment. Magnetic and electric field values ​​at specific locations (such as the near-field probe position and the far-field antenna position) are extracted from this data as predicted data for direct comparison with physical test data. Ensuring that these predicted values ​​have the same time reference as the input data is a prerequisite for the accuracy of subsequent comparative analysis. Simulation software typically provides post-processing functions or data export interfaces. After each single-step calculation, through programming instructions or scripts, the instantaneous values ​​of the magnetic field and electric field strength at preset monitoring points (e.g., the virtual positions of the near-field probe array and the far-field antenna) are extracted from the simulation results and associated with the current timestamp. Alternatively, this can be achieved through a custom data processing module. After the simulation engine completes a single-step calculation, this module automatically accesses the model's output data structure, extracts the required near-field magnetic field and far-field electric field values ​​based on predefined spatial coordinates and physical quantity types, and encapsulates them into a data packet with a timestamp for subsequent processing.

[0094] Through the above technical solution, this application establishes a strict simulation timing control loop and keeps its step size synchronized with the timestamp increment of the multiphysics synchronous data stream. This ensures that the digital twin model is strictly aligned with the physical test data in the time dimension, effectively avoiding data drift or accumulated errors caused by inconsistent time steps. At each simulation step, the excitation and boundary condition values ​​at the corresponding moment are obtained in real time from the joint excitation source and dynamic boundary conditions, ensuring accurate matching between the model input and the current state of the physical system and avoiding data delay or misalignment. These real-time obtained excitation and boundary condition values ​​are loaded into the multiphysics digital twin model of the electric drive system, and single-step multiphysics coupling calculation is performed. This allows the model to perform accurate physical field response simulation based on the most accurate real-time input, generating prediction results that match the physical state at a specific moment. The predicted near-field magnetic field and far-field electric field values ​​matched with the corresponding moment are output as predicted near-field radiation data and predicted far-field radiation data, thus ensuring that the model prediction output is directly and accurately aligned with the measured data in time. This series of rigorous synchronization control mechanisms fundamentally solves the problem of time alignment between digital twin model prediction data and physical test data under dynamic operating conditions, improving the accuracy and reliability of model prediction. It provides a solid foundation for subsequent high-precision time-frequency domain comparative analysis, model correction, and shaft current contribution calculation, thereby improving the accuracy and reliability of the entire integrated testing method.

[0095] In some of the solutions described above in this application, a deviation spectrum is obtained by comparing the predicted near-field radiation data and predicted far-field radiation data with the measured near-field radiation data and measured far-field radiation data in the time and frequency domains. This deviation spectrum is then used to correct the multi-physics digital twin model of the electric drive system. However, in this process, the deviation spectrum may contain mixed errors from the coupling of multiple physical fields, and it is impossible to effectively separate the error components that are strongly correlated with the dynamic behavior of the mechanical structure. This results in inaccurate model correction and affects the reliability of subsequent shaft current contribution analysis.

[0096] To address this, this application further proposes a time-frequency domain comparative analysis of the predicted near-field radiation data, the predicted far-field radiation data, and the corresponding measured near-field radiation data and measured far-field radiation data in the multiphysics synchronous data stream, to obtain the deviation spectrum. (See [link to relevant documentation]). Figure 5 The process includes: 501. Perform time-domain cross-correlation calculation and frequency-domain complex transfer function calculation on the predicted near-field radiation data and the measured near-field radiation data, as well as the predicted far-field radiation data and the measured far-field radiation data, respectively, to obtain a set of time-domain correlation functions and a set of frequency-domain complex transfer functions.

[0097] 502. Using this set of time-domain correlation functions and this set of frequency-domain complex transfer functions, and introducing the spatial topological relationship between the measured near-field radiation data and the measured far-field radiation data, a composite error data structure is constructed that simultaneously represents the consistency of model prediction errors in the time domain, the amplitude and phase in the frequency domain, and the spatial distribution characteristics.

[0098] 503. Based on the energy of the characteristic frequency band in the vibration data of the multi-physics synchronous data stream, the error components strongly correlated with the dynamic behavior of the mechanical structure are separated from the composite error data structure to generate a deviation spectrum. This deviation spectrum is used to correct the structural dynamic coupling parameters in the multi-physics digital twin model of the electric drive system.

[0099] Specifically, time-domain cross-correlation calculations and frequency-domain complex transfer function calculations are performed on the predicted near-field radiation data, predicted far-field radiation data, and measured near-field radiation data, respectively. This aims to quantify the similarity between the predicted and measured data in terms of time series and the differences in frequency components. Time-domain cross-correlation calculations assess the similarity of two signals at different time delays by sliding one signal and calculating its sum with another signal. This can be efficiently implemented using algorithms based on the Fast Fourier Transform (FFT) or through direct convolution operations. Frequency-domain complex transfer function calculations reveal the system's response characteristics to signals at different frequencies, including amplitude gain and phase lag. This can be obtained, for example, by dividing the Fourier transform of the output signal by the Fourier transform of the input signal, or by estimating it in the frequency domain using system identification techniques. These calculation results form a set of time-domain correlation functions and a set of frequency-domain complex transfer functions, providing comprehensive foundational data for subsequent error analysis.

[0100] This paper utilizes a set of time-domain correlation functions and a set of frequency-domain complex transfer functions, and introduces the spatial topological relationship between measured near-field radiation data and measured far-field radiation data, to construct a composite error data structure that simultaneously represents the temporal consistency, frequency-domain amplitude and phase, and spatial distribution characteristics of model prediction errors. The aim is to integrate error information from different dimensions (time, frequency, and space) into a unified framework. The spatial topological relationship refers to the geometric arrangement and relative position information of the sensors (such as magnetic near-field probe arrays and broadband antennas) used to collect measured radiation data in space. For example, it can be a list containing the coordinates of each sensor, or a matrix describing the distance and direction between sensors. The composite error data structure is a multi-dimensional data carrier capable of simultaneously storing and representing the correlation of errors in the time domain, the amplitude and phase deviations in the frequency domain, and the spatial distribution characteristics. For example, it can be a high-dimensional tensor whose dimensions correspond to time, frequency, and spatial location, respectively, or a composite object containing multiple substructures (such as a time-domain error matrix, a frequency-domain error matrix, and a spatial weight matrix).

[0101] This method uses the energy of characteristic frequency bands in vibration data from a multi-physics synchronous data stream as a basis to separate error components strongly correlated with the dynamic behavior of the mechanical structure from a composite error data structure, generating a deviation spectrum. The aim is to locate and extract model prediction errors caused by mechanical vibration. The energy of the characteristic frequency bands in the vibration data refers to the concentrated vibration energy regions exhibited by the electric drive system at specific mechanical resonance frequencies or operating modes. These energy-concentrated frequency bands can be identified, for example, by performing spectral analysis (such as Fourier transform or wavelet transform) on the vibration data. Separating error components can be achieved in various ways. For instance, a frequency domain weighting factor can be generated based on the energy distribution of the vibration data within the characteristic frequency bands and applied to the frequency domain error components in the composite error data structure, thereby highlighting errors related to mechanical vibration. Alternatively, machine learning algorithms can be used to train and identify the correlation between vibration modes and radiation errors, thereby intelligently extracting mechanically related error components from the composite error data structure. The generated deviation spectrum is a spectrum specifically designed to reflect the radiation prediction errors caused by the dynamic behavior of the mechanical structure. It will serve as a direct basis for subsequently correcting the structural dynamic coupling parameters in the multi-physics digital twin model of the electric drive system.

[0102] Through the above technical solution, this application effectively solves the problem that existing deviation spectra cannot effectively separate mechanical structure-related errors. By performing comprehensive time-domain cross-correlation calculations and frequency-domain complex transfer function calculations on predicted and measured radiation data, the performance of model prediction errors in time series and frequency components can be quantified in detail, providing a solid data foundation for subsequent error analysis. By combining these time-frequency domain error information with the spatial location and topological relationships of measured radiation data, a composite error data structure is constructed, making the error representation more comprehensive and three-dimensional, avoiding the omission of error information that may be caused by single-dimensional analysis. More importantly, by introducing vibration data from multi-physics synchronous data streams and using the energy of their characteristic frequency bands as a basis, error components strongly correlated with the dynamic behavior of mechanical structures can be separated from the complex composite error data structure, thereby generating a deviation spectrum with clear physical meaning. This deviation spectrum is specifically used for the targeted correction of structural dynamic coupling parameters in the multi-physics digital twin model of electric drive systems, making the model correction process more targeted and efficient, and improving the accuracy and fidelity of the digital twin model under dynamic conditions. This model correction provides a more reliable basis for the subsequent quantitative analysis of the contribution of shaft current to total radiated emission, avoiding analytical biases caused by model inaccuracies, thereby improving the reliability and practicality of the entire testing method.

[0103] In some of the embodiments described above in this application, a deviation spectrum is obtained through time-frequency domain comparative analysis to correct the multi-physics digital twin model of the electric drive system. However, in its implementation, the spatial positional relationship between the measured near-field radiation data and the measured far-field radiation data is ignored, which causes the model prediction error to fail to accurately reflect the spatial distribution characteristics of the electromagnetic field, thereby affecting the accuracy of the model correction.

[0104] To address this, this application further proposes a composite error data structure that utilizes a set of time-domain correlation functions and a set of frequency-domain complex transfer functions, and introduces the spatial topological relationship between measured near-field and far-field radiation data, to simultaneously represent the model prediction error's consistency in the time domain, amplitude and phase in the frequency domain, and spatial distribution characteristics. Specifically, this method includes: The set of time-domain correlation functions and the set of frequency-domain complex transfer functions are organized into a matrix to form a time-frequency error characteristic matrix with the measurement position as the row and the time-domain delay unit and frequency-domain spectral line as the column.

[0105] Based on the spatial topological relationship between the measured near-field radiation data and the measured far-field radiation data, a spatial correlation weight matrix is ​​constructed to reflect the electromagnetic field coupling strength between different measurement locations.

[0106] The spatial correlation weight matrix is ​​used to perform spatial weighted fusion and transformation on the time-frequency error feature matrix to obtain the composite error data structure.

[0107] The set of time-domain correlation functions and the set of frequency-domain complex transfer functions are key indicators used to quantify the time and frequency differences between predicted and measured radiation data. The time-domain correlation function reflects the similarity or delay relationship between two signals on the time axis, while the frequency-domain complex transfer function reveals the amplitude ratio and phase difference of the signals at different frequency components. They are fundamental to evaluating the accuracy of model predictions. In practice, the time-domain correlation function can be calculated using cross-correlation algorithms; for example, for discrete signals, it can be obtained by summing products within a sliding window. The frequency-domain complex transfer function can be obtained by comparing the Fourier transforms of the predicted and measured signals, where the complex form contains amplitude and phase information. Furthermore, advanced signal processing techniques such as wavelet transform or Hilbert transform can be used to extract time-frequency domain features, thereby constructing functions that reflect signal similarity.

[0108] The spatial topological relationship between the measured near-field radiation data and the measured far-field radiation data refers to the specific spatial arrangement of the magnetic near-field probe array and broadband antenna during radiated emission testing, and their geometric relationships. This topological relationship is crucial for understanding the propagation and distribution of electromagnetic fields in space, as the intensity and phase of the electromagnetic field change with spatial location, and there is a physical correlation between measurement results at different locations. This topological relationship can be precisely represented by recording the Cartesian coordinates (x, y, z) of each probe or antenna, and further recording the probe's orientation and polarization information. Alternatively, the spatial layout of these measurement points can be visualized and managed by establishing a three-dimensional model, which can include the distances and angles between measurement points and their relative positions to the electric drive system under test.

[0109] This composite error data structure is a comprehensive data representation that integrates the characteristics of model prediction errors across the time, frequency, and spatial dimensions. Its purpose is to provide a comprehensive, high-fidelity view of the error, enabling subsequent digital twin model corrections to more accurately capture and correct various types of errors. This data structure can be a multidimensional array or tensor, where different dimensions represent time, frequency, spatial location, and error type (e.g., amplitude error, phase error). Alternatively, it can be a structure or object containing multiple fields, each storing error information from a different dimension, with internal links or indexes maintaining their correlation.

[0110] This set of time-domain correlation functions and frequency-domain complex transfer functions are organized into a matrix to form a time-frequency error feature matrix with measurement locations as rows and time-domain delay units and frequency-domain spectral lines as columns. This aims to structurally integrate scattered time-domain and frequency-domain error information. Matrixing allows for a unified representation of error characteristics at different measurement locations, facilitating subsequent spatial dimension processing. For example, the time-domain correlation function (e.g., its peak value, width, and delay time) and frequency-domain complex transfer function (e.g., amplitude difference and phase difference at a specific frequency point) for each measurement location can be extracted as feature vectors. These feature vectors can then be stacked according to measurement locations to form a matrix. Alternatively, the time-domain correlation function curve and frequency-domain spectral lines (amplitude spectrum and phase spectrum) for each measurement location can be directly used as rows or columns of a matrix. Through appropriate sampling and discretization, a high-dimensional time-frequency error feature matrix can be constructed.

[0111] Based on the spatial topological relationship between the measured near-field and far-field radiation data, a spatial correlation weight matrix is ​​constructed to reflect the electromagnetic field coupling strength between different measurement locations. This is to quantify the degree of mutual influence between electromagnetic fields at different measurement points. Since electromagnetic fields couple as they propagate in space, the error at one point may be correlated with the errors at adjacent points. By constructing a spatial correlation weight matrix, this spatial dependence can be reflected more accurately, thus assigning reasonable weights to the errors at different locations during error fusion. This matrix can be determined by calculating the reciprocal of the distance between different measurement locations or based on an electromagnetic field propagation model (such as a free-space propagation loss model). Points that are closer or more strongly coupled have greater weights. Alternatively, electromagnetic simulation of the test environment can be performed to pre-calculate the mutual coupling coefficients between different measurement points and use them as elements of the spatial correlation weight matrix.

[0112] By utilizing the spatial correlation weight matrix to perform spatial weighted fusion and transformation on the time-frequency error feature matrix, a composite error data structure is obtained. This is a crucial step in integrating error information from the time, frequency, and spatial dimensions. Through spatial weighted fusion and transformation of the time-frequency error feature matrix, a comprehensive composite error data structure can be generated. This structure not only includes the time-frequency characteristics of the error but also reflects its spatial distribution and interrelationships. For example, the spatial correlation weight matrix can be multiplied or convolved with the time-frequency error feature matrix to achieve weighted fusion. A weighted average can then be applied to each column of the time-frequency error feature matrix (representing a time-domain delay unit or frequency-domain spectral line), with the weights provided by the spatial correlation weight matrix. Alternatively, dimensionality reduction techniques such as Principal Component Analysis (PCA) or Independent Component Analysis (ICA) can be used, combined with the spatial correlation weight matrix, to transform the time-frequency error feature matrix and extract the most representative spatial-time-frequency error components, thus forming the composite error data structure.

[0113] Through the above technical solution, this application effectively solves the problem that the model prediction error cannot accurately reflect the spatial distribution characteristics of the electromagnetic field in the process of correcting the multi-physics digital twin model of an electric drive system due to neglecting the spatial positional relationship between the measured near-field radiation data and the measured far-field radiation data. Specifically, by organizing the time-domain correlation function and the frequency-domain complex transfer function into a matrix, a time-frequency error feature matrix is ​​formed. This integrates the variation characteristics of the error in time and frequency in a structured way, avoiding the information loss caused by isolated analysis. On this basis, a spatial correlation weight matrix is ​​constructed based on the spatial positional topological relationship of the measured radiation data. This matrix considers the electromagnetic field coupling strength between different measurement points, so that the error representation can truly reflect the interaction in physical space. The time-frequency error feature matrix is ​​weighted and transformed in the spatial dimension using the spatial correlation weight matrix to generate a composite error data structure. This composite error data structure integrates the error characteristics of all dimensions, providing high-fidelity basic data for the subsequent generation of the deviation spectrum, thereby ensuring that the deviation spectrum can more comprehensively capture the dynamic changes of the error in the time-domain consistency, frequency-domain amplitude and phase, and spatial distribution characteristics. This improves the accuracy and reliability of the correction of the multiphysics digital twin model of the electric drive system, enabling the corrected model to more accurately predict the radiated emission behavior of the electric drive system, and providing a more refined and comprehensive basis for the diagnosis and optimization of electromagnetic compatibility issues.

[0114] In some of the solutions described above in this application, a deviation spectrum is proposed to be generated by separating error components strongly correlated with the dynamic behavior of mechanical structures from a composite error data structure based on vibration data, in order to directionally correct the structural dynamic coupling parameters in the digital twin model. However, in its implementation, since the composite error data structure contains multiple error sources and involves time domain, frequency domain and spatial distribution characteristics, direct separation may not be able to accurately identify and enhance the error components strongly correlated with the dynamic behavior of mechanical structures, resulting in an inaccurate deviation spectrum, which in turn affects the accuracy of the digital twin model correction and the reliability of the subsequent shaft current contribution analysis.

[0115] In response, this application further proposes to use the energy of characteristic frequency bands in the vibration data of a multi-physics synchronous data stream as a basis to separate error components strongly correlated with the dynamic behavior of the mechanical structure from the composite error data structure and generate a deviation spectrum. The steps include: Spectral analysis of the vibration data identifies one or more characteristic frequency bands corresponding to the resonant frequency of the mechanical structure of the tested electric drive system.

[0116] Calculate the energy distribution of the vibration data within one or more characteristic frequency bands, and generate a frequency domain weighting factor based on the energy distribution.

[0117] The frequency domain weighting factor is convolved with the composite error data structure in the frequency domain to enhance the extraction of error components corresponding to the one or more feature frequency bands in the composite error data structure, thereby obtaining the deviation spectrum.

[0118] Specifically, spectral analysis of the vibration data aims to convert the time-domain vibration signal into a frequency-domain signal to reveal its frequency components and their intensity. The core of this step is identifying specific frequency ranges related to the resonance of the mechanical structure of the tested electric drive system. These frequency ranges typically correspond to frequencies at which mechanical components are prone to large-amplitude vibrations under specific excitation. Identifying these characteristic frequency bands is fundamental for subsequent accurate separation of error components, ensuring the targeted nature of the correction process. In practice, the Fast Fourier Transform (FFT) algorithm can be used to process the acquired vibration data to obtain the spectrum of the vibration signal. By analyzing the frequency points or frequency ranges with concentrated energy or prominent amplitudes in the spectrum, combined with the structural design parameters and material properties of the electric drive system, one or more characteristic frequency bands corresponding to the resonance frequency of the mechanical structure can be identified manually or through a preset algorithm. Alternatively, time-frequency analysis methods such as wavelet transform and Hilbert-Huang transform (HHT) can be used to simultaneously observe the changes in the frequency components of the vibration signal over time in the time and frequency domains, thereby more accurately capturing transient or non-steady-state resonance phenomena. By setting energy thresholds or peak detection algorithms, frequency ranges with increased energy within a specific time period can be automatically identified as characteristic frequency bands.

[0119] Based on this, the energy distribution of the vibration data within one or more characteristic frequency bands is calculated, and a frequency domain weighting factor is generated based on this energy distribution. The energy distribution reflects the intensity or activity level of the vibration signal within a specific frequency range. Generating a frequency domain weighting factor based on this energy distribution aims to quantify the activity level of the mechanical structure in different resonant frequency bands, providing a quantitative basis for the subsequent enhancement of error components. Higher energy frequency bands may have a greater impact on radiation errors due to their corresponding mechanical dynamic behavior, thus requiring higher weights. For example, for each identified characteristic frequency band, the integral area or root mean square value of the vibration spectrum within that band can be calculated as a representative of the energy of that band. These energy values ​​are then normalized, or their relative magnitudes are directly used as frequency domain weighting factors. Linear or nonlinear mappings (such as logarithmic mappings) can be used to convert the energy values ​​into weighting factors between 0 and 1. Alternatively, more complex statistical methods can be used, such as calculating the power spectral density (PSD) of the vibration signal within the characteristic frequency band and integrating the PSD to obtain the energy. When generating the weighting factor, a time window function can be introduced to dynamically weight the energy distribution over different time periods, reflecting the transient characteristics of the vibrational behavior. The weighting factor can be designed to reach its maximum at the energy peak and gradually decrease at lower energy levels.

[0120] Furthermore, the frequency-domain weighting factor is convolved with the composite error data structure in the frequency domain to enhance the extraction of error components corresponding to one or more feature frequency bands in the composite error data structure, thus obtaining the deviation spectrum. Frequency-domain convolution is a signal processing technique. When applied to this scenario, its function is to "filter" or "enhance" the composite error data structure using the frequency-domain weighting factor, thereby highlighting error components strongly correlated with the dynamic behavior of the mechanical structure while suppressing other uncorrelated or weakly correlated error components. In this way, the model prediction error caused by mechanical vibration can be more accurately focused on, generating a more targeted deviation spectrum. For example, the frequency-domain weighting factor can be regarded as a frequency-domain filter, directly multiplied point-by-point with the composite error data structure (which may already contain frequency-domain information) (this is equivalent to time-domain convolution in the frequency domain). Specifically, if the composite error data structure is represented as E(f) in the frequency domain and the frequency-domain weighting factor is represented as W(f), then the enhanced error component can be expressed as E_enhanced(f) = E(f) × W(f). This method is simple and efficient, enabling direct weighted enhancement in the frequency domain. Alternatively, the composite error data structure can be decomposed into different frequency domain components (if its original form is not purely frequency-domain), and then frequency-domain weighting factors can be applied to these components. For example, if the composite error data structure is a time-frequency error feature matrix, each frequency spectral line can be weighted with its corresponding frequency-domain weighting factor before reconstruction.

[0121] Through the above technical solution, this application effectively solves the problem of inaccurate identification caused by multiple error sources and complex time-frequency spatial characteristics when separating error components strongly correlated with the dynamic behavior of mechanical structures from composite error data structures. Specifically, by performing spectral analysis on vibration data, one or more characteristic frequency bands corresponding to the resonant frequency of the mechanical structure of the tested electric drive system can be identified, thereby precisely locking the analysis focus on the key influence area of ​​mechanical behavior and avoiding interference from irrelevant frequency bands. On this basis, the energy distribution of vibration data in these characteristic frequency bands is calculated and a frequency domain weighting factor is generated, quantifying the actual contribution intensity of mechanical dynamic behavior in different resonant frequency bands, providing an objective and quantitative basis for subsequent error enhancement. By performing frequency domain convolution operation on the composite error data structure with the frequency domain weighting factor, the effective enhancement and extraction of error components corresponding to the characteristic frequency bands is achieved, while suppressing interference from other error sources. This allows the generated deviation spectrum to more accurately reflect the model prediction error caused by the dynamic behavior of the mechanical structure, improving the accuracy and pertinence of digital twin model correction, thereby enhancing the reliability of subsequent shaft current contribution analysis and providing a more accurate diagnostic basis for the electromagnetic compatibility optimization of the electric drive system.

[0122] In some embodiments described above in this application, a dynamic correction of critical path impedance parameters of a digital twin model based on deviation spectrum is proposed to obtain a high-fidelity model consistent with the test state. However, model parameter correction is a complex inverse problem. Directly using deviation spectrum for parameter adjustment may face the following challenges: First, the lack of a systematic correction objective and optimization framework leads to a blind and inefficient correction process. Second, the dynamic variation range of critical path impedance parameters (such as bearing impedance) is subject to physical constraints from multi-physics states (such as vibration and temperature). Without these constraints, the correction results may deviate from physical reality, leading to model distortion. Third, the lack of a convergence judgment mechanism makes it impossible to ensure that the predicted accuracy of the corrected model meets the expected requirements, affecting the reliability of subsequent analysis.

[0123] To address this, this application further proposes a method for dynamically correcting the critical path impedance parameters in the multiphysics digital twin model of the electric drive system based on the deviation spectrum, thereby obtaining a high-fidelity digital twin model consistent with the current test state. (See [link to relevant documentation]). Figure 6 Specifically, it includes: 601. Taking the deviation spectrum as the optimization objective, the critical path impedance parameter in the multiphysics digital twin model of the electric drive system is defined as the variable to be optimized, and a model parameter inversion problem is constructed.

[0124] 602. Using the vibration data and temperature data in the multiphysics synchronous data stream, constrain the solution space of the model parameter inversion problem to limit the dynamic variation range of the critical path impedance parameter within a physically reasonable range.

[0125] 603. An optimization solver is used to iteratively solve the parameter inversion problem of the model. In each iteration, the critical path impedance parameter is updated and the predicted radiation data is recalculated until the norm of the deviation spectrum between the predicted radiation data and the measured radiation data is less than a preset threshold. At this point, the high-fidelity digital twin model is obtained.

[0126] Specifically, the deviation spectrum is used as the optimization objective. This deviation spectrum is a quantitative representation of the differences between predicted near-field radiation data, predicted far-field radiation data, and measured near-field and far-field radiation data. It includes composite error information encompassing temporal consistency, frequency domain amplitude and phase, and spatial distribution characteristics. Using it as the optimization objective means that the correction process aims to minimize the difference between model predictions and actual measurements, thereby enabling the model to more accurately reflect the behavior of the physical system. For example, the L2 norm (mean square error) or L1 norm (sum of absolute errors) of the deviation spectrum can be used as the objective function to quantify the degree of fit between the model and reality. Simultaneously, the critical path impedance parameters in the multiphysics digital twin model of the electric drive system are defined as variables to be optimized. These critical path impedance parameters refer to parameters in the multiphysics digital twin model of the electric drive system that affect shaft current conduction and radiated emission, such as the equivalent resistance and equivalent capacitance of the bearing oil film, and the contact impedance of the housing grounding path. These parameters may dynamically change due to factors such as wear, temperature, and vibration during actual operation. Defining these parameters as variables to be optimized is to improve the model's accuracy by adjusting them to adapt to the current physical state. For example, the initial value and range of variation of the bearing's equivalent resistance can be set, or the contact resistance of the grounding path can be used as an adjustable parameter. Based on the above definition, a model parameter inversion problem is constructed, which infers unknown parameters within the system (critical path impedance parameters in this case) by observing the system's output (the deviation spectrum). Constructing this problem typically involves defining an objective function (based on the deviation spectrum) and determining the parameters to be optimized and possible constraints. For example, a least-squares problem can be constructed to minimize the sum of squares of the deviation spectrum between predicted and measured radiation data, where the critical path impedance parameter is the independent variable.

[0127] To ensure the physical accuracy of the correction results, this application utilizes the vibration and temperature data from the multiphysics synchronous data stream to constrain the solution space of the model parameter inversion problem, thereby limiting the dynamic variation range of the critical path impedance parameter within a physically reasonable range. Vibration and temperature data are important physical quantities reflecting the operating state of an electric drive system. For example, the vibration state of a bearing affects its contact impedance, while temperature changes affect the conductivity and dielectric constant of the material. Real-time acquisition of vibration and temperature data provides a basis for the physical variation range of the critical path impedance parameter. For example, when vibration is severe, the bearing oil film may break down, leading to a decrease in equivalent resistance. When the temperature rises, the resistivity of the conductor increases. This constraint is to ensure the physical accuracy of the model correction results. Without physical constraints, optimization algorithms might find a parameter combination that mathematically minimizes the deviation spectrum but is physically unreasonable. Vibration and temperature data allow for setting dynamic upper and lower limits or trends for the critical path impedance parameter. For example, the allowable variation range of the bearing's equivalent resistance can be dynamically adjusted based on the bearing's vibration amplitude and frequency. Alternatively, the range of material conductivity can be dynamically adjusted based on the temperature of the motor housing.

[0128] Based on this, an optimization solver is used to iteratively solve the parameter inversion problem of the model. An optimization solver is an algorithmic tool used to find the optimal combination of parameters that maximizes the objective function. Iterative solving refers to the process of gradually approaching the optimal solution through a series of repeated computational steps. For example, gradient descent, Newton's method, genetic algorithms, or particle swarm optimization can be used as optimization solvers. These solvers adjust the parameters in each iteration based on the gradient information or heuristic rules of the objective function. In each iteration, the optimization solver generates a new set of critical path impedance parameters based on the current objective function value and parameter gradients, and updates these critical path impedance parameters. These new parameters are input into the multiphysics digital twin model of the electric drive system, the simulation is rerun, and the predicted radiation data is recalculated. This process ensures the gradual optimization of model parameters and continuous improvement of prediction results. The iterative process continues until the norm of the deviation spectrum between the predicted and measured radiation data is less than a preset threshold. This is the convergence condition for terminating the iteration. The norm of the bias spectrum (e.g., the L2 norm) quantifies the overall difference between the prediction and the measured data. When this difference is less than a preset, acceptable threshold, it indicates that the model has fitted the actual data accurately enough, and the iteration can stop. The preset threshold can be determined based on the accuracy requirements of the actual application; for example, it can be set as a certain percentage of the average amplitude of the measured radiation data. When the iteration meets the convergence condition, the current critical path impedance parameter value is considered to be the parameter that fits the current test state, at which point the high-fidelity digital twin model is obtained.

[0129] Through the above technical solution, this application systematically solves the accuracy and rationality issues of dynamic model correction by constructing and solving a physically constrained model parameter inversion problem. It explicitly defines the goal of "making model predictions consistent with measured data" as a "model parameter inversion problem" with the deviation spectrum as the objective, and uses key impedance parameters as optimization variables. This provides a clear mathematical framework and automated solution path for the correction process, overcoming the subjectivity and inefficiency of manual trial and error. It creatively utilizes synchronously acquired vibration and temperature data to impose physical constraints on the solution space of the parameter inversion problem. This step is crucial, ensuring that during the optimization process, the correction values ​​of parameters such as bearing contact impedance can only vary within a reasonable physical range determined by the mechanical vibration state and temperature conditions. This prevents the generation of physically unreliable parameter solutions simply for data fitting, ensuring the physical authenticity and extrapolation prediction ability of the corrected model. By iteratively solving and setting a convergence threshold based on the deviation spectrum norm, this scheme achieves automatic iteration and precise termination of the correction process. When the deviation between the predicted and measured data reaches a preset accuracy, a high-fidelity digital twin model that closely matches the current complex test conditions is automatically obtained, laying a reliable foundation for subsequent accurate calculation of the shaft current contribution. This scheme deeply integrates data-driven approaches with physical constraints, enabling model correction to move from "empirical adjustment" to "quantitative optimization."

[0130] In some of the embodiments described above in this application, a dynamic correction of the critical path impedance parameters in the multiphysics digital twin model of an electric drive system based on the deviation spectrum is proposed. However, in its implementation, without an effective iterative solution mechanism, the correction process may not converge, be inefficient, or lack accuracy, thus failing to reliably obtain a high-fidelity digital twin model consistent with the current test state.

[0131] To address this, this application further proposes an iterative solution using an optimization solver to solve the model parameter inversion problem, updating the critical path impedance parameters and recalculating the predicted radiation data in each iteration. This iterative solution process specifically includes: setting an optimization tolerance threshold and a maximum number of iterations; using the critical path impedance parameters to drive a multiphysics digital twin model of the electric drive system to calculate the predicted radiation data for the current iteration; determining the bias spectrum and objective function value for the current iteration based on the predicted and measured radiation data; and determining whether the objective function value is less than the optimization tolerance threshold or whether the maximum number of iterations has been reached. If so, the iteration terminates; otherwise, a new set of critical path impedance parameters is generated based on the optimization solver's strategy, and the next iteration begins.

[0132] Specifically, in the initial stage of iterative solution, it is necessary to pre-set the optimization tolerance threshold and the maximum number of iterations. The optimization tolerance threshold defines the convergence criterion for model correction; that is, when the objective function value (e.g., the norm of the bias spectrum) is less than or equal to this preset value, the optimization is considered to have reached sufficient accuracy, and the iteration can be terminated. For example, an absolute error threshold, such as 0.01, can be set, indicating that the iteration stops when the L2 norm of the bias spectrum is less than this value. Alternatively, a relative error threshold, such as 1%, can be set, indicating that the iteration stops when the rate of change of the objective function value between two consecutive iterations is less than this percentage. The maximum number of iterations is used to prevent the optimization process from getting stuck in an infinite loop or converging too slowly, thereby ensuring computational efficiency and reasonable resource utilization. For example, an integer value, such as 100 or 500 iterations, can be directly set based on experience or estimated computational complexity. Alternatively, the maximum number of iterations can be dynamically adjusted based on available computation time or hardware resources.

[0133] In each iteration, the current critical path impedance parameters are used to drive the multiphysics digital twin model of the electric drive system, calculating the predicted radiation data for the current iteration. This step aims to rerun the digital twin model using the updated critical path impedance parameters from the current iteration to generate new predicted radiation data. This data will be compared with measured data to evaluate the effectiveness of the current parameter corrections. For example, the new critical path impedance parameter values ​​generated by the optimizer can be directly substituted into the corresponding parameterization module in the multiphysics digital twin model of the electric drive system, and then the forward simulation calculation of the model can be performed, outputting predicted near-field and predicted far-field radiation data. Alternatively, the critical path impedance parameters can be passed as input variables to the model through the application programming interface (API) or interface of the digital twin model. The model internally adjusts its behavior based on these parameters, performs simulation calculations, and returns predicted radiation data.

[0134] Based on the predicted and measured radiation data for the current iteration, the bias spectrum and objective function value for the current iteration are determined. The core of this step is to quantify the difference between the current model prediction and the actual measurement, providing direction for the optimization algorithm. The bias spectrum is the error distribution obtained after comparing and analyzing the predicted and measured radiation data in the time and frequency domains. It reflects in detail the inaccuracies of the model at different frequencies and time points. The objective function value is a single numerical value used to quantitatively evaluate the bias spectrum, usually a norm (such as L2 norm or L1 norm) or a weighted average of the bias spectrum, used to measure the overall fit between the model prediction and the measured data. For example, following the above method, a time-frequency domain comparison analysis can be performed on the predicted and measured radiation data for the current iteration to obtain the bias spectrum, and then the L2 norm of the bias spectrum can be used as the objective function value. Alternatively, depending on specific needs, different weights can be assigned to the errors in different frequency bands or different spatial locations of the bias spectrum, and then the norm of the weighted bias spectrum can be calculated as the objective function value.

[0135] The iteration process determines whether the objective function value is less than the optimization tolerance threshold or whether the maximum number of iterations has been reached. If so, the iteration terminates; otherwise, a new set of critical path impedance parameters is generated based on the solver's strategy, and the next iteration begins. This step is the control logic of the iterative optimization process, used to decide when to stop optimization and how to adjust parameters to continue. If the current objective function value meets the preset accuracy requirement (less than the optimization tolerance threshold) or has reached the preset maximum computational limit (the maximum number of iterations has been reached), the optimization process stops. If the termination condition is not met, the solver calculates a new set of critical path impedance parameters based on its internal algorithm and the current iteration's bias spectrum or objective function value, aiming to further reduce the objective function value in the next iteration. For example, a gradient descent-based optimization strategy can be used to calculate the gradient of the objective function with respect to the critical path impedance parameters, and then update the parameters along the negative direction of the gradient, such as using the Adam optimizer or stochastic gradient descent (SGD) optimizer. Alternatively, for non-convex or high-dimensional optimization problems, heuristic algorithms such as genetic algorithms, particle swarm optimization (PSO), or simulated annealing can be used to search for optimal parameters.

[0136] Through the above technical solution, this application provides an efficient, robust, and adaptive model parameter correction mechanism. This mechanism ensures that the multiphysics digital twin model of the electric drive system converges stably during dynamic correction and achieves the expected accuracy at a controllable computational cost. This enables the obtained high-fidelity digital twin model to more accurately reflect the real physical behavior of the electric drive system under the current test conditions, providing a solid foundation for subsequent shaft current contribution analysis and improving the reliability and practicality of the entire integrated testing method.

[0137] In some embodiments described above in this application, a method for calculating the radiation contribution of shaft current based on a high-fidelity digital twin model is proposed. This requires obtaining two sets of radiation data—one including the influence of shaft current and one excluding the influence of shaft current—as a basis for comparison. However, obtaining these two sets of data directly from physical testing is technically infeasible because it is impossible to physically "turn off" the shaft current without changing other states of the system. If the "shaft current-free" state is simulated solely through theoretical estimation or simplified calculation, the calculation results will be severely deviated from the actual system state because real multiphysics coupling boundary conditions (such as real-time vibration and temperature) are not considered. This results in an unreliable benchmark for contribution calculation and distorted quantification results.

[0138] To address this, this application further proposes a method for obtaining benchmark simulation results and comparative simulation results, see [link to relevant documentation]. Figure 7 Specifically, it includes: 701. In the high-fidelity digital twin model, a first set of input conditions is configured, which includes the shaft current data, the vibration data, and the temperature data in the multi-physics synchronous data stream.

[0139] 702. Based on the first input condition set, drive the high-fidelity digital twin model, perform the first simulation calculation, and obtain the benchmark simulation result.

[0140] 703. In the high-fidelity digital twin model, the component parameters used to represent the shaft current conduction path are modified to a high impedance state, and a second input condition set that does not include the shaft current data is configured. The second input condition set includes the vibration data and the temperature data.

[0141] 704. Based on the modified high-fidelity digital twin model driven by the second input condition set, perform the second simulation calculation to obtain the comparative simulation results.

[0142] To better understand the above technical solutions, the technical features involved will be explained in detail below.

[0143] This high-fidelity digital twin model is an accurate mapping of the electric drive system in virtual space, capable of simulating its multiphysics behavior under real-world operating conditions. Its function is to serve as a platform for virtual experiments, and its high-fidelity characteristics ensure that the simulation results are highly consistent with the behavior of the actual physical system. This model can be constructed in various ways. For example, it can be built based on numerical methods such as finite element analysis (FEA) or boundary element analysis (BEM), combined with a system-level equivalent circuit model. Alternatively, it can be learned and built from a large amount of measured data using machine learning or data-driven methods to capture complex nonlinear relationships. Another approach is a hybrid modeling method combining physical modeling and data-driven methods, using physical equations to describe known mechanisms and data-driven methods to compensate for unknown or difficult-to-model parts.

[0144] The first input condition set is a set of time-domain synchronized input data containing all key physical quantities (shaft current data, vibration data, and temperature data) used to drive the digital twin model. Its function is to simulate the complete operating state of the electric drive system under real dynamic conditions, serving as input for subsequent benchmark simulations. This condition set can, for example, directly use shaft current data, vibration data, and temperature data from a multiphysics synchronous data stream acquired synchronously from physical tests. Alternatively, it can be used as input to the model after preprocessing these data (e.g., filtering, interpolation, resampling).

[0145] The shaft current, vibration, and temperature data in this multiphysics synchronous data stream are real-time data reflecting the electromagnetic, mechanical, and thermal states of the tested electric drive system, collected under a unified time base. Their function is to serve as a joint excitation source and dynamic boundary condition for the digital twin model, ensuring a high degree of alignment between simulation and physical testing in terms of time, operating conditions, and states. Specifically, shaft current data can be acquired using high-frequency current probes (such as Hall effect probes or Rogowski coils) in the shaft grounding loop. Vibration data can be acquired using vibration accelerometers (such as piezoelectric accelerometers) placed at key locations such as bearing housings and casings. Temperature data can be acquired using temperature sensors (such as thermocouples or resistance temperature detectors) installed on key heat-generating components such as motor windings, bearings, and casings.

[0146] The process of driving the high-fidelity digital twin model to perform the first simulation calculation involves loading the first set of input conditions into the high-fidelity digital twin model and starting the model's solver to perform multiphysics coupled simulation. Its function is to reproduce the multiphysics response of the electric drive system under real operating conditions in a virtual environment, thereby obtaining the total radiated emissions, including the contribution of shaft current. This can be achieved by utilizing the API interfaces or scripting functions of commercial simulation software (such as ANSYS Maxwell, COMSOL Multiphysics, and Abaqus) to map the input data to the model's boundary conditions and excitation sources, and then performing time-domain or frequency-domain solutions. Alternatively, a customized simulation platform can be developed, integrating a multiphysics solver to achieve real-time loading of input data and simulation calculations.

[0147] The baseline simulation result is the simulation data obtained from the first simulation calculation, reflecting the total radiated emissions of the electric drive system under real operating conditions. Its purpose is to serve as a benchmark for subsequent differential calculations with the comparison simulation results, representing the combined effect of all radiation sources, including shaft current. This result can be time-domain or frequency-domain data such as far-field electromagnetic field strength, power density, or near-field magnetic field strength directly output by the simulation software. Alternatively, it can be post-processed simulation data, such as Fourier transform and spectral analysis, to obtain spectral data suitable for comparison.

[0148] Modifying the component parameters representing the shaft current conduction path to a high-impedance state refers to adjusting the parameters of equivalent circuit components (such as resistors, inductors, and capacitors) related to shaft current conduction in the digital twin model to make them equivalent to an open circuit or a highly attenuated path. Its function is to "shield" or "cut off" the shaft current conduction path in the virtual environment, thereby simulating an ideal state without shaft current contribution, while keeping the effects of other physical fields (vibration, temperature) unchanged. This can be achieved by setting the parameter values ​​representing the bearing oil film, shaft ground brush, and the equivalent resistance of the bearing inner and outer ring contact in the model to an extremely large value (e.g., much larger than 10^6 times the actual impedance), making it approximately an open circuit in the circuit. Alternatively, equivalent circuit branches or physical field coupling terms related to the shaft current conduction path can be directly removed or disabled in the model.

[0149] The second input condition set, which includes the vibration and temperature data, refers to a set of time-domain synchronized input data containing only vibration and temperature data used to drive the digital twin model. Its function is to simulate the mechanical and thermal state of the electric drive system under real dynamic operating conditions while "shielding" the contribution of shaft current, serving as input for subsequent comparative simulations. This condition set can be extracted from a multi-physics synchronous data stream acquired synchronously during physical testing, using only vibration and temperature data as input to the model. Alternatively, based on the first input condition set, the shaft current data portion can be zeroed or removed.

[0150] The modified high-fidelity digital twin model, after undergoing a second simulation, involves loading a second set of input conditions into the modified model and initiating a multiphysics coupling simulation using the model's solver. Its purpose is to reproduce the multiphysics response of the electric drive system under real-world conditions in a virtual environment, without shaft current contribution, thus obtaining the radiated emissions caused only by vibration, temperature, and other factors. This is similar to the first simulation, but uses the modified model parameters and the second set of input conditions, ensuring consistency in time step, solver settings, etc., between the two simulations to guarantee comparability of results.

[0151] The comparative simulation results are obtained from the second simulation calculation and reflect the radiated emission simulation data of the electric drive system under real operating conditions, but without shaft current contribution. Their purpose is to serve as the other party in subsequent differential calculations with the baseline simulation results, representing the combined effect of radiation sources other than shaft current. These results can be time-domain or frequency-domain data such as far-field electromagnetic field strength, power density, or near-field magnetic field strength directly output by the simulation software. Alternatively, they can be post-processed simulation data, such as Fourier transform and spectral analysis, to obtain spectral data suitable for comparison.

[0152] By employing the aforementioned technical solution and utilizing a modified high-fidelity digital twin model, a pair of highly comparable simulation results are generated under identical multiphysics dynamic environments using the "controlled variable method," thus creating ideal conditions for accurately quantifying the radiation contribution of shaft current. By configuring a first input condition set containing all real data (shaft current data, vibration data, and temperature data) in the model and executing the simulation, a benchmark simulation result accurately reflecting the "actual total radiation state" is obtained. This result is equivalent to a "perfect measurement" that is physically impossible to achieve. While keeping other parts of the model unchanged, only the component parameters representing the shaft current conduction path are modified to a high-impedance state, and the shaft current data in the input is removed, forming a second input condition set, and the simulation is repeated. This step, while perfectly reproducing the real mechanical vibration and thermal state, cleverly "virtually shields" the shaft current, thus obtaining a comparative simulation result under the same operating conditions and boundary conditions, but lacking only the shaft current contribution. By comparing the simulation results from two simulations run under completely identical model frameworks and dynamic boundary conditions, the differences can be strictly and uniquely attributed to the effect of the shaft current, thus providing an absolutely reliable data foundation for subsequent unambiguous frequency domain difference and contribution calculations. This method achieves an "ideal control experiment" in virtual space that is impossible in the physical world, solving the problem of obtaining radiation data "without the influence of shaft current" in physical testing, avoiding distortion caused by theoretical estimations or simplified calculations, and ensuring the benchmark reliability and accuracy of the contribution calculation results. Furthermore, through specific data mapping, simulation execution, and result conversion steps, the problem of inaccurate data-model alignment in benchmark simulations is solved, ensuring that the simulation results truly reflect the electromagnetic response under dynamic conditions. Mapping shaft current data, vibration data, and temperature data to the model's excitation source interface, mechanical boundary interface, and thermal boundary interface respectively ensures that the input data corresponds to different physical field interfaces of the model, avoiding simulation distortion caused by data misalignment. The driving model performs multiphysics coupled time-domain simulations using mapped data as input. It utilizes synchronously acquired real-world operating data as a foundation, ensuring that the simulation process covers dynamic load changes within a complete cycle or a preset duration, thereby improving the simulation's realism and consistency. The time-domain electric field response at preset far-field observation points is extracted from the simulation results and converted into a benchmark simulation spectrum. This conversion transforms the time-domain response into a frequency-domain representation, providing directly comparable spectral data for subsequent frequency-domain differential calculations, supporting the accurate quantification of the shaft current contribution.

[0153] In some of the embodiments described above in this application, a first simulation calculation is proposed to be performed on a high-fidelity digital twin model driven by a first input condition set to obtain a benchmark simulation result, which is used to calculate the contribution spectrum of the shaft current to the total radiated emission. However, in this process, due to the lack of a specific data mapping mechanism and simulation execution details, the input data may not match the model interface, and the simulation process may not be able to accurately simulate the time-domain dynamic characteristics of the real working condition, thereby causing deviation in the benchmark simulation result and affecting the accuracy and reliability of the subsequent contribution spectrum calculation.

[0154] To address this, this application further proposes specific steps for driving the high-fidelity digital twin model based on the first input condition set, performing a first simulation calculation, and obtaining the benchmark simulation result. These steps include: mapping the shaft current data, vibration data, and temperature data from the first input condition set to the corresponding excitation source interface, mechanical boundary interface, and thermal boundary interface of the high-fidelity digital twin model, respectively, to obtain first mapping data. Driving the high-fidelity digital twin model, using the first mapping data as input, performs a complete cycle or a preset duration of multiphysics coupled time-domain simulation. From the results of the multiphysics coupled time-domain simulation, the time-domain electric field response at a preset far-field observation point is extracted, and this time-domain electric field response is converted into a benchmark simulation spectrum as the benchmark simulation result.

[0155] Specifically, before performing the first simulation calculation, the shaft current data, vibration data, and temperature data from the first input condition set need to be mapped to the corresponding excitation source interface, mechanical boundary interface, and thermal boundary interface of the high-fidelity digital twin model, respectively. This mapping process aims to ensure that the multi-physics synchronous data collected during physical testing can be accurately and seamlessly input into the digital twin model. This mapping can be implemented in various ways. For example, the raw data can be formatted and units standardized through a data parsing module to conform to the input specifications of the model interface. Then, the processed data can be directly assigned to the predefined excitation source, mechanical boundary, and thermal boundary parameters in the model through a programming interface (API) or scripting language. Alternatively, a data adaptation layer can be constructed, which is responsible for monitoring the input data stream in real time and dynamically routing different types of data to the corresponding physics solver or sub-model interface in the digital twin model according to preset mapping rules. Through this precise mapping, simulation errors caused by incompatible data formats or mismatched interface definitions can be effectively avoided, ensuring that the model receives excitations and boundary conditions consistent with the real physical state.

[0156] After obtaining the first mapping data, the high-fidelity digital twin model is driven to perform a full-cycle or preset-duration multiphysics coupled time-domain simulation using this first mapping data as input. Driving the digital twin model means activating its computational engine, enabling it to solve for the physical fields based on the input excitations and boundary conditions. This can be achieved in various ways. For example, a graphical user interface or batch command provided by a professional simulation software platform (such as Ansys, COMSOL, etc.) can be used to load the model and input data and start the simulation calculation. Alternatively, a customized simulation scheduling system can be developed to automate the simulation process by calling the model's underlying solver library or API. This multiphysics coupled time-domain simulation refers to the model continuously solving for the interactions between multiple physical fields, such as electromagnetic fields, mechanical vibrations, and thermal distribution, in the time dimension. Executing a full-cycle or preset-duration simulation is to comprehensively capture the transient response and steady-state characteristics of the electric drive system under dynamic operating conditions. For example, the simulation duration can be set to cover a complete motor operation cycle, or a sufficiently long preset time period can be set according to specific analysis needs to ensure that all important dynamic processes and coupling effects can be fully simulated.

[0157] From the results of this multiphysics coupled time-domain simulation, the time-domain electric field response at a preset far-field observation point is extracted, and this time-domain electric field response is converted into a benchmark simulation spectrum, which serves as the benchmark simulation result. Extracting the time-domain electric field response refers to filtering the electric field intensity data of a specific spatial location (i.e., the preset far-field observation point) over time from the large amount of data output by the simulation. This preset far-field observation point is usually determined based on electromagnetic compatibility testing standards or actual application scenarios, such as a specific direction at a certain distance from the electric drive system. The extraction process can be performed through the post-processing module of the simulation software, or by parsing the simulation result file using a custom script. Converting the time-domain electric field response into a benchmark simulation spectrum typically employs signal processing techniques such as Fourier transform (e.g., Fast Fourier Transform, FFT) to convert the time-domain signal to the frequency domain, thereby obtaining the amplitude and phase information of different frequency components. For example, FFT processing can be performed on the extracted time-domain electric field response data to obtain its spectral distribution at different frequencies; this spectrum is the benchmark simulation spectrum.

[0158] Through the above technical solution, this application effectively solves the problem of input data mismatch and model interface discrepancy caused by the lack of specific data mapping mechanism and simulation execution details when calculating the contribution spectrum of shaft current to total radiated emission, resulting in the inability of the simulation process to accurately simulate the real working conditions' time-domain dynamic characteristics, thus causing deviations in the benchmark simulation results. By accurately mapping shaft current data, vibration data, and temperature data to the corresponding interface of the high-fidelity digital twin model, a high degree of consistency between physical test data and model input is ensured, avoiding data misalignment or distortion, thereby providing accurate initial and boundary conditions for subsequent simulations. Executing a complete cycle or a preset duration of multi-physics coupled time-domain simulation enables the model to comprehensively capture the transient changes of the electric drive system under dynamic conditions and the complex coupling effects between multiple physics fields, greatly improving the simulation results' ability to reproduce the real physical process. Extracting the time-domain electric field response at a preset far-field observation point from the simulation results and converting it into a benchmark simulation spectrum provides standardized and quantifiable benchmark results, facilitating subsequent accurate frequency domain difference calculations with comparative simulation results. These measures work together to improve the accuracy and reliability of the benchmark simulation results, laying a solid foundation for the accurate calculation of the subsequent shaft current contribution spectrum, thus making the quantitative analysis of the radiation emission sources of the electric drive system more effective.

[0159] In some of the embodiments described above in this application, a second simulation calculation is performed based on a high-fidelity digital twin model with modified driving parameters of the second input condition set to obtain comparative simulation results. However, in this process, there may be operational ambiguities and uncertainties in how to accurately map vibration and temperature data to the corresponding interfaces of the model, and how to ensure that the simulation is executed completely and the electric field response is accurately extracted and converted into a spectrum. This will affect the reliability of the comparative simulation results and the accuracy of subsequent contribution calculations.

[0160] In response, this application further proposes a high-fidelity digital twin model based on the modified driving parameters of the second input condition set, performs a second simulation calculation, and obtains comparative simulation results, including: The vibration data and temperature data in the second input condition set are mapped to the corresponding mechanical boundary interface and thermal boundary interface of the high-fidelity digital twin model after parameter modification, respectively, to obtain the second mapping data.

[0161] The high-fidelity digital twin model with modified driving parameters, using the second mapping data as input, performs a full cycle or a preset duration of multiphysics coupled time-domain simulation.

[0162] From the results of the multiphysics coupled time-domain simulation, the time-domain electric field response at the preset far-field observation point is extracted, and the time-domain electric field response is converted into a comparative simulation spectrum as the comparative simulation result.

[0163] Specifically, when mapping the vibration and temperature data from the second input condition set to the corresponding mechanical and thermal boundary interfaces of the modified high-fidelity digital twin model to obtain the second mapping data, this step aims to accurately input the actual vibration and temperature data collected during physical testing into the digital twin model as its dynamic boundary conditions. Mapping ensures the synchronization and consistency between the model simulation and the actual physical process in these key physical quantities. This can be implemented through a data parsing module, which reads the vibration and temperature data from the second input condition set and converts them into a model-recognizable input format according to a predefined interface protocol or data format. For example, vibration data can be mapped to the instantaneous displacement or acceleration of a specific structural node in the model, and temperature data can be mapped to the instantaneous temperature field of a specific region in the model. Alternatively, mapping can be performed using scripts or application programming interfaces (APIs). By writing specific scripts or calling the model's API, the vibration and temperature data from the data stream can be directly injected into the mechanical and thermal boundary interfaces of the digital twin model in real-time or near real-time. For example, for mechanical boundary interfaces, vibration data can be converted into force or displacement boundary conditions. For thermal boundary interfaces, temperature data can be converted into heat flow or temperature boundary conditions.

[0164] In the high-fidelity digital twin model with modified drive parameters, this step is the core of the comparative simulation when performing a full cycle or a preset duration of multiphysics coupled time-domain simulation using the second mapping data as input. It utilizes the modified digital twin model, with the shaft current path shielded, to simulate the multiphysics behavior of the electric drive system under the influence of actual vibration and temperature data. Executing a full cycle or a preset duration ensures the comprehensiveness and representativeness of the simulation results, covering various states of the system under dynamic conditions. This can be achieved through an integrated simulation platform. This platform can coordinate the operation of different physics solvers (such as structural mechanics solvers, heat conduction solvers, and electromagnetic field solvers) and use the second mapping data as a unified input. The simulation duration can be set according to the duration of the actual operating conditions or the specific dynamic process of interest, such as a complete vehicle driving cycle or the duration of an acceleration / deceleration transient process. Alternatively, a co-simulation approach can also be used. The digital twin model is decomposed into multiple sub-models (such as mechanical, thermal, and electromagnetic sub-models), each handled by a dedicated solver. Through a co-simulation interface, these sub-models exchange data and perform iterative calculations driven by the second mapping data, collectively completing a multiphysics coupled time-domain simulation. The simulation duration can be flexibly configured as needed, for example, set to the entire process from startup to stable operation of the electric drive system, or the duration under a specific fault mode.

[0165] In the process of extracting the time-domain electric field response at a predetermined far-field observation point from the results of the multiphysics coupled time-domain simulation, and converting this time-domain electric field response into a comparative simulation spectrum as the comparative simulation result, this step aims to accurately obtain far-field electric field information directly related to radiative emission from the complex simulation results and convert it into frequency domain form for comparison with the benchmark simulation results. Extracting the response at the predetermined far-field observation point ensures the standardization and comparability of the results. In implementation, simulation software typically provides a post-processing module where users can define far-field observation points (e.g., specific spatial locations at a certain distance from the electric drive system). After the simulation is complete, the post-processing module can automatically extract the time-domain electric field intensity data at these observation points. The time-domain electric field response is then converted into a frequency domain spectrum using a Fast Fourier Transform (FFT) or other spectral analysis algorithms to obtain the comparative simulation spectrum. Alternatively, this can be achieved through a programming interface or a custom script. During or after the simulation, the electric field data stream at the predetermined far-field observation point can be directly obtained by calling the model's output interface. Use functions from signal processing libraries (such as MATLAB and Python's SciPy library) to perform spectral analysis on these time-domain data. For example, use the Welch method or Short Time Fourier Transform (STFT) to generate contrast simulation spectra with good resolution.

[0166] Through the above technical solution, the vibration and temperature data are accurately mapped to the corresponding mechanical and thermal boundary interfaces of the modified high-fidelity digital twin model. This ensures that, after shielding the shaft current path, the internal mechanical vibration and thermal distribution of the digital twin model can realistically and accurately reflect the actual dynamic conditions in the physical test. This avoids distortion of simulation results due to input data mismatch or inaccurate mapping, thereby improving the reliability of the comparative simulation results. Simultaneously, by using the modified high-fidelity digital twin model with the second mapped data as input, a complete cycle or a preset duration of multiphysics coupling time-domain simulation is performed, ensuring the comprehensiveness and representativeness of the simulation process. This allows the model to fully capture the radiative emission behavior of the electric drive system under dynamic conditions, without the influence of shaft current, caused by vibration, temperature, and other factors, providing a complete and reliable benchmark for subsequent contribution calculations. Furthermore, the time-domain electric field response at the preset far-field observation point is extracted from the results of this multiphysics coupling time-domain simulation, and this time-domain electric field response is converted into a comparative simulation spectrum. This process standardizes the output format of the simulation results. It directly provides quantifiable frequency domain data, offering accurate and consistent input for subsequent frequency domain differential calculations with benchmark simulation results. This effectively supports the accurate quantitative analysis of the contribution of shaft current to total radiated emission. Combined with benchmark simulations, this scheme, through precise comparative simulations, can clearly separate the influence of non-axis current factors (such as mechanical vibration and thermal effects) on radiated emission. This provides a solid foundation for subsequent frequency domain differential calculations of shaft current contribution, greatly improving the accuracy and reliability of the quantitative analysis of shaft current contribution.

[0167] In some of the schemes described above in this application, frequency domain differential calculation is proposed to quantify the contribution of shaft current to total radiated emission. However, in this process, the model simulation itself may have baseline errors, which may lead to inaccurate calculated shaft current radiation component spectrum, thereby affecting the reliability of the contribution spectrum.

[0168] To address this, this application proposes a method for calculating the frequency domain difference between the benchmark simulation results and the comparative simulation results to obtain the spectrum of the contribution of the shaft current to the total radiated emission. (See [link to relevant documentation]). Figure 8 Specifically, it includes: 801. Perform point-by-point amplitude subtraction in the frequency domain on the benchmark simulation spectrum and the comparison simulation spectrum contained in the comparison simulation result to obtain the axis current radiation component spectrum.

[0169] 802. Based on the benchmark simulation spectrum, the amplitude of the axis current radiation component spectrum is calibrated to eliminate the baseline error in the model simulation itself, and the calibrated axis current radiation component spectrum is obtained.

[0170] 803. Calculate the point-by-point amplitude ratio between the calibrated axis current radiation component spectrum and the reference simulation spectrum to obtain the contribution spectrum of the axis current to the total radiative emission.

[0171] Specifically, a point-by-point amplitude subtraction operation is performed in the frequency domain between the baseline simulation spectrum and the comparative simulation spectrum to obtain the axis current radiation component spectrum. This aims to initially separate the radiation contribution of the axis current through direct spectral differences. This step can be implemented by directly subtracting the amplitude value of the comparative simulation spectrum from the amplitude value of the baseline simulation spectrum at each frequency point, thus obtaining an amplitude spectrum representing the individual effect of the axis current. Alternatively, when considering the complex form of the spectrum, a complex subtraction operation can be performed first, and then the amplitude of the result can be taken as the axis current radiation component spectrum.

[0172] Furthermore, based on the benchmark simulation spectrum, the amplitude of the axis current radiation component spectrum is calibrated to eliminate the baseline error inherent in the model simulation, resulting in a calibrated axis current radiation component spectrum. This calibration step is crucial to this scheme, used to correct the inherent systematic biases of the simulation model. One approach is to analyze the characteristics of the benchmark simulation spectrum in a specific frequency band (e.g., a band where the axis current theoretically contributes very little), identify and quantify the baseline error in the model, and then subtract this error from the axis current radiation component spectrum. Another approach is to construct a calibration function or filter based on the benchmark simulation spectrum and apply it to the axis current radiation component spectrum to compensate for the inherent biases of the model at different frequencies, thereby obtaining a more accurate calibrated axis current radiation component spectrum.

[0173] Based on this, the amplitude ratio of the calibrated shaft current radiation component spectrum to the reference simulated spectrum is calculated point by point to obtain the contribution spectrum of the shaft current to the total radiated emission. This step is used to quantify the relative importance of the shaft current in the total radiation. Specifically, at each frequency point, the amplitude value of the calibrated shaft current radiation component spectrum can be divided by the amplitude value of the reference simulated spectrum to obtain a ratio between 0 and 1, representing the contribution ratio of the shaft current at that frequency point. Alternatively, for a more intuitive representation, this ratio can be converted to a percentage form.

[0174] Through the above technical solution, this application can effectively solve the baseline error problem in model simulation and improve the accuracy and reliability of quantitative analysis of the contribution of shaft current to total radiated emission. By performing point-by-point amplitude subtraction in the frequency domain, the radiated component of the shaft current can be initially separated. Using this benchmark simulation spectrum to calibrate the amplitude of the shaft current radiated component spectrum can identify and eliminate inherent systematic biases in the simulation model, ensuring that the obtained shaft current radiated component spectrum more realistically reflects the actual contribution of the shaft current and avoiding misjudgments caused by model errors. By calculating the point-by-point amplitude ratio, the proportion of the calibrated shaft current radiated component in the total radiation can be accurately quantified, providing engineers with a clear and reliable diagnostic basis. This method enables the location and quantification of shaft current radiation sources within the electric drive system under dynamic operating conditions based on a high-fidelity digital twin model, thereby guiding more efficient and targeted electromagnetic compatibility rectification measures, significantly reducing trial-and-error costs, and improving the overall electromagnetic compatibility performance of the electric drive system.

[0175] In some of the schemes mentioned above in this application, amplitude calibration is proposed to eliminate the baseline error inherent in the model simulation itself. However, in this process, due to the inherent error that may be introduced by the model simulation, the calibrated axis current radiation component spectrum is not accurate enough, and the contribution of the axis current to radiated emission cannot be accurately quantified.

[0176] To address this, this application further proposes a method for amplitude calibration of the shaft current radiation component spectrum based on a benchmark simulation spectrum. This calibration eliminates baseline errors inherent in the model simulation, resulting in a calibrated shaft current radiation component spectrum. The specific steps include: calculating the cutoff frequency of shaft current conduction based on the equivalent circuit parameters of the shaft current conduction path in the multiphysics digital twin model of the electric drive system, and defining the frequency band above the cutoff frequency as the reference frequency band. The average amplitude of the shaft current radiation component spectrum within the reference frequency band is calculated as a baseline error estimate. Finally, the baseline error estimate is subtracted from the amplitude of the shaft current radiation component spectrum across the entire frequency band to obtain the calibrated shaft current radiation component spectrum.

[0177] In the multiphysics digital twin model of this electric drive system, the equivalent circuit parameters of the shaft current conduction path refer to the set of parameters characterizing the electrical characteristics of the shaft current conduction path within the electric drive system. These parameters typically include equivalent resistance, equivalent inductance, and equivalent capacitance, which together determine the conduction characteristics of the shaft current at different frequencies. These parameters can be obtained by modeling and analyzing the geometric structure and material properties of physical components such as the shaft, bearings, and grounding brushes, or extracted by impedance measurement and system identification of the actual system. For example, a lumped parameter model can be used to represent the bearing oil film as a capacitor and resistor in parallel, and the shaft as an inductor and resistor in series, thus obtaining the complete equivalent circuit parameters.

[0178] The cutoff frequency for shaft current conduction, as calculated in this context, refers to the frequency point in the frequency response characteristics of the shaft current conduction path where energy transfer efficiency decreases or characteristics change, determined based on the equivalent circuit parameters mentioned above. This cutoff frequency is an inherent physical property of the shaft current conduction path, reflecting its ability to suppress or allow currents of different frequencies. The calculation method can be to analyze the transfer function of the equivalent circuit to find its -3dB frequency point, or to identify the frequency at which the amplitude response begins to decay significantly by simulating the frequency response curve of the equivalent circuit. For example, for an RC circuit, its cutoff frequency can be calculated using the product of the resistance and capacitance.

[0179] Determining a frequency band above the cutoff frequency as a reference band involves selecting one or more frequency ranges that are higher than the calculated shaft current conduction cutoff frequency. This band is chosen because the direct conduction effect of the shaft current is typically weaker or more predictable within it; therefore, the radiation components within this band are more likely to reflect inherent errors in the model itself rather than the actual contribution of the shaft current. The reference band can be determined by including all frequency bands above the cutoff frequency, or by selecting a certain bandwidth (e.g., an octave) above the cutoff frequency as a reference.

[0180] The spectrum of the radiated component of the shaft current is obtained by frequency domain difference calculation based on the baseline simulation results and the comparison simulation results. It quantifies the contribution of the shaft current to the total radiated emission at different frequencies.

[0181] The average amplitude of the axis current radiation component spectrum within the reference frequency band, used as a baseline error estimate, is calculated by statistically averaging the amplitude values ​​of the aforementioned axis current radiation component spectrum across all frequency points within the defined reference frequency band. This average is considered representative of the inherent error of the model in frequency bands where there is no direct contribution from the axis current. The average can be calculated using an arithmetic mean (the sum of all amplitude values ​​divided by the number of frequency points), a weighted average (assigning higher weights to amplitudes at specific frequency points), or a root mean square (RMS) average to better reflect energy characteristics.

[0182] The calibrated axis current radiation component spectrum is obtained by subtracting the baseline error estimate from the amplitude value of the axis current radiation component spectrum across the entire frequency band. This involves subtracting the baseline error estimate from the amplitude value at each frequency point of the axis current radiation component spectrum. This operation aims to eliminate the inherent systematic bias of the model, allowing the calibrated spectrum to more accurately reflect the true contribution of the axis current to the total radiated emission at each frequency. This subtraction is typically a simple point-by-point arithmetic subtraction, subtracting a scalar error value from the entire spectrum.

[0183] Through the above technical solution, this application effectively solves the problem of inaccurate shaft current radiation component spectrum caused by inherent errors introduced in model simulation. By calculating the equivalent circuit parameters of the shaft current conduction path in the multiphysics digital twin model of the electric drive system based on the equivalent circuit parameters, a physically meaningful cutoff frequency is determined, and a reference frequency band is established based on this, ensuring the accuracy and physical rationality of the baseline error estimation. The average amplitude of the shaft current radiation component spectrum within the reference frequency band is calculated as the baseline error estimate, effectively capturing the inherent bias of the model in frequency bands where the shaft current contribution is small. Subtracting this baseline error estimate from the shaft current radiation component spectrum across the entire frequency band achieves systematic calibration of the entire spectrum. This calibration method based on physical parameters avoids subjective assumptions, resulting in a more accurate calibrated shaft current radiation component spectrum. This allows for a more reliable quantification of the shaft current's contribution to total radiated emission, providing a solid data foundation for subsequent electromagnetic compatibility problem diagnosis and optimization.

[0184] In some of the solutions described above in this application, the cutoff frequency of shaft current conduction is calculated to determine the reference frequency band for baseline error calibration. However, in this process, how to ensure the accuracy and reliability of the cutoff frequency calculation is a key issue to avoid calibration errors. Specifically, inaccurate acquisition of equivalent circuit parameters or lack of standardization in the calculation process may lead to a deviation in the selection of the reference frequency band, thereby affecting the accuracy of baseline error estimation and contribution spectrum.

[0185] To address this, this application further proposes a method for calculating the cutoff frequency of shaft current conduction based on the equivalent circuit parameters of the shaft current conduction path in a multiphysics digital twin model of an electric drive system. The method includes: obtaining equivalent resistance and equivalent capacitance parameters representing the shaft current conduction path from the multiphysics digital twin model of the electric drive system; calculating the first-order RC circuit time constant of the shaft current conduction path based on the equivalent resistance and equivalent capacitance parameters; and calculating the cutoff frequency based on the first-order RC circuit time constant using a preset transformation relationship.

[0186] Specifically, when obtaining the equivalent resistance and equivalent capacitance parameters representing the shaft current conduction path from the multiphysics digital twin model of the electric drive system, these parameters are lumped parameters used to simplify and characterize the electrical properties of complex circuits or physical systems. They represent the resistance encountered by the current flowing through the path and the charge storage capacity, respectively. These parameters are key components of the internal construction of the multiphysics digital twin model of the electric drive system, reflecting actual physical characteristics such as bearings, lubricating oil films, parasitic capacitance between the shaft and the housing, and contact resistance. One implementation approach is that, during the construction of the multiphysics digital twin model of the electric drive system, these equivalent parameters may have already been theoretically calculated, finite element simulations, or experimentally measured and stored in the model's parameter database. The system can then directly query and extract these predefined parameter values ​​from this database. Another approach is that, for more complex digital twin models, the equivalent parameters may not be fixed values, but rather dynamically calculated based on the model's current state (such as temperature, vibration, etc.). For example, the model may contain a sub-module that calculates the equivalent resistance and capacitance in real time based on inputs such as the bearing's geometry, material properties, and lubricating oil film thickness.

[0187] When calculating the time constant of a first-order RC circuit for an axial current conduction path based on the equivalent resistance and equivalent capacitance parameters, the first-order RC circuit time constant (τ) is a key parameter for measuring the transient response speed of the RC circuit, defined as the product of resistance R and capacitance C (τ = R × C). It represents the time required for the capacitor to charge or discharge to approximately 63.2% of its voltage. Calculating this time constant in an axial current conduction path helps quantify the path's response characteristics to high-frequency signals. One approach is to obtain the equivalent resistance parameter R and the equivalent capacitance parameter C, and then calculate the time constant using a simple multiplication operation R × C. This is suitable for cases where the equivalent circuit model is a standard series or parallel RC structure. Another approach is that if the equivalent circuit of the axial current conduction path is more complex, such as containing multiple RC branches, it may be necessary to first simplify the complex circuit to an equivalent first-order RC circuit, or determine the dominant time constant by analyzing its transfer function, for example, by calculating the Thevenin equivalent resistance and Norton equivalent capacitance of the circuit before performing the multiplication operation.

[0188] When the cutoff frequency is calculated based on the time constant of a first-order RC circuit using a preset conversion relationship, the cutoff frequency (f_c) is an important indicator of the circuit's or system's frequency response, typically defined as the frequency at which the output power drops to half the input power (i.e., the -3dB point). For a first-order RC circuit, there is a fixed mathematical relationship between its cutoff frequency and time constant; the preset conversion relationship refers to this known, standardized mathematical formula. One implementation is to use the most common preset conversion relationship f_c = 1 / (2 × π × τ), which directly converts the calculated time constant τ into the corresponding cutoff frequency. Another implementation is that, in certain specific applications, considering non-ideal factors or specific frequency response definitions, a modified or empirical conversion formula may be used. For example, if the focus is on a specific signal attenuation threshold rather than the -3dB point, the conversion relationship may be adjusted, but its essence is still based on calculations using the time constant.

[0189] Through the above technical solution, this application effectively solves the accuracy and reliability problems in cutoff frequency calculation. By directly obtaining the equivalent resistance and equivalent capacitance parameters representing the shaft current conduction path from the multiphysics digital twin model of the electric drive system, the realism and completeness of the digital twin model are fully utilized, avoiding the arbitrariness of parameter estimation, thereby ensuring the accuracy and physical consistency of the input data. Based on these accurate equivalent parameters, the time constant of the first-order RC circuit of the shaft current conduction path is calculated, transforming the physical parameters into the time constant of the standard circuit model, ensuring the physical correctness and consistency of the calculation basis. The cutoff frequency is calculated through a preset conversion relationship (such as a standard formula), which provides a repeatable and predictable conversion mechanism, eliminating subjective errors in the calculation process and ensuring the reliability and verifiability of the cutoff frequency result. Overall, this method, through a step-by-step approach, directly obtains key parameters from the model and applies standard calculation rules, effectively solving the uncertainty in cutoff frequency calculation and providing solid support for subsequent baseline error calibration. This enables baseline error estimation based on an accurate and reliable reference frequency band when calibrating the shaft current radiation component spectrum, thereby obtaining a more accurate calibrated shaft current radiation component spectrum. This improves the accuracy of the contribution spectrum of shaft current to total radiated emission and provides more reliable data for electromagnetic compatibility diagnosis and optimization of electric drive systems.

[0190] In some of the solutions described above in this application, methods are proposed to quantify the contribution of shaft current to radiated emission under dynamic operating conditions. However, in this process, there is a lack of systematic storage and utilization of historical diagnostic data, which means that the entire process needs to be executed from scratch for each test. It is impossible to use previous experience to optimize test condition planning or model initialization parameters, thereby affecting test efficiency and accuracy.

[0191] To this end, this application further proposes a method that includes: identifying one or more sensitive frequency bands dominated by the shaft current's contribution spectrum to the total radiated emission; extracting the contribution value corresponding to the sensitive frequency band from the contribution spectrum; and generating a structured diagnostic record and storing it in a radiation contribution case library based on the sensitive frequency band, the corresponding contribution value, and the operating condition and multiphysics state data extracted from the multiphysics synchronization data stream that are associated with the occurrence time of the sensitive frequency band. When subsequently testing the same or similar electric drive systems, the radiation contribution case library is queried, and the test condition planning or initialization model correction parameters are optimized based on the matching historical diagnostic records.

[0192] Specifically, after obtaining the contribution spectrum of shaft current to total radiated emissions, it is necessary to identify one or more sensitive frequency bands where shaft current dominates radiation. Sensitive frequency bands refer to the frequency range where the contribution of shaft current to total radiated emissions is higher than other frequency bands. The purpose of identifying these frequency bands is to focus on the problem, make accurate diagnoses, and avoid blind rectification. For example, a contribution threshold can be set, defining the continuous frequency range in the contribution spectrum exceeding that threshold as a sensitive frequency band. Alternatively, peak detection can be performed on the contribution spectrum to identify frequency points with local maxima and their surrounding areas as sensitive frequency bands. Furthermore, expert experience or pre-defined electromagnetic compatibility standards can be combined to delineate specific frequency ranges requiring special attention within the contribution spectrum as sensitive frequency bands.

[0193] The contribution value corresponding to the sensitive frequency band is extracted from the contribution spectrum. Extracting the contribution value quantifies the radiation impact of the shaft current within the sensitive frequency band, providing concrete data support for subsequent diagnosis and optimization. For example, the contribution value at each frequency point within the sensitive frequency band can be directly read, or statistics such as the average contribution and maximum contribution within the band can be calculated. Alternatively, the contribution spectrum within the sensitive frequency band can be integrated to obtain the total contributed energy or power of that band, which can then be used as its contribution value.

[0194] Based on this, a structured diagnostic record is generated and stored in the radiation contribution case library, based on the sensitive frequency band, the corresponding contribution value, and the operating conditions and multiphysics state data extracted from the multiphysics synchronous data stream that are associated with the time of occurrence of the sensitive frequency band. The structured diagnostic record organizes and stores the key information identified in this test (sensitive frequency band, contribution, related operating conditions, and multiphysics state) in a unified and standardized format, forming a knowledge base that can be queried and reused. The radiation contribution case library is the database that stores these structured diagnostic records. For example, a structured diagnostic record may contain fields such as: test date, model of the tested electric drive system, sensitive frequency band range, peak / average shaft current contribution, corresponding operating parameters (such as speed, torque, temperature), vibration data characteristics, near-field / far-field radiation data characteristics, and possible preliminary diagnostic conclusions. The case library can be a relational database (such as MySQL, PostgreSQL) or a non-relational database (such as MongoDB) to store and manage these records. Alternatively, the structured diagnostic record can also be defined using data formats such as XML or JSON to ensure data consistency and parsability. The radiation contribution case library can be a file system that stores each record as a separate file and manages it through an index file.

[0195] When performing subsequent tests on the same or similar electric drive systems, the radiation contribution case library is queried, and test condition planning or model correction parameters are optimized based on matching historical diagnostic records. This step aims to use historical data to guide future testing and model correction, avoiding repetitive work and improving efficiency and accuracy. For example, during the query, similar historical diagnostic records can be retrieved from the case library based on the model of the electric drive system under test, the test target, or preliminary test results (such as exceeding the standard frequency band). Optimizing test condition planning can refer to adjusting the test condition parameters of subsequent tests based on the conditions that lead to high contribution in the historical records, making them more targeted. Initializing model correction parameters can refer to using the parameters of successfully corrected models in the historical records as the initial values ​​for this model correction, accelerating convergence. Alternatively, the matching algorithm can use similarity calculation based on feature vectors (such as cosine similarity) to compare the features of the current test with the historical records in the case library. Optimizing test condition planning can include suggestions for more detailed testing under specific speed, load, or temperature conditions. Initializing model correction parameters can use the corrected critical path impedance parameters in the historical records as the starting point for this iterative solution.

[0196] Through the above technical solution, this application addresses the problem of insufficient historical data utilization and optimizes subsequent testing processes by introducing a diagnostic record generation and case library utilization mechanism. Specifically, sensitive frequency bands are identified based on the contribution spectrum of axis current to total radiated emission, enabling focus on radiation problem areas dominated by axis current and avoiding generalized analysis. Contribution values ​​are extracted from the contribution spectrum, providing quantitative indicators for accurate diagnosis. Structured diagnostic records are generated based on sensitive frequency bands, contribution values, and relevant operating conditions and multi-physics state data, integrating key test information to form systematic historical data. This data is stored in a radiation contribution case library, establishing a queryable database that supports data reuse. During subsequent testing, the case library is queried, and test condition planning or model initialization parameters are optimized based on matching records. Historical experience is utilized to reduce repetitive work and improve testing efficiency and accuracy. This not only improves testing efficiency and reduces unnecessary testing steps and time but also helps to more accurately understand the root causes of axis current radiation by providing rich contextual information, thereby optimizing test condition planning to better reflect actual problems. Meanwhile, by using successful model correction parameters from historical records as initial values, the calibration and optimization process of the digital twin model is accelerated, the high fidelity of the model is improved, and an effective way is provided for enterprises to accumulate and pass on their experience and practices in diagnosing shaft current radiation in electric drive systems.

[0197] The following example will provide a more detailed explanation of the above technical solution: During the research and development phase of an electric drive system for a new energy vehicle, it is necessary to assess its electromagnetic compatibility under actual operating conditions, especially the radiated emissions caused by shaft current. In existing technologies, shaft current measurement and radiated emission testing are usually conducted independently, resulting in data that cannot be aligned in time and state, making it difficult to quantify the contribution of shaft current to total radiation.

[0198] This method provides an integrated testing and analysis solution. The electric drive system under test is mounted on a dynamometer bench. This bench can simulate real road conditions. For example, by parsing a comprehensive road cycle condition file, the vehicle speed curve, road gradient curve, and load demand curve are extracted, and the demand torque time-series curve for driving the electric drive system is calculated based on the vehicle dynamics model. According to this demand torque time-series curve, the dynamometer bench applies dynamic torque loads, including acceleration, deceleration, and load abrupt changes, to the electric drive system.

[0199] While applying dynamic loads, multiple physical quantities are simultaneously acquired. Specifically, shaft current data is acquired using a high-frequency current probe configured in the shaft grounding loop of the electric drive system. Near-field radiation data is acquired using a magnetic near-field probe array positioned at key locations on the electric drive system housing. Far-field radiation data is acquired using a broadband antenna positioned at a standard test distance. Vibration and temperature data are acquired using vibration acceleration and temperature sensors mounted on the bearing housing and housing of the electric drive system, respectively. The output signals of all these sensors are connected to the same multi-channel synchronous data acquisition system, and synchronous sampling and recording are performed under the control of a unified time base clock, thereby generating a time-aligned multi-physics synchronous data stream. Unlike existing technologies where test items are independent and data cannot be aligned, this method ensures that all key physical quantities are acquired synchronously under the same dynamic operating conditions, laying the foundation for subsequent correlation analysis.

[0200] The multiphysics synchronous data stream is used to drive the operation of a multiphysics digital twin model of the electric drive system. This digital twin model is a parameterized multiphysics coupling model, constructed to include a complete coupling path from the common-mode voltage source inside the motor, the distributed parameter network between the windings and the housing, the bearing conductive loop, and the shaft current conduction to spatial radiation within the motor housing structure. The model establishes multiphysics coupling relationships between electromagnetic response, mechanical vibration response, and temperature distribution. For example, vibration deformation leads to dynamic changes in structural and electromagnetic parameters, and temperature changes lead to dynamic changes in the electromagnetic properties of the material. The shaft current data, vibration data, and temperature data from the synchronous data stream are input into the model as a set of time-domain synchronized joint excitation sources and dynamic boundary conditions. For example, shaft current data segments are mapped to the time-varying amplitude of the high-frequency common-mode voltage source in the model, and vibration and temperature data segments are mapped to the instantaneous vibration state and local temperature state of the corresponding structural components, respectively. Under simulation timing control loops, the model generates predicted near-field radiation data and predicted far-field radiation data in real time, synchronized with physical testing.

[0201] To ensure the high fidelity of the digital twin model, it needs to be corrected. A time-frequency domain comparative analysis is performed on the predicted near-field radiation data, predicted far-field radiation data, and corresponding measured near-field and far-field radiation data from the multi-physics synchronous data stream. This includes performing time-domain cross-correlation calculations and frequency-domain complex transfer function calculations on the predicted and measured data respectively, and constructing a composite error data structure by combining the spatial topological relationships of the measured radiation data. Based on the energy of characteristic frequency bands in the vibration data, error components strongly correlated with the dynamic behavior of the mechanical structure are separated from the composite error data structure to generate a deviation spectrum. This deviation spectrum serves as an optimization objective, used to dynamically correct critical path impedance parameters in the digital twin model (such as the equivalent circuit parameters of the bearing oil film and the housing grounding path, and the coupling coefficient of vibration on contact impedance modulation). Through iterative optimization of the solver, under the constraints of vibration and temperature data, the model parameters are continuously updated until the norm of the deviation spectrum between the predicted and measured radiation data is less than a preset threshold, thus obtaining a high-fidelity digital twin model consistent with the current test state. This correction process resolves the issue of discrepancies between the model and reality in existing technologies, ensuring the accuracy of subsequent analyses.

[0202] Once a high-fidelity digital twin model is obtained, the contribution of shaft current can be quantified. A first set of input conditions is configured in the model, including shaft current data, vibration data, and temperature data. This drives the model to perform a first simulation calculation, yielding a baseline simulation result (i.e., the total radiation spectrum). The component parameters representing the shaft current conduction path are then modified to a high-impedance state (simulating the case where the shaft current is shielded), and a second set of input conditions (containing only vibration and temperature data) without shaft current data is configured. This drives the modified model to perform a second simulation calculation, yielding a comparative simulation result (i.e., the radiation spectrum without shaft current).

[0203] Frequency domain difference calculations were performed on the baseline simulation results and the comparison simulation results. Specifically, point-by-point amplitude subtraction was performed on the baseline simulation spectrum and the comparison simulation spectrum to obtain the axis current radiation component spectrum. To eliminate the baseline error of the model simulation itself, the amplitude of the axis current radiation component spectrum was calibrated based on the baseline simulation spectrum. For example, by calculating the cutoff frequency of the axis current conduction path, the frequency band above the cutoff frequency was determined as the reference frequency band, and the average amplitude of the axis current radiation component spectrum within this reference frequency band was calculated as the baseline error estimate. Then, the amplitude value of the entire frequency band was subtracted from this estimate. The point-by-point amplitude ratio of the calibrated axis current radiation component spectrum and the baseline simulation spectrum was calculated to obtain the contribution spectrum of the axis current to the total radiated emission. This spectrum intuitively shows the specific contribution ratio of the axis current to the total radiated emission at different frequencies, solving the problem that existing technologies cannot quantify the contribution of interference sources, and providing engineers with a quantitative basis for targeted rectification.

[0204] Furthermore, based on this contribution spectrum, sensitive frequency bands dominated by one or more shaft currents can be identified, and the corresponding contribution values ​​can be extracted. Combining this with operating condition and multiphysics state data extracted from the multiphysics synchronous data stream and associated with the occurrence time of the sensitive frequency bands, structured diagnostic records are generated and stored in a radiation contribution case library. When subsequently testing the same or similar electric drive systems, this case library can be queried to optimize test condition planning or initialize model parameters based on matching historical diagnostic records, further improving the efficiency of testing and analysis.

[0205] All of the above-mentioned optional technical solutions can be combined in any way to form the optional embodiments of this application, and will not be described in detail here.

[0206] Figure 9 This is a schematic diagram of an integrated shaft current and radiated emission testing system for an electric drive system provided in an embodiment of this application. See also... Figure 9 The system includes: The acquisition module 901 is used to apply dynamic load excitation based on real road conditions to the electric drive system under test and perform synchronous acquisition of multiple physical quantities to obtain a synchronous multi-physical field data stream. The synchronous multi-physical field data stream includes time-aligned shaft current data, measured near-field radiation data, measured far-field radiation data, vibration data, and temperature data.

[0207] The running module 902 is used to drive the multi-physics digital twin model of the electric drive system to run based on the shaft current data, vibration data and temperature data in the multi-physics synchronous data stream, and generate predicted near-field radiation data and predicted far-field radiation data synchronized with physical testing. The multi-physics digital twin model of the electric drive system contains a complete coupling link from shaft current conduction to space radiation.

[0208] Analysis module 903 is used to perform time-frequency domain comparative analysis of the predicted near-field radiation data, the predicted far-field radiation data, and the corresponding measured near-field radiation data and measured far-field radiation data in the multi-physics synchronous data stream to obtain the deviation spectrum. Based on the deviation spectrum, the critical path impedance parameters in the multi-physics digital twin model of the electric drive system are dynamically corrected to obtain a high-fidelity digital twin model consistent with the current test state.

[0209] Simulation module 904 is used to obtain benchmark simulation results and comparative simulation results based on the high-fidelity digital twin model by performing benchmark simulations containing shaft current paths and comparative simulations of shielded shaft current paths in the model.

[0210] The calculation module 906 is used to perform frequency domain difference calculations on the benchmark simulation results and the comparison simulation results to obtain the spectrum of the contribution of the shaft current to the total radiated emission.

[0211] It should be noted that the integrated shaft current and radiated emission testing system for electric drive systems provided in the above embodiments is only illustrated by the division of the functional modules described above. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the computer device can be divided into different functional modules to complete all or part of the functions described above. Furthermore, the integrated shaft current and radiated emission testing system for electric drive systems provided in the above embodiments and the integrated shaft current and radiated emission testing method embodiments for electric drive systems belong to the same concept. The specific implementation process is detailed in the method embodiments and will not be repeated here.

[0212] Figure 10 This is a schematic diagram of a system structure provided in an embodiment of this application. The system 1000 can vary significantly due to different configurations or performance. It may include one or more Central Processing Units (CPUs) 1001 and one or more memories 1002. The one or more memories 1002 store at least one computer program, which is loaded and executed by the one or more processors 1001 to implement the methods provided in the various method embodiments described above. Of course, the system 1000 may also have wired or wireless network interfaces, a keyboard, and input / output interfaces for input and output. The system 1000 may also include other components for implementing device functions, which will not be elaborated upon here.

[0213] In an exemplary embodiment, a computer-readable storage medium is also provided, such as a memory including a computer program that can be executed by a processor to perform the integrated shaft current and radiated emission testing method for an electric drive system described in the above embodiments. For example, the computer-readable storage medium may be a read-only memory (ROM), a random access memory (RAM), a compact disc read-only memory (CD-ROM), magnetic tape, floppy disk, or optical data storage device, etc.

[0214] In an exemplary embodiment, a computer program product or computer program is also provided, which includes program code stored in a computer-readable storage medium. A processor of a computer device reads the program code from the computer-readable storage medium and executes the program code, causing the computer device to perform the above-described integrated test method for shaft current and radiated emission of an electric drive system.

[0215] In some embodiments, the computer program involved in the present application embodiments may be deployed and executed on a computer device, or executed on multiple computer devices located in one location, or executed on multiple computer devices distributed in multiple locations and interconnected through a communication network. Multiple computer devices distributed in multiple locations and interconnected through a communication network may constitute a blockchain system.

[0216] Those skilled in the art will understand that all or part of the steps of the above embodiments can be implemented by hardware or by a program instructing related hardware. The program can be stored in a computer-readable storage medium, such as a read-only memory, a disk, or an optical disk.

[0217] The above are merely optional embodiments of this application and are not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A method for integrated testing of shaft current and radiated emission in an electric drive system, characterized in that, The method includes: A dynamic load excitation based on real road conditions is applied to the electric drive system under test, and multi-physical quantity synchronous acquisition is performed to obtain a multi-physical field synchronous data stream. The multi-physical field synchronous data stream includes time-aligned shaft current data, measured near-field radiation data, measured far-field radiation data, vibration data, and temperature data. Based on the shaft current data, vibration data, and temperature data in the multiphysics synchronous data stream, the multiphysics digital twin model of the electric drive system is driven to run, generating predicted near-field radiation data and predicted far-field radiation data synchronized with the physical test. The multiphysics digital twin model of the electric drive system includes a complete coupling link from shaft current conduction to space radiation. A time-frequency domain comparative analysis is performed on the predicted near-field radiation data, the predicted far-field radiation data, and the corresponding measured near-field radiation data and measured far-field radiation data in the multi-physics synchronous data stream to obtain a deviation spectrum. Based on the deviation spectrum, the critical path impedance parameters in the multi-physics digital twin model of the electric drive system are dynamically corrected to obtain a high-fidelity digital twin model consistent with the current test state. Based on the high-fidelity digital twin model, benchmark simulation results and comparative simulation results are obtained by performing benchmark simulations containing shaft current paths and comparative simulations with shielded shaft current paths in the model. Frequency domain difference calculations are performed on the baseline simulation results and the comparison simulation results to obtain the spectrum of the contribution of the shaft current to the total radiated emission.

2. The method according to claim 1, characterized in that, The process of applying dynamic load excitation based on real road conditions to the tested electric drive system and performing synchronous acquisition of multiple physical quantities to obtain a synchronous data stream of multiple physical fields includes: On the dynamometer test bench, dynamic load excitation is generated based on the comprehensive road cycle working condition file, and the dynamometer test bench is controlled to apply dynamic torque load including acceleration, deceleration and load change to the tested electric drive system. While applying the dynamic torque load, the shaft current data is collected by a high-frequency current probe configured in the shaft grounding circuit of the electric drive system under test, the measured near-field radiation data is collected by a magnetic near-field probe array arranged at key positions in the housing of the electric drive system under test, the measured far-field radiation data is collected by a broadband antenna set at a standard test distance, and the vibration data and temperature data are collected by a vibration acceleration sensor and a temperature sensor installed on the bearing housing and housing of the electric drive system under test, respectively. The output signals of the high-frequency current probe, the magnetic near-field probe array, the broadband antenna, the vibration acceleration sensor, and the temperature sensor are connected to the same multi-channel synchronous data acquisition system. Under the control of a unified time base clock, they are synchronously sampled and recorded to generate the multi-physics synchronous data stream.

3. The method according to claim 1, characterized in that, The method, based on the shaft current data, vibration data, and temperature data in the multiphysics synchronous data stream, drives the multiphysics digital twin model of the electric drive system to run, generating predicted near-field radiation data and predicted far-field radiation data synchronized with the physical test, including: A multi-physics digital twin model of the electric drive system is constructed. The multi-physics digital twin model of the electric drive system is a parameterized multi-physics coupling model that can be used to represent the coupling relationship between shaft current conduction, mechanical vibration, heat distribution and spatial radiation field. The shaft current data, vibration data, and temperature data in the multiphysics synchronous data stream are used as a set of joint excitation sources and dynamic boundary conditions synchronized in the time domain, and are input into the multiphysics digital twin model of the electric drive system. The multiphysics digital twin model of the electric drive system is synchronously driven to run, so that the multiphysics digital twin model of the electric drive system generates, in real time, the predicted near-field radiation data and the predicted far-field radiation data with the same time reference as the multiphysics synchronous data stream under the drive of the joint excitation source and the dynamic boundary conditions.

4. The method according to claim 3, characterized in that, The construction of the multiphysics digital twin model of the electric drive system includes: A model architecture is constructed that includes a complete coupling path from shaft current conduction to spatial radiation. The complete coupling path includes at least the common-mode voltage source inside the motor, the distributed parameter network between the winding and the housing, the bearing conductive loop, and the motor housing structure. In the model architecture, a multi-physics coupling relationship is established between the electromagnetic response of the complete coupling path, the mechanical vibration response of the motor housing structure, and the temperature distribution of the electric drive system. The complete coupling path and the multiphysics coupling relationship are parameterized to form the parameterized multiphysics coupling model.

5. The method according to claim 4, characterized in that, The establishment of multi-physics coupling relationships between the electromagnetic response of the complete coupling path, the mechanical vibration response of the motor housing structure, and the temperature distribution of the electric drive system within the model architecture includes: In the model architecture, a first coupling relationship is established between the electromagnetic response and the mechanical vibration response. The first coupling relationship is used to represent the dynamic changes of structural parameters and electromagnetic parameters in the complete coupling path caused by the vibration deformation of the motor housing structure. A second coupling relationship is established between the electromagnetic response and the temperature distribution, the second coupling relationship being used to represent the dynamic change of the material's electromagnetic properties in the complete coupling path caused by the change in the temperature distribution; Based on the first coupling relationship and the second coupling relationship, a coupled field equation is constructed to describe the mutual modulation effect among the electromagnetic response, the mechanical vibration response and the temperature distribution, so as to complete the establishment of the multi-physics coupling relationship.

6. The method according to claim 1, characterized in that, The time-frequency domain comparison analysis of the predicted near-field radiation data, the predicted far-field radiation data, and the corresponding measured near-field radiation data and measured far-field radiation data in the multi-physics synchronous data stream yields the deviation spectrum, including: For the predicted near-field radiation data and the measured near-field radiation data, as well as the predicted far-field radiation data and the measured far-field radiation data, time-domain cross-correlation calculation and frequency-domain complex transfer function calculation are performed respectively to obtain a set of time-domain correlation functions and a set of frequency-domain complex transfer functions. By utilizing the set of time-domain correlation functions and the set of frequency-domain complex transfer functions, and introducing the spatial topological relationship between the measured near-field radiation data and the measured far-field radiation data, a composite error data structure is constructed that simultaneously represents the consistency of model prediction errors in the time domain, the amplitude and phase in the frequency domain, and the spatial distribution characteristics. Based on the energy of characteristic frequency bands in the vibration data of the multi-physics synchronous data stream, error components strongly correlated with the dynamic behavior of the mechanical structure are separated from the composite error data structure to generate a deviation spectrum. The deviation spectrum is used to directionally correct the structural dynamic coupling parameters in the multi-physics digital twin model of the electric drive system.

7. The method according to claim 1, characterized in that, The process of dynamically correcting the critical path impedance parameters in the multiphysics digital twin model of the electric drive system based on the deviation spectrum to obtain a high-fidelity digital twin model consistent with the current test state includes: Using the deviation spectrum as the optimization objective, the critical path impedance parameter in the multiphysics digital twin model of the electric drive system is defined as the variable to be optimized, and a model parameter inversion problem is constructed. By using the vibration data and temperature data in the multiphysics synchronous data stream, the solution space of the model parameter inversion problem is constrained to limit the critical path impedance parameter to a physically reasonable range of dynamic changes. An optimization solver is used to iteratively solve the model parameter inversion problem. In each iteration, the critical path impedance parameters are updated and the predicted radiation data is recalculated until the norm of the deviation spectrum between the predicted radiation data and the measured radiation data is less than a preset threshold. At this point, the high-fidelity digital twin model is obtained.

8. The method according to claim 1, characterized in that, Based on the high-fidelity digital twin model, benchmark simulation results and comparative simulation results are obtained by performing a benchmark simulation including the shaft current path and a comparative simulation with a shielded shaft current path in the model, including: In the high-fidelity digital twin model, a first set of input conditions is configured, which includes the shaft current data, the vibration data, and the temperature data in the multi-physics synchronous data stream; Based on the first input condition set, the high-fidelity digital twin model is driven to perform the first simulation calculation and obtain the benchmark simulation results; In the high-fidelity digital twin model, the component parameters used to represent the shaft current conduction path are modified to a high-impedance state, and a second input condition set that does not include the shaft current data is configured. The second input condition set includes the vibration data and the temperature data. Based on the high-fidelity digital twin model with modified driving parameters driven by the second input condition set, a second simulation calculation is performed to obtain the comparative simulation results.

9. The method according to claim 1, characterized in that, The frequency domain difference calculation of the baseline simulation results and the comparison simulation results to obtain the contribution spectrum of the axis current to the total radiated emission includes: A point-by-point amplitude subtraction operation in the frequency domain is performed on the benchmark simulation spectrum contained in the benchmark simulation result and the comparison simulation spectrum contained in the comparison simulation result to obtain the axis current radiation component spectrum. Based on the benchmark simulation spectrum, the amplitude of the axis current radiation component spectrum is calibrated to eliminate the baseline error in the model simulation itself, and the calibrated axis current radiation component spectrum is obtained. The contribution spectrum of the axial current to the total radiative emission is obtained by calculating the point-by-point amplitude ratio between the calibrated axis current radiation component spectrum and the reference simulation spectrum.

10. An integrated testing system for shaft current and radiated emission in an electric drive system, characterized in that, The system includes: The acquisition module is used to apply dynamic load excitation based on real road conditions to the electric drive system under test and perform synchronous acquisition of multiple physical quantities to obtain a synchronous multi-physical field data stream. The synchronous multi-physical field data stream includes time-aligned shaft current data, measured near-field radiation data, measured far-field radiation data, vibration data, and temperature data. The running module is used to drive the multi-physics digital twin model of the electric drive system to run based on the shaft current data, vibration data and temperature data in the multi-physics synchronous data stream, and generate predicted near-field radiation data and predicted far-field radiation data synchronized with physical testing. The multi-physics digital twin model of the electric drive system contains a complete coupling link from shaft current conduction to space radiation. The analysis module is used to perform time-frequency domain comparative analysis on the predicted near-field radiation data, the predicted far-field radiation data and the corresponding measured near-field radiation data and measured far-field radiation data in the multi-physics synchronous data stream to obtain the deviation spectrum. Based on the deviation spectrum, the critical path impedance parameters in the multi-physics digital twin model of the electric drive system are dynamically corrected to obtain a high-fidelity digital twin model consistent with the current test state. The simulation module is used to obtain benchmark simulation results and comparative simulation results based on the high-fidelity digital twin model by performing a benchmark simulation containing the shaft current path and a comparative simulation of the shielded shaft current path in the model. The calculation module is used to perform frequency domain difference calculation on the benchmark simulation results and the comparison simulation results to obtain the spectrum of the contribution of the shaft current to the total radiated emission.

Images