Hybrid-Driven Digital Twin Modelling Framework for an EV Propulsion Drive System

Abstract

Digital twin (DT) plays a vital role across various applications, notably in electric vehicles (EVs). It serves as a virtual counterpart to physical systems. Repurposing legacy EV propulsion systems—those developed prior to the rise of connected vehicle technologies—can reduce electronic waste and support sustainability goals. However, adapting DTs in such systems is challenging due to limited connectivity, incomplete schematics, and performance degradation over time. This paper presents a hybrid-driven DT modelling framework for an EV repurposed with a legacy propulsion system, where some components, like the drive system, have uncertain parameters. A physics-based model is developed for the motor, leveraging its well-defined parameters, while a data-driven model is applied to the drive system due to its uncertainty. The data-driven model was developed using a nonlinear autoregressive neural network with exogenous inputs (NARX). It was trained on physical test bench data and achieved a validation RMSE of 0.04 on unseen data. A hybrid-driven model, combining the NARX-based drive system with a physics-based motor model, was first validated offline in MATLAB/Simulink, then deployed on a speedgoat baseline target machine for real-world testing. The deployment enabled validation under real world vehicle conditions beyond the test bench and assessment of its real-time capability. Real-time testing demonstrated high steady-state accuracy and reliable performance, with an average execution cycle of 8 ms, 60% CPU load, and 300 MB memory usage. Communication via user datagram protocol confirmed the model's real-time readiness and suitability for practical DT integration.

1. Introduction

Digital twin (DT) plays a pivotal role across various engineering and industrial applications, particularly in the field of electric vehicles. Grieves [1] defines digital twin as a set of virtual information that fully describes a potential or actual physical manufactured product from the micro to the geometrical level. This virtual counterpart can contain all information about the product that could be obtained during its lifecycle phases—design, production, operation, and recycling [2]. In recent years, digital twins have become a cornerstone of smart manufacturing and industrial IoT, enabling real-time monitoring, predictive maintenance, and simulation [3].

Electric vehicles represent a key application domain for digital twins. With the global shift towards sustainable transportation, EV adoption has increased significantly [4]. However, many legacy propulsion systems—those developed prior to the rise of connected vehicle technologies—lack the ability to directly integrate with modern digital twin platforms. These systems suffer from limited connectivity, incomplete technical documentation, and performance degradation over time, making their repurposing a significant challenge [5].

This paper presents a hybrid-driven digital twin modelling framework that combines physics-based and data-driven modelling to overcome the challenges of repurposing legacy propulsion systems. A physics-based model is developed for the motor, where parameters are well-defined, while a data-driven model (NARX neural network) is employed for the drive system with uncertain parameters. The hybrid model was first validated via MATLAB/Simulink simulation, then deployed on a Speedgoat platform for real-time testing under actual vehicle conditions.

2. Hybrid Framework Methodology

The proposed framework is based on integrating two complementary models: a physics-based model (PBM) for the motor, and a data-driven model (DDM) for the drive system. This hybrid approach aims to leverage the advantages of both methods: the physical accuracy of the physics-based model, and the adaptive flexibility of the data-driven model in handling uncertain systems.

2.1 Physics-Based Motor Model

The physics-based model was developed using conventional motor equations, leveraging its well-defined parameters. The model incorporates the electrical and mechanical equations of a permanent magnet synchronous motor (PMSM), the most common type in EV applications. Parameters were derived from the motor datasheet and laboratory measurements.

2.2 Data-Driven Drive System Model

The data-driven model was developed using a nonlinear autoregressive neural network with exogenous inputs (NARX). NARX was chosen for its ability to model nonlinear dynamic systems while maintaining a relatively simple architecture. The network was trained on physical test bench data including voltage, current, torque, and speed measurements under different operating conditions. The network achieved a validation RMSE of 0.04 on unseen data, demonstrating high predictive accuracy.

2.3 Model Integration

The two models were integrated into a single hybrid framework, where the data-driven model (drive system) supplies the necessary inputs to the physics-based model (motor). The integrated framework was validated offline in MATLAB/Simulink before deployment on actual hardware.

3. Experimental Setup and Test Platform

Experiments were conducted on a physical test bench consisting of a permanent magnet synchronous motor (PMSM) and associated drive system. The platform included sensors for measuring voltage, current, speed, and torque, along with a data acquisition system for communicating with the MATLAB/Simulink environment.

The test setup comprises:

• A 4 kW PMSM with integrated drive system
• Speedgoat Baseline Target Machine for real-time operation
• UDP communication interface for data transfer between the physical system and virtual model
• Host computer running MATLAB/Simulink for development and monitoring

4. Results and Validation

Real-time testing results demonstrated excellent performance of the proposed hybrid model. The model was validated under various operating conditions, including steady-state and transient cases, as well as scenarios simulating real driving conditions.

4.1 Real-Time Performance

The hybrid model demonstrated excellent real-time capability with an average execution cycle of only 8 ms. CPU load was approximately 60%, with memory usage of 300 MB—well within acceptable limits for real-time digital twin applications on embedded hardware.

Execution Cycle = 8 ms | CPU Load = 60% | Memory = 300 MB

4.2 Steady-State Accuracy

The model exhibited high steady-state accuracy, with the hybrid model outputs closely matching actual measurements from the physical system. The average speed estimation error was less than 2%, while the torque estimation error was less than 3%. This strong performance is attributed to the effective combination of physics-based and data-driven modelling.

4.3 Transient Performance

Under transient conditions—such as sudden acceleration or load changes—the hybrid model showed fast and accurate response. The hybrid model outperformed both standalone models (physics-only and data-only) in capturing the system's dynamic behaviour, confirming the effectiveness of the hybrid approach in modelling complex systems with uncertain parameters.

5. Discussion

The proposed hybrid framework offers several advantages over traditional approaches. First, the hybrid approach allows repurposing of legacy propulsion systems without complete replacement, contributing to reduced electronic waste and supporting sustainability goals. Second, the combination of physics-based and data-driven modelling provides a balance between accuracy and flexibility, where the physics-based model provides a sound physical foundation while the data-driven model adapts to system uncertainties.

However, some limitations must be acknowledged. The performance of the data-driven model depends on the quality and quantity of training data. In cases where training data is limited, the data-driven model may not achieve the desired performance. Additionally, the hybrid model requires careful calibration of the physics-based model to ensure result accuracy.

6. Conclusions and Future Research

This paper presented a hybrid-driven digital twin modelling framework for legacy EV propulsion drive systems. The framework combines physics-based motor modelling with data-driven (NARX) drive system modelling. The framework was validated through simulation and real-time testing on a Speedgoat platform, demonstrating high accuracy and reliable performance in both steady-state and transient conditions.

Future research directions include:

• Extending the framework to include additional EV propulsion components such as battery and charger
• Exploring advanced neural network architectures, such as CNNs or LSTMs, to improve data-driven modelling accuracy
• Integrating the hybrid model into a full digital twin lifecycle management system for comprehensive predictive maintenance
• Evaluating framework performance across a fleet of EVs under diverse operating conditions and environments
• Developing a graphical user interface to facilitate hybrid model configuration and deployment

Acknowledgments

The authors acknowledge the supporting institutions that contributed to the completion of this research, including laboratory facilities and funding organizations. We also thank the editors and reviewers for their valuable feedback.

References

[1] Grieves, M. (2014). Digital Twin: Manufacturing Excellence through Virtual Factory Replication. White Paper.

[2] Tao, F., Zhang, H., Liu, A., & Nee, A.Y.C. (2019). Digital twin in industry: State-of-the-art. IEEE Transactions on Industrial Informatics, 15(4), 2405–2415.

[3] Rasheed, A., San, O., & Kvamsdal, T. (2020). Digital twin: Values, challenges and enablers from a modeling perspective. IEEE Access, 8, 21980–22012.

[4] International Energy Agency (2024). Global EV Outlook 2024.

[5] Ibrahim, M., Rassõlkin, A., Vaimann, T., & Kallaste, A. (2022). Overview on Digital Twin for Autonomous Electrical Vehicles Propulsion Drive System. Sustainability, 14(2), 601.

[6] Ibrahim, M., Rassõlkin, A., Vaimann, T., Kallaste, A., Zakis, J., Hyunh, V.K., & Pomarnacki, R. (2023). Digital Twin as a Virtual Sensor for Wind Turbine Applications. Energies, 16(17), 6246.

[7] Ibrahim, M., Rjabtšikov, V., & Gilbert, R. (2023). Overview of Digital Twin Platforms for EV Applications. Sensors, 23(3), 1414.