The aim of this work is to develop a modular digital twin of an electric powertrain based on novel machine learning (ML)-based model structures and a systematic, component-oriented architecture. This results in a modular, real-time-capable digital twin that precisely maps the efficiency characteristics of the entire drivetrain and can be used in development and testing processes. The further goal here is to enable virtual testing, which can be used for frontloading and thus both prevent errors and increase the speed of product development. Based on a comprehensive set of measurement data from a test bench, a multi-stage procedure is implemented that integrates data acquisition, physically informed feature selection, modeling at the component and subsystem level, and hybrid coupling strategies. The digital twin captures inverter, electric machine, and mechanical transmission stages and generates physically consistent predictions of key variables such as torque, speed, power factors, and (subsystem) efficiencies. The methodology developed encompasses various model variants, from white-box and black-box to simplified bright-grey-box approaches.
The article addresses existing limitations of classic simulation models and conventional data-driven approaches, which often achieve either short computation times or high model complexity, but only limited prediction quality. The methodology provided uses new model structures that explicitly integrate physical dependencies while also using ML models to map nonlinear effects. The hybrid architectures presented have been shown to significantly reduce the measurement effort while achieving nearly identical model quality and surpassing purely physics-based models in terms of accuracy, robustness, and real-time capability. The study shows that bright-grey-box-models in particular offer an optimal compromise between information content, error quality, and synchronization rate, thus representing a methodological advance over conventional digital twins, which are often created at the component level. The shown methodology provides a directly implementable framework for digital twins of electric powertrains.
The results allow the thesis and show based on experience that the use of the developed digital twin support key objectives including increased efficiency, real-time capability, frontloading in the development process, and the transferability of digital methods to industrial test environments. In addition, the modular modeling concept opens up perspectives for condition monitoring, virtual sensor technology, and adaptive operating strategies.