Digital Twins for the Thermal Analysis of Permanent Magnet Synchronous Motors

Abstract

Thermal phenomena hold vital significance in automotive applications involvingpermanent-magnet synchronous motors (PMSMs). The accumulation of heat invulnerable regions, such as magnets or end windings, has the potential to impair device performance and inflict severe damage to components under extreme circumstances.Obtaining precise knowledge of the device temperatures is vital forimplementing preemptive measures. However, conducting direct measurements tomonitor thermal effects during production is often unfeasible. As an alternative,the development of real-time digital twins (DTs), capable of accurately representingthe motor’s thermal effects using a limited set of sensor data, emerges as a viablesolution. One of the approaches to building these electric motor DTs for thermalanalysis is lumped parameter thermal networks (LPTNs). LPTNs feature an excellenttradeoff between computational cost and accuracy and are an effective optionto model thermal phenomena in these devices and serve as the starting point forthe development of DTs. The key to obtaining a high-fidelity LPTN model residesin the identification of its parameters. This task is addressed with the developmentof custom LPTN architectures and methods to identify their parameters. Diversemodels have been built and their topology has evolved driven by the requirementsof the application of these models for DT formulation for thermal analysis. Additionally,some new alternatives arose in the context of this work, which were alsoinvestigated.However, the underlying models in the DTs may experience modifications overtime. Thermal ageing is one of the main causes of the gradual deterioration of theinsulation system in electric motors, which leads to changes in the thermophysicalproperties of its dielectric materials. This effect considerably decreases the expectedlifetime of the insulation. Hence, thermal ageing should be considered in order to increasethe fidelity of the LPTN models used in DTs of electric motors. This challengewas tackled with the execution of two experimental campaigns and the generationof a deterioration model to replicate the variations in the eMotor throughout itslifecycle.Moreover, Model-Based System Testing (MBST) is gaining traction in the developmentcycle, especially in the automotive sector. This innovative approachcombines computer simulations and real-world testing to detect and address issuesearly on. By minimizing resource expenditure on problem-solving and design revisions,MBST enables the timely detection of flaws before finalizing a full vehicleprototype. Under the vision of this approach, cyber-physical test benches (CPTBs)emerge as useful platforms for DT implementation. In the context of this thesis,two CPTBs have been built, one in GKN Automotive Zumaia and another one inthe Laboratorio de Ingenier´ıa Mec´anica at the University of La Coru˜na. These environmentsare vital for the tuning of DTs and are intensely advantageous for latervalidation during the operation of eMotors.