Framework for Analytical-based Digital Twin Development of Electric Drives Using Slotless PM Machines

Abstract

This Ph.D. thesis investigates and develops analytical models and techniques to build the foundation for the digital twin development of electric drives using slotless permanent magnet (PM) machines. The idea is to rely on analytical models’ intrinsic computational efficiency to meet the digital twins’ fundamental requirement. Despite the required multi-disciplinarity hidden behind the development of the digital replica of an electrical drive, this research is limited to developing a generalized electromagnetic model of inverter-driven slotless machines. To implement a generally applicable model using accurate state-of-the-art (SotA) modeling techniques, several challenges were highlighted during implementation. Therefore, the focuswas on ensuring the uttermost reproducibility of the proposed modeling techniques. Nevertheless, the presented models can offer a sound tool in the design phase of inverter-driven slotless machines, thus highlighting a promising potential for digital twin development. The analytical electromagnetic models of slotless PM machines are based on solutions to Maxwell’s equation through Fourier series decomposition. These models give a noteworthy balance between accuracy and computational time, making them a desirable solution for design and analysis purposes and digital twin implementation. The different formulations allow for magneto-static and magneto-quasi-static modeling. The resulting field formulations have been heavily exploited throughout the research, giving a solid foundation for design purposes. The magnetostatic field formulation, in particular, is employed to estimate torque, inductance, and induced voltage. The same formulation is also employed for iron loss estimate in the stator core by using four different laminations’ loss models with the support of experimental measurements; moreover, an improved optimization algorithm includes the possibility of accounting for magnet demagnetization. On the other hand, the magneto-quasi-static formulations were investigated to calculate the induced eddy currents in the magnets. Every field formulation has been thoroughly validated through FEA to ensure the best accuracy achievable by the analytical models. The same validation step highlighted some significant flaws in the mathematical expressions if directly implemented as executable code causing accuracy reduction due to numerical inaccuracies. These numerical challenges are fully addressed in this work, and the field formulations show to deliver results with the same accuracy as FEA. In support of future development around these modeling techniques, all the implemented programs and algorithms are made available as complementary accessories to the presented work.