Permanent Magnet Synchronous Machine Flux and Inductance Estimation Using Experimental Data and Gaussian Process Regression

Accurate flux and inductance estimation is crucial for high-fidelity modelling and emulation of interior permanent magnet synchronous machines (IPMSMs). This paper presents a systematic workflow for extracting d-and q-axis flux and inductance look-up tables (LUTs) from full-factorial experimental data using voltage equations derived from the IPMSM equivalent circuit model. The workflow begins with data acquisition from an IPMSM testbed, capturing current, voltage, speed, torque, and temperature across a wide operating range. Using the IPMSM voltage equations, the d- and q-axis flux linkages and inductances are computed while accounting for temperature-dependent resistance variations. Gaussian Process Regression (GPR) is then employed to interpolate and extrapolate flux values over an extended operating range, ensuring accurate LUT generation. The final flux and inductance LUTs are formatted for direct integration into electric machine emulators, enabling real-time validation and optimisation of electric drive control strategies. Experimental validation confirms the accuracy and reliability of the proposed approach, demonstrating its potential for hardware-in-the-loop (HIL) testing, virtual prototyping, and control system development in electrified powertrains.