Accurate fitting of Preisach model parameters using GMM with non-uniform grid partition

Abstract The inverse Preisach hysteresis model plays an important role in simulating the hysteresis characteristics of magnetic materials and calculating their losses. The uniformly distributed centered cycle (UDCC) method is commonly used to determine the distribution function of the inverse Preisach model, but its accuracy is low when simulating hysteresis characteristics at low magnetic densities. This paper proposes an accurate fitting method of Preisach model parameters using Gaussian mixture model (GMM) with non-uniform grid partition. This method includes two innovative works: first, a hybrid sensitivity function associated with the curvature and slope of the curve is constructed to determine key regions. Additionally, the method proposes using GMM to approximate the Preisach density function and fit the limit rotational curve. Genetic algorithms are used to optimize the GMM parameters. Compared to UDCC method, this approach does not require more measured hysteresis loop data and reduces the computational load of the distribution function matrix. Experimental and simulation results show that the proposed method achieves higher fitting accuracy in identifying the inverse Preisach model distribution function compared to UDCC method, validating the effectiveness of the proposed method.