Surrogate Vibration Modeling Approach for Design Optimization of Electric Machines

Abstract

This article presents a surrogate modeling approach to predict the vibrational behavior during electric machine design. The methodology used in identifying and implementing the surrogate model allows a time-efficient integration of the surrogate model in multiphysics geometric optimization of electric machines, especially in scenarios of drop-in replacement. Based on the state-of-the-art review, the use of analytical or numerical techniques alone considering motor frame and support are not adequately accurate, time efficient, and flexible. The proposed method uses the 3-D finite-element (FE) structural simulations for identifying characteristic response profiles in the surrogate model and does not use structural FE simulations in the optimization loop. Experimental tests are performed on a 6/10 switched reluctance machine to validate the 3-D FE model used to identify the surrogate model. The rational fraction polynomial (RFP) method is used to fit the vibration impulse responses for a number of machines geometries defined by three stator dimensions. The impulse response of any geometry can be determined using linear interpolation of the RFP coefficients. Using the radial forces determined from the FE electromagnetic simulation, the deformation is determined using superposition.