Parameter identification of a brushless motor drive system using a modified version of the fast simulated diffusion algorithm

The implementation of an adaptive version of the fast simulated diffusion (FSD) algorithm, as a practical optimization tool in system identification, for optimal dynamical parameter extraction of a high performance brushless motor drive system (BLMDS) with a multiminima objective function is presented. A considerable reduction in the number of least squares cost evaluations is achieved with the modified FSD search for global minimum convergence resulting in substantial savings in computation time. The requirement for an accurate physical model of BLMDS nonlinear operation including inverter blanking as part of the identification strategy is briefly discussed. The restriction of the search procedure to quantized parameter space, because of false `local´ minima proliferation with cost surface noise near the global extremum, is briefly examined. The effectiveness of the modified FSD search technique, with remote initialization, in reaching the global minimizer for a range of known drive inertial loads is demonstrated. This is achieved by the use of step response current feedback, responsible for genuine local minima plurality, in the objective function formulation. Frequency and phase coherence of the drive model simulation, based on returned FSD parameter estimates, with BLMDS test data attest to FSD convergence accuracy which validates this parameter identification method