Compensatory fuzzy neural network control with dynamic parameters estimation for linear voice coil actuator

The object of this study is to develop an intelligent control strategy, which comprises a compensatory fuzzy neural network (CFNN) controller with a dynamic particle swarm optimization (DPSO) based estimator, for on-line parameter estimation and control of a linear voice coil actuator (VCA). Because the plant Jacobian of the VCA is nonlinear and time-varying, it is difficult to derive the learning algorithm for the CFNN by using the conventional back-propagation (BP) method directly. Therefore, it is strongly desirable that an on-line manner can provide a reasonably good estimation of the plant Jacobian in the practical applications. In this study, the operating principle and dynamic analysis of the VCA are introduced first. Subsequently, the algorithms of the DPSO and CFNN are given where the DPSO and CFNN are utilized to obtain the control signal and estimate the plant Jacobian, respectively. Moreover, a convergence analyses is given to derive specific learning rates for ensuring the convergence of the control error. Finally, the proposed control strategy is implemented on a 32-bit floating-point digital signal processor (DSP) for experimental verification. Experimental results demonstrate the improved tracking performance and robustness of the proposed CFNN-DPSO controller with online Jacobian estimation compared with the conventional CFNN controller with constant one, for the VCA control system.