Uncertain nonlinear system identification using Jaya-based adaptive neural network

The piezoelectric actuator has been receiving tremendous interest in the past decade, due to its broad applications in areas of micro-robotics, neurosurgical robot, MEMS, exoskeleton, medical applications, and other applications. However, the hysteresis nonlinearity widely existing in smart materials yields undesirable responses, which make the hysteresis control problem even more challenging. Therefore, many studies based on artificial neural networks have been developed to cope with the hysteresis nonlinearity. However, the back-propagation algorithm which is popular in training a neural network model often performs local optima with stagnation and slow convergence speed. To overcome these drawbacks, this paper proposes a new training algorithm based on the Jaya algorithm to optimize the weights of the neural NARX model (called Jaya-NNARX). The performance and efficiency of the proposed method are tested on identifying two typical nonlinear benchmark test functions and are compared with those of a classical BP algorithm, particle swarm optimization algorithm, and differential evolution algorithm. Forwardly, the proposed Jaya-NNARX method is applied to identify the nonlinear hysteresis behavior of the piezoelectric actuator. The identification results demonstrate that the proposed algorithm can successfully identify the highly uncertain nonlinear system with perfect precision.