Identification of Acceleration Harmonics for a Hydraulic Shaking Table by Using Hopfield Neural Network

The paper aims to develop a harmonic identification scheme for a hydraulic shaking table's sinusoidal acceleration response. Nonlinearities are inherent in a hydraulic shaking table. Some of them are dead zone of servo valve, backlash and friction between joints, and friction in actuator. Nonlinearities cause harmonic distortion of the system shaking response when it corresponds to a sinusoidal excitation. This lowers the system control performance. An efficient, time-domain acceleration harmonic identification is developed by using Hop field neural network. Due to the introduction of energy function used to optimize the computation for the identification harmonic method, the fully connected, single-layer feedback neural network does not require training in advance and is able to identify harmonic amplitudes and phase angles. Each harmonic, as well as the fundamental response, can be directly obtained. Simulations and experiments show very promising results that the proposed scheme is really applicable to identify harmonics with high precision and good convergence. Comparisons between the presented method and another method are carried out to further demonstrate its efficiency. (C) 2018 Sharif University of Technology. All rights reserved.