Abstract:
This work proposes a two stage prediction approach for the estimation of non-stationary machine variables through an optimum and generalized model imbibing real time data uncertainties. The prediction of machine speed and controller set point has been made using the proposed model for a three-phase induction motor operating on a single loop speed control with AC drive and PI controller. The trend of the machine variables has been extracted and added upon the Auto Regressive Moving Average (ARMA) time series prediction at stage one. ARMA prediction has been carried out using different combinations of Auto Regressive (AR) and Moving Average (MA) methods in order to obtain prediction results with less Mean Squared Error (MSE). The resulting prediction error indicates the inadequacy of the model to estimate the data characteristics which has been resolved at the subsequent stage by cascading an adaptive Least Mean Square (LMS) FIR filter to the time series model. The adaptive filter receives the predicted output including training data and iteratively adjusts its coefficients for zero error convergence. This has been tested for different parameter settings of step size and iterations at a specified filter length. The inclusion of adaptive filter in cascade also models the unknown real time factors influencing the system operation in an optimum and adaptive manner from the data available rather than the physical or fixed assumptions. The prediction accuracy of the model proposed has been compared with the existing technique of linear adaptive filter prediction using MSE as a comparison index. The wide difference in the MSE values of the prediction results obtained from the proposed and existing methods substantiates the efficiency of the proposed model in predicting time varying machine variables for better maintenance.
