Backlash detection of two-mass rotation system using Artificial Neural Network (AAN) / Muhammad Amier Rizal Khairudin

In this study, NFTOOL with Levenberg-Marquardt (trainlm) algorithm has been used to detect the presence of backlash in a vehicle driveline system model. The database consists of data with and without backlash that was collected from ECP Model 220 industrial plant emulator. The same data was first filtered using a Butterworth filter then was used as input and target to develop ANN models. The model was trained, tested and validated using MATLAB R2013a software. The result was analyzed based on the mean square error (MSE) and regression (R) values. MSE and R value is obtained for different sampling time and compared. It was found that MSE value for dataset without backlash is smaller than dataset with backlash. While R value for dataset without backlash is better than dataset with backlash.