Prediction of Time-series Friction Data using ANFIS

Modelling is frequently used in science and industry. Friction, wear, and corrosion issues are the main design criteria in peanut kernel grading machines. In this study, the time-series of friction force data is modelled with adaptive neuro-fuzzy inference system (ANFIS). Machine learning focuses on developing models for prediction and classification without explicit programming. The data on the friction force is obtained from a simulation based on the discrete element method. The simulation takes 63 days, 18 hours and 27 minutes to calculate the real time of 60 seconds. A Takagi-Sugeno type ANFIS network is constructed. The network is clustered using grid partitioning method. ANFIS helps to optimise machine performance by modelling friction data. In the obtained peanut kernel classification model, the correlation value is 0.799 and the root of the mean square error is 0.514 N. The percentage of the mean absolute error is found to be 1.666%. 100 iterations are run. Calculations take 20.7 seconds. The model has a high linear relationship. It is also observed that the ANFIS network eliminates the need for any pre-processing of the data. Background of the network used, its hyper-parameters, and the prediction performance are presented in the study.