Atomistic study on the effect of the size of diamond abrasive particle during polishing of stainless steel

Abstract

Prediction of material removal in any machining process is usually based on the input machining parameters. However, apart from controllable parameters, there are various other parameters that needs to be monitored in real time to ensure better prediction of accuracy, especially in random processes. Hence, real time data monitoring using appropriate sensors in machining processes is extremely important as the input parameters cannot predict the output with high efficiency. In Micro EDM (MEDM), real time signal monitoring can yield various time domain features of individual current and voltage pulses that can help to enhance the prediction accuracy of material removal. In this study, an attempt has been made to predict the material removal in single spark MEDM based on two different modelling approaches i.e. multiple linear regression (MLR) and classification and regression tree (CART). A total number of 21 experiments were conducted on a specially designed single spark MEDM machine with input parameters viz. voltage and capacitance. Material removal measurements was carried out using Coherent Correlation Interferometer. Open source software “R-3.4.0” was used for building and prediction of the model. A total of 14 predictors (2-input and 12-time domain extracted predictors) and a single output i.e. material removal was used for prediction. Prediction model by multiple linear regression (MLR) showed root mean square error of 5.82 whereas that by CART showed 12.07. Hence, material removal in single spark MEDM can be predicted by MLR with better accuracy as compared to CART.