Modeling of tool wear in machining of AISI 52100 steel using artificial neural networks

Abstract

Abstract Machining is an arduous process for both the cutting tool and workpiece material. Conducting a series of machining tests involves a lot of money and time. It is important to prevent time consuming runs and prioritize technology. Hence, in this investigation, artificial neural network (ANN) approach is employed to estimate the cutting tool wear. From the experimentally obtained values, the input variables chosen for ANN modeling are cutting speed, feed and depth of cut corresponding to the output variable tool wear. The ANN modeling comprises Levenberg-Marquardt (trainlm) and tangent sigmoid (tansig) as training and testing functions respectively. Optimum architecture 3–6-6–1 is obtained based on the mean sum squared error (MSE) and average error (AE) of input data. Tool flank wear is assessed in three different machining environments such as dry machining, wet machining (flood cooling) and minimum quantity lubrication (MQL) machining. The emulsifier oil based cutting fluid is used as cutting fluid in wet and MQL machining. ANN model predicted values are compared with experimental results. Based on the correlation coefficient factor, it was observed that the prediction of ANN is high, showing a significant harmony between predicted response and experimental output.