Evaluation of Cutting Tool Vibration and Surface Roughness in Hard Turning of AISI 52100 Steel: An Experimental and ANN Approach

Abstract

Hardened steels are being extensively used in aerospace, automobile industries, bearing and die industries. High hardness of such steels makes it difficult to machine. During turning process of such materials, the cutting tool is subjected to heavy mechanical load and thus creates vibrations throughout machining process. It affects the surface quality of machined part and provokes higher rate of tool wear with lowering tool life. Therefore, measurement and prediction of vibration induced is of prime importance. The aim of this paper is to evaluate vibration acceleration and surface roughness with varying machining parameters such as cutting speed, feed and depth of cut to develop predictive mathematical model. The central composite rotatable design (CCRD) method is used in designing the experimental runs. The experimental results are further used to develop mathematical models using regression analysis. It is performed using Design Expert tool. The ANN model is developed using MATLAB tool and the predictions are obtained with acceptable deviations. The comparison of predictive model with experimentation is performed to report the deviation. The examination of the outcomes revealed that the cutting conditions are having prominent and mixed-type effect on vibration signals. The regression and ANN models have been found to be acceptable for prediction of vibration induced and the surface roughness. The coefficient of regression ($$R^{2}$$) is found to 0.92 which shows that the developed mathematical models have a good approximation in correlating the effect of cutting parameters on vibration of a cutting tool. The obtained correlations are verified by conformity test and have reported the close degree of agreement with respect to experimental values. It registered a lowest deviation of 3.3%. The ANN model is effective in reproducing experimental results through simplifying the complex machining process. The investigation reports predictions of ANN are more accurate than regression analysis. The surface roughness predictions agreed well with experimental results and registered the acceptable deviation of 4.33% using regression analysis and 1.37% using ANN approach.