Optimization of different non-traditional turning processes using soft computing methods

In this study, different non-traditional turning operations were investigated using various soft computing methods. In these operations, cutting speed, machining method, material type and tool overhang lengths were used as machining inputs. Surface roughness, stable cutting depths and maximum cutting tool temperatures were considered as machining outputs. In the first stage, artificial neural network, classification and regression tree (CART) and support vector machine models were developed to predict these outputs. In the second stage, an optimization study (regression analysis) was conducted. CART model produced better prediction results compared to the other methods. In CART models; 0.991, 0.998 and 0.959 values of correlation coefficients were calculated for the prediction of surface roughness, stable cutting depth and maximum cutting tool temperatures, respectively. In the optimization study, ultrasonic assisted/hot ultrasonic assisted turning methods, a tool overhang length of 60 mm and a cutting speed of 10 m/min provide optimum conditions. The proposed soft computing models will help to understand the effect of various parameters in non-traditional machining methods. These models will give a preliminary idea before the experiments. These models can be used as an alternative instead of 2D finite element machining simulations. Less analysis time is required compared to the finite element simulations.