Neural Network Models on the Prediction of Tool Wear in Turning Process: A Comparison Study

M.S. Alajmi (UK), S.E. Oraby (Kuwait), and I.I. Esat (UK)


: Tool wear, Neural Networks, Turning Process


A reliable estimation of tool life in metal cutting process is the aim of many researchers. In this study tool life prediction modelled using different types of neural networks. A comparison study is conducted for Feed forward Backpropagation Neural Network (FFBPNN), General Regression Neural Networks (GRNN), Elman Network (EN), and Radial Base Function Networks (RBFN) to find the best model. Tool life data used in the study was obtained previously, which included cutting speed, feed rate, depth of cut, cutting time, feed force (Fx), radial force (Fy), vertical force (Fz), notch wear, nose wear, and flank wear. A graphical study of the data reveals high non-linearity and early experiments carried out in this study using simple backpropagation network gave only marginally acceptable results. All neural networks models in this study are trained by the same experimental data acquired using Central Composite Design (CCD), which is one of the methods of experimental design (DOE). The report presents a competitive study of the performance of these networks for the tool life prediction problem. It is shown that FFBP is the best model among the others.

Important Links:

Go Back