A hybrid machining simulator based on predictive machining theory and neural network modelling

A machining simulation system based on a hybrid machining model integrating the predictive machining theory developed by Oxley and neural network mode ls for predicting machining characteristic factors is presented in this pap er. The model consists of two components, an analytical component and a neu ral network component. The analytical component uses Oxley's predictive mac hining theory, from which the essential machining characteristics such as c utting forces, temperature in the cutting region and chip geometry can be p redicted from the input data of the fundamental properties of the workpiece material, tool geometry and cutting conditions, taking into account the ef fect of strain, strain rate and temperature on chip formation. The neural n etwork component predicts machining characteristics that are difficult to m odel analytically, such as tool wear, machined workpiece surface roughness and chip breaking ability from the essential machining characteristic facto rs. The neural network component operates on the essential machining charac teristics to make its predictions. The analytical component not only predic ts the essential machining characteristics for direct output but also machi ning characteristic factors for the neural network component which uses the se to predict tool wear, machined workpiece surface roughness and chip brea king ability. The tool wear and surface finish are modelled based on their dependence on the analytically predictable machining characteristic factors such as cutting forces and temperature. The chip-breaking ability is defin ed using a chip packaging density index that is modelled with analytically determined factors: forces, flow stress at the shear plane, chip flow angle , chip thickness and chip width. The accuracy of the hybrid machining simul ator has been verified with extensive experimental tests. The simulator, im plemented within Microsoft Windows, is capable of predicting results in bot h numerical and graphical form. (C) 1999 Elsevier Science S.A. All rights r eserved.