Predication of Kerf Width and Surface Roughness in Waterjet Cutting using Neural Networks

Abstract

Abrasive waterjet cutting is one of the unconventional methods used to cut some of the difficult to cut materials. In certain materials this method has proved to give better results compared to the conventional methods. In this model the water jet cutting is done on hastelloy using the three parameters which are abrasive mass flow rate, traverse speed, and the stand-off distance. The mathematical modelling to predict the kerf width based on these three input parameters is discussed in this paper. As the relation between the input and the output parameter is non-linear in nature neural network back propagation algorithm is used for the prediction. Here the experiment is conducted using waterjet cutting machine and the data's like surface roughness, metal removal rate, kerf width and the kerf angle data are collected. Both the input and the output parameters are fed to the neural network toolbox programmed in the MATLAB. After 1000 iterations it has been found that the prediction is closer to the actual value. The mathematical constants which is the weight matrix is used to test the new set of data for accuracy. It has been found that the prediction is more accurate compared to the conventional methods. This experiment is based Taguch's design which uses the above three parameters to cut the material. This paper also discusses on the predication of surface roughness of hastelloy created due to the variation in these parameters.