Monitoring of drill flank wear using fuzzy back propagation neural network

Abstract

The present work deals with developing a fuzzy back propagation neural network scheme for prediction of drill wear. Drill wear is an important issue in the manufacturing industries, which not only affects the surface roughness of the hole but also influences the drill life. Therefore, replacement of drill at an appropriate time is of significant importance. Flank wear in a drill which depends upon the input parameters like, speed, feed rate, drill diameter, thrust force, torque and chip thickness. Therefore sometimes it becomes difficult to have a quantitative measurement of all the parameters and a qualitative description becomes easier. For this kind of situations, a fuzzy back propagation neural network model has been trained in the present work and has been shown to predict drill wear with reasonable accuracy. In the present case a left and right (LR) type fuzzy neuron has been used. The proposed model is composed of various modules like fuzzy data collection at input fuzzy neuron, defuzzyfication of input data to get output, calculation of mean square error (MSE) and feeding back to update the network. Results from the present work show a very good prediction of drill wear from the present fuzzy back propagation neural network model.