Neural network for prediction of holes diameters and surface roughness in drilling process

Abstract

Several systems are currently tested in order to obtain a feasible and safe method for automation and control of drilling process. This work aims to predict the final diameters and surface roughness of titanium (Ti-6A1-4V) and aluminum (2024 T3) alloy during the machining process in a drilling machine. Acoustic emission, vibration, electrical motor power and force signals were acquired by a commercial data acquisition system. These signals were digitally processed through known statistics methods and were used as input data for an artificial neural network (newff), which estimates the current diameter and surface roughness. After this procedure other neural network (newfftd) was used for predicting the next hole diameter and surface roughness based on the output information from the first neural network. The neural network newff, the mathematical logical method that interprets the signals acquired, was used for estimating the actual hole diameter and surface roughness. The neural network newfftd is the most straightforward dynamic network, which consists of a feed-forward network with a tapped delay line at the input. The neural network newfftd predicts the next hole diameter and the surface roughness one step forward. The results from the neural networks were compared with the actual diameters and surface roughness taken from the worpiece, and showed a good accuracy and a sophisticated method for monitoring and controlling the drilling process, especially the one step drilling process.