On-line sensing of drill wear using neural network approach

A 9X4X1 neural network is used for online sensing of drill wear. The input vector of the neural network is obtained by processing the signals of the thrust and torque. The outputs are wear states and drill wear measurements. The learning process of the neural network can be performed by backpropagation. The results of a 9X14X1 neural network with and without adaptive activation-function slopes are compared. The 9X14X1 neural network with adaptive activation-function slopes can converge much faster than the conventional neural network. This modified neural network can achieve a success rate of 100% for online classification of drill wear, even when the drilling condition has been changed. The neural network is also capable of measuring the drill wear accurately, with an average error of 7.73%