Towards the identification of worn picks on cutterdrums based on torque and power signals using Artificial Neural Networks

The paper presents an attempt to identify the status of cutters working as an assembly on a multi-tool head. Initial tests covered machining with a single radial tool ( Gajewski and Jonak., 2007] and Jonak and Gajewski., 2007]). The time courses of mining power and torque for a multi-tool head with installed radial and tangent-rotational tools were recorded. The tests covered mining with new - sharp - cutters and partially worn cutters. In order to limit the variables influencing the mining process, a model rock block was used for the experiment.The received time courses were used as input variables for the Artificial Neural Network (ANN). For this purpose, mining power and torque signals statistical parameters were established: variance, skewness, and kurtosis. The status of mining cutters (sharp or worn) was the input variable Artificial Neural Network. Multilayer perceptron (MLP) structure networks, verified in the previous identification tests ( Gajewski and Jonak, 2006] and Gajewski, 2005]), were used for analysis.