Bartolini A.

Nowadays, in the current situation of global market, it becomes more and more important to introduce into the market machines with a concrete added value, for example a system able to detect the wear of the tool for the marble cutting in order to improve the surface finishing and to prevent the tool breakage that means a long downtime. That means also economical benefits because the reduction of maintenance interventions caused by failures, reduction of maintenance time and optimization of the spare parts. The target is to increase the availability of the machine and so the production level. The machine under test was designed for the cutting of marble blocks in some slices, the time necessary to cut every slice is four or five hours, so we have to assure that the tool is good for the whole time of working. In this study we analyzed the most significant field signals (torque, current, velocity and so on) as inputs to a neural network. We considered also the acoustic signals as important indicator of the tool wear in order to identify the frequency, or the frequencies involved in the wear process by FFT analysis. The use of neural network is particularly suitable in this case, because we have a large amount of experimental data at our disposal in order to carry on the training of the neural network itself.