Automatic monitoring and diagnosis of the dressing operation through the classification of textural patterns in acoustic maps

Abstract

A competitive manufacturing enterprise depends on a high level performance of its processes. In the metal- mechanic industry, such a requirement is pursued through the development of reliable monitoring systems. These systems must assure reliable information about the process itself, and about the machine’s parameters. This paper proposes an efficient strategy for the automatic monitoring and diagnosis of dressing operations. The proposed system is based on Artificial Intelligence (AI) techniques, like neural networks, support vector machines, and decision trees, to classify textural features of an image, the acoustic map, which represents the interaction between the dresser and the grinding wheel. The classification indicates if the dressing operation should stop or not, what implies in a better use of the grinding wheel and costs reduction. The results obtained in the performed simulations are very promising, with 100% of right matches with the best tested classifiers. Such initial results point out to an increase in the production velocity, and the reducing in the number of defective pieces.