Intelligent fault detection and diagnosis of a rotary cutoff in a corrugator

In this paper an artificial intelligence based framework for fault detection and diagnosis to support supervision of the cardboard production is presented. Cutting accuracy significantly affects the quality of the product and because there are many different causes of errors, their identification requires a sound knowledge and experience of the service staff. The authors observed that the sources of errors can be characterized by a probability density function (pdf) of these errors. Therefore, they proposed a diagnostic method based on classification of sources of disturbances via the analysis of pdf calculated by a kernel density estimator. The multilayer perceptron is proposed as a classifier. Classification procedure is discussed with emphasis on generalization properties of the classifier. The application for data acquired from a real industrial process is presented.