Detección y diagnóstico de fallas en sistemas complejos de manufactura empleando técnicas de softcomputing

The main goal of a fault detection and diagnosis of manufacturing systems is to prevent an undesired stop of the production line and its corresponding cost. Modern manufacturing systems are a very complex domain due to all the physical variables involved and their correlations. This represent a very challenging task for monitoring systems. The present article shows a fault detection and diagnosis system based on the history data process. The proposal is composed by two phases. In the first phase it learns the normal operation behavior of the system using an autoassociative neural network which carries out the detection process. In the second phase the final diagnosis is given using 35 Research in Computing Science 116 (2016) pp. 35–49; rec. 2016-03-14; acc. 2016-05-13 a multiclass support vector machine, which classifies the type of fault present and gives its time of occurrence.