Detection of abnormal energy patterns pointing to gradual conveyor misalignment in a factory automation testbed

This paper presents a method to detect gradual deterioration of behavior in discrete manufacturing pieces of equipment dedicated to transportation of parts between the processing workstations. Monitored energy patterns are correlated with workload data (number of pallets occupying conveyor segments at a time). Detection of gradual conveyor misalignment relies on checking whether the output of a rule based engine mapping input workload data into classes match with the output of an Support Vector Machine classifier operating on the energy values.