Embedded service oriented diagnostics based on energy consumption data

This paper presents a method to identify at an early stage incipient faults in pieces of equipment typically used for transportation of parts in discrete manufacturing settings. The method relies on knowledge of the typical energy consumption of the described devices using different workloads (the energy signature of the machines). At runtime, Support Vector Machines (SVM) are used to classify the input energy data into two pattern descriptions, typical of regular workloads expected on the devices. Consecutive mismatches between the output of the SVM and the workloads observed in real time indicate possibility of incipient failure at device level. The method is implemented using Web Services deploying Service Oriented Architecture in a real multirobot factory automation industrial testbed, originally used for assembly of mobile phones.