Gradient-based Reconfiguration of Cyber-Physical Production Systems
Cyber-physical production systems (CPPS) are susceptible to various faults like failing components or leaking connections. Even though CPPS have many possibilities to adapt to faults like using an alternative path or redundant hardware, the full potential is not exploited: Nowadays, the control software of CPPS is static and not able to adapt to unforeseen situations like an unknown fault. Hence, CPPS are not able to adapt to faults but a machine operator needs to reconfigure the system manually. Reconfiguration is the task of recovering valid system behavior after a fault has occurred. To enable CPPS to adapt autonomously to faults, automated reconfiguration is necessary. But since CPPS usually are dynamic systems consisting of interconnections of various components, identifying the necessary changes for reconfiguration is challenging: The effects of changes may only manifest after some time and lead to deviations in components far away from their root. This paper presents an approach on automated reconfiguration for CPPS, i.e. the automated recovery from faults. The approach is based on the usage of a logical calculus that reasons about the consequences of a fault and the possible adaptions. Therefore, the system is modeled in terms of logic. Gradient information is integrated to capture the system dynamics. The applicability of the algorithm is shown using a benchmark system from process engineering. Thus, CPPS are enabled to adapt to faults autonomously.