The requirements for cyber-physical production systems (CPPS) grow steadily, leading to increased complexity of such systems. To address issues due to reconfiguration or failures, we present an approach to creating a context model of the CPPS components during runtime and to detect deviations from the planned production processes. Furthermore, we present a concept for modelling production processes that allows for dynamic process adaption during runtime. We then utilize this process adaption and the context model to cope with unforeseen failures and increase the resilience of the system. Finally, we evaluate the feasibility of our approach by applying it to an Industry 4.0 production system.