ABSTRACT
Cyber-Physical Production Systems (CPPSs) as distributed Systems of Systems (SoS) are at the center of attention from different industries. CPPSs face different categories of errors. These errors will cause failures of the entire production chain. To handle this concern, production systems should be converted into fault-tolerant production systems. To present such systems, a fault tolerance approach was developed to help possible faults prediction and detection of faults causes in this study. Also, the increasing complexity and uncertainty of CPPS call for Digital Twin (DT)-based fault tolerance approach. The proposes approach uses an extraction module to extract the faults signatures efficiently. Based on all extracted faults, appropriate responses could be generated through reliable faults patterns prediction. This method is provided using Fault Tree Analyzer (FTA), Zero-suppressed Decision Diagram (ZDD), and Support Vector Machine-Adaptive Neuro-Fuzzy Inference System (SVM-ANFIS) structure. The results based on digital twin-based CPPS of the food production system as a use case show that the proposed approach can predict reliable faults signatures to prevent failures and make a much reliable production system. Also, this method can guarantee that CPPS is up and running with optimal levels at all times.
