Machine learning based manufacturing control system for intelligent Cyber-Physical Systems

Abstract

Cyber-physical systems are often misunderstood to be just any embedded systems. The real cyber-physical system should have both physical and digital (computational-communication-control) parts inter-connected in each part and process, and the system itself should have the capacity to change its own behaviour to adapt to changing requirements. This paper presents an architecture of an intelligent cyber-physical system where a reconfigurable manufacturing system is supported by a machine learning algorithm to provide enhanced decision-making to the manufacturing control system. Experiment results are presented showing the machine learning module can help the control system adjust itself with changing requirements provided externally (by a user) in the form of training examples. The result is an architecture of an intelligent cyber-physical system, with physical and digital parts always working in synchronization, enabling change in the system's behaviour in terms of manufacturing process-flow in order to adapt to any change in the production planning.