Reconfigurable manufacturing process monitoring systems

Abstract

Performance measurement is indispensable to manufacturing, due to the fact that if the efficiency of an activity cannot be measured it could not be effectively controlled. Recent trends in process monitoring systems point towards a transformation from static centralisation to dynamic decentralisation. This change has been motivated by the need to enable reconfigurable systems that are internally flexible to production requirements, and externally adaptive to multiple processes. The convergence of industrial systems with advanced computing, low-cost sensing, and new levels of connectivity permitted by network technology has been the catalyst for this transformation. Furthermore, these decentralised cloud manufacturing systems are being combined with advanced analytics and artificial intelligence, to form cyber-physical production systems. This research work explores the dynamics of decentralised software architecture within field-level manufacturing process monitoring systems. The need for this understanding has been driven by the prediction that these cyber-physical systems will create the next generation of innovative intelligent machines. This research investigates the capability of decentralisation design to provide the core fundamental functionality of process monitoring systems in a new reconfigurable format. The embodiment of this investigation is the design and development of a decentralised architecture for the creation of a reconfigurable process monitoring system within field-level manufacturing. An investigation into available data interoperability systems and field-level manufacturing process monitoring system requirements, resulted in the identification of a research opportunity. Evidently, current academic and commercial mediums could not provide for the high communication speed, high data capacity, and heterogeneous data requirements present in field-level manufacturing systems. Through a combination of decentralised modelling, and state-of-the-art technologies and techniques, a new data interoperability architecture was developed. The resultant architecture, namely the ARC, is tested in respect to speed, capacity, and correlation accuracy. The results showed a; <= 1 ms communication speed, 1 Hz to 1MHz data capacity, and 99.95% correlation accuracy. Evidently, the ARC is an effective data interoperability medium for utilisation in field-level manufacturing systems, beyond the capability of all previously reviewed systems. Furthermore, the ARC was adapted to monitor multiple process variables from a CNC turning machine tool, such as: tool force, spindle and axis motor current, spindle and turret vibration. The ARC provided a platform to evaluate the migration of signal process techniques, and time and frequency domain analytics, within a decentralised architecture. The results from this work represent a first case migration of fundamental manufacturing process monitoring steps within a cyber-physical system. Furthermore, an advanced cyber-physical system was created for autonomous process performance characterisation. In order to investigate the industrial application of the ARC, a study was undertaken into the variation in dry CNC turning machining, thereby evaluating the capability of the ARC signal processing techniques and analytics to achieve process insight. The result of which, was the successfully implementation of the ARC to achieve multi-scalable data acquisition, signal processing, and process performance analysis of a CNC turning machine tool. The generic building blocks present within the ARC were configured to produce unique signal feature extraction across multiple process variables. Process insight was evident in the multi-perspective view of machine actions, via the time and frequency domain analysis, of tool force, spindle and axis motor current, and spindle and turret vibration.