Novel analysis and modelling methodologies applied to pultrusion and other processes
Often a manufacturing process may be a bottleneck or critical to a business. This thesis focuses on the analysis and modelling of such processest, to both better understand them, and to support the enhancement of quality or output capability of the process. The main thrusts of this thesis therefore are: To model inter-process physics, inter-relationships, and complex processes in a manner that enables re-exploitation, re-interpretation and reuse of this knowledge and generic elements e.g. using Object Oriented (00) & Qualitative Modelling (QM) techniques. This involves the development of superior process models to capture process complexity and reuse any generic elements; To demonstrate advanced modelling and simulation techniques (e.g. Artificial Neural Networks(ANN), Rule-Based-Systems (RBS), and statistical modelling) on a number of complex manufacturing case studies; To gain a better understanding of the physics and process inter-relationships exhibited in a number of complex manufacturing processes (e.g. pultrusion, bioprocess, and logistics) using analysis and modelling. To these ends, both a novel Object Oriented Qualitative (Problem) Analysis (OOQA) methodology, and a novel Artificial Neural Network Process Modelling (ANNPM) methodology were developed and applied to a number of complex manufacturing case studies- thermoset and thermoplastic pultrusion, bioprocess reactor, and a logistics supply chain. It has been shown that these methodologies and the models developed support capture of complex process inter-relationships, enable reuse of generic elements, support effective variable selection for ANN models, and perform well as a predictor of process properties. In particular the ANN pultrusion models, using laboratory data from IKV, Aachen and Pera, Melton Mowbray, predicted product properties very well.