UKSim 2006

Modelling of thermo-mechanical processes in the metal industry has attracted much research interest over many years. As more and more insight is gained, there is a growing demand for through-process integrated thermo-mechanical process models to predict the microstructure evolution and the mechanical properties of the product during the whole production cycle. The main benefit of such an integrated models is to enable a more precise control of the overall process to achieve a ‘right at first time’ production. The benefits of using intelligent systems based paradigms have been highlighted by several research groups in the UK and elsewhere in order to tackle the too-often difficult tasks of modelling, control, and optimisation relating to the current complex and highly integrated systems. The term "intelligent systems" is usually taken to cover the application of machine intelligence. This attempts to understand and replicate, using computers, phenomena that we, humans, associate with "intelligence", i.e. the generalised, flexible, learning and adaptive behaviour that we see in the human brain. The author of this invited presentation will highlight his own experiences and those of his Research Group, namely the Institute for Micro-structural and Mechanical Process Engineering: the University of Sheffield (IMMPETUS), in introducing several techniques which pertain to the above paradigm, in particular those falling within neural networks, fuzzy logic and evolutionary computing to solve some of the acute problems that the metal industry faces in terms of downstream and upstream processes. 1 This is a multi-disciplinary research centre established in May 1996 to study thermo-mechanical processing, to model the events taking place, and to develop improved planning and control of industrial processes. The modelling approaches require a number of techniques to which the Department of Automatic Control and Systems Engineering currently contributes and include Intelligent Systems, Biologically-Inspired Computing, Hybrid Modelling, Modelbased Predictive Control, and Cellular Automata. IMMPETUS is based within three host Departments, Automatic Control and Systems Engineering, Engineering Materials, and Mechanical Engineering. SCIENTEFIC PROGRAMME EVOLVED TOPOLOGY GENERALIZED MULTI-LAYER PERCEPTION (GMLP) FOR JOINT CONSTRAINT MODELLING GLENN JENKINS, PAUL ANGEL School of Computing, University of Glamorgan, Pontypridd (Cardiff), CF37 1DL, Wales Email: @glam.ac.uk ABSTRACT The accurate simulation of anatomical joint models is becoming increasingly important for both medical and animation applications. We propose the use of Artificial Neural Networks to accurately simulate joint constraints based on recorded data. This paper describes the application of Genetic Algorithm approaches to neural network training in order to model corrective piece-wise linear / discontinuous functions required to maintain valid joint configurations. The results show that Artificial Neural Networks are capable of modelling continuous boundary shapes for a range of constraint sizes.