Towards Computational Design Methods for Conceptual and Parametric Design
The subject matter of this dissertation is computational design methods for mechanical design. The term computational design method here denotes a method by which a computer program can perform one or more steps in a design process. Two particular design problems are addressed in the thesis:
The first problem is the conceptual design of energy-transforming systems. Examples of such are systems whose main purposes are to transfer power, serve as auxiliary energy sources, or process information, if the information carrier is an energy flow. The work reported is focused towards computational methods for synthesis of alternative solutions to design problems formulated in terms of the input and output energy flows of a "black box". The approach adopted is based on a function vocabulary for energy-transforming systems that is based on the modelling concepts of bond graphs. The basic idea is to use bond graphs for modelling the physical principles on which the modes of action of machines are founded. An automated synthesis procedure for energy-transforming systems and its computer implementation are presented. This synthesis procedure is based on the developed function vocabulary, and consists of two steps: First, alternative process (function) structures are generated. Alternative design concepts are then composed by combining partial solutions into total solutions. The dynamic behaviour of a proposed design solution may also be simulated.
The second problem is the modelling and optimization (parametric design) of complex products. Complex products are here defined as products that can be described as a hierarchically organized structure of interrelated components. General models for complex products and the components they consist of are presented. Special reference is made to the knowledge needed to apply a system optimization approach towards the optimization of the complex product. A system for parametric design of complex products, based on these general product models, is described. The system optimization method employed is the Model Coordination Method. A module of the system supports interactive optimization. Special attention is also called to the fact that optimization algorithms require an exact problem formulation, whereas in engineering problems the values of the input parameters, the constraint boundaries etc., are rarely exactly known. The utility of sensitivity analysis when identifying crucial uncertainties in the problem statement is demonstrated.