The robust design of complex systems

Robust Engineering Design has evolved as an important methodology for the integration of quality with the process of design. The methodology encompasses the disciplines of experimental design, model building and optimization. First an experiment is conducted on a system (or a simulation of the system), second a model is built to emulate the system and finally the emulation model is used to optimize the system design. Applying these methods to large problems can be difficult and time-consuming because of the complexity of most design problems. It is the goal of this thesis to introduce methods which reduce problem complexity and so make the application of Robust Engineering Design (RED) methodology easier for large design problems. By drawing from methods used in systems theory and circuit optimization several techniques are presented with the aim of reducing the complexity of performing experiments for Robust Engineering Design. A common framework for experimentation is created by combining a commercial circuit simulator with established methods for experimental design and model building. This provides the basis for experimentation in subsequent chapters. A method of design optimization with respect to quality is presented to complete the model-based Robust Engineering Design cycle. Three approaches to reducing problem complexity are adopted. First a method of system decomposition is applied directly to an electronic circuit to reduce the size of experiment required for RED. Second a method of modelling system response functions is described which integrates the action of the circuit simulator with the model building process. Third information about system topology is used in the design of experiments to enhance the model-building process. Conclusions are drawn about the effectiveness of the approaches described with respect to the impact on problem complexity