The integration of mechanical engineering, electrical engineering and information technology in one mixed system has found vast applications in industry and everyday life. This interdisciplinary field, known as Mechatronics, has attracted a great deal of attention, particularly in the context of optimal design of multi-domain systems. To this end, the present thesis represents an original investigation into the development of formal and systematic methodologies for the optimal design and design evaluation of mechatronic systems.

This work presents a new philosophy and approach for the optimal design of mechatronic systems. It takes into account that an optimal mechatronic design requires concurrent, integrated, and system-based thinking with regard to all design parameters and criteria involved in the mechatronic system. The mathematical model and axioms which support this thinking are presented in the thesis. A design evaluation index has been presented in the present work which supports above statement. It is based on the concept of mechatronic design quotient (MDQ). MDQ is a multi-criteria index reflecting the overall degree of satisfaction of the design criteria for a mechatronic system. In this thesis, a nonlinear fuzzy integral is used for the aggregation of the various design criteria and for handling possible correlations among them. For an existing mechatronic system, MDQ is a useful index for evaluating its design as well, and determining the potential for improvement. In different stages of design, it can be used as an index for the purposes of optimization and/or decision making.

In the present work, a new systematic mechatronic design methodology based on the concept of MDQ maximization is presented. The design procedure is treated in multiple stages. In the conceptual stage, MDQ provides guidance to the designer in selecting the best design choices and making effective decisions about the essential structure of the design. In the next stage, which concerns detailed design, a niching genetic algorithm is employed to find the elite design alternatives for all possible configurations and combinations of system parts. Since a full MDQ assessment is computationally expensive, it is not practical to consider all MDQ attributes in the course of an evolutionary optimization. Only the essential criteria which have a veto effect on the MDQ evaluation are considered in the process of the niching genetic algorithm. A full and detailed MDQ assessment is then employed to find the best choice among the elite representatives.

The reliability assessment of mechatronic systems is studied as well in the thesis. A new reliability assessment methodology is developed which has two practical advantages over the available methodologies. First, in view of dynamic interactions that may exist in a mechatronic system, the developed method uses a Petri-net simulation of the dynamic behavior of the system. Here, all possible events and conditions of the real operation of the system and all possible interactions can be accurately modeled in the level of detail that the designer prefers. Second, the severity of the failure modes is considered in the reliability evaluation methodology. The developed reliability evaluation approach contributes in the mechatronic design process by revealing information about the performance of various design choices.

A bond graph mechatronic simulation tool is developed in this work. It is based on a new matrix-based formulation, which is presented in the thesis. The bond graph simulation tool is integrated with genetic programming to form a unified evolutionary mechatronic tool. As a new contribution, this synergic integration is extended for the general case of nonlinear mechatronic problems. This tool can concurrently optimize both the topology and "size" of a bond graph model of a mechatronic system in order to achieve the best fitness for the solution. It can be used in any mechatronic problem, provided that an effective fitness evaluation is established for the problem. In particular, the application of this tool for automated system identification of nonlinear mechatronic systems is presented in the thesis.

The methodologies developed in this research are validated by applying them to the modeling and redesign process of an industrial fish processing machine, called the Iron Butcher - a complex electromechanical system which falls into the class of mixed or multi-domain systems.