No abstract

In the design and optimization of artifacts requiring both mechanical and control design, the process is typically divided and performed in separate steps. The physical structure is designed first, a control strategy is selected, and the actual controller is then designed. This article examines how this separation could affect the overall system design and how the combination of the separate problems into a single decision model could improve the overall design, using an electric DC motor as a case study. The combination is challenging since the two problems often have different design criteria, decision objectives, and mathematical model properties. Furthermore, the two problems are often fully coupled in that the physical structure depends on the controller and the controller depends on the physical structure. A Pareto analysis is suggested as a rigorous way to compare a variety of design scenaria.

In the design and optimization of artifacts requiring both mechanical and control design, the process is typically divided and performed in separate steps. The physical structure is designed first, a control strategy is selected, and the actual controller is then designed. This paper examines how this separation could affect the overall system design and how the combination of the separate problems into a single decision model could improve the overall design, using an electric DC motor as a case study. The combination is challenging since the two problems often have different design criteria and objectives and mathematical model properties. A Pareto analysis is suggested as a rigorous way to compare a variety of design scenaria.

Optimal embodiment design and control systems can be solved using several different strategies. This paper considers whether those strategies will find the true system optimum. Optimality conditions determine whether a point is an optimum of a system. This paper presents a development and comparison of the optimality conditions for several methods of optimizing fully coupled embodiment design and control systems. The results demonstrate that the methods do not possess the same optimality conditions and therefore will not converge to the same optimal solutions, often leading to sub-optimal solutions. Since these strategies have been used successfully to solve various examples, criteria are developed to determine when the various methods will find the true system optimum.