An integrated optimization environment for structural configuration design.

摘要

The relatively recent efforts for designing lightweight structures have motivated increasing use of formal mathematical optimization techniques for selecting the shape and sizes of structural members. How to obtain rigorous optimal topologies has been an elusive goal until the very recent introduction of the homogenization method.;This dissertation describes a system for designing complete structures starting only with an available space and loading and boundary specifications. A three-phase design process is followed: generate an approximate topological image with homogenization (Phase I), process the image to obtain a practical realistic structure (Phase II), and refine the final topology by detailed shape and size optimization (Phase III). An overview of the system is given together with several practical examples. The emphasis of this dissertation is on the appropriate design and implementation of the image interpretation module in Phase II. Image processing and computer vision techniques are used in conjunction with insights from mechanics and knowledge about fabrication processes to implement the system.;The main concentration of this dissertation is on two-dimensional structures. In Phase II, two types of models are distinguished for two-dimensional structures, namely, skeletal and plain stress/plain strain models. Different approaches are taken for the treatment of both models in Phase II. For skeletal structures their skeleton is useful, whereas for solid structures the information on their boundaries is pertinent. Techniques to extract and process this information are described. For each type of structure a fabrication domain is chosen, and rules are devised whose application to the representations in Phase II generates designs that are more easily manufacturable. For solid structures casting is chosen as the fabrication domain and, for skeletal structures weldments of trusses are explored. Extraction of other useful properties, such as, symmetry and immobility, for two-dimensional structures is discussed.;Interpretation of three-dimensional images is more challenging, and the techniques described for two-dimensional structures cannot easily be extended to apply to three-dimensional structures. A novel algorithm is devised that converts the spatial enumeration scheme generated in Phase I into a constructive solid geometry (CSG) scheme. The CSG representation is at a higher level of abstraction and can be used for further manipulation of the design.