A Product Data-Driven Methodology for Automating Variable Topology Multi-Body Finite Element Analysis Reference Koo, D.(2000) A Product Data-Driven Methodology for Automating Variable Topology Multi-Body Finite Element Analysis, Masters Thesis, Georgia Institute of Technology, Atlanta. Abstract CAD-CAE integration focuses on shortening the transition time between a design model and its analysis models. The multi-representation architecture (MRA) developed in recent years emphasizes CAD-CAE integration to make automated simulation-based design ubiquitous. To date the MRA has successfully automated product data-driven finite element analysis (FEA) using the fixed topology template methodology. While this class of problems is important, variable topology analysis models are often needed to support product design. Yet creating FEA models that have many bodies (material regions) is a manually intensive effort in current practice. This thesis presents a new methodology to better handle such cases. The variable topology multi-body (VTMB) methodology introduced here creates algorithms that automatically transform design models into FEA models. Such an algorithm uses the MRA to extract idealized attributes from a specific class of detailed design models. It then identifies basic shapes and assembly information, recognizes VTMB FEA issues, and performs procedures that are normally done manually (e.g., decomposition). Finally, it creates FEA tool inputs, executes the tool, and extracts relevant results for further processing by its MRA context. To test the methodology, algorithms were developed for several classes of problems in the electronic packaging industry. Test cases are given for one 2D regular shape case (printed wiring board warpage), two 3D regular shape cases (thermal resistance analysis of ball grid array packages), and one 3D irregular shape case (thermal resistance analysis of quad flat packs). Multiple designs were analyzed for each of these cases. These experiences highlighted areas where the MRA needs extensions to reduce algorithm development time and broaden algorithm applicability. However, results show that the present VTMB methodology can produce algorithms that reduce FEA model creation time by a factor of 10 or more (from days/hours to minutes). Test cases also demonstrate methodology applicability to a diversity of problems. Ultimately this automation leads to better designs by enabling more analysis and better understanding.