Ph.D. Proposal Presentation by Sai Zeng
Thursday, April 1, 2004

( Dr. Suresh Sitaraman, Chair)

"A Knowledge-based FEA Modeling for Highly Coupled Variable Topology Multi-body Problems"


As engineering systems are becoming more and more complex, information models of the systems must be improved accordingly. Extensive research is currently underway to develop engineering data management capabilities and to understand the role of information as a systems integrator. In the area of CAD/CAE applications, an information gap exists between idealized design models and analysis models because of the engineers’ different usage views towards product information models, such as designers’ views of high fidelity solid models and analysts’ views of approximate FEA mesh model. The gap impedes the communication and interoperability among different design disciplines within product design. Moreover with increasing complexity and variability in materials, geometric shapes, and connectivity configurations, etc. in applications such as chip package, FEA modeling is a tedious and time-consuming activity. Every time when the package design is modified, the FEA modeling has to be repeated manually by experienced engineers. Typically finite element modeling takes hours or even days to complete an analysis for a single chip package design. To resolve this problem, it is necessary to present a method that can bridge this information gap; thus in the dissertation,

it is proposed to develop a generic process model and corresponding rich information models to link the idealized design models and FEA models in order to realize a truly automated design-analysis integration for Highly Coupled Variable Topology Multi-Body (HCVTMB) problems.

To realize this research goal, a generic method is presented consisting of three steps, process partition; process detail design and process input and output information representation. At first, a generic design-analysis process partition and corresponding process model are presented to reduce the complexity of the overall process and increase the relative transparency of the relationships between inputs and outputs. Second, within the process design steps, a decomposition based inter-body mesh algorithm is proposed to support the geometric conversion between the idealized design models and FEA models. And a linking mechanism is designed to support geometric transformation with maintained information associativity. Third, after decomposing the process, the input and output information of the individual sub-process is represented using Object Oriented modeling approach, thus the input and output information is constructed which is semantically rich and reusable. Finally, the efficacy of the proposed method is validated in the context of real-world engineering problems such as chip packages thermo-mechanical FE analysis and aerospace structural FE analysis, whose idealized models are characterized as HCVTMB problems.