The increasingly competitive market is forcing the industry to develop higher-quality products more quickly and less expensively. Engineering analysis, at the same time, plays an important role in helping designers evaluate the performance of the designed product against design requirements. In the context of automated CAD/FEA integration, the domain-dependent engineers different usage views toward product models cause an information gap between CAD and FEA models, which impedes the interoperability among these engineering tools and the automatic transformation from an idealized design model into a solvable FEA model. Especially in highly coupled variable topology multi-body (HCVTMB) problems, this transformation process is usually very labor-intensive and time-consuming. In this dissertation, a knowledge-based FEA modeling method, which consists of three information models and the transformation processes between these models, is presented. An Analysis Building Block (ABB) model represents the idealized analytical concepts in a FEA modeling process. Solution Method Models (SMMs) represent these analytical concepts in a solution technique-specific format. When FEA is used as the solution technique, an SMM consists of a Ready to Mesh Model (RMM) and a Control Information Model (CIM). An RMM is obtained from an ABB through geometry manipulation so that the quality mesh can be automatically generated using FEA tools. CIMs contain information that controls the FEA modeling and solving activities. A Solution Tool Model (STM) represents an analytical model at the tool-specific level to guide the entire FEA modeling process. Two information transformation processes are presented between these information models. A solution method mapping transforms an ABB into an RMM through a complex cell decomposition process and an attribute association process. A solution tool mapping transforms an SMM into an STM by mimicking an engineers selection of FEA modeling operations. Four HCVTMB industrial FEA modeling cases are presented for demonstration and validation. These involve thermo-mechanical analysis scenarios: a simple chip package, a Plastic Ball Grid Array (PBGA), and an Enhanced Ball Grid Array (EBGA), as well as a thermal analysis scenario: another PBGA. Compared to traditional methods, results indicate that this method provides better knowledge capture and decreases the modeling time from days/hours to hours/minutes.