Numerical Substructuring Methods in Finite Element Analysis

Abstract

The proper coupling of substructures into system assemblies is a key part of achieving acceptable model fidelity and accuracy. Component model reduction leads to the system becoming stiffer and thus residual flexibilities at load and boundary interfaces must be included in the system-level model. The condition of the substructure system matrices can offer insight into properly re-conditioning the system to reduce the residual error associated with model reduction. Numerical experimentation will demonstrate how different substructure matrix conditioning affects the entire system assembly and solution. By taking the simulated results of the substructure, corrupting the mode shapes, and assembling the new model, the differences in solutions can be evaluated and then compared. The methods of numerical experimentation are presented and their results will be used to gain insight into ways to recondition the matrices associated with the residual structural model to minimize model errors. The methods of reconditioning the matrices associated with the residual structural model are being developed to apply to system assemblies where some of the components are defined entirely from experimental data and where it is difficult to pre-assess the condition of the parameters of the matrices associated with experimentally defined components.

This work was performed at New Mexico Tech and supported in part by the Federal Aviation Authority Center of Excellence for Commercial Space Transportation under Research Grant 293, 2012.