부산대학교 도서관

This thesis addresses idealisation error control for the nonlinear finite element method. The focus is on accurate failure prediction of mid-size aerospace structures. The objective is the development of technologies that shorten the certification process of new airplanes, by replacing expensive and time consuming testing with reliable calculation methods. The SAFESA (Safe Structural Analysis) approach was applied to the collapse analyses of stiffened metal panels. ABAQUS/Standard was thereby the utilised nonlinear solver. Because the original SAFESA procedure is tailored for linear analyses, the methodology needed an update. The first analysis case is a stiffened panel compression test which was arranged as a lecture demonstration at Cranfield University. The analysis behaviour is highly nonlinear due to the thin-walled properties of the panel. The second analysis investigates an Airbus compression panel. Until failure, the panel behaves geometrically less complicated because the major load bearing parts are thick-walled and bend smoothly. The main research work is the critical analysis of important modelling assumptions concerning the used material model, boundary conditions and geometrical imperfections. In both cases, the method helped to identify idealisation errors and to build a reliable FEM model. In order to deal with the nonlinear error sources, minor extensions to the original method had to be made. The major achievement is the development of the first expert system which applies the idealisation error control methodology. CAD data import, geometry visualization, a knowledge-based decision making advisor and audit trail functionality were implemented. The expert system leads the user through a step-by-step idealisation process. Each decision is documented and a confidence level must be supplied. This way, every uncertainty is flagged out as potential error source. An interactive interface was created, which provides the user with expert advice on how to treat the idealisation errors. The software has been validated and shown to meet the program objectives.