SummaryThis paper describes a procedure to predict the elasto-plastic responses of structures within the framework of the finite element method when design variables are changed. Reanalysis technique was used applying the Taylor Series ExpansionMethod with the results of sensitivity analysis.In the incremental finite element analysis, the load magnification factor is usually introduced to control the load increments so that yielding of individual elements takes place exactly. In sensitivity analysis, sensitivity of the load magnification factor with respect to the design variable was newly derived. In this connection, the sensitivity of elasto-plastic responses with respect to the yield stress of the material was also derived.Some sample calculations were performed on plane truss, plane frame and plane stress problems. The results proved the rationality of the proposed method. The reanalysis solution was much improved by considering the sensitivity of the load magnification factor although the computational time for sensitivity analysis increases. The limit of changes in design variables was also discussed to get rational reanalysis solutions.
IntroductionIn the optimal design of structures, it is essential to know the change in objective and constraint functions with respect to design variables. For this purpose, sensitivity analysis and reanalysis are generally performed, which occupy a major part in optimization procedures.A number of research works on sensitivity analysis and reanalysis have been performed regarding the linear elastic problems. Contrary to this, those for the nonlinear problems are very few."' Regarding elasto-plastic plane stress problems, the authors developed the fundamental procedure of sensitivity analysis within the framework of the finite element method' In that paper, the coordinates of nodal points and the thickness were taken as the design variables, and the first and second-order sensitivities of stress components and nodal displacements were derived. A series of reanalyses was also performed applying the Taylor Series Expansion Method. The first and second-order sensitivities were evaluated with good accuracy. However, there existed large differences between the results of reanalysis and the exact solutions especially when the design variables were largely changed.