WIT Press

Structural Design Optimization Using Regression Techniques


Free (open access)

Paper DOI









552 kb


G. Steenackers, P. Guillaume & S. Vanlanduit


A structure excited by forces can result in excessive vibration levels if not properly designed. To overcome such problems, a structural design optimization approach will be presented. Most updating techniques require many finite element calculations, resulting in high calculation times. This paper concentrates on a regressive finite element updating approach.During the optimization the finite elementmodel is replaced by a multivariate regression model, resulting in a significant reduction of the computation time. To guarantee a reliable final solution, the multivariate regressive model is improved and updated during the optimization process. The effectiveness of the suggested procedure will be illustrated by means of the case study of an airplane component. 1 Introduction Finite Element (FE) analysis is widely employed in todays Computer Aided Engineering (CAE) to model realworld structures in an early design stage. FE models are mostly considered deterministic, implicitly assuming that all design parameters are precisely known and that the manufacturing process produces identical structures. Finite element model updating, also known as inverse modelling, is an inverse problem to identify and correct uncertain modelling parameters that lead to better predictions of the dynamic behavior of a target structure. The procedure of updating a finite element model uses measurement data as exact reference data in order to update the chosen model parameters (ref. [1] and [2]). Assume that one wants to reduce the maximum displacement of a structure under a given dynamic load. Two possible ways of doing this is by changing mass or stiffness of the structure. Adding for instance one or more ribs to the original