Applying The Differential Evolution Algorithm To The Optimisation Of Cross Sections Of Steel Columns
Juha Kilkki, Jouni Lampinen, H. Martikka · WIT transactions on the built environment · 2001
The suitability of the evolution algorithm for nlixed number optimisation is presented. The cross section of a welded or edged steel column is optimised using the total mass as the object function. The optinlisation is repeated using different effective lengths of the column. The thickness of the material is chosen from the commonly available plate thicknesses. The widths and heights of the plate fields are used as continuous optinlisation parameters. The discrete optimisation parameters are the thickness of the steel plates, the number of plate fields, and material properties. The selection of the population size, the mutation probability and the value of the differential factor have a strong effect on both the ability of the algorithm to achieve the global optimum and on the solution tme. An attempt is made to automate the selecting of these parameters of the algorithm. It is important to find a reliable automated method for setting the parameters of the algorithm in the practical task of optirmsing steel structures. Optimal cross sections of the columns are presented, together with the effect of the parameters of the optimisation algorithm. The suitability of the differential evolution algorithm for mised number optimisation in the current case is evaluated. Autonlating the selection of the parameters of the algorithm is also evaluated and discussed critically.