Reliability-Based Optimum Design of a Square Box Column Constructed from Cellular Plates Luis M.C. Simões 1 , József Farkas 2 , and Károly Jármai 2 1 Dep. Civil Eng., University of Coimbra, Portugal lcsimoes@dec.uc.pt 2 University of Miskolc, Miskolc, Hungary {altfar,altjar}@uni-miskolc.hu Abstract. Cellular plates can be calculated as isotropic ones, bending moments and deflections being determined by using the classic results for various loads and support types. A cantilever stub column of a square box section composed of welded cellular plates is optimized. The column is subject to compression and bending and is constructed from four equal cellular side plates. The con- straints on overall buckling are formulated according to the Det Norske Veritas design rules. The horizontal displacement of the column top is limited. The cost function to be minimized includes the costs of the materials, assembly, welding and painting. Randomness is considered both in loading and material properties. A level II reliability method (FORM) is employed. Keywords: cellular plates, reliability, optimization, box column. 1 Introduction Box beams and columns of large load-carrying capacity are widely applied in bridges, buildings, highway piers and pylons. Since the thickness required for an unstiffened box column can be too large, stiffened plate elements or cellular plates should be used. The strength is considerably larger than that of a plate stiffened on one side by open section ribs because of the larger torsional stiffness of the cellular plate (Farkas and Jármai 2007). The stiffening presented here consists of rectangular hollow sections (RHS) applied as an orthogonal grid. In this work a maximum probability of failure is stipulated for design and the reliability is evaluated by using a level II procedure FOSM (first order second order reliability method) (Hasofer and Lind 1974), the sensitivity information being obtained analytically. The overall probability of failure which account for the interac- tion by correlating the modes of failure is considered. A branch and bound strategy coupled with a entropy-based algorithm is used to solve the reliability-based optimization. The entropy-based procedure is employed to find optimum continuous design variables giving lower bounds on the decision tree and the discrete solutions are found by implicit enumeration.