ABSTRACT A recently calibrated strength formulation under combined loading which minimizes the model uncertain factor is applied in the optimization study of a comer column of a TLP structure. The various limit states for the orthogonally stiffened cylindrical shell are used and they include bay instability, frame bending, frame tripping, and minimum requirements for area and rigidity of ring frames. Multiple attribute decision making technique based on adaptive generation of nondominated designs is used to achieve better, well balanced designs compared to the original ones. Design variables selected for the examples include shell thickness, structural scantlings of stiffeners, number of stringers, spacing of ring frames. Design attributes include structural weight, structure cost, minimal reliability index for various failure modes. INTRODUCTION Orthogonally stiffened cylindrical shell structural components are used as columns or pontoons for tension leg platforms or other floating production systems. These structural elements have to be designed to resist axial, bending and shear forces, radial pressures and other similar load effects and should have adequate but not excessive safety. A recent joint industry project1,2,3reviews the various limit state formulations for orthogonally stiffened structural components by comparison with available experimental results to establish model uncertainty factors associated with these formulations. Better, robust strength formulations are recommended which minimize the model uncertainty factor when compared with existing codes and experimental results. The various limit states that are used for this study include bay instability, tripping, etc. The effect of initial imperfections and residual stress are included implicitly in these formulations. An optimized design should fulfill different design objectives and satisfy limit state criteria corresponding to different loading conditions. These complex, often conflicting requirements can best be solved using multiple criteria decision making techniques and in this case a design procedure is developed based on the multiple-attribute decision making approach4. Formulation of the design model includes identification of design variables and design criteria (attributes, constraints). Design variables selected for the example design include structural scantlings, number of stringers, etc. Design attributes include: structural weight structural cost and reliability index for various failure modes. In the process of selection of preferred design, intra-attribute preference may be defined through a membership grade functions based on the fuzzy set theory. Inter-attribute relative significance can be obtained via an eigen vector of a subjective decision matrix representing subjective choices of a particular designer. The design procedure for stiffened cylinders includes an analysis and synthesis module: The analysis module contains: design load definition, structural response calculation, feasibility check via ultimate strength criteria; reliability analysis; attribute value calculation. The reliability analysis uses an Advanced First Order Second Moment (AFOSM) reliability calculation procedure which is very efficient and accurate for the level of reliability bounds adopted for the design.