Reliabillity-based Structural Optimazation Based On The Response Surface Method

Traditional structural reliability analysis methods are difficult to be applied to practicalengineering because of heavy demand of FEM calculations. To address this problem, thethesis researches on the Response Surface Method (RSM) and applies RSM to structuralreliability analysis and reliability-based structural optimization. The main contents andconclusions in this thesis are listed as following:1. Improve RSMThis thesis gives3improved response surface methods, as following:a) The improved RSM based on the control pointThe thesis raises the concept of control point which could obtain fine design point ascentral point and constructs RSM1with control point. Based on RSM1, RSM2is constructedto improve the accuracy of failure probability evaluation. Both of RSM1and RSM2are moreefficient than classical RSM and provide accurate reliability index, RSM2provides accuratefailure probability. But RSM1and RSM2cannot be applied to problems with multi designpoints because only one control point is constructed.b) Improved RSM based on the reference pointThe RSM based on reference point obtains experiment point by the information ofreference point. The obtained experiment point locates close to design point and increases theaccuracy of Limit State Function (LSF) approximation. The failure probability obtained bythe improved RSM is very close to that of Monte Carlo Simulation (MCS).c) The improved RSM based on reasonable experiment pointThe thesis proposes an improved RSM based on the concept of reasonable arrangementof experiment point. The proposed method can consider the nonlinear trend of LSF, andovercome the weakness of classical RSM. The reliability of a stiffened composite panel underaxial compression load is analyzed with the proposed method. The results show that theaccuracy and the efficiency of the proposed method are both satisfactory. The number ofrandom variables can be greatly reduced according to actual engineering, and uncertainties of geometry and Young’s modulus E11should be top-priority in the buckling reliability analysisof a composite stiffened panel.2. Reliability-based structural optimizationThis thesis presents the progress of reliability-based structural optimization usingresponse surface method. The improved response surface methods are applied toreliability-based structural optimization. The numerical examples illustrate the proposedmethods have high accuracy and efficiency in reliability-based structural optimization.3. ConclusionsA large number of examples are given to demonstrate the efficiency and the accuracy ofthe proposed methods. It is shown that the proposed methods have good accuracy and highcomputational efficiency, and can significantly reduce the number of evaluation of actual limitstate function. The thesis provides feasible methods for structural reliability analysis andreliability-based structural optimization. The proposed methods can solve the problem that theefficiency of reliability analysis and reliability-based structural optimization for large andcomplex structures using traditional reliability methods is low.