Abstract: When the classical first-order reliability methods are used to solve the reliability problem of nonlinear structure, the accuracy and efficiency of the solution cannot meet the requirements because the calculation process stability is reduced by using the local information of current iterative design point merely. A self-adaptive conjugate aonlinear approximation method is proposed in this work. Firstly, based on the two-point adaptive nonlinear approximation method, a nonlinear approximation model of real limit state function is constructed by using the limit state function value and gradient value of the current and previous iteration design points. Furthermore, on the basis of two-point adaptive nonlinear approximation method, the conjugate search direction is constructed at the current iteration design point by using the limit state function value and gradient value information of the previous two iteration design points. With the development of the iterative process, the nonlinear exponential and the search direction in the model can be adjusted adaptively to improve the approximate accuracy of the model and alleviate the oscillation in the iterative process, then improving the stability of the calculation. Finally, three examples are given to illustrate the applicability of the proposed method.

Key words: structuralreliability, two-pointadaptivenonlinearapproximation, conjugatesearchdirection, self-adaptiveiteration, first-orderreliabilitymethod