Aircraft Structure Reliability Calculation Based On Kriging Model

The calculation of reliability is of great significance to the design of modern aircraft structures.In general,it is difficult to show the explicit expression of limit state function.In this case,surrogate model has significant advantages.In this paper,on the premise of both calculation accuracy and calculation cost,some researches are made on surrogate model,which are as follows:（1）An optimal subset of the Kriging surrogate model considering the global trendAn improved Kriging model is proposed in this paper and its details are as follows.The trend alternative functions are determined by using the first-order to third-order response surface methods.The order of the trend alternative function with the best fitting effect is selected as the order of the trend item in Kriging model.The unnecessary items of the trend function is optimized by genetic algorithm.The calculation accuracy of Kriging model is further improved by the active-learning function.The results show that in the case of similar sample size,the proposed method has better calculation accuracy than other Kriging model methods,with the calculation accuracy increased by 2.985%to 22.498%.（2）Reliability calculation based on improved Kriging modelThe reliability of structures with single failure and multiple failure modes are calculated respectively.Among them,in the calculation of structural reliability of multi-failure modes,an optimal subset of the Kriging surrogate model considering the global trend,only conducts one-time modeling for all functional functions,the correlation between each failure mode can be considered during the modeling process.The numerical examples and the actual engineering structure cases are compared and analyzed.The results show that,in the case of the same sample size,compared with the classical Kriging model which does not consider the trend item,the calculation accuracy of an optimal subset of the Kriging surrogate model considering the global trend has been improved slightly,the calculation precision is increased by 0.087%to 5.160%.The result shows a smaller fluctuation range,with the fluctuation range reduced by 9.515%to 29.333%.