UNCERTAINTY ANALYSIS METHOD BASED ON A COMBINATION OF THE MAXIMUM ENTROPY PRINCIPLE AND THE POINT ESTIMATION METHOD

Uncertainty is inevitable in product design processes. Therefore, to make reliable decisions, uncertainty analysis incorporating all kinds of uncertainty is needed. In engineering practice, due to the incomplete knowledge, the distribution of some design variables can not be determined. Furthermore, the performance function is highly nonlinear, therefore, the high order moments of the performance function are needed to calculate the probability of failure accurately. In this paper, an uncertainty analysis method combining the maximum entropy principle and the bootstrapping method is proposed. Firstly, the bootstrapping method is used to calculate the confidence intervals of the first four moments for mixed random variables and sample variables. Secondly, the high order moments of limit state functions are estimated using the reduced dimension method. Thirdly, to calculate the probability density function (PDF) and cumulative distribution function (CDF) of the limit state functions, an optimization model based on the maximum entropy principle is formulated. In the proposed method, the assumptions that the distribution of the random variables are known and the calculation of the sensitivity for limit state function with respect to the Most Probable Point (MPP) are avoided. Finally, comparisons of results from the proposed methods and the MCS method are presented and discussed with numerical examples.