A multivariate distribution model that describes the relevance of variables has a wide range of engineering applications. Based on the multiple linear autoregressive model and the sorting algorithm

a generation algorithm of multidimensional random variable with the specified edge distribution and the specified structure is studied. The algorithm was implemented using MATLAB. The proposed approach

a direct sampling Monte Carlo simulation method

is applied to determine the structural reliability and slope reliability. The problem related to correlative variables

difficult to solve by traditional Carlo Monte simulation

is solved by the proposed method. The numerical analysis shows that this method is reliable in calculation. The method can loosen the idealized restrains of theory model

and realistically reflect the practical problem. Compared the results from first-order second-moment method

it is found that the first-order second-moment method usually gives a high estimation in reliability. Failure probability is sensitive to the distribution type of the variable. Same safety factor maybe correspond to different failure probabilities. The proposed approach can be applied to calculation of complex structural system reliability and to simulation with dependent random variables in other fields.