STOCHASTIC SENSITIVITY ANALYSIS METHOD BASED ON SUPPORT VECTOR MACHINE AND MONTE CARLO

The SVM (support vector machine) which possesses significant learning capacity at a small amount of information and generalization is used to regress the explicit function of structural responses. Based on the explicit function, the sensitivity coefficients of random variables are calculated. And combined with the Monte Carlo simulation, the stochastic analysis of structures can be done. The adaptive hybrid particle swarm is applied to optimize the parameters of the SVM so as to improve the computational efficiency. In order to verify the feasibility of this method, two engineering examples are analyzed so as to contrast the effect of the training samples method on the calculation accuracy. The results from these examples indicate that the complementary sampling method can achieve a more precise stochastic result in the extraction of the training samples, and the fitting probability density curves can better reflect the true situation. Meanwhile, the structure response sensitivity of the two examples is studied by the sensitivity coefficients of random variables.