Improved stochastic response surface method based on stepwise regression analysis
Yang Lu-feng · Jisuan lixue xuebao · 2013
An improved stochastic response surface method was proposed based on the stepwise regression analysis.The correlated random variables given the marginal cumulative distribution function were transformed into the independent standard normal variables according to the Nataf transformation,while the stochastic response of complex structure was expressed as the polynomial chaos expansion of independent standard normal variables.The optimal probabilistic collocation points were selected based on the linear independent criterion and the insignificant expansion terms were eliminated efficiently by the stepwise regression analysis.The numerical results show that the proposed method can reduce the probabilistic collocation points and unknown coefficients of polynomial expansion significantly on the condition of ensuring calculation accuracy.