回归法在实时混合模拟数值子结构不确定性分析中的应用

Application of regression method in uncertainty analysis for numerical substructure in real-time hybrid simulation

为了分析实时混合模拟试验数值子结构不确定性对试验的影响,采用多项式混沌展开建立不同时滞情况下数值子结构不确定参数与位移最大值的代替模型.多项式混沌展开的系数采用回归法计算,并基于模型误差对多项式混沌模型进行评价.模拟结果表明,时滞和不确定性参数均方差越大,位移最大值的均值和方差越大.当不确定性参数均方差为均值的0.05、0.10和0.20倍时,模型误差的变化范围分别为0.002%~0.02%、0.02%~0.4%和0.5%~5%,说明基于多项式混沌展开的实时混合模拟数值子结构不确定性分析精度受不确定性参数方差的影响.

To analyze the influence of uncertainty of numerical substructures on the experimental results in real-time hybrid simulation, the polynomial chaos expansion was used to establish the alternative model between uncertainty parameters in the numerical substructure under different time delays and maximum displacements. The coefficients of the polynomial chaos expansion were calculated by the regression method, and the polynomial chaos model was evaluated based on the model errors. The simulation results indicate that the mean and the variance of the maximum displacement increase with the increase of the time delay and the stand variance of the uncertainty parameters. The model error varies from 0.002% to 0.02%, 0.02% to 0.4% and 0.5% to 5% when the ratios between the variance and the mean of the uncertainty parameters are 0.05, 0.10 and 0.20, respectively. The accuracy of uncertainty analysis for the numerical substructure in real-time hybrid simulation using polynomial chaos expansion is affected by the variance of the uncertainty parameters.