基于多模型的参数空间预测推断

当建立多个模型对工程结构进行数值仿真时,为了得到更可靠的预测结果,需要综合考虑模型选择不确定性和模型形式不确定性对预测结果的影响. 联合贝叶斯方法与实验数据计算不同模型的可信度,采用调节因子方法传播模型选择不确定性得到系统响应置信区间,并叠加模型形式不确定性的影响获得综合模型计算结果的置信区间,再通过插值得到关心量在预测点的置信区间. 最后通过某飞行器气动力系数的预测推断检验了该方法的可行性.

PREDICTION BASED ON MULTI-MODELS IN THE PARAMETRIC SPACE

In some engineering problems, multi-models are used to simulate the structural behavior. In order to obtain reliable prediction results, the model selection uncertainty and the model form uncertainty should be considered. In this paper, different model's degrees of confidence are computed by combining the Bayesian method with the test data. The adjustment factor approach is used to pass the model selection uncertainty into the prediction of a system response, and then the response confidence interval is obtained from the results of the synthesis model. The final simulation results are obtained by combining the confidence interval of the model form uncertainty with the results of the synthesis model. The confidence interval at the prediction position is calculated by the interpolation method. Finally, the methodology is applied to the prediction of the aerodynamic coefficient of some aircraft subjected to different angles of attack. The agreement between the prediction results and the test results shows that the method in the paper is feasible in the engineering simulation.