Validation and updating in a large automotive vibro-acoustic model using a P-box in the frequency domain

详细信息    查看全文

• 作者：Dooho Lee ; Nam H. Kim ; Hyeon-Seok Kim
• 关键词：Aleatory and epistemic uncertainties ; Large vibro ; acoustic FE model ; Model validation ; Parameter prioritization ; Probability ; box ; Area metric
• 刊名：Structural and Multidisciplinary Optimization
• 出版年：2016
• 出版时间：December 2016
• 年：2016
• 卷：54
• 期：6
• 页码：1485-1508
• 全文大小：
• 刊物类别：Engineering
• 刊物主题：Theoretical and Applied Mechanics; Computational Mathematics and Numerical Analysis; Engineering Design;
• 出版者：Springer Berlin Heidelberg
• ISSN：1615-1488
• 卷排序：54

文摘

In this paper, a model validation framework is proposed and applied to a large vibro-acoustic finite element (FE) model of a passenger car. The framework introduces a p-box approach with an efficient quantification scheme of uncertainty sources and a new area metric which is relevant to the responses in the frequency domain. To prioritize the input uncertainties out of the enormous FE model, the experts’ knowledge is utilized to select candidate input parameters which have large potential influences on the response of interests (ROI) among several thousands of input parameters. Next, a variance-based sensitivity analysis with an orthogonal array is introduced in effort to quantify the influence of the selected input parameters on the ROIs. The employment of the eigenvector dimension reduction method and orthogonal combinations of interval-valued input parameters provides the p-box of the ROI even if the size of the FE model is very large. A color map and the u-pooling of the p-boxes over the frequency band as well as the p-box at different frequencies are introduced to assess the model error and quantitative contributions of the aleatory and the epistemic input uncertainties to the overall variability of the ROIs in the frequency domain. After assessing the model error, the FE model is updated. It was found that the sensitivity results and the experts’ knowledge about the associated components effectively determine the modifications of the component models and the input parameter values during the updating process.