Scholarlink

OptimizationOptimization under uncertainties is achieving more and more agreement in the industrial designDesign community. In fact, most of the industrial processes are permeated by uncertainties, including for instance dimensional tolerances and fluctuations in the operating conditionsConditions. These uncertainties are commonly transferred to the performance of the systemSystems, which cannot be determined by a single value, but rather by a statistical distribution of results. To deal with industrial problemsProblem characterized by a large number of uncertainties and expensive simulationSimulation timeTime, it is particularly important to develop methodologies to obtain accurate results, and rely on a reduced number of sample evaluationsEvaluation at the same timeTime. In this paper two different methodologies, applied to aeronautical test cases, are presented. The first one takes advantage of multi-fidelity metamodels, consisting of combining few high-fidelityHigh-fidelity simulationsSimulation (HF) with low-fidelityLow-fidelity ones (LF), to evaluate the propagationPropagation of the uncertainties using Cokriging algorithmAlgorithms. The method is particularly efficient for large number of uncertainties since it provides quite a linearLinear correlation of the samplesSample with the uncertainties number, while other methods, such as the ones based on Polynomial Chaos Expansion, may require at least a quadratic correlation. The second methodology is instead based on the UQ application of Reduced OrderReduced-order ModelsModel (ROM), that define an equivalent CFD modelModel in functionFunction of a given set of parameters, by interpolating a sampling series of CFD snapshots. SurrogateSurrogate modelsSurrogate-model like Deep LearningLearning can be used for the interpolationInterpolation, or as alternative POD (Proper Orthogonal Decomposition) methods which are instead based on the interpolationInterpolation of a reduced number of modes (the principal components). In both cases, the ROM model can be used to instantly evaluate a Monte Carlo DOE at the variationVariational of the uncertain parameters, allowing the computation of the uncertaintyUncertainty propagationPropagation ofUncertainty-propagation the vectorial fieldField of interest.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Uncertainty Quantification Based on Multi-fidelity Metamodels and Reduced Order Models

  • 章节

  • 发表日期：

    2026年

查看全文
  • Alberto Clarich,
  • Luca Battaglia,
  • Carlo Poloni,
  • Livia Trambaiolo

摘要

OptimizationOptimization under uncertainties is achieving more and more agreement in the industrial designDesign community. In fact, most of the industrial processes are permeated by uncertainties, including for instance dimensional tolerances and fluctuations in the operating conditionsConditions. These uncertainties are commonly transferred to the performance of the systemSystems, which cannot be determined by a single value, but rather by a statistical distribution of results. To deal with industrial problemsProblem characterized by a large number of uncertainties and expensive simulationSimulation timeTime, it is particularly important to develop methodologies to obtain accurate results, and rely on a reduced number of sample evaluationsEvaluation at the same timeTime. In this paper two different methodologies, applied to aeronautical test cases, are presented. The first one takes advantage of multi-fidelity metamodels, consisting of combining few high-fidelityHigh-fidelity simulationsSimulation (HF) with low-fidelityLow-fidelity ones (LF), to evaluate the propagationPropagation of the uncertainties using Cokriging algorithmAlgorithms. The method is particularly efficient for large number of uncertainties since it provides quite a linearLinear correlation of the samplesSample with the uncertainties number, while other methods, such as the ones based on Polynomial Chaos Expansion, may require at least a quadratic correlation. The second methodology is instead based on the UQ application of Reduced OrderReduced-order ModelsModel (ROM), that define an equivalent CFD modelModel in functionFunction of a given set of parameters, by interpolating a sampling series of CFD snapshots. SurrogateSurrogate modelsSurrogate-model like Deep LearningLearning can be used for the interpolationInterpolation, or as alternative POD (Proper Orthogonal Decomposition) methods which are instead based on the interpolationInterpolation of a reduced number of modes (the principal components). In both cases, the ROM model can be used to instantly evaluate a Monte Carlo DOE at the variationVariational of the uncertain parameters, allowing the computation of the uncertaintyUncertainty propagationPropagation ofUncertainty-propagation the vectorial fieldField of interest.