Scholarlink

<p>Realizing Integrated Computational Materials Engineering (ICME) from a design perspective involves co-considering the interactions between decision-makers from manufacturing, materials, and product disciplines. Each discipline introduces its own sources of uncertainty, which arise from experimental variability, model assumptions, and limited data. These uncertainties propagate through the process chain and influence the final product performance. Therefore, the effective realization of the product–material–manufacturing system from an ICME standpoint requires a design approach that enables the coordination of complex interactions across the disciplines while accounting for different sources of uncertainties. In this paper, we present a data-driven framework that integrates the compromise Decision Support Problem (cDSP) construct with Bayesian Optimization (BO) for design decision support in ICME. Using the framework, we carry out data-driven inverse robust design exploration of ICME design problems under uncertainty. Within this framework, Gaussian Process surrogate models are utilized to establish Processing–microStructure–Property–Performance linkages and to quantify predictive uncertainty. While BO is used to adaptively guide data acquisition to mitigate model uncertainty in an information-efficient manner, robust design constructs are used within the cDSP formulation to manage parametric, interpolation, and experimental uncertainties by identifying robust satisficing solutions that are relatively insensitive to variability. Through this integration, uncertainty is quantified, managed, and mitigated in a unified decision-support framework. The effectiveness of the presented cDSP-based BO approach is demonstrated using a mathematical example, which illustrates how adaptive sampling progressively concentrates within robust regions of the design space rather than converging solely toward optimal solutions. The framework is then applied to an industry-representative hot rod rolling case study, which considers the interactions between the manufacturing, material, and product disciplines. The results demonstrate that the presented framework facilitates convergence toward robust satisficing solutions for a multidisciplinary problem while considering the interactions between the disciplines. Finally, the framework is structured in a problem-agnostic manner; therefore, it applies to a broad class of engineering systems characterized by multidisciplinary interactions and requires the accounting of uncertainty.</p>

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

Design for ICME: A Data-Driven Decision Support Framework for Quantifying and Managing Uncertainty

  • 期刊论文

  • 发表日期：

    2026年7月6日

查看全文
  • H M Dilshad Alam Digonta,
  • Maryam Ghasemzadeh,
  • Anton van Beek,
  • Anand Balu Nellippallil

摘要

Realizing Integrated Computational Materials Engineering (ICME) from a design perspective involves co-considering the interactions between decision-makers from manufacturing, materials, and product disciplines. Each discipline introduces its own sources of uncertainty, which arise from experimental variability, model assumptions, and limited data. These uncertainties propagate through the process chain and influence the final product performance. Therefore, the effective realization of the product–material–manufacturing system from an ICME standpoint requires a design approach that enables the coordination of complex interactions across the disciplines while accounting for different sources of uncertainties. In this paper, we present a data-driven framework that integrates the compromise Decision Support Problem (cDSP) construct with Bayesian Optimization (BO) for design decision support in ICME. Using the framework, we carry out data-driven inverse robust design exploration of ICME design problems under uncertainty. Within this framework, Gaussian Process surrogate models are utilized to establish Processing–microStructure–Property–Performance linkages and to quantify predictive uncertainty. While BO is used to adaptively guide data acquisition to mitigate model uncertainty in an information-efficient manner, robust design constructs are used within the cDSP formulation to manage parametric, interpolation, and experimental uncertainties by identifying robust satisficing solutions that are relatively insensitive to variability. Through this integration, uncertainty is quantified, managed, and mitigated in a unified decision-support framework. The effectiveness of the presented cDSP-based BO approach is demonstrated using a mathematical example, which illustrates how adaptive sampling progressively concentrates within robust regions of the design space rather than converging solely toward optimal solutions. The framework is then applied to an industry-representative hot rod rolling case study, which considers the interactions between the manufacturing, material, and product disciplines. The results demonstrate that the presented framework facilitates convergence toward robust satisficing solutions for a multidisciplinary problem while considering the interactions between the disciplines. Finally, the framework is structured in a problem-agnostic manner; therefore, it applies to a broad class of engineering systems characterized by multidisciplinary interactions and requires the accounting of uncertainty.