Semiconductor: COLLABORATIVE RESEARCH: Hierarchical Modeling of Yield and Defectivity to Improve Factory Operations

这笔赠款为开发模型提供资金，以了解半导体制造环境中子工艺输出对产品性能特征的影响和贡献，如良率、规格符合率、缺陷、客户退货或其他质量指标。对制造过程和制造后评估之间的关系进行建模非常重要。然而，像这样的高维复杂系统在数据分析方面提出了许多独特的挑战。将开发方法论方法来解决半导体制造中出现的复杂建模问题，方法是利用分层结构中的广义线性模型(GLM)。这将通过首先开发整合数据结构的基础来实现，这些数据结构随后将用于建立和评估分层模型。这项研究还将把平面线性模型的理论扩展到层次结构。如果成功，这项研究的结果将导致作战建模和层次模型的改进。在操作层面上确定这些关系的特征将提高生产率，识别和验证质量控制参数，并能够评估相互竞争的过程模型。这项研究的结果将对制造业务产生影响，因为它将提高技术资源的利用率，并提供检查当前运营方法和指标的有效性和有效性的手段。将在以下几个领域取得方法学上的进展：通过分层建模，提高使用数据估计高维系统的最新水平；探索数据处理策略与大型和复杂制造系统建模之间的相互作用；结合非线性统计建模的进展，开发高维建模算法。

This grant provides funding for developing models to understand the influences and contributions of sub-process outputs on product performance characteristics, such as yield, conformance to specifications, defectivity, customer returns, or other quality measures in a semiconductor manufacturing environment. Modeling relationships between fabrication processes and post-fabrication assessment is important. However, high dimensional complex systems, such as this, present a number of unique challenges in data analysis for understanding. Methodological approaches will be developed to solve the complex modeling problems that arise in semiconductor manufacturing by utilizing Generalized Linear Models (GLMs) in a hierarchical structure. This will be accomplished by first developing the foundation for integrating data structures that will be subsequently used in building and evaluating hierarchical models. This research will also extend the theory of flat, linear models into a hierarchical structure.If successful, the results of this research will lead to improvements in operational modeling and hierarchical models. Characterizing the relationships at the operational level will improve productivity, identify and validate quality control parameters, and enable evaluation of competing process models. The results of this research will impact manufacturing operations in that it will improve the utilization of technical resources and provide the means to examine the validity and effectiveness of current operating methods and metrics. Methodological advances will be made in several areas: advance the state of the art in using data to estimate high-dimensional systems by means of hierarchical modeling; exploring the interaction between data manipulation strategies and modeling large and complex manufacturing systems; and developing high dimensional modeling algorithms in combination with advances in nonlinear statistical modeling.