Collaborative Research: Collaborative Degradation Analysis for Enterprise-Level Maintenance Management via Dynamic Segmentation

在许多运行大量单元的应用程序中，出现了对企业级管理的新需求，这需要彻底了解它们的降级模式。虽然最近传感技术的进步提供了前所未有的数据收集机会，但开发所需的企业级框架仍面临一些挑战。一种常见的做法是确定一个具有代表性的退化模型，该模型假设所有单元在其整个使用寿命期间都是同质的。这种方法捕获平均特征，但忽略了单元之间的差异和不同单元所采取的不同退化路径。另一种选择是对每个单位进行个性化的管理操作，考虑到企业层面上涉及的单位的绝对数量，这种方法要么难以处理，要么代价高昂。该项目将导致一个可实施的集成框架，用于学习大量单元的异构退化过程，并指导有限的监测和维护资源的分配。这项研究的结果将使经营大量工作单位的各种美国制造或生产企业受益。这项研究与教育工作相一致，通过将代表性不足的本科生指导整合到高级研究中，K-12外展计划结合基本和高级工程设计活动，为学生提供与工业现场工程师互动的机会，并与国际合作者合作，为制造业企业准备下一代工程劳动力。该项目的目标是为制造企业创建一种协作预测和健康管理方法。综合框架将通过研究单个单元之间的异质退化过程来对大量单元的异质退化过程进行建模：总体特征将由形成企业知识库的可管理数量的规范模型来表示，而个体退化特征将通过动态分割来捕获，该动态分割将每个单元的退化模式与规范模型之间的相似性建模。研究结果将有助于以下科学进展：1)建立一种新的协同降解模型方法，该方法可以同时表征种群水平和个体水平的降解机制异质性；2)一种可扩展的感知方法，该方法可以将统计预测信息和分段结构结合起来，从而有效地监测大量单元；3)在进行多单元维修时，利用成本结构的相互依赖性，使总体成本最小化的企业级维修决策。

There is an emerging need for enterprise-level management in many applications where a large number of units operate, which requires thorough understanding of their degradation patterns. While recent advancements in sensing technology provide unprecedented data collection opportunities, developing the desired enterprise-level framework, however, faces several challenges. A common practice is to identify a representative degradation model that assumes the homogeneity of all units throughout their operational life. Such approaches capture average characteristics, but ignore differences among the units and the different degradation paths taken by different units. Another alternative, that of individualizing management operations for each unit, is either intractable or unrealistically costly, given the sheer number of units involved at the enterprise level. This project will lead to an implementable integrated framework for learning heterogeneous degradation processes of a large number of units and guiding the allocation of limited monitoring and maintenance resources. The results from this research will benefit a variety of US manufacturing or production enterprises that operate massive number of working units. This research aligns well with the educational efforts to prepare the nation's next-generation engineering workforce for manufacturing enterprises via integration of underrepresented undergraduate student mentoring into advanced research, K-12 outreach programs incorporating basic and advanced engineering design activities and opportunities for students to interact with field engineers in industry and to partner with international collaborators.The objective of this project is to create a collaborative prognostics and health management methodology for manufacturing enterprises. The integrative framework will model the heterogeneous degradation processes of a large number of units by investigating the differences and similarities among individual units: the population characteristics will be represented by a manageable number of canonical models forming an enterprise knowledge base, whereas the individual degradation characteristics will be captured via dynamic segmentation that models the resemblance between each unit's degradation pattern with the canonical models. The results will contribute to the following scientific advancements: 1) a new collaborative degradation modeling method which can characterize both population-level and individual-level heterogeneities in their degradation mechanism; 2) a scalable sensing method which can incorporate both statistical prognostics information and segment structure for effectively monitoring a large number of units; 3) an enterprise-level maintenance decision-making which can minimize the overall costs by exploiting the interdependency of the cost structure while conducting multi-unit repairs.