GOALI: An Integrated Framework for Stiction Detection and Compensation in Control Loops

Raghuratha Rengasamy和Randy Miller研究所：克拉克森大学提案编号：0553992标题：控制回路中粘滞检测和补偿的集成框架项目摘要这个NSF目标项目是克拉克森大学和美国霍尼韦尔为开发和实施控制器性能监控技术而共同努力的项目。过去十年见证了在线控制回路性能监测技术的出现和商业化。任何业绩监测工具的主要目标都可以归类为：(1)检测业绩下降，(2)诊断业绩下降的原因，以及(3)纠正行动。虽然第一个目标在文献中得到了全面的论述，但关于其他两个目标的工作还很缺乏。在本项目中，将重点研究振动诊断、粘滞检测和粘滞补偿。将探索解决这些问题的定量和定性技术。通过与霍尼韦尔的工业合作，所有开发的技术将在100万个工业控制回路上进行广泛验证。智能优点：将开发基于形状和基于模型的算法用于粘滞诊断。这些算法在部署过程中的可扩展性问题将通过在一百万个工业控制回路数据集上进行基准测试来研究。展望了无模型和基于模型的粘滞补偿算法的发展。初步结果表明，可以利用良性的气门运动来设计高度优化的补偿信号来进行粘滞补偿。更广泛的影响：这项工作将产生可供美国流程工业利用的工具，以保持其竞争优势。该项目产生的结果将为霍尼韦尔提供一个性能评估工具，该工具可以基于大型回路数据数据库进行基准测试，并随后在各种控制回路上实施。据估计，仅检测和诊断控制回路退化就可以使整个流程工业的能源成本降低1%，每年可能高达3亿美元。这可能会推动其他控制供应商接受并开发自己的控制器性能监控工具。传播活动包括：在档案期刊上发表文章，在针对院士和行业参与者的会议上发表演讲，以及将材料纳入国际和平研究所共同编写的一本关于故障诊断的书。克拉克森大学将对实验室实验进行升级，并将用于向本科生传授控制器性能评估的概念。PI将与Clarkson的教育管道计划密切合作，从Clarkson与其建立了衔接协议和/或谅解备忘录的大学招收有才华的未被充分代表的学生。

ABSTRACTPI: Raghunatha Rengasamy and Randy Miller Institution: Clarkson UniversityProposal Number: 0553992Title: An Integrated Framework for Stiction Detection and Compensation in Control LoopsProject SummaryThis NSF GOALI project is a joint effort between Clarkson University and Honeywell, USA for the development and implementation of controller performance monitoring techniques. The last decade has seen the emergence and commercialization of on-line control loop performance-monitoring techniques. The primary objectives of any performance-monitoring tool can be categorized as: (i) detection of performance degradation, (ii) diagnosis of the cause for performance degradation, and (iii) corrective action. While the first objective has been addressed comprehensively in the literature, work on the other two objectives is lacking. In this project, the focus will be on oscillation diagnosis, stiction detection, and stiction compensation. Quantitative and qualitative techniques will be explored for solving these problems. Through industrial collaboration with Honeywell, all the techniques that are developed will be extensively validated on one million industrial control loops.Intellectual Merit: Shape-based and model-based algorithms will be developed for stiction diagnosis. The scalability issues in the deployment of these algorithms will be studied through benchmarking on a million industrial control loop dataset. The development of both model-free and model-based stiction-compensation algorithms are envisaged. Preliminary results suggest that highly optimal compensation signals can be designed with benign valve movements for stiction compensation. Broader Impact: This work will generate tools that can be leveraged by the US process industries to maintain their competitive edge. Results arising out of the project will provide Honeywell with a performance assessment tool that can be benchmarked on a large database of loop data and subsequently implemented on a wide variety of control loops. It has been estimated that detection and diagnosis of control loop degradation alone could reduce energy cost of the overall process industry by 1% which could amount to as much as $300 million per year. This could be a driving force for other control vendors to embrace and develop their own controller performance monitoring tools. The dissemination activities include: publication in archival journals, presentation at conferences targeted at academicians and industry participants, and inclusion of material in a book on fault diagnosis that the PI is co-authoring. A laboratory experiment will be upgraded at Clarkson University and will be used to teach undergraduate students the concept of Controller Performance Assessment. The PI will work closely with Clarkson's Pipeline of Education Program to recruit talented underrepresented students from universities with which Clarkson has established articulation agreements and/or memorandums of understanding.