Integrating System Physics with Sensor Data for Health Prognostics of Complex Engineered Systems

将系统物理与传感器数据相结合，用于复杂工程系统的健康预测

• 批准号: 1904165
• 负责人: Xiao Liu
• 金额: $ 39.99万
• 依托单位: University of Arkansas
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2019
• 资助国家: 美国
• 起止时间: 2019-08-01 至 2022-07-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1904165&HistoricalAwards=false
• 关键词: Integrating; System; Physics; Sensor; Data

This award will advance the Nation's economic welfare by improving the reliability and energy-efficiency of modern Data Center (DC) operations. As computational needs increase in the era of artificial intelligence (AI) and big data, DCs are becoming larger consumers of energy and currently reflect 2 percent of the US energy market. Within the DC, the cooling systems make up 30-40 percent of the total energy consumption. Cooling system failures are critical and cause interruptions on massive cloud/mobile services provided by large corporations such as Microsoft, IBM, Amazon, Netflix, etc. This award will investigate the integration of physics-based and data-driven methods for the modeling and prognostics of complex systems as represented by DC cooling systems. The project includes activities to broaden interest, especially among underrepresented groups, in STEM fields including applied probability and statistics, reliability engineering and data-analytics through hands-on activities and internship opportunities provided by the industry collaborators. Integrating research outcomes into education and support of graduate students will nurture a pool of next-generation data scientists and engineers for the northwest Arkansas and rural Virginia regions. This project will develop and validate novel methods to enable the integration of fundamental system physics and sensor monitoring data for the modeling and prognostics of complex engineering systems. The project consists of three connected research thrusts. Thrust 1 will establish a flexible two-layer physical-statistical modeling framework which enables the integration of system physics into the modeling and interpretation of sensor monitoring data. The connection between Gaussian Process (GP) regression, dynamical models, and Linear Time-Invariant (LTI) Stochastic Difference Equations (SDE) will be established in order to exploit advantages in modeling, computation and interpretation. Thrust 2 will develop a new class of multivariate stochastic models to capture the complex dynamics and dependency among multiple system health state variables. Thrust 3 will perform comprehensive validation, testing and continuous improvement of the project's methods based on real datasets provided by IBM, Arkansas High Performance Computing Center (AHPCC-UARK), and Virginia Tech Advanced Research Computing facilities (ARC-VT). This research is expected to lead to interpretable data-driven models, actionable engineering insights and explainable operational decisions with tangible impacts on the health prognostics of complex engineering systems.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

该奖项将通过提高现代数据中心（DC）运营的可靠性和能源效率来促进国家的经济福利。 随着人工智能（AI）和大数据时代计算需求的增加，DC正在成为更大的能源消费者，目前占美国能源市场的2%。 在DC中，冷却系统占总能耗的30- 40%。 冷却系统故障至关重要，会导致微软、IBM、亚马逊、Netflix等大公司提供的大规模云/移动的服务中断。该奖项将研究基于物理和数据驱动方法的集成，用于以直流冷却系统为代表的复杂系统的建模和自动化。 该项目包括扩大兴趣的活动，特别是在代表性不足的群体中，在STEM领域，包括应用概率和统计，可靠性工程和数据分析，通过实践活动和行业合作者提供的实习机会。将研究成果融入研究生的教育和支持中，将为阿肯色州西北部和弗吉尼亚州农村地区培养一批下一代数据科学家和工程师。该项目将开发和验证新的方法，使基础系统物理和传感器监测数据的复杂工程系统的建模和动力学的集成。该项目包括三个相互关联的研究重点。推力1将建立一个灵活的两层物理统计建模框架，使系统物理集成到传感器监测数据的建模和解释。高斯过程（GP）回归，动态模型和线性时不变（LTI）随机差分方程（LTI）之间的连接将被建立，以利用建模，计算和解释的优势。 Thrust 2将开发一类新的多变量随机模型，以捕获多个系统健康状态变量之间的复杂动态和相关性。Thrust 3将根据IBM、阿肯色州高性能计算中心（AHPCC-UARK）和弗吉尼亚理工大学高级研究计算设施（ARC-VT）提供的真实的数据集，对项目方法进行全面验证、测试和持续改进。该研究有望产生可解释的数据驱动模型、可操作的工程见解和可解释的运营决策，并对复杂工程系统的健康特性产生切实影响。该奖项反映了NSF的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

项目成果

Reliability analysis of artificial intelligence systems using recurrent events data from autonomous vehicles

• DOI: 10.1111/rssc.12564
• 发表时间: 2021-02
• 期刊: Journal of the Royal Statistical Society: Series C (Applied Statistics)
• 作者: Yili Hong;Jie Min;Caleb King;W. Meeker

Building degradation index with variable selection for multivariate sensory data

• DOI: 10.1016/j.ress.2022.108704
• 发表时间: 2021-10
• 期刊: Reliab. Eng. Syst. Saf.
• 作者: Yueyao Wang;I-Chen Lee;Yili Hong;Xinwei Deng

Modeling bivariate geyser eruption system with covariate-adjusted recurrent event process

• DOI: 10.1080/02664763.2021.1910937
• 发表时间: 2021-03
• 期刊: Journal of Applied Statistics
• 影响因子: 1.5
• 作者: Zhongnan Jin;Lu Lu-Lu;K. Bedair;Yili Hong

A physics-regularized data-driven approach for health prognostics of complex engineered systems with dependent health states

一种物理正则化数据驱动方法，用于对具有相关健康状态的复杂工程系统进行健康预测

• DOI: 10.1016/j.ress.2022.108677
• 发表时间: 2022
• 期刊: Reliability Engineering & System Safety
• 影响因子: 8.1
• 作者: Hajiha, Mohammadmahdi;Liu, Xiao;Lee, Young M.;Ramin, Moghaddass
• 通讯作者: Ramin, Moghaddass

Degradation under dynamic operating conditions: Modeling, competing processes and applications

• DOI: 10.1080/00224065.2020.1757390
• 发表时间: 2020-05
• 期刊: Journal of Quality Technology
• 影响因子: 2.5
• 作者: M. Hajiha;Xiao Liu;Yili Hong