Collaborative Research: Multi-Level Data Fusion for Real-Time Prognostic Health Management of Hierarchical Systems

该奖项的研究目标是为分级工程系统开发一种新的实时预测和健康管理(PHM)方法。这一新方法的基础是融合从系统不同层面上的组件和子系统收集的数据。第一个努力是建立一个离线模型系统，从数学上集成独立的部件级生存模型，每个模型代表环境和压力等因素对部件故障风险的影响。然后，将离线模型系统与在线系统性能测量结合以形成状态空间模型，该状态空间模型将系统性能降级描述为传统上不可观测的危险状态演变的观测。在状态空间模型的基础上，将统计过程控制和非线性滤波技术相结合，发展了在线监测和预测方法。此外，还将根据监测和预测的结论进行实时维修调度和资源分配的优化决策。这些方法将通过行业支持者提供的真实数据、案例研究和试验床进行验证和实施。如果成功，本研究成果将为层次化工程系统的实时PHM过程提供新的概念、标准和算法，从而推进最先进的方法学。开发的方法可以应用于为满足关键任务行业的某些高度复杂和高级功能需求而设计的系统，如交通、能源、基础设施和制造。研究成果的传播将大大提高对这类系统故障的理解和预测。因此，可以预期改进的实时PHM实践，实现了提高系统可用性和降低维护成本。此外，这种合作研究的跨学科性质将使学生受益，使他们接触到新的课程模块和研究机会，涉及学习和应用可靠性、数据挖掘和统计学方面的先进方法。

The research objective of this award is to develop a new real-time prognostics and health management (PHM) methodology for hierarchical engineering systems. This new methodology is based on the fusion of data collected from components and subsystems on different levels of a system. The first effort is to establish an offline model system mathematically integrated from separate component-level survival models, each of which represents the impacts of factors, such as environments and stress, on a component's failure hazard. The offline model system is then incorporated with the online system performance measurements to formulate a state space model that describes the system performance degradation as an observation of conventionally unobservable hazard state evolvement. Based on the state space model, methodologies of online monitoring and prognostics will be developed by combining statistical process control and nonlinear filtering techniques. In addition, real-time decision of maintenance scheduling and resource allocation will be optimally conducted according to the monitoring and prognostics conclusions. These methodologies will be validated and implemented with the real data, case studies and testbed provided by industry supporters. If successful, the results of this research will advance the state-of-the-art methodologies by contributing new concepts, criteria and algorithms to the course of real-time PHM of hierarchical engineering systems. The developed methodology can be applied to systems that are designed to meet certain highly complex and advanced functional demands in mission-critical industries, such as transportation, energy, infrastructure, and manufacturing. The dissemination of the research results will significantly improve the understanding and prediction of failures of such systems. Consequently, improved real-time PHM practice can be expected, achieving increased system availability and reduced maintenance cost. In addition, the interdisciplinary nature of this collaborative research will benefit students by exposing them to new course modules and research opportunities that involve learning and applying advanced methodologies in reliability, data mining and statistics.