Information Theoretic Approaches for Equipment Condition Assessment, Diagnostics and Maintenance

该项目旨在发展一种信息理论方法来评估电力系统设备状况，以提高电力系统的整体完整性。鉴于对国家民用基础设施系统质量的关注，这项研究被认为特别重要。在整个电力系统的可靠性和安全性方面，单个设备的状态起着至关重要的作用。电力公司在检测加速老化和防止灾难性设备故障方面投入了大量精力。传统的维护和状态监测计划主要基于制造商推荐的固定时间间隔。近年来，由于更大的经济压力，公用事业公司已经开始转向更长的维护间隔，二是设备负荷大大超过制造商额定。在正常情况下，有足够的设计余量来允许减少维护和一些过载。最近的传感器发展提供了通过监测某些关键变量进一步扩展设备限制的能力。不幸的是，大多数设备的故障机制仅部分了解，因此开发的设备故障早期检测方法不精确且不可靠。为了克服这种印象并进行有效的诊断，对某一特定技术的广泛经验是必要的。此外，大多数设备的预期寿命相对较长，因此可能没有重要的故障日期来执行有意义的故障统计模型。综上所述，公用事业公司在对系统可靠性的长期影响了解有限的情况下采取了一些做法。他们可能不完全了解他们的设备可以被推到的极限。此外，他们缺乏计算框架来将他们的数据和经验包含在一个可行的模型中。这项工作的重点是开发一个信息理论框架，用于对设备故障的不完全知识进行建模，并能够随着经验的获得而轻松地合并新信息。长期目标是在一个模型中允许统计、过程和启发式数据，从而允许对维护间隔、诊断结论和可靠性评估提出改进建议。该项目将重点关注电力系统设备在开发该模型时的基本原理。＊＊＊

9527302 Tomsovic This project is aimed at developing an information theoretic approach for assessing power system equipment condition with the goal of improving the overall integrity of the power system. This research is felt to be particularly important in light of concerns with the quality of the nation's civil infrastructure system. In terms of overall power system reliability and security, individual equipment condition plays a critical role. Electric utilities direct significant effort towards detecting accelerated aging and preventing catastrophic equipment failures. Traditional maintenance and condition monitoring schedules were based primarily on manufacturer recommended fixed time intervals. In recent years due to greater economic pressures, utilities have begun, one, to shift towards longer maintenance intervals and two, to load equipment significantly beyond manufacturer ratings. Under normal conditions, there is adequate design margin to allow both reduced maintenance and some overloading. Recent sensor developments offer the ability to further stretch equipment limits by monitoring certain key variables. Unfortunately, the breakdown mechanism of most equipment are only partially understood so that developed methods for early detection of equipment failures are imprecise and unreliable. Broad experience with a particular technique is necessary to overcome this impression and perform effective diagnosis. In addition, most equipment has relatively long expected lifetimes so that there may not be significant failure date to perform meaningful statistical models of failures. Together the above factors, suggest that utilities are following practices with limited knowledge of the long term impacts on system reliability. They may not fully understand the limits to which their equipment can be pushed. Further, they lack the computational framework to include their data and the experience in a workable model. This work focuses on developing an info rmation theoretic framework for modeling the imperfect knowledge of equipment breakdown with the ability to easily incorporate new information as experience is gained. The long term objective is to allow both statistical, process and heuristic data in a model which allows for improved suggestions on maintenance intervals, diagnostic conclusions and reliability evaluation. This project will focus on underlying general principles of power system equipment in developing this model. ***