Quickest change detection, statistical learning and nonlinear filtering of jet engine data

现代喷气发动机是飞机上最昂贵的部件之一，其设计非常可靠，成本也很高(每个发动机高达2500万美元)。然而，他们仍然可能经历意想不到的灾难性失败，导致悲剧和生命损失。为了监控发动机的健康状况和性能，每台发动机配备了大约100个传感器，这些传感器测量发动机性能参数，从发动机气路的压力和温度到旋转部件的振动。拟议的合作研究的目标是开发新的数学和统计方法，利用发动机的新传感能力来(I)构建新的异常检测方案和最快变化检测算法；以及(Ii)发现不稳定极限，从而防止对发动机部件的损坏并延长其使用寿命。本文提出的研究结合基于模型和数据驱动的理论，针对我们的具体问题，开发了用于快速变化检测、统计学习和非线性滤波的高效数值算法。一旦发动机状态发生变化，必须尽快检测到该变化，同时将错误检测降至最低。该提案的主题2侧重于适用于喷气发动机产生的数据的数据驱动方法。当没有明确的动力学模型可用时，系统知识归结为实时测量，可能由过程历史补充。与行业合作伙伴TECSIS公司合作取得的进展将对用于检测燃气轮机发动机突然变化或异常的预后和健康管理(PHM)技术产生持久影响。拟议的以数据为中心的方法将使整个航空航天工业受益，并开辟新的研究视角和领域，这些领域严重依赖测量来进行系统监测和控制。

Modern jet engines are some of the most expensive components on an aircraft and are engineered to be extremely reliable at great cost (up to $25 million per engine). Yet they can still experience unexpected catastrophic failure, resulting in tragedy and loss of life. In order to monitor engine health and performance, each engine is equipped with about 100 sensors that measure engine performance parameters, from pressure and temperature of the engine gas path to vibration of the rotating components. The goals of the proposed collaborative research are to develop novel mathematical and statistical methods that exploit new sensing capabilities on engines for (i) constructing new anomaly-detection schemes and quickest-change detection algorithms; and (ii) discovering instability limits, thus preventing damage to engine components and lengthening their life-span. The research proposed here uses both model-based and data-driven theories to develop efficient numerical algorithms for quickest change detection, statistical learning and nonlinear filtering targeted at our specific problem.Theme 1, the proposed quickest-change detection, will be a vital procedure for engine performance monitoring. Once a change in an engine state has occurred, that change must be detected as soon as possible, while minimizing false detections. Theme 2 of the proposal focuses on the data-driven methods appropriate for data generated by jet engines. When no explicit dynamical model is available, system knowledge boils down to real-time measurements, possibly complemented by process history. Advances made in collaboration with the industry partner, TECSIS Corporation, will have a lasting impact on the Prognosis and Health Management (PHM) techniques for detecting abrupt changes or anomalies in gas turbine engines. The proposed data-centric methods will benefit the aerospace industry at large as well as open new research perspectives and domains that rely heavily on measurements for system monitoring and control.