Homi Bhabha National Institute, Mumbai 400094, India; Electronics Division, Bhabha Atomic Research Centre, Mumbai 400085, India.
Nuclear Power Corporation of India Limited, Mumbai 400094, India.
ISA Trans. 2019 Sep;92:180-190. doi: 10.1016/j.isatra.2019.02.011. Epub 2019 Feb 28.
Sensor real-time monitoring is an indispensable to achieve reliable plant operation along with stricter safety and environmental measures. This paper presents a statistical algorithm for sensors time-varying incipient fault detection and isolation. The proposed approach formulates the fault detection index and fault signature using the extended Kalman filter. Algorithm relaxes assumption on a monitored system stability and a priori knowledge of the fault profile. Further, fault decision statistics has been devised using Kullback-Leibler Divergence (KLD) and mixed with an Exponential Weighted Moving Average (EWMA) control chart. Pressurized water reactor nuclear power plant temperature and neutron flux sensors incipient fault detection and isolation have been demonstrated to illustrate the effectiveness of proposed methodology.
传感器实时监测对于实现可靠的工厂运行以及更严格的安全和环境措施是不可或缺的。本文提出了一种用于传感器时变初始故障检测和隔离的统计算法。所提出的方法使用扩展卡尔曼滤波器来构造故障检测指标和故障特征。该算法放宽了对被监测系统稳定性和故障轮廓先验知识的假设。此外，还使用 Kullback-Leibler 散度 (KLD) 设计了故障决策统计量，并将其与指数加权移动平均 (EWMA) 控制图混合。已经演示了压水堆核动力厂温度和中子通量传感器的初始故障检测和隔离，以说明所提出方法的有效性。
