Online Stress Corrosion Crack Monitoring in Nuclear Reactor Components Using Active Ultrasonic Sensor Networks and Nonlinear System Identification - Data Fusion Based Big Data Analytics Approach

摘要

The current state of the art nondestructive evaluation (NDE) techniques used in nuclear reactor structural inspection are manual labor intensive, time consuming, and only used when the reactor has been shut down. Also, despite periodic inspection of plant components, a failure mode such as stress corrosion crack can initiate in between two scheduled inspections and can become critical before the next scheduled inspection. In this context, real time monitoring of nuclear reactor components is necessary for continuous and autonomous monitoring of component structural health. In this research, an active ultrasonic based on-line monitoring (OLM) framework is developed which can be used for real-time monitoring of degradation of nuclear power plant systems, structures, and components. Nonlinear system identification technique such as Bayesian Gaussian Process technique method is investigated to estimate the structural degradation in realtime. Active broadband ultrasound input is used for damage interrogation and a multi-sensor configuration is implemented to improve spatial resolution of state estimation. The damage index at any particular time is computed using nonlinear techniques such as Gaussian Process probabilistic modeling and the necessity of sensor data fusion is evaluated. The framework was demonstrated through the monitoring of an anomaly trend in a nuclear reactor steam generator tube undergoing stress corrosion cracking (SCC) testing.