Loading...

Detecting of Faulty Sensor and Its Value Reconstructing in Small Break Loss of Coolant Accident using Neural Networks

Mohammad Parchavi, Zeinab | 2024

0 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 57590 (46)
  4. University: Sharif University of Technology
  5. Department: Energy Engineering
  6. Advisor(s): Vosoughi, Naser; Ghaffari, Mohsen
  7. Abstract:
  8. Sensors are one of the most important parts of safety and control systems in nuclear power plants. Failure to calibrate the sensors or their failure in a power plant and displaying incorrect values can lead to wrong and sometimes disastrous decisions. The first idea to check the sensors is to physically visit them. This method requires the reactor to be shut down and increase the radiation exposure of the operators and reduce the reliability and useful life of the equipment. Another approach is to use hardware redundancy. In this method, additional sensors are used to measure a variable, and the average value of the sensors is used as a reference for cross-checking. The limited available space in the reactor and the imposition of economic costs due to the increase of sensors practically make the use of this method inappropriate. Another approach is to use analytical redundancy, which analytically estimates the output of a sensor according to other parameters related to it. This approach is divided into 2 categories: 1- signal-based and 2- data-oriented. Neural networks are one of the data-oriented methods, whose application in the field of identifying damaged sensors and reconstructing its signal value has been proven in many studies. In this research, using the multi-layer perceptron neural network to identify the broken sensor and using the adversarial generator neural network and the multi-layer perceptron neural network and auto-encoder neural network to reconstruct its signal value in the event of the loss of the cooling fluid of the first circuit due to a small failure. (SBLOCA) has been paid in Bushehr power plant. The input data required for training the neural network is the data received from the simulation of the accident in Bushehr power plant in the RELAP code
  9. Keywords:
  10. Sensor Fault ; Neural Networks ; Multi-Layer Perceptron (MLP) ; Generative Adversarial Networks ; Autoencoder Neural Networks ; Autoencoder ; Faulty Sensor Signal Reconstruction ; Cooling System Accident Loss

 Digital Object List

 Bookmark

No TOC