Detection of fuel failure in pressurized water reactor with artificial neural network. (1st June 2020)

• Record Type: Journal Article
• Authors: Dong, Bing; Xiao, Wei; Yin, Junlian; Wang, Dezhong
• Is Part Of: Annals of nuclear energy. Volume 140(2020)
• Publication Date: 2020-06-01
• Subjects: Artificial neural network -- Fuel failure detection -- Isotopic ratios method -- Booth-type diffusion model -- First-order kinetic model; Nuclear energy -- Periodicals; Nuclear engineering -- Periodicals; 621.4805
• DOI: 10.1016/j.anucene.2019.107104
• Languages: English
• ISSNs: 0306-4549

Abstract

Highlights: An artificial neural network (ANN) method for fuel failure detection is proposed. The training set is generated by Booth-type diffusion model and the first kinetic model. Comparing with the isotopic ratios method, the ANN is more responsive to detect the fuel failure. Abstract: During the reactor operation, it is necessary to detect the degree of the fuel cladding failure in real time, which is helpful to determine whether discharging the defective fuel rod to prevent the radioactivity being released into the environment. Traditionally, the most commonly used method for fuel failure detection is the isotopic ratios method. In this work, an artificial neural network (ANN) method for fuel failure detection is proposed. The inputs of the ANN are specific activities of fission products in the primary coolant. The outputs of the ANN are 6 degrees of the fuel cladding failure. The value of neurons in the output layer represents the probability of the corresponding degree. The training set is generated by Booth-type diffusion model and the first-order kinetic model. The performance of the ANN is presented in the paper. The validation results show that the ANN works well for fuel failure detection. Comparing with the isotopic ratios method, the ANN is more responsive for fuel failure detection. Furthermore, the benchmark using the real reactor monitoring data shows that the ANN is able to capture the fuel failure onset in time.