Activity inversion of radioactive source term based on convolutional neural network

The radioactive source terms are often distributed inside the detection target, which are difficult to be located and measured directly. In order to monitor the dose level of radioactive source terms during the waste treatment and nuclear facilities decommissioning. This research proposes an activity inversion method of radioactive source terms based on convolutional neural network, and establishes the correlation between the radiation field flux and the source term activity. In this method, source terms calculation models are established for waste steel barrel, concrete barrel with built-in filter, square box with built-in air filter and pippeline in primary circuit of reactor, and the samples are learned based on convolutional neural network for source term activity inversion. Compared the activity inversion results with the simulated measured values, the average error can be controlled at about 10%, which verifies the effectiveness of the proposed method. In addition, the accuracy of prediction and stability performance of this method are superior to other activity inversion methods. Which proves that the combination of convolutional neural network and source term activity inversion has high application value in the fields of waste disposal and nuclear facilities decommissioning.