Radioactive source item survey runs through the whole life of nuclear power plants. The investigation results of radioactive source item are the important basis for selecting decommissioning strategy, making decommissioning plan, determining dismantling and decontamination process, estimating the production and cost of radioactive waste and design of radiation protection. There are many radioactive items involved in nuclear power plants in service. In order to ensure the representativeness of sedimentary radioactive source items, this paper preliminarily determines the scope of investigation based on the generation mechanism of sedimentary radioactive source items, and then uses a HPGe detector to carry out a large range of  in-situ non-destructive measurement experiments on 4 PWR nuclear power units, and uses machine learning algorithm to analyze the experimental data. According to the analysis results, the selection of investigation objects is divided into three levels, which are refined to specific items step by step. Finally, a set of selection methods for sedimentary radioactive source items of nuclear power plants in service is established, which can provide useful reference for the investigation of radioactive source items of nuclear power plants for decommissioning.

nuclear power plant decommissioning; Radioactive source item survey; Radiation detection; Intelligent algorithm