An optoelectronic system for automatic inspection of the external view of fuel pellets

Results of the development and testing of a system for automatic inspection of the external view of fuel pellets for fuel elements of reactors of nuclear power plants are presented. The system operates on the basis of the optoelectronic method for producing high-contrast images of the pellet surface and subsequent computer-aided processing of the obtained images. During processing of information for identifying defective pellets, characteristics of the local dispersion of image density, training algorithms, and logical rules for making decisions are used. A database containing over 1000 images of pellets and simulators has been created for debugging software and training the system. The testing of the test specimen of the inspection system showed that the probability of detecting defective pellets is high (over 95%) for the test rate of 1 item per second.