This article presents a new automatic identification technique of structural failures in nuclear green fuel pellet. This technique was developed to identify failures occurred during the fabrication process. It is based on a smart image analysis technique for automatic identification of the failures on uranium oxide pellets used as fuel in PWR nuclear power stations. In order to achieve this goal, an artificial neural network (ANN) has been trained and validated from image histograms of pellets containing examples not only from normal pellets (flawless), but from defective pellets as well (with the main flaws normally found during the manufacturing process). Based on this technique, a new automatic identification system of flaws on nuclear fuel element pellets, composed by the association of image pre-processing and intelligent, will be developed and implemented on the Brazilian nuclear fuel production industry. Based on the theoretical performance of the technology proposed and presented in this article, it is believed that this new system, NuFAS (Nuclear Fuel Pellets Failures Automatic Identification Neural System) will be able to identify structural failures in nuclear fuel pellets with virtually zero error margins. After implemented, the NuFAS will add value to control quality process of the national production of the nuclear fuel.