The pressurizer is one of the critical components in a nuclear reactor, whose main function is to maintain the primary system pressure within specific boundaries. The pressurizer is a large vessel that is made of carbon or low‐alloy steel shell, clad internally with a layer of stainless steel. It works with high temperature, high humidity and high pressure. The crack defects that would probably appear in the internal surface of the vessel must be detected and recorded, as well as changes to any existing known cracks. This paper presents an implementation of a mobile robot, which can sense and transmit the image of the internal surface to a remote analysis center, where the image is processed and recorded. A neural network‐based pattern classifier is employed to assist inspectors to detect the flaws that appear in the surface of the vessel. Because the real‐life defects rarely exist in the pressurizer, the training of the network becomes very difficult. A new algorithm is exploited to solve the problem. General considerations about the robot design are also presented.
MCB UP Ltd
Copyright © 2000, MCB UP Limited