부산대학교 도서관

A steam generator in NPS (Nuclear Power Station) usually has hundreds or even thousands of heat-transfer pipes. The flowing of liquid medium inside the pipes and the vibration outside the pipes can give rise to the flaws as wear or corrosion, which will influence the safety of the nuclear power facilities. It has been paid attention to search algorithms for detecting flaws all the time. In this paper, a method of mixing with multi-frequency eddy current signals based on NNT (Neural Network Technology) cancellation is put forward, in which, utilizing the measurement parameters about the calibration tubes, some techniques such as rotating the signal matrix, zeroing the mean of the columns of the signal matrix and NNT cancellation, are used for mixing signals, and then, the results are shown by Lissajous curve, through which the information about the flaws can be found. The practical detection result shows that, compared with the traditional mixing, this algorithm is more stable and the mixing result is relative ideal without structure information about the calibration tube.