Frequency Scanning System by Multi-frequency Excitation Spectrogram Method

本课题的目的之一是提出一种新的用于涡流多频系统仿真和缺陷识别的神经网络模型。本课题采用多频激励与谱图涡流系统（MFES）和逆神经网络模型对蒸汽发生器管中的自然缺陷进行检测和识别。结果表明，应用动态神经网络模型的电痉挛传感器具有很高的运算速度，并保证了识别结果的可靠性。其次，我们给出了在磁场中产生的巴克豪森噪声的混沌行为。这种噪声是由磁畴突然脱蚀引起的磁化跳变产生的。研究了噪声在无损检测技术中的应用。结果表明，在旋转磁场中可以连续产生巴克豪森噪声，并且该信号在可能范围内具有与被测材料中被检测或流动有关的信息。研究了利用嵌入的方法使巴克豪森噪声中混沌的存在从重构吸引子中消除。然后，我们为该系统开发了便携式仪器。

One purpose of this project is to present new neural network model used for eddy current multi-frequency system simulation and flaws identification. In this project a Multi-Frequency Excitation and Spectrogram Eddy Current System (MFES) and an inverse neural model were used to detect and identify natural flaws in steam generator tubes. It is shown that the applied dynamic neural model of the ECT sensor offers very high speed of operation and guarantees reliability of the recognition results.Next, we present chaotic behavior of the Barkhausen noise generated in magnetic fields. This noise is generated by the magnetization jumps caused by abrupt depinning of magnetic domains. We have studied the application of the noise to nondestructive testing techniques. The results showed that the Barkhausen noise could be continuously generated in a rotational magnetic field and it was within the bounds of possibility that the signal had information referent to detects or flows in measuring materials. We investigated to make the existence of Chaos in the Barkhausen noise clear from the reconstructed attractor by Embedding.Then we developed the portable apparatus for this system.