首页> 外文会议>Annual Symposium on Quantitative Nondestructive Evaluation; 19980719-24; Snowbird,UT(US) >A NOVEL METHOD FOR AUTOMATIC DEFECT CLASSIFICATION USING ARTIFICIAL NEURAL NETWORKS
【24h】

A NOVEL METHOD FOR AUTOMATIC DEFECT CLASSIFICATION USING ARTIFICIAL NEURAL NETWORKS

机译:基于人工神经网络的自动缺陷分类的新方法

获取原文
获取原文并翻译 | 示例

摘要

In this paper, we proposed an intelligent UNDE detection and classification (IUDC) system, an important application of artificial neural networks and joint time-frequency analysis for automatic defect classification in ultrasonic nondestructive evaluation systems using estimated defect signatures. These defect signatures are then presented to a WVD computer for estimating their time-frequency representations to preserve their nonstationary nature. The large redundancy in the WVD images is optimally reduced using the PCA analyzer block. The extracted features are finally fed to a multilayer perceptron for defect classification. The superior performance of the proposed modular IUDC system has been demonstrated using synthetic ultrasonic data operating at low signal-to-noise ration levels. Important new architectures for modular IUDC systems based on the emerging multiresolution wavelet transforms and genetic algorithms are currently under investigation.
机译:在本文中,我们提出了一种智能的UNDE检测和分类(IUDC)系统,这是人工神经网络和联合时频分析在使用估计缺陷特征的超声无损评估系统中进行自动缺陷分类的重要应用。然后将这些缺陷签名提供给WVD计算机,以估计其时频表示,以保持其非平稳性。使用PCA分析器模块可以最佳地减少WVD图像中的大量冗余。最后将提取的特征馈送到多层感知器以进行缺陷分类。使用在低信噪比水平下运行的合成超声数据已经证明了所提出的模块化IUDC系统的优越性能。目前正在研究基于新兴多分辨率小波变换和遗传算法的模块化IUDC系统的重要新架构。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号