机构地区: 华南理工大学机械与汽车工程学院
出 处: 《现代制造工程》 2006年第4期98-100,共3页
摘 要: 超声检测缺陷识别一直是无损检测领域研究的热点。提出采用小波包分析方法来提取缺陷的故障特征,应用集成神经网络对缺陷进行识别。实验表明,集成神经网络的识别效果明显优于单神经网络,该方法为缺陷模式分类问题的解决提供了一条有效途径。 Recognition of defects in ultrasonic testing is always a hot-topic in nondestructive test domain. A way of wavelet packet analysis is proposed to extract fault features, then recognize them by using Integrated Neural Networks (INN). The result shows that the INN method's recognition rate is higher than the single neural networks, so that it provides effective means to defect classification in ultrasonic testing.
领 域: [自动化与计算机技术] [自动化与计算机技术]