摘要
为了克服BP算法的缺陷与超声检测混凝土材料缺陷时收敛慢、精度低等问题,采用了蚁群优化算法与BP神经网络融合的方法,建立了蚁群神经网络的数学模型,实现了蚁群神经网络的训练,并通过实例验证了该方法的有效性。由试验得知蚁群神经网络识别混凝土缺陷时,对位置的识别比对尺寸更有效。
In order to overcome the deficiency of slower convergence speed and low accuracy of BP algorithm and BP neural network, a combination of Ant Colony optimization algorithm and BP neural network training was used and the validity of the method was verified. It was concluded that the identification of the defect location was shown more effective as compared with the defect size when the ant Colony neural network was used to identify concrete defects.
出处
《无损检测》
2013年第5期4-7,共4页
Nondestructive Testing
基金
河南省科技厅重点科技资助项目(11210231056)
南阳师范学院校级专项基金资助项目(ZX2009002)
关键词
蚁群神经网络
混凝土结构
超声检测
Ant colony neural network
Concrete structure
Defect detection
