摘要
针对传统的BP算法存在收敛性能差,易陷入局部最小的缺点,在涡流无损检测的缺陷快速识别中,提出采用遗传算法(GA)作为神经网络的学习算法。为避免网络的过早收敛,对传统的遗传BP网络进行了改进,应用自适应算法选择遗传算子值。结果表明,与BP神经网络相比,改进GA神经网络的收敛性能和推广能力都有了显著提高。
For eddy current testing, genetic algorithm （GA） was adopted, which could overcome the disadvantages of back propagation （BP） artificial neural network （ANN）, such as slow convergence and possibility of being trapped on locally minimum value. Moreover, genetic operators were selected by adaptive algorithm to avoid unwanted early convergence. Compared with BP- ANN, the precision and generalization of GA -ANN were improved remarkably.
出处
《无损检测》
北大核心
2005年第9期469-471,共3页
Nondestructive Testing
基金
河北省自然科学基金资助项目(602378)
河北省教育厅博士基金资助项目(B2001206)
关键词
涡流检测
遗传算法
收敛
自适应算法
Eddy current testing
Genetic algorithm
Convergence
Adaptive algorithm
