摘要
将模糊理论与神经网络技术相结合,建立了基于模糊神经网络的拉索耐久性评价模型;网络训练采用改进的梯度下降动量BP算法,经过265次学习,确定了网络各层参数值,得到了网络的收敛解;最后通过5个校验样本验证了拉索耐久性模型的正确性。研究结果表明:拉索构件的耐久性评价指标分为拉索索力、锚固系统、拉索保护层、减震装置4种;采用动态BP算法对网络进行计算时,网络的收敛速度优于常用的BP算法;训练好的模糊神经网络很好地获得并储存了评价专家的知识、经验和判断,可将网络应用于拉索构件的耐久性评价。
Combining fuzzy theory and neural network theory,a durability evaluation model based on fuzzy neural network is proposed. The network are trained with dynamic BP algorithm, and the network parameters of each layer are confirmed with 265 times learning. Finally, the fuzzy-neural network is verified by five check-up samples ,.The results show that the four evaluation indexes of cable durability are cable tension, anchored device, cable cover and damping device. Dynamic BP algorithm is prior to common BP algorithm in training speed of network. The trained fuzzy-neural network can store the experts＇ knowledge and experiences, and can be used in durability evaluation of cable module.
出处
《华东交通大学学报》
2010年第2期8-12,共5页
Journal of East China Jiaotong University
基金
江西省自然科学基金(2007GZC0855)
江西省教育厅科学技术研究项目(GJJ-09143)
关键词
桥梁工程
拉索
模糊神经网络
耐久性评价
评价指标
bridge engineering
cable
fuzzy neural network
durability evaluation
evaluation index
