修剪型神经网络在锚杆锚固缺陷识别中的应用

Application of pruning type neural network in defect identification of blot anchoring systems

锚杆在桥梁、隧道、建筑等方面应用越来越广泛。在施工的过程中,由于地质条件、材料和施工等因素的影响,锚固系统会产生许多缺陷。这些缺陷都会对锚杆的寿命和安全性能造成影响,所以对锚杆的缺陷识别是一项很有价值的研究。人工神经网络作为一个智能的分类器,可以对锚杆的缺陷进行识别分类,提出一种自适应阈值前馈神经网络修剪算法,其实质是通过判断隐含层神经元在学习过程中对输出的贡献值,利用显著性指数作为指标来删除网络中的冗余节点,实现网络结构的动态优化调整。结果表明,该方法能够降低网络结构的复杂度,同时提高了锚杆缺陷分类识别的精度。

Rock bolts are widely applied in bridges,tunnels,and buildings etc.In construction processes,due to influences of geological conditions,materials,architectures and other factors,there are many defects in anchoring systems.All these defects affect the life and safety of rock bolts,so it is very valuable to identify defects of anchor bolts.Artificial neural network can be used as an intelligent classifier to identify and classify defects of anchor bolts.Here,an adaptive threshold feed-forward neural network pruning algorithm was proposed,its essence was to judge the contribution value of hidden layer neurons in their learning processes to output,take the significant exponent as an index to delete redundant nodes of the network,and realize dynamic optimization and adjustment of the network structure.The simulation results showed that the proposed method can not only reduce the complexity of the network structure,but also improve classification and identification accuracies of anchor bolt defects.