基于BP神经网络的盾构前方障碍物预测研究

Research About the Predict of Obstruction in Shield Construction Based on BP Neural Network

通过对盾构机的千斤顶总推力,刀盘马达总扭矩和推进速度进行BP神经网络构建,实现了对无障碍物以及素混凝土桩、钻孔灌注桩、工法桩等障碍物的分类和预测,特别是以三种参数同时作为输入特征进行识别时,平均准确率达到99.45%。通过在实际工程中的应用,证明基于BP神经网络的盾构前方障碍物分类预测方法有效,可以为后续的障碍物处理措施提供可靠依据。

This article realized the classification and prediction when shield tunnelling machine faced with no obstacles, plain concrete pile, bored pile, joist steel, etc. , which through the structure of BP neural network with input parameter of shield tunnelling machine ： the total thrust of jack, the total torque of cutter disc motor and propulsion speed. The average accuracy of identification result reached 99.45% in three parameters as the input characteristics. Through the application in practical engineering, it can be proved that the classification and prediction method based on BP neural network is effec- tive and can provide reliable basis for the subsequent process of obstruction.