中国工程机械学报

盾构施工过程中滚刀磨损是影响施工进度和成本的关键因素。为避免盾构机滚刀磨损量超过阈值而难以及时发现的情况，提出一种基于思维进化算法(MEA)优化反向传播(BP)神经网络(MEA-BP)的盾构机滚刀磨损量预测模型，利用MEA选择BP神经网络的最优初始权值和阈值，提升BP神经网络收敛速度，降低其陷入局部极小值的概率。依据某过江隧道工程中盾构机的施工数据，建立包含盾构推力F、刀盘扭矩Q、掘进速度V、刀盘转速r、滚刀安装半径R_i和单环磨损量ω的数据样本集，对盾构机滚刀磨损量进行预测。结果表明：通过MEA-BP神经网络模型对盾构机滚刀磨损量进行预测，所测样本中81%的数据集决定系数R²≥0.80,90%的数据集预测误差ε≤5%，能够更准确地预测盾构机滚刀磨损量，为盾构施工中选择合理的开仓换刀时间提供依据。

The wear of the disc cutters is a key factor affecting the progress and cost during shield tunnelling. In order to solve the problem that detect the disc cutters wear exceeding the threshold timely, a prediction model of the disc cutter wear based on BP neural network was proposed, which optimized by mind evolutionary algorithm(MEA). The MEA was used to select the optimal initial weights and thresholds of the BP neural network, so as to improve the convergence speed and reduce the probability of it falling into the local minimum. According to the construction data of the shield machine in a river crossing tunnel project, a data sample set including shield thrust F, cutterhead torque Q, excavation speed V, cutterhead speed r, disc cutter installation radius R_i, and single ring wear ω was established to predict the wear of the disc cutters. The results show that the MEA-BP neural network prediction model is used to predict the wear of the disc cutters, 81% of the datasets have a determination coefficient R² ≥ 0. 80; 90% of the datasets have a prediction error ε≤ 5%, which can more accurately predict the wear of the disc cutters and provide a basis for selecting a reasonable opening and changing time for shield tunnelling.