基于WOA优化神经网络的斜坡道拱顶沉降预测研究

为了更准确地预测地下矿山中斜坡道拱顶沉降的趋势，并控制预测精度，以保障矿山安全。本文提出鲸鱼算法优化神经网络的斜坡道拱顶沉降预测方法。主要步骤为：首先采取邻点中值平滑处理的方法对原始数据进行处理，将处理好的监测数据作为输入样本对BP、Elman神经网络进行训练、测试；再利用鲸鱼算法对初始权值和阈值优化，最后通过不同模型输出预测值。通过仿真实验研究表明:鲸鱼优化后的BP、Elman神经网络模型相比优化前均能更准确地预测斜坡道拱顶沉降；WOA-Elman模型的决定系数为0.948，优于WOA-BP模型0.941，但WOA-Elman模型运行时间耗费671.214 s远超于WOA-BP模型307.226 s，WOA-Elman耗费了更多的训练时间换取了少量的精度提升，大幅降低了训练效率；结合工程实例实测值、预测值的分析比较，鲸鱼算法(WOA)优化后的BP神经网络表现出了更高效且准确的斜坡道拱顶沉降预测能力。

Research on Optimization of Neural Network Based on Whale Algorithm for Prediction of Vault Subsidence in Slope

In order to more accurately predict the trend of vault subsidence in underground mine ramps and control the prediction accuracy to ensure mine safety, this paper proposes a whale optimization algorithm (WOA) enhanced neural network method for vault subsidence prediction. The main steps are as follows: firstly, the original data is processed using the method of adjacent point median smoothing, and the processed monitoring data is used as input samples for training and testing BP and Elman neural networks; then, the WOA is used to optimize the initial weights and thresholds, and finally, different model output predictions are obtained. Simulation experiments show that the BP and Elman neural network models optimized by the whale optimization algorithm can more accurately predict the vault subsidence compared to before optimization. The determination coefficient of the WOA-Elman model is 0.948, which is superior to the WOA-BP model at 0.941, but the running time of the WOA-Elman model, 671.214 s, far exceeds that of the WOA-BP model, 307.226 s. The WOA-Elman model consumes more training time for a small improvement in accuracy, significantly reducing training efficiency. Combined with the analysis and comparison of measured values and predicted values from engineering examples, the WOA-optimized BP neural network exhibits more efficient and accurate vault subsidence prediction capability.