中文摘要：

基于时间序列分析理论，建立了地下工程支护效果的自回归滑动平均模型（ARMA），将顶板累积垂直岩移量和锚杆轴力作为评价指标，提取监测数据趋势项及其平稳残差时序，对顶板支护效果进行预测。现场应用表明，ARMA模型预测的顶板累积垂直岩移量及锚杆轴力值与实测值相比，相对误差分别不超过3％和2％；与GA—BP神经网络法相比，ARMA模型预测结果的精度显著提高。

英文摘要：

Based on time series analysis, an ARMA model was proposed to predict the supporting effect in underground engineering. By taking cumulative vertical movement of roof rock and axial force of anchor bolt as evaluation indicators, the supporting of roof underground was predicted by applying the trend term of residual time series. Field application showed that the cumulative vertical movement of roof rock and axial force of anchor bolt predicted by ARMA model statistical testing were compared with the practical measurement, with the relative error not less than 3% and 2%, respectively. Compared with the accuracy of GA-BP neural network, ARMA model has greatly improved the accuracy of prediction.