基于多变量混沌时间序列的煤矿斜井TBM施工动态风险预测

Risk analysis on long inclined-shaft construction in coalmine by TBM techniques based on multiple variables chaotic time series

在全面分析煤矿长斜井TBM(盾构)施工动态风险特点的基础上,利用多变量混沌时间序列预测方法对其进行预测.利用主成分分析法,确定影响煤矿长斜井TBM施工风险的主要成分.对煤矿长斜井TBM施工风险多变量时间序列进行相空间的重构,确定时间延迟τi和嵌入维数mi,采用小数据量法计算煤矿长斜井TBM施工多变量风险时间序列的最大Lyapunov指数,证明了其具有混沌特性,提出了一阶局域法与双隐层神经网络的组合预测模型,该模型能够对多变量风险时间序列随时间的变化进行预测.仿真实验表明,该预测模型误差小于单变量时间序列的预测误差,具有较强的预测能力和较好的预测效果,可为煤矿长斜井TBM施工风险分析与评估提供一种新的途径.

Multi-variable chaotic time series are used to predict the long inclined-shaft construction in coalmine construction by TBM techniques, and principal component analysis （PCA） is used to determine the main factors that impact risk （shield） of the long inclined-shaft construction in coalmine by TBM techniques. Phase space of risk time series for construction by TBM are reconstructed; time delay and embedding dimension are determined. Maximum Lyapunov indexes of risk are obtained by using small data quantity method;it is found that the time series have characteristics of chaos. Prediction model is established using the combination of first-order local method and double hidden layer neural network. Simulation experiments show that the combined model has a strong ability of prediction and achieves better effect. As a result, it provides a new way for long inclined-shaft construction in coalmine by TBM techniques.