基于数学形态学多尺度分析的顶板下沉量预测

利用实测顶板沉降序列预测顶板下沉量是分析顶板稳定性的有效手段。提出基于数学形态学（MM）的预测方法，将沉降序列视为一组信号，用多尺度形态学算法将其解耦为若干分量，用最小二乘支持向量机（LSSVM）算法对各分量进行预测，叠加还原得到最终预测结果。根据某金矿762采场预控顶板AIII测点实测沉降时序，建立MM-LSSVM模型进行预测研究。研究结果表明，MM-LSSVM的解耦处理强化了各分量内部规律、弱化了相互干扰。与其他传统预测算法相比，MM-LSSVM的平均绝对百分误差和均方根误差大幅下降，预测精度显著提高。

Roof settlement predicting based on multi-scale mathematical morphological analysis

Predicting roof settlement according to measured series is the main measure of roof stability analysis. Based on mathematical morphology (MM), a new method is proposed to improve predicting accuracy. Settlement series is treated as a group of signal which is decoupled to some components by using multi-scale mathematical morphological method. Least squares support vector machine (LSSVM) algorithm is used to predict each component. Rebuilding these components turns to be the final predicting result. According to the measured settlement time series on AIII monitor point of #762 stope pre-controlling roof in a gold mine, an MM-LSSVM predicting model is established to test the new method. The predicting results show that the components present more regularity and less interference than the original series. Comparing with the traditional predicting methods, the mean absolute percentage error (MAPE) and root mean square error (RMSE) of MM-LSSVM have a great decrease; and the predicting accuracy is improved obviously.