最优子集回归在岩体分级中的应用

将最优子集回归分析引入到岩体质量分级的问题中，以工程实测岩体质量数据作为回归样本，在众多的回归方程中选择了最优的回归方程，建立了岩体质量分级公式。对模型进行了回归系数显著性、复共线性和残差图检验以及预测能力的检验，研究结果表明：最优子集分析建立的最优回归方程是合理的；回归模型性能良好，回判估计的误判率很低，预测精度高。相比距离判别分析模型，最优子集回归模型在现场岩体分级更加方便，评判因子更少，为工程建设节省了因为收集多余因素而浪费的人力、物力和时间，因此最优子集回归分析模型是一种值得推广运用的岩体分级方法。

Application of Optimal Subset Regression to Rockmass Classification

The optimal subset regression analysis is introduced into rock mass classification. A group of rock samples is used as statistical samples to establish rock classification equations, among which the optimal regression equation is selected to get a formula for the site where the samples are from. The optimal subset regression analysis model is examined in terms of regression coefficient significance, multi-collinearity and residual map, and predictive ability. Results show that the optimal equation is appropriate for rock mass clarification; besides, the model has good predictive performance. Compared with the distance discriminant analysis model, the optimal subset regression analysis model is better to classify rockmass on site with less evaluation factor and less devotion of personnel, material resources and time for test due to collecting parameters of samples, and can be used to establish rock classification equations.