深部硬岩岩爆评判的加权马氏距离判别法

A Weighted Mahalanobis Distance Discriminant Analysis for Predicting Rock-Burst in Deep Hard Rocks

距离判别法是一种按多指标性能参数进行分类的方法,已被广泛应用在岩爆等级分类中去,但传统的距离判别分析方法对各个指标的重要性是同等对待的,夸大了一些微小变化指标的作用.采用加权马氏距离对样本中各指标进行加权处理以区分其重要性.首先对岩爆主控因子进行主成分分析,计算出一组新的指标参数和与其对应的反应指标信息的贡献率,以贡献率组成对角矩阵作为权值矩阵,然后把数据进行标准化处理,计算加权马氏距离判别岩爆等级,并与马氏距离判别法进行了比较.比较结果表明:加权距离判别法具有更高的判别能力,判别效果更符合实际情况.利用建立好的加权马氏距离模型,对马路坪硬岩矿山和铜陵冬瓜山深埋硬岩矿山岩爆进行预测.实例研究表明:预测结果与工程实例实际岩爆情况相符合.

Distance discriminant analysis method,as one of the classification methods to estimate multi-index performance parameters,is now widely used for predicting and classifying rock-burst.However,the traditional Mahalanobis distance discriminant method redards the importance of all parameters as the same.So it exaggerates the role of some those parameters which have only small changes.According to the weighted Mahalanobis distance approach,the importance of each parameter is distinguished through the concrete application.Firstly,by means of analyzing the principal components of the parameters of rock-burst,this method calculates a new group of parameters and their corresponding percentage contributions,which are formed into the weighted matrix as a diagonal matrix of the percentage contributions.Then,by setting data to standardization,the weighted Mahalanobis distance can be calculated to classify rock-burst to be applied to practical project.Through contrasting the Mahalanobis distance discriminant method,the proposed method has a high discriminant ability and more accurate prediction results.Finally,rock-burst of hard rock in the Maluping and Dongguashan mines were predicted and evaluated under this model,whose results are almost consistent with the in-situ test results.