基于粒子群算法优化支持向量机的岩爆预测研究

Prediction of Rock Burst by Improved Particle Swam Optimizationbased Support Vector Machine

由于地下工程的复杂性,岩爆的发生受到多种因素的影响,目前尚没有一种可靠的预测方法来对其进行预报,进而有针对性地进行工程灾害的风险控制。笔者提出将应力强度比(σ_θ/σ_c)、脆性系数(σ_c/σ_t)和弹性能量指数(Wet)作为影响岩爆的主要指标,并根据粒子群优化算法的参数选取和收敛速度快的优势及支持向量机的小样本、高维度、非线性的特性,提出了用粒子群优化算法对影响支持向量机分类性能的两个主要参数进行优化,进而获得优化的支持向量机分类器。利用PSO-SVM对在建二广九标茅田界隧道深埋变质砂岩岩爆发生情况进行预测,定量地判断该标段不存在岩爆现象,预测结果与茅田界隧道的实际情况基本相符。

Due to the complex features of rock burst hazard assessment systems, such as multi-variables, strong coupling and interference, it can not be exactly predicted for the space-time distribution of rockburst. The study proposed stress strength coefficient（σθ/σc）, brittleness coefficient（σc/σt） and elastic energy index （Wet） as inputs for the prediction of rockburst. The particle swam optimization （ PSO） was then selected to improve the training of support vector machine （SVM）. In this method, PSO is efficient in parameter selection and quick in solution convergence. Supports vector machine has advantages of small sample, high latitude and nonlinear. A rockburst dataset of 19 samples was employed to evaluate the current method for predicting rockburst, and the validity of the PSO-SVM model was carried out. The performance of the PSO-SVM model has been utilized to predict the rockburst tendency of metamorphic sandstone in a deep buried section of the Maotianjie tunnel under construction.