基于人工智能方法的岩爆预测系统  

Research on Prediction System for Rockburst Based on Artificial Intelligence Application Methods

作　　者：彭琦[1] 钱爱国[2] 肖钰 

机构地区：[1]四川大学水利水电学院,四川成都610065 [2]中国水电顾问集团华东勘测设计研究院,浙江杭州310014 [3]浙江省水利水电勘测设计院,浙江杭州310002

出　　处：《四川大学学报（工程科学版）》2010年第4期18-24,共7页Journal of Sichuan University (Engineering Science Edition)

基　　金：国家基础研究发展规划(973计划)资助项目(2002CB412705);国家自然科学基金创新群体项目(50221402);国家自然科学基金委雅砻江水电开发联合研究基金重点项目(50639100);国家自然科学基金资助项目(50579042)

摘　　要：通过理论分析预测法和现场实测法,建立了一套长期预测和短期预测相结合的岩爆预测系统。长期预测模型是基于神经网络思想,运用国内外相关工程的岩爆资料作为训练样本,建立了小波神经网络预测模型,对工程范围内岩爆发生趋势进行了预测。短期预测模型首先针对监测到的声发射建立小波神经网络模型,对声发射时间序列进行拟合预测;再运用突变理论对预测的声发射建立了岩爆突变预测模型,进而对监测点附近岩爆发生情况进行准确的预测。两种预测模型都运用到了目前人工智能方法中比较新颖的小波神经网络理论,提高了收敛速度,容错能力,保证了预测的效果。通过工程实际运用,建立的岩爆预测系统预测精度高,预测结果与现场情况一致。两套预测模型可以适用于不同的工程阶段,互相验证,具有很好的工程实用性。Based on theoretical analysis and on-the-spot monitoring methods,a prediction system for rockburst consisting of long-term and short-term predicting models was proposed. The long-term predicting model adopted a wavelet neural network predicting model by using the rockburst materials of underground projects at home and abroad,so as to forcast the trend of rockburst. In the short-term prediction model,a wavelet neural network model based on the Acoustic Emission（ AE） monitored was established to forecast the future AE firstly,and then a catastrophe prediction model for rockburst was founded based on AE forecasted in order to forcast the rockburst near the monitoring site accurately. The two models both used wavelet neural network theory,and can enhance the rate of convergence and fault-tolerant capability,and assure the effects of prediction. A practical example showed that the prediction system has high accuracy,and the prediction results accord with the field performances.

关 键 词：岩爆 预测系统 长期预测模型 短期预测模型 声发射 小波神经网络 突变理论 

分 类 号：TU443[建筑科学—岩土工程]