SOM-SVM Prediction Model Of Coal And Gas Burst

Coal production is a mining industry with high risk. In the production process potentialsafety problems, for example, gas, floods and landslides always with it. What s more, coal andgas bursting is one of the most serious and devastating safe accident, so mine needs toconcentrate much manpower and material resources into the outburst prevention measures. Inorder to reduce the blindness of outburst prevention measures and take reasonable measures, itis important to research some predictive methods about coal and gas bursting. With thesemethods, we can improve the precision and efficiency of prevention measures and providescientific proof for carrying out the outburst prevention measures.Self-organizing feature map imitates the self-organizing mapping function of brain nervesystem, from doing this, it can pick up brain nerve system and some inherent law from acombination of statistics, then classify according to the discrete time way. This classificationreflects the essential difference from the sample set, and it is an unsupervised learningalgorithm. Support vector machine is based on statistics theory; it is a supervised learningalgorithm. According to structural risk minimization principle, it increases the generalizationof machine learning to the greatest extent. It has many special advantages on solving smallsample, misalignment and high dimensional pattern recognition.In unsupervised learning algorithm, classification results are often not satisfactory.However in supervised learning algorithm, the classification accuracy depend on the trainingsamples, which need to be chosen by manual, and these samples directly affect theclassification accuracy. What s more, this method s training efficiency will greatly reduce formass data set. For this reason, this paper presents a new pattern classification method of SOMcombining withSVM. It cannot only meet SVM s requirement of the training sample size, butalso ensure the classification accuracy. This paper will take it used in the coal and gas outburstprevention.This paper is based on the mechanization and the general situation of coal and gasoutburst, understanding the influencing factors first, then comparing and improvement the existing classification methods according to the current research status of outstanding forecastat home and abroad and the research result. Because coal and gas outburst prediction is a multiclass classification problem, this paper applies the Matlab and Libsvm software to write thisclassification algorithm, then sets up SOM-SVM prediction model of coal and gas burst.According to field measurement data, this paper carries out the coal and gas outburst riskprediction. Results indicate that using SOM-SVM model can forecast rapidly and accuratelythe outstanding level of risk in the mining area. And then workers can reasonably plan themining scale and progress. On the one hand, it can save the unnecessary outburst costs andimprove economic efficiency; on the other hand, it can ensure safe production. So thecombination of two algorithms produces an effective prediction of the danger of coal and gasoutburst. Compared with traditional prediction methods, this method produces quickerclassification speed, higher fault tolerance and better accuracy.