煤与瓦斯突出预测的Fisher判别分析法及应用

Forecasting coal and gas outburst based on Fisher discrimination analysis method and its application

在综合分析煤与瓦斯突出多种影响因素的基础上,将Fisher判别分析应用到煤与瓦斯突出预测中,结合我国典型煤与瓦斯突出煤矿17个突出实例,建立了煤与瓦斯突出预测的Fisher判别分析模型,模型回代预测的误差率为0。应用该模型对云南恩洪煤矿8个突出实例进行预测,并与单项指标法、综合指标法、BP网络进行比较。结果表明,Fisher判别分析模型具有较高的可靠性和精确性,能对煤与瓦斯突出进行有效预测。

Based on the principle of Fisher discrimination analysis, this paper intends to establish a Fisher discrimination analysis model to predict and forecast the probability of different types of coal and gas outbursts. According to the coal weight gained from the coal seam, we have divided coal and gas outbursts into 4 categories, in which there exist four discriminant functions in correspondence with the four types of coal and gas outbursts. When doing the prediction and forecast, it is necessary to work out the four function values. And, the five effective indexes may include the initial speed of methane diffusion, the consistent coefficient of coal, the gas pressure, the destructive manner of the coal, and the mining depth, which can be regarded as the discriminant factors of Fisher discriminant analysis model. As to the discriminant functions, they can be obtained through training a large set of samples study from the instances of typical coal and gas outburst mines. The ratio of the mistake discrimination may seem to be considerably depressed zero after the Fisher discriminant analysis model is trained. As compared with the single index method and comprehensive index method, the Fisher discriminant analysis model can not only help to predict accurately if any coal and gas outburst will happen, but also can reflect the risk degree. In addition, the application of the Fisher discriminant analysis model can also help to get rid of the problem of local minima of the objective function and the slow convergence speed when using BP network training samples. Thus, we have adopted the coal and gas outburst instances of Yunnan Enhong mine to verify this model, showing that the Fisher discriminant analysis model has an excellent performance and a high prediction accuracy. The results show that the Fisher discriminant analysis model enjoys a high credibility in assessing coal and gas outbursts. Thus, it can be taken as a new approach to forecasting the coal and gas outburst, which can be applied to practical engineering projects.