摘要
利用BP神经网络具有的“信息影射”特性,实现了影响煤与瓦斯突出的因素与突出事件之间的特定刺激—反应式感知和识别,进而挖掘和捕捉二者之间的内在相关规律,并将其应用于煤与瓦斯突出区域预测。对于原始BP神经网络收敛速度慢、易陷入局部假饱和以及产生震荡等问题,通过调整学习系数、惯量系数等进行了改进,改进的BP神经网络在煤与瓦斯突出预测精度、预测效率方面明显优于传统预测方法。
With the powerful 'information mapping' function and set-study capability, back propagation neural network (BP network for short) makes it possible to perceive, distinguish the correct response to any particular input pattern, which could be used to capture the interrelated pattern of the outburst-disaster with the outbursts-factors really and to learns the laws and characteristics (especially the hides) objectively . So it could be helpful to predict coal and gas outbursts. Three methods of momentum strategy, bold driver and improved cost function are used in practice to overcome of three shortcomings of the primary BP network: slow convergence, fault saturation, and oscillation problem. The above improved BP network was applied to Jiaozuo colliery for prediction of coal and gas outburst, and better results,were obtained.
出处
《辽宁工程技术大学学报（自然科学版）》
CAS
北大核心
2004年第1期9-11,共3页
Journal of Liaoning Technical University (Natural Science)
基金
河南省自然科学基础研究基金资助(974071005)
焦作工学院博士基金(49772130)
