基于异步迭代算法的冲击地压预测
|
摘要
采用三层BP神经网络方法对冲击地压建立了数学模型。网络的训练算法采用具有松弛因子的动态异步迭代法,该算法在克服网络的麻痹现象及局部极小问题上都优于目前常用的训练方法,因而,采用此算法对网络进行了训练及震级的预报。基于新汶矿务局华丰煤矿1999～2000年的冲击地压现场监测数据,把冲击地压的能量、产生波的幅值、频次做为输入数据,相应期间的最大震级为输出数据,组成神经网络的训练样本及测试样本,对原始数据进行了数学预处理,网络结构采用了输入层3个结点,中间层7个结点,输出层1个结点的前向神经网络;网络最终的训练误差为0.06,预测结果的相对误差率平均为 9.2 %,预测效果比较理想。
A rockburst mathematical model is established by three layers neural network; and a new algorithm which is called asynchronous iterative algorithm with relaxation factor is trained the model. The algorithm has a better speed and avoids local minima and paralysis in some degree. As a results, maximum magnitude of rockburst is forecasted. The paper adopts 19992000抯 field rockburst data of Huafeng Coal Mine of Xinwen Mining Company and turns rockburst energy, wave amplitude and frequency into neural network input, turns related maximum magnitude of rockburst with above data into neural network output. So, we set up three layers forward neural network with three input nodes, seven hiden layers nodes and one output node. The paper selects 35 groups from 40 groups in 19992000 year and turns it into training groups, the remains is regarded as testing groups. The training program is designed by Matlab 6.1. After 50 000 iterations, the network error is 0.06 and the value is under the network抯 error precision. So, the network is convergence. At last, average error between the network forecast value and testing value is 9.2﹪,the result verifies that the method is successsful.
A rockburst mathematical model is established by three layers neural network; and a new algorithm which is called asynchronous iterative algorithm with relaxation factor is trained the model. The algorithm has a better speed and avoids local minima and paralysis in some degree. As a results, maximum magnitude of rockburst is forecasted. The paper adopts 19992000抯 field rockburst data of Huafeng Coal Mine of Xinwen Mining Company and turns rockburst energy, wave amplitude and frequency into neural network input, turns related maximum magnitude of rockburst with above data into neural network output. So, we set up three layers forward neural network with three input nodes, seven hiden layers nodes and one output node. The paper selects 35 groups from 40 groups in 19992000 year and turns it into training groups, the remains is regarded as testing groups. The training program is designed by Matlab 6.1. After 50 000 iterations, the network error is 0.06 and the value is under the network抯 error precision. So, the network is convergence. At last, average error between the network forecast value and testing value is 9.2﹪,the result verifies that the method is successsful.
引文
[1]RumelhartD E,HintonWilliamsR.Learning representation by back propagation errors[J].Nature,1986,323:533-536.
[2]OctavianStan,EdwardKamen.A local linearized least squares algorithm for training feed forward neural networks[J].Institution ofElectricalElectronics,TransationNeuralNetworks,2000,11(2):487-495.
[3]AbidS,FnaiechF,NajimM.A fast feed forward training algorithm using a modified form of the standard backpropagation algorithm[J].Institution ofElectricalElectronics,TransationNeuralNetworks,2001,12(2):424-429.
[4]王淑坤.冲击地压机理研究[J].岩石力学与工程学报,1996,15:500-503.
[5]潘一山.冲击地压发生和破坏过程研究[博士学位论文D].北京:清华大学,1999.
[6]石成钢,刘西拉.人工神经网络在震害预测中的应用[J].地震工程与工程振动,1991,11(2):39-47.
[7]王虎栓.基于人工神经网络的峰值地震动物理参数的智能判别[J].地震学报,1993,15(2):208-216.
[8]王炜,蒋春曦,张军.BP神经网络在地震综合预报中的应用[J].地震,1999,19(2):118-126.
[9]陈海军,郦能惠,聂德新等.岩爆预测的人工神经网络模型[J].岩土工程学报,2002,24(2):229-232.
[10]朱宝龙,陈强,胡厚田.基于人工神经网络的岩爆预测方法[J].地质灾害与环境保护,2002,13(3):56-59.
[11]陈刚,煤岩材料后破坏的理论研究及预测[博士学位论文D].阜新:辽宁工程技术大学,2002.
[2]OctavianStan,EdwardKamen.A local linearized least squares algorithm for training feed forward neural networks[J].Institution ofElectricalElectronics,TransationNeuralNetworks,2000,11(2):487-495.
[3]AbidS,FnaiechF,NajimM.A fast feed forward training algorithm using a modified form of the standard backpropagation algorithm[J].Institution ofElectricalElectronics,TransationNeuralNetworks,2001,12(2):424-429.
[4]王淑坤.冲击地压机理研究[J].岩石力学与工程学报,1996,15:500-503.
[5]潘一山.冲击地压发生和破坏过程研究[博士学位论文D].北京:清华大学,1999.
[6]石成钢,刘西拉.人工神经网络在震害预测中的应用[J].地震工程与工程振动,1991,11(2):39-47.
[7]王虎栓.基于人工神经网络的峰值地震动物理参数的智能判别[J].地震学报,1993,15(2):208-216.
[8]王炜,蒋春曦,张军.BP神经网络在地震综合预报中的应用[J].地震,1999,19(2):118-126.
[9]陈海军,郦能惠,聂德新等.岩爆预测的人工神经网络模型[J].岩土工程学报,2002,24(2):229-232.
[10]朱宝龙,陈强,胡厚田.基于人工神经网络的岩爆预测方法[J].地质灾害与环境保护,2002,13(3):56-59.
[11]陈刚,煤岩材料后破坏的理论研究及预测[博士学位论文D].阜新:辽宁工程技术大学,2002.
版权所有：© 2023 中国地质图书馆 中国地质调查局地学文献中心