论文部分内容阅读
建立了一个神经网络模型来预测球团矿的冷压强度,该网络模型采用三层前向BP神经网络,网络结构为12-12-1,12个输入变量分别为给料率、料层高度、焙烧温度、干透点温度、COREX煤气单耗、膨润土的添加量、生球水分、生球碳含量以及成品球的FeO、MgO、Al2O3含量和碱度;隐层含有12个神经元;输出为成品球团冷压强度;神经元激活函数选择双曲正切函数;神经网络学习算法使用的是带惯量项的误差反向传播学习算法(BP学习算法)。选取353组数据来训练和测试神经网络,其中247组数据用于训练网络,其余数据用于测试网络。测试结果表明,该网络的预测结果与实际结果的误差在3%以内,同时通过敏感性分析得出以下结论:①膨润土添加量、生球碳含量以及成品球的FeO、MgO、Al2O3含量和碱度对球团矿的冷压强度有重要影响;②增加膨润土添加量、成品球碱度、MgO含量、焙烧温度、干透点温度、COREX煤气单耗有助于改善球团矿的冷压强度;③增加FeO含量、生球碳含量、Al2O3含量、料层高度、给料率将使球团矿的冷压强度迅速下降;④增加生球水分会降低冷压强度;⑤提高球团矿冷压强度的参数设置(膨润土的添加量:0.86%～0.92%;wFeO<0.5%;生球碳含量:1.00%～1.10%;MgO含量:0.39%～0.44%);⑥在0.3～0.7范围内增加碱度不能显著改善球团矿的冷压强度。
A neural network model is established to predict the cold compression strength of pellets. The network model uses three layers of forward BP neural network with 12-12-1 network structure. The input variables are feed rate, material height, Roasting temperature, dry-point temperature, unit consumption of COREX gas, amount of bentonite added, green ball moisture, green ball carbon content and finished ball FeO, MgO, Al2O3 content and alkalinity; hidden layer contains 12 neurons; Cold compression strength of finished pellets; hyperbolic tangent function of neuron activation function; neural network learning algorithm uses error back propagation learning algorithm (BP learning algorithm) with inertia term. Select 353 sets of data to train and test the neural network, of which 247 sets of data for training network, the rest of the data used to test the network. The test results show that the error between the predicted result and the actual result is within 3%, and the following conclusions are reached through sensitivity analysis: ① The content of bentonite, the content of eutectic carbon and the content of FeO, MgO and Al2O3 Degree of cold compressive strength of pellets have an important impact; ② increase the amount of bentonite, finished ball alkalinity, MgO content, calcination temperature, dry-point temperature, COREX gas consumption will help improve the pellet cold compressive strength ; ③ increase the content of FeO, green ball carbon content, Al2O3 content, the height of the material layer, the feed rate will make the pellet cold strength decreased rapidly; ④ increase the ball of water will reduce the cold compressive strength; The parameters of strength were set (addition amount of bentonite: 0.86% ~ 0.92%; wFeO <0.5%; green carbon content: 1.00% -1.10%; MgO content: 0.39% -0.44%); Alkalinity can not significantly improve the cold crushing strength of pellets.