中文摘要：

为更好地预测煤的成浆性,以大量煤种成浆浓度试验数据为基础,建立了3个输出因子的神经网络成浆浓度预测模型,模型采用L-M算法,对输入数据进行数据预处理,最后对比分析了神经网络预测模型与回归分析模型的预测结果。结果表明,以Ad、哈氏可磨性指数HGI和氧含量O为输入因子的模型预测结果平均绝对误差为0.63%,以M（ad）、HGI和O为输入因子的模型预测结果平均绝对误差为0.60%,以M（ad）、HGI和氧碳比O/C为输入因子的模型预测结果平均绝对误差为0.40%,3种组合的模型结果均小于回归分析模型的平均绝对误差1.15%。因此神经网络模型比回归分析模型有更好的预测能力,其中以M（ad）、HGI和O/C为输入因子的神经网络模型预测结果最好。

英文摘要：

In order to improve prediction accuracy,based on experimental data of coal slurrying,BP neural network model with three input factors was set up for slurry concentration prediction. The BP neural networks＇ algorithm was Levenberg- Marquardt algorithm. The input data was treated in order to get accurate results. The A-）d,HGI,O input factors neural network model＇s mean absolute errors was 0. 63%,the M-）（ad）,HGI,O model＇s mean absolute error was 0. 60% and the M-）（ad）,HGI,O/C model＇s mean absolute error was 0. 40%,but the exist regression model＇s mean absolute error was 1. 15%,so the neural network models were effective in predicting the slurrying,and the M-）（ad）,HGI,O / C model was the best among the three prediction models.