煤矸石-水泥颗粒群匹配与性能关系的人工神经元网络

ARTIFICIAL NEURAL NETWORK OF THE RELATIONSHIP BETWEEN THE MATCHING OF PARTICLE SIZE DISTRIBUTION OF COAL REFUSE CEMENT AND ITS PROPERTIES

建立煤矸石水泥胶砂强度与影响煤矸石水泥胶砂强度的主要因素(如:水泥细度、煤矸石细度以及煤矸石与水泥的细度匹配)间的量化预测模型。采用以反向传播学习算法,即神经网络算法(backpropagationarithmetic,BP)调整网络中各权值,对煤矸石水泥体系的胶砂强度与其影响因子建立了BP神经网络模型。用另一套非建模数据进行检验。结果表明:预测值与实测值比较接近,相对误差不超过2%。这说明BP神经网络模型在本研究系统的建立是成功的,它从一些杂乱无章的数据中找出了隐含其中的规律,较好地反映了煤矸石水泥颗粒群特征参数与其胶砂强度的非线性函数映射,为有效激发煤矸石水泥强度提供了颗粒群匹配的方法。

The quantitative predictive model for the relationship between refuse cement strength and its influence factors, such as the fineness of coal refuse cement, matching of particle size distribution of coal refuse-cement and volume content of coal refuse was set up. Back propagation arithmetic (BP) was used to adjust mass values of the net. The BP artificial neural network model for the relationship between the strength of coal refuse cement and its influence factors was established. The model was investigated with another suit of data. Results of forecast value and tested value are very closed and the relative errors are less than 2%. It indicates that the BP artificial neural network model is useful the rule hidden in disorderly data was found out. It reflects the non-linear relationship between the matching of particle size distribution of coal refuse cement and its mortar strength. The work we had done will offer the method of particle size distribution for improving the coal refuse cement strength is provided.