Abstract: By changing operating conditions, such as conveying pressure P₁, total differential pressure ΔP_T and fluidizing gas flow rate Q_f, and by changing properties of pulverized coal, such as particle size d_p, moisture content W and coal category, experimental investigations on solids flux ψ of pulverized coal were conducted on a pilot scale experimental setup of high-pressure and dense-phase pneumatic conveying. Considering the complexity of high-pressure and dense-phase solids-gas flow, artificial neural network(ANN)was introduced based on sufficient experiments.Original BP neural network and two kinds of improved algorithm were used and their simulation and prediction results were compared.The results showed that ψ increased with P₁ and ΔP_T, and it increased firstly and then decreased with Q_f.ψ reduced with the increase of d_p and W, and it was also affected by coal category.The two kinds of improved algorithm in BP neural network could successfully simulate and predict ψ of pulverized coal, however, convergence speed and prediction accuracy could not be satisfied simultaneously.The work in this paper can guide the control, operation, simulation and prediction of key characteristic parameters for high-pressure and dense-phase pneumatic conveying system.

Key words: high-pressure, pneumatic conveying, solid flux, neural network

摘要： 在高压密相气力输送中试试验台上,研究了输送压力P₁、总差压ΔP_T和流化风量Q_f等操作条件,以及煤粉平均粒径d_p、含水率W和煤粉种类等物性参数对煤粉输送速率通量ψ的影响规律。鉴于高压密相气固两相流的复杂性,在充分试验的基础上,先后采用原始BP神经网络和两种改进算法,对ψ进行模拟和预测,并比较各算法的优劣。研究结果表明:ψ随着P₁和ΔP_T的增大而增大,随着Q_f的增大而先增大后减小;ψ随着d_p和W的增大而减小,也受煤粉种类的影响;两种改进算法的BP网络可以对研究对象进行较好的模拟和预测,但收敛速度和预测精度不可兼得。试验结果将对高压密相气力输送系统的操作运行以及关键特征参数的模拟预测起到一定的指导作用。

关键词: 高压, 气力输送, 输送速率通量, 神经网络