CAJ | 학술논문

水力旋流器分离过程复杂,其性能指标与影响因素之间属于典型的多维非线性关系。以往旋流器分离过程的理论和经验模型大多在特定的简化条件下得到,且预测单一。为了全面预测分离器性能指标,建立了三层BP神经网络模型,通过输入结构参数和操作参数,模拟输出分离粒径、生产能力、底流质量分数等多个分离性能指标。以生产能力为例,分析了神经网络与理论和经验模型计算值的预测精度。结果表明：在各传统预测公式中,庞学诗法的预测精度最高,误差为20.88%,与其相比,BP神经网络的预测误差仅为16.64%,优于其他各模型的预测精度,且能够实现性能指标的全面预测。人工神经网络是预测水力旋流器分离性能的可靠方法。
수력선류기분리과정복잡,기성능지표여영향인소지간속우전형적다유비선성관계。이왕선류기분리과정적이론화경험모형대다재특정적간화조건하득도,차예측단일。위료전면예측분리기성능지표,건립료삼층BP신경망락모형,통과수입결구삼수화조작삼수,모의수출분리립경、생산능력、저류질량분수등다개분리성능지표。이생산능력위례,분석료신경망락여이론화경험모형계산치적예측정도。결과표명：재각전통예측공식중,방학시법적예측정도최고,오차위20.88%,여기상비,BP신경망락적예측오차부위16.64%,우우기타각모형적예측정도,차능구실현성능지표적전면예측。인공신경망락시예측수력선류기분리성능적가고방법。
Aimed at addressing complex separating process of hydrocyclone which suffers from a typical multidimensional nonlinear relationship between the influencing factors and the performance indexes,compounded by the previous theoretical and empirical models available often under simplifying some conditions and limited in prediction capability,this paper features a three-layers BP neural network model capable of predicting separated particle size,production capacity,the underflow concentration and so on,with the structure and operating parameters,for comprehensive prediction of the separator performance index.Comparison between the results derived from the BP network and the previous model associated with the production capacity shows that BP neural network boasts the prediction precision of 16.64%,comparing favourably with 20.88% for Pang Xueshi law,the best of all traditional prediction formula.The BP neural theoretical model proves a reliable way for predicting classification performance of hydrocyclones.