摘要
针对径向式导叶多级泵内部流动状态复杂多变而导致其水力性能曲线难以精确测量的技术难题,采用遗传算法(genetic algorithm,GA)迭代优化反向传播(back propagation,BP)神经网络的权值与阈值,构建了基于GA-BP神经网络的径向式导叶多级泵水力性能预测模型,以MD500-57型径向式导叶多级泵为研究对象,建立了输入层为13个神经元、隐含层为10个神经元、输出层为2个神经元的GA-BP神经网络,采用正交试验方法设计了试验参数的正交试验方案,运用数值模拟计算方法对正交试验方案进行求解,获得了试验参数的训练样本,并对神经网络进行训练与测试,计算了过流部件关键几何参数的最优组合方案。试验结果表明:优化后该多级泵在设计工况下扬程增加了2.4 m,效率提高了3.34%,且高效区范围变宽。
To solve the technical problem that the flow state of radial diffuser multistage pump is complex and changeable,which makes it difficult to accurately measure its hydraulic performance curve,a genetic algorithm was adopted to optimize the weight and threshold of back propagation(BP)neural network.A hydraulic performance prediction model of radial diffuser multistage pump based on GA-BP neural network was constructed.Based on MD500-57 as the research object,a GA-BP neural network was built with the input layer to 13 neurons,hidden layer to 10 neurons,and output layer to 2 neurons.The test scheme designed by using the orthogonal test method was solved by numerical simulation method.The GA-BP neural network was trained and tested based on training samples.Then,the optimal combination scheme of the key geometric parameters in the overcurrent component was calculated.The tests show that the head of the radial diffuser multistage pump is increased by 2.4 m,the efficiency is increased by 3.34%,and the range of high efficiency zone is widened.
作者
王延锋
张连军
段海鹏
WANG Yan-feng;ZHANG Lian-jun;DUAN Hai-peng(School of Mechanical Electronic&Engineering,China University of Mining and Technology(Beijing),Beijing 100083,China;Mining Products Safety Approval and Certification Center Co.,Ltd.,Beijing 100013,China;CCTEG Tangshan Research Institute,Tangshan 063000,China)
出处
《科学技术与工程》
北大核心
2021年第4期1375-1381,共7页
Science Technology and Engineering
基金
国家重点研发计划(2017YFC0804100)。
关键词
径向式导叶多级泵
水力性能
遗传算法
反向传播(BP)神经网络
性能预测
radial diffuser multistage pump
hydraulic performance
genetic algorithm
back propagation neural network
performance prediction
