PDF《基于人工神经网络的管道泵进水流道性能优化》

2

0 1

8 年9月农业机械学报第49 卷第9期doi:10.

6041 / j. issn. 1000鄄1298. 2018. 09.

015 基于人工神经网络的管道泵进水流道性能优化 裴摇 吉摇 甘星城摇 王文杰摇 袁寿其摇 唐亚静 (江苏大学国家水泵及系统工程技术研究中心, 镇江 212013) 摘要: 立式管道泵是一种具有进口弯管的单级单吸离心泵,常被应用于安装空间受限的地方. 由于进口的特殊结 构,该泵不可避免地产生了一定程度的能量损失,从而降低了整体的效率. 为了提高管道泵的性能,基于人工神经 网络进行了肘形进水流道的优化研究. 进水流道的形状可由流道中线和各截面的形状控制,选择五阶贝塞尔曲线 拟合流道中线,三阶贝塞尔曲线拟合截面控制参数沿流道中线的变化趋势. 考虑到泵实际安装需求,选取进水流 道的

11 个参数为优化变量,泵效率为优化目标. 采用拉丁方试验设计方法设计了

149 个进水流道方案,应用人工 神经网络建立了泵效率与

11 个设计变量间的高精度非线性数学表达式,采用粒子群算法对数学表达式进行了优 化,得到了肘形进水流道的最优参数组合. 研究结果表明:计算结果与试验结果在小流量和设计流量下呈现出很 好的一致性;

人工神经网络(ANN)能够准确反映泵效率和设计变量之间的关系,优化后预测值与计算值之间的偏 差为 0郾32% ;

优化后的模型相对于原始模型效率提高了 1郾17 个百分点,扬程提高了 0郾23 m,高效运行区得到拓宽;

相比于原始进口管,优化后进口管内流动得到改善. 关键词: 管道泵;

进水弯管;

自动优化;

人工神经网络;

粒子群算法 中图分类号: TH311;

O357郾1文献标识码: A 文章编号: 1000鄄1298(2018)09鄄0130鄄08 收稿日期:

2018 02 27摇 修回日期:

2018 07

05 基金项目: 国家自然科学基金项目(51879121)、 青蓝工程冶项目和江苏大学 青年骨干教师培养工程冶项目 作者简介: 裴吉(1984―),男,副研究员,主要从事离心泵不稳定流动及水力优化设计研究,E鄄mail: jpei@ ujs. edu. cn Hydraulic Optimization on Inlet Pipe of Vertical Inline Pump Based on Artificial Neural Network PEI Ji摇GAN Xingcheng摇WANG Wenjie摇YUAN Shouqi摇TANG Yajing (National Research Center of Pumps, Jiangsu University, Zhenjiang 212013, China) Abstract: Vertical inline pump is a single鄄stage single suction centrifugal pump with a bent pipe before the impeller, which is usually used in where the constraint is installation space such as pumphouses. But these unavoidable bents before the impeller inlet also result in the hydraulic losses at the entry of the pump and the decrease of efficiency. In order to improve the performance of a vertical inline pump, an optimization on inlet pipe was proposed based on artificial neural network ( ANN) and particle swarm optimization (PSO). The profile of inlet pipe was controlled by the mid curve and the shape of cross sections. The shape of mid curve was fitted by using a fifth ordered Bezier curve and the trend of parameters of cross sections along the mid curve were fitted by third ordered Bezier curves. Considering the real installation of the pump, totally

11 design parameters of inlet pipe were set as the design variables and the efficiency of the pump was set as the objective function. In order to build high鄄precision ANN model between the objective function and the

11 design variables, totally

149 groups of sample data were created by using Latin hypercube sampling. After that, the ANN model was solved for the optimum solution of the design variables of inlet pipe by using particle swarm optimization. The result showed that there was a good agreement between computational results and experimental results;