基于改进神经网络的离心泵性能预测模型构建  

Construction of Centrifugal Pump Performance Prediction Model based on Improved Neural Network

作　　者：孔琳[1] Kong Lin(Aviation Manufacturing Engineering School, Xi＇an Aeronautical Polytechnic Institute （Xi＇an, Shaanxi, 710089, China)

机构地区：[1]西安航空职业技术学院航空制造工程学院,陕西西安710089

出　　处：《小型内燃机与车辆技术》2018年第4期60-63,共4页Small Internal Combustion Engine and Vehicle Technique

摘　　要：针对当前离心泵运转中存在效率低以及容易出现驼峰等现象,进而造成离心泵运行不稳定的问题,结合当前的智能算法,提出一种基于改进BP算法的离心泵性能预测方法。对此,首先就离心泵的主要性能参数进行了分析,从而为后续的研究奠定基础理论知识;其次,对当前的BP神经网络算法进行分析,并结合遗传算法在全局搜索方面的优势,对改进算法的步骤进行设计;最后,以52组离心泵运转数据作为训练样本,以5组数据作为测试样本,并设定9组输入参数,以扬程和效率作为输出,对上述的模型进行验证。结果表明,无论是在预测趋势,还是在迭代次数和收敛方面,设计的算法都具有很强的优势,进而验证了改进算法的有效性。In view of the problem of low efficiency and hump prone in the operation of centrifugal pumps, and causing the unstable operation of centrifugal pumps, combined with the current intelligent algorithm, a performance prediction method of centrifugal pump based on improved BP algorithm is proposed. This arti- cle first analyzes the main character of the centrifugal pump parameters, which lays the foundation for the follow-up study of the theory of knowledge; secondly, the BP neural network algorithm at present are ana- lyzed. Combined with the advantages of genetic algorithm in the global search, the improved algorithm steps were designed. Finally, with operation data of 52 groups of centrifugal pumps as the training sample, with 5 sets of data set as the test sample, with 9 sets of input parameters, with head and efficiency set as output, the above model is verified. The results show that the algorithm designed in this paper has strong advantages both in prediction trend, iteration times and convergence, which further verify- the effectiveness of the improved algorithm.

关 键 词：BP神经网络 离心泵 遗传算法 迭代次数 仿真 

分 类 号：TH311[机械工程—机械制造及自动化]