GARBF网络法预测水泵全特性曲线

发布时间：2018-11-15 22:12

【摘要】：水泵应用于国民经济和社会发展的各个领域,其技术性能的好坏影响着应用效果的可靠性,故获得水泵的各种性能参数是很有必要的。在一般含泵装置的工程应用中,通过水泵的性能曲线来研究水泵的运行情况,但在泵系统的水力过渡过程(如水锤的计算及分析)的研究中,就需要用能反映在任意可能运行条件下的泵运行特性的全特性曲线。虽然有很多研究人员做过对水泵全特性曲线的预测工作,但结果还不够精确,不能满足某些实际工程应用的需要,且没有较实用的泵全特性预测软件。所以对水泵全特性曲线资料的预测在理论研究和实际工程应用价值方面均具有重要意义。本文的研究内容和成果有： 1、对水泵全特性曲线的几种表现方法进行研究分析后,发现x-WH及x-WM坐标水泵全面特性曲线在运用中广泛使用。对比几种神经网络模型和算法,提出遗传算法和径向基函数神经网络相结合的GARBF神经网络模型。 2、运用GARBF神经网络模型,开发出以MATLAB为设计平台的预测水泵全特性曲线的工程软件。以已知的几个比转数的水泵全特性曲线资料做为样本,进行训练预测,可以得到任意比转数泵的全特性曲线资料(以x-WH及x-WM曲线资料为表现方法)。 3、提出比较水泵全特性曲线的几种预测方法的评价系统,并运用相关系数等评价指标对用GARBF神经网络法的预测结果与已知样本资料进行拟合度判断。结果显示该方法的预测效果较好。并利用该评价系统对用GARBF网络法与四项式拟合,及BP神经网络法预测的水泵全特性曲线资料进行拟合度判断,结果表明在预测水泵全特性曲线资料(尤其对离心泵)方面,GARBF网络法比其他预测法准确度高。
[Abstract]:Water pump is applied in various fields of national economy and social development, and its technical performance affects the reliability of application effect, so it is necessary to obtain various performance parameters of water pump. In the general engineering application of pump equipment, the performance curve of pump is used to study the running condition of pump, but in the hydraulic transition process of pump system (such as the calculation and analysis of water hammer), It is necessary to use a full characteristic curve that can reflect the pump operating characteristics under any possible operating conditions. Although many researchers have done the prediction of the full characteristic curve of the pump, the results are not accurate enough to meet the needs of some practical engineering applications, and there is no practical software for predicting the full characteristics of the pump. Therefore, the prediction of the full characteristic curve of water pump is of great significance in both theoretical research and practical engineering application value. The main contents and achievements of this paper are as follows: 1. After studying and analyzing several performance methods of water pump's full characteristic curve, it is found that the x-WH and x-WM coordinate water pump's overall characteristic curve is widely used in application. By comparing several neural network models and algorithms, a GARBF neural network model combining genetic algorithm and radial basis function neural network is proposed. 2. Using the GARBF neural network model, the engineering software of predicting the full characteristic curve of water pump based on MATLAB is developed. Based on the known total characteristic curves of pumps with specific rotation number as samples and training prediction, the full characteristic curve data of arbitrary specific rotation pumps can be obtained (represented by x-WH and x-WM curves). 3. The evaluation system of several forecasting methods for comparing the full characteristic curve of water pump is put forward, and the fitting degree between the prediction result of GARBF neural network method and the known sample data is judged by using correlation coefficient and other evaluation indexes. The results show that this method has good prediction effect. The evaluation system is used to judge the fitting degree of the whole characteristic curve of water pump predicted by GARBF network method and quaternion method and BP neural network method. The result shows that the whole characteristic curve data of pump (especially for centrifugal pump) are predicted. The accuracy of GARBF network method is higher than that of other prediction methods.
【学位授予单位】：长安大学
【学位级别】：硕士
【学位授予年份】：2011
【分类号】：TH38;TP319

【参考文献】