The Complete Characteristics Predictions Of Pumps Based On GARBFNN

Pumps are used on various aspects of National Economy and areas of social developments. Their technical characteristics make a big difference in reliability of useful result. So it is very necessary to get the various Characters Function Curves. On some pump used projects, we study the working mood through the Characters Function Curves. If we want to figure out the Hydraulic transition process of Pumping System (for example: calculation and analysis of Water Hammer), we'll need the Complete Performance Curve of Water Pump which can represent any working condition. Though many researchers have measured the Complete Performance Curve of Water Pump, it is so inaccurate that can't satisfy some real project and there is no practical software can figure it out. So it is very important to measure the Complete Performance Curve of Water Pump. The research contents of this assay are as follows:1. Have studied and analyzed the expression methods of the Complete Performance Curve of Water Pump, it appeared that they are x-WH and x-WM coordinates were widely used. Compared with some Neural Network Models and Algorithms, it came out GARBF Neural Network Model that combined Genetic Algorithm with Radial Basis Function Neural Network.2. On the GARBF Neural Network Model, there came out the software measuring the Complete Performance Curve of Water Pump which based on the MATLAB platform. Take the data of the pump with known Specific Speed as model, we can get the data of the Complete Performance Curve of Water Pump of the pump with any Specific Speed(Expressed by x～WH and x～WM curve).3. Provide the evaluation system of prediction methods of the Complete Performance Curve of Water Pump. And judge the Goodness for Fit of predicted result of GARBF Neural Network Model and the known sample data, according to the evaluation index just like Coefficient of Correlation, etc. It turned out that it has got a better prediction result. With this evaluation system, when fitting the GARBF Neural Network with quarternomials and BP Neural Network, it was proved that GARBF Neural Network has got the more accurate results on the aspect of predicting the data of the Complete Performance Curve of Water Pump, especially for centrifugal pump.