Research On The Discriminates System Of The Monitoring On-line Thermodynamic Parameters Of Turbomachinery

In this thesis, a software system, which real-time discriminates on-line monitoring thermodynamic parameters of turbomachinery, is developed, applying the artificial neural networks method and based on the work of Evaluations to Thermodynamic Parameters of Centrifugal Compressors. This system can on-line discriminate the thermodynamic parameters such as pressure and temperature at each inlet and outlet, mass flow and rotary speed. In the thesis, it is thoroughly analysed the mechanism of different parts of BP neural network algorithm and made several improvement on the basic algorithm based on integrating former researches. At last the software system, discriminating on-line monitoring thermodynamic parameters of turbomachinery on the fundament of BP neural network algorithm, is designed and completed, which can solved all kinds of restrains due to traditional analysis redundancy method and have excellent discriminating performance in both steady and dynamic conditions of turbomachinery. At the same time, the computing and analyzing instances, concerned with the GB2O1 and 1Ol-J compressor sets, are presented, of which the results show that the BP neural networks method can be successfully applied on discriminating thermodynamic parameters of turbomachinery. The application of the software system is undoubtedly significant to reduce the maintenance efforts of the data sampling system and improve the reliability of the successive set performance analysis and fault diagnosis to the thermodynamic parameters.