Study On Neural Network Based Modeling For Axial-flow Fan With Adjustable Vanes

According to the static performance curve feature of an axial-flow fan with adjustable vanes, the relation between different parameters are analyzed. A method of neural network is adopted to set up the static performance model of an axial-flow fan. Several different modeling results with traditional BP network, improved BP network and RBF network are analyzed and compared. The results show that each neural network model can reflect the parameter mapping very well. Based on the neural network static model, a real-time dynamic simulation model for an axial-flow fan is set up by using the relevant laws and principles of fans. The real-time dynamic simulation algorithm for an axial-flow fan is developed with Visual Fortran language under the System Analysis and Diagnosis System (SADS) software platform. A small process system model is built for testing the algorithm under different cases. The results demonstrate that the model is with higher static accuracy and good dynamic performance. It can meet the engineering application in both simulation and performance analysis and on-line monitoring for an axial flow fan with adjustable vanes.