An improved method of one-dimensional meanline model for axial flow compressor performance prediction

One-dimensional methods have the advantage of fast computation and convergence in predicting the performance of axial flow compressors. However, the empirical models used have limited applicability and are therefore problematic regarding accuracy. In this study, a new method for correcting one-dimensional meanline model for axial compressors is established. The empirical coefficients of the air flow angle model and the total pressure loss model in the performance prediction model are optimized by Particle Swarm, targeting experimental data. Inlet velocity coefficients, Mach number with correction coefficients for each operating point are used as inputs and outputs for training the BP neural network. The coefficient calculated from neural network are then combined with the 1D meanline method to predict overall performance of the compressor. As a result, this method has a small error in the calculation of the interpolated rotation speed, with an average error less than 1.7%. When performing prediction of extrapolated rotation speed, the error is slightly higher because the inl et velocity coefficient or Mach number exceeds the network learning range, but the average error still stays below 3.8%. It is confirmed that the correction method has high accuracy and can be used in the performance prediction of axial compressor.