Design Optimization and Analysis of Exit Rotor with Diffuser Passage based on Neural Network Surrogate Model and Entropy Generation Method

In this paper,a diffuser passage compressor design is introduced via optimization to improve the aerodynamic performance of the exit rotor in a multistage axial compressor.An in-house design optimization platform,based on genetic algorithm and back propagation neural network surrogate model,is constructed to perform the optimization.The optimization parameters include diffusion angle of meridian passage,diffusion length of meridian passage,change of blade camber angle and blade number.The impacts of these design parameters on efficiency and stability improvement are analyzed based on the optimization database.Two optimized diffuser passage compressor designs are selected from the optimization solution set by comprehensively considering efficiency and stability of the rotor,and the influencing mechanisms on efficiency and stability are further studied.The simulation results show that the application of diffuser passage compressor design can improve the load coefficient by 12.1%and efficiency by 1.28%at the design mass flow rate condition,and the stall margin can be improved by 12.5%.According to the local entropy generation model analysis,despite the upper and lower endwall loss of the diffuser passage rotor are increased,the profile loss is reduced compared with the original rotor.The efficiency of the diffuser passage rotor can be influenced by both loss and load.At the near stall condition,decreasing flow blockage at blade root region can improve the stall margin of the diffuser passage rotor.

An effective description of a periodic one-dimensional hopping model

The periodic one-dimensional hopping model is useful for studying the motion of microscopic particles in the thermal noise environment. Based on the explicit formulations of mean velocity, mean first passage time and effective diffusion constant, a general N internal states or even infinite internal states model can be approximated by a one state model that retains the basic properties of the original process. This effective description aids the analysis of biochemical and biophysical problems. This effective description also implies that, to some extent, many processes can be well described by simple two-state models, or even one-state models.