CFD Simulation Of Complex Flows In Turbomachinery And Robust Optimization Of Blade Design

In moden industrial engineering, turbomachinery is a class of widely used power generation equipment which usually has complex configuration. The performance of turbomachinery is related closely to the complicated internal flow. A good understanding of the internal flow is the keystone of the high performance design. Moreover, advanced optimal design methodologies are also essential in modern aerodynamic designs due to the increasing rigorous performance demands. Therefore, the main research work of the present thesis is contributed to these two points.The first part of this thesis focuses on the numerical analysis of the film cooling flow and unsteady interactions between rotor and stator, which are two hot research issues in recent decades. For the film cooling flow, firstly, a relatively simple model of a flat plate with a single square cooling hole on it is adopted to investigate the effect of the blowing ratio on the local flow structure nearby a cooling hole. Both RANS simulations and DES simulations are performed. The simulated results denomstrate the development of the vortices system nearby the cooling hole, which indicates the existence of the "critical blowing ratio". The vortices shed in a stable symmetric manner when the blowing ratio is low; with the increase of blowing ratio, the vortex structure becomes unstable gradually and evolves into a stable asymmetric structure. While, the horseshoe legs and the cooling flow close to the flat plane are warped into the dominated CVP resulting a large decreasement of the cooling effect. Therefore, the blowing ratio should be lower than the critical blowing ratio. Based on the profound understanding of the local structure of the cooling flow, further steady RANS simulations are performed on the AGTB planar cascades with film cooling at the leading edge. Three different cooling holes configurations of slots, straightforward holes and radial inclined holes are simulated under different blowing ratios, respectively. Based on the validation to experiment data, the analysis mainly focuses on the impacts of the cooling flow on the main flow and the related additional losses. The simulation results show a large separation flow exists downstream the slots which leads to the deterioration of cooling effect and large additional losses; The cooling flows issue from the inclined holes are distributed more uniformly than that from the straightforward holes. However, the additional losses of the former are a little bit larger than that of the latter.For 3D unsteady effect of the rotor/stator interaction on the flows whihin blade passages, unsteady simulations using nonlinear harmonic (NLH) method are performed on a low speed axial turbine from AIST. The simulated results show the effectiveness of the NLH method on the treatment of unsteady R/S interactions. Moreover, NLH method shows high advantages in both computational labor and time to the full unsteady method and phase-lag method. Increasing the number of harmonics can improve the simulated results. However, if only one R/S interface exists, a satisfied result could be obtained by using 2 harmonics could in general. The discussions mainly focus on the unsteady properties of flow within blade passages based on the validation to experiment data. Due to the extremely low rotating speed and inlet Reynolds number, a large laminar separation region is observd on the suction side of the stator. The validation also inditcates that the compressibility of the fluid used in simulations has great influence on the simulated results. The simulated result using incompressible fluid predictes more accurate position of the separated flow than that using compressible fluid.In the second part of the present thesis, a coupled optimization framework for the aerodynamic design of turbomachinery is developed. The most widely used multi-objective genetic algorithm NSGA-â…¡which is based on Pareto optimal concept is used and CFD simulations are employed to evaluate the aerodynamic performance of turbomachinery. The effect of uncertainties on the performance are also taken into account by introducing the stochastic analysis into the CFD simulation. Artificial neural network is also used as surrogate model to reduce the computation labor of CFD. In order to verify the convergence of the coupled method, a set of classical mathematical problems, ZDT problems, are tesed. It is shown that the optimized results of NSGA-â…¡are stuck easily into a local optimum due to the predication error of ANN. Therefore, an improved crowding distance is proposed to amend the original selecting strategy and enhance the capability of jumping out the local optimum. Meanwhile, a novel coarse-to-fine iteration strategy among numerical simulations, training of ANN and optimization is employed to improve the quality of the training samples. Then the prediction accuracy of ANN will be improved which drives the optimization results to approach the real Pareto front gradually. The improved results of testing problems illustrated the effects of improving strategies.A practical industrial application of the well validated multi-objective optimizaiton framework is performed on the multi working points optimization on the NASA Rotor37. A trade-off solution is chosen using multi-criteria decision method from the converged Pareto front, which shows better performance than that of the single objective optimization. A robust optimization is also performed on rotor37 under stochastic outlet static pressures. A non-intrusive probabilistic collocation method is coupled with CFD simulations to quantify the uncertainty of total performance and flow field. The overall performance of the optimized results is improved obviously compared with that of the origin design, and the sensitiveness to fluctuations of the outlet static pressure is decreased simultaneously. Then, the robustness of the design is enhanced.The final chapter of this thesis summarizes the conclusions of the present PhD research and discusses future challenges for high credible CFD analysis and robust optimization of turbomachinery, which are very promosing research topics.