Design of a Three-Passage Low Reynolds Number Turbine Cascade with Periodic Flow Conditions

• Main Author: Rogers, Daniel R.
• Format: Others
• Published: BYU ScholarsArchive 2008
• Subjects: computational fluid dynamics; optimization; wind tunnel; low Reynolds number; Mechanical Engineering
• Online Access: https://scholarsarchive.byu.edu/etd/1630; https://scholarsarchive.byu.edu/cgi/viewcontent.cgi?article=2629&context=etd

A numerical method for modeling a low Reynolds number turbine blade, the L1M, is presented along with the pitfalls encountered. A laminar solution was confirmed to not accurately predict the flow features known in low Reynolds number turbine blade flow. Three fully turbulent models were then used to try to predict the separation and reattachment of the flow. These models were also found to be insufficient for transitioning flows. A domain was created to manually trip the laminar flow to turbulent flow using a predictive turbulence transition model. The trip in the domain introduced an instability in the flow field that appears to be dependent on the discretization order, turbulence model, and transition location. The method was repeated using the Pack B blade and the same obstacles were apparent. The numerical method developed was then used in an optimization technique developed to design a wind tunnel simulating periodic flow conditions using only 2 blades. The method was first used to predict a c_p distribution for the aft loaded L1A research blade provided by the U.S. Air Force. The method was then extended to a larger domain emulating the 2 blade, 2D wind tunnel. The end-wall geometry of the tunnel was then changed using previously defined control points to alter the distribution of c_p along the suction surface of the interior blades. The tunnel c_p's were compared to the computationally acquired periodic solution. The processed was repeated until an acceptable threshold was reached. The optimization was performed using the commercially available software iSIGHT by Engineous Solutions. The optimization algorithms used were the gradient based Successive Approximation Method, the Hooke Jeeves, and Simulated Annealing.