CFD simulation of complex rotor equipment using FloEFD: Part V – Diffusor shock waves


Karl Du Plessis

In Part IV we showed how accurately FloEFD can predict the thrust produced by the KJ-66 micro gas turbine engine. In this blog we are looking at the same engine again, but to keep it interesting we will consider the effects of operating the engine well outside of it’s duty envelope and even consider the effect of a poorly designed diffusor, where the main intent is to investigate the formation of shock waves between the vanes of the diffusor. Please note that any assumptions and modifications made are not based on actual data for this turbine engine and are purely implemented to “force” shock waves in the diffusor. To simulate the operation outside of the duty envelope the compressor inlet temperature was reduced to an extreme low of -40°C. Furthermore to simulate a poor diffusor design, the diffusor geometry was also altered such that the passages between the vanes have been narrowed slightly. For this analysis we will again rely on the sliding mesh rotation functionailty, since the rotor stator interaction is expected to be strong due to pressure build-up at the inlet of the diffusor that would impose on the impeller blades passing by, eliminating the averaging rotation option.
The analysis was simplified somewhat by not including the combustion chamber and turbine regions in the coputational domain, to limit the size of the model purely for illustative pusposes. The Rotating Region component for the compressor is defined exactly the same as in Part IV of the series. The computational domain along with the boundary conditions and rotating region are illustrated in the image below. Local mesh refinement regions were placed at the inlet to the diffusor where the shock waves are expected to form, i.e. at the minimum flow area as the flow passes through the diffusor vanes in a radial direction.


Figure 1: Boundary conditions, Compressor rotating region and shock local mesh refinement.

We have discussed FloEFD’s CAD geometry handling and meshing functionalities in Part I & II, thus we will not duplicate the discussion here again. We have also demonstrated the possibility of FloEFD to provide accurate results with relatively course meshes in Part IV. However, when compressibility effects become severe especially when shocks are involved, a high level of mesh resolution is required to capture the extreme gradients of fluid parameters across shocks, i.e. pressure, temperature and density gradients. Hence the local refinement regions at the inlet of the diffusor, of which the resulting mesh density is illustrated in Figure 2 below. In general one would prefer to refine the mesh within the vicinity of the shocks even further than shown here, to produce sharp or crisp shocks. FloEFD is very strong in compressible flow and the reader is urged to take a look at the [compressible flow with FloEFD] blog article for a really interesting and insightful read. Nonetheless, the current mesh and model setup does produce the desired effects.


Figure 2: Mesh resolution in the expected shock locations.

The animation below clearly shows the shocks forming transiently at the inlet of the diffusor vanes as the impeller blades pass by. What is also very interesting is that the shocks form on opposite sides of the diffusor and travel around the diffusor in an opposite direction to the impeller rotation. Note that the video shows one revolution of the impeller and at 80,000RPM one full revolution only takes 0.75ms. The time step size used for the analysis was 5µs. Obviously the presence of shocks in the diffusor is not desired since the flow is choked, limiting the flow rate of air into the engine and diminishing performance. Comparing the mass flow rate between the original diffusor and the altered “poor design” diffusor shows the reduction of air flow from 0.3kg/s to 0.25kg/s under the same conditions. Additional to the loss in performance a considerable amount of noise and vibration may also be induced due to the fluctuations of pressure and velocity. At 80,000RPM this can be detrimental and one could soon see engine parts flying through the air.

This blog article is Part V of a series of blog articles on the CFD simulation of complex rotor equipment.
Part I: LEGO Technic Aero Hawk
Part II: Apache Helicopter
Part III: Rotating Radar Dish
Part IV: KJ-66 Micro Gas Turbine Thrust Prediction
Part V: Diffusor Shock Waves in a Centrifugal Compressor