# Evaluation

## Comparison of CFD calculations with Experiments

Figure 5Figure 10 provide some key comparisons between computed and measured flow quantities. The main quantity used to judge the competency of the CFD simulations is the chord-wise static pressure coefficient distribution. The pressure coefficient is defined as

${\displaystyle C_{p}={\frac {p-p_{1}}{1/2\rho V_{1}^{2}}}}$

where the subscript 1 denotes the inlet.

The existence of a vast number of measured parameters for this test case facilitates the comparisons between experimental and computational results. So, velocity and turbulence intensity profiles at the suction side, the variation of the exit flow angle and pressure losses with the inlet flow angle, as well as distributions of boundary layer parameters have been presented in the relevant papers. Losses are computed using the mass average values of total pressure at inlet and exit, as

${\displaystyle \omega ={\frac {{\overline {p_{t1}}}-{\overline {p_{t2}}}}{{\overline {p_{t1}}}-{\overline {p_{1}}}}}}$

Finally, contours of the stream function and/or velocity vectors have been employed in order to confirm the existence of the bubble and the separation region at the suction side and measure its extent.

From the comparison of the computational and experimental flow quantities the following remarks appear in the relevant papers. Kang et al. (1995) reported:

• A reasonable prediction of exit angle.

• A fair estimation of the development of the shear layers.

• Confirmed the existence of the bubble existence in high flow angles but no separation was predicted.

• The use of PTMM instead of the standard version of the low-Reynolds-number turbulence model resulted in the thickening of the bubble and the appearance of separation.

• The refinement of the numerical technique and turbulence models is required to achieve close agreement with experiment.

Lien et al. (1996) stressed that:

• The H-O block grid gave a better resolution of quantities near the stagnation point as a result of a better quality grid that was generated.

• The linear models failed entirely to resolve the bubble, the non-linear responded much more sensitively.

Tselepidakis (1996) stated:

• In turbomachine blades, a turbulence model must be able both to return a good representation of the fully turbulent boundary layer and to resolve transitional characteristics at the leading edge.

• The slightly modified k–ω model has been capable of capturing the laminar leading edge separation on the suction side of the blade. In addition the size of the resulted bubble responded to incidence angle increase in close agreement to the experiment.

• The details of the interaction between free-flow turbulence and the boundary layer development on the blades are not in full agreement with the experiment in the off-design test cases.

Finally Biswas et al. (1997) reported:

• The method manages to capture the boundary layer growth as indicated by the variation of shape factor.

• Good agreement is observed between the measured and computed surface pressure distributions, the wake profiles and the wake minimum velocity.