Impingement and impingement/effusion cooling of gas turbine components: conjugate heat transfer predictions

Conjugate heat transfer (CHT) and computational fluid dynamics (CFD) were combined in this work using ICEM meshing and ANSYS Fluent software. Block-structured grids with hexahedral elements were used to investigates the key features of impingement cooling of gas turbine metal surfaces, with applications to combustor wall, nozzle and turbine blade cooling. Only flat wall cooling was investigated and not any influence of surface curvature. Combustor wall and turbine blade flank cooling both approximate to a flat wall as the hole diameter and pitch are all small in relation to the combustor or blade curvature. Also the experimental data base on impingement cooling predominantly uses a flat wall. The aim was to validate the computations against experimental data from hot metal wall research facilities and then to use the validated computational methodology to predict improved cooling geometries. Experimental investigations that used hot wall rigs at 770 K cross-flow temperature and 293 K coolant were modelled to predict the overall cooling effectiveness for impingement cooling. The impingement cooling of the metal surface with an equivalent heat flux was modelled, at a hot gas value equals to 100 kW/m2 and is an input relevant to real gas turbine combustor applications of 250 kW/m2K heat transfer coefficient (HTC). Much of the experimental data base with metal walls used electrically heated metal wall experiments with relatively low wall temperatures. These were also modelled using a constant hot gas side temperature and the thermal gradient through the thickness and between impingement and effusion holes were predicted.
The work was confined to the internal wall heat transfer and did not investigate the combined film effusion cooling that is often used in combination with impingement cooling. However, the interaction of internal wall effusion cooling with impingement cooling was investigated, so that the whole internal wall cooling could be predicted. The heat transfer in a metal wall with a square array of 90o holes is a subcomponent of impingement and effusion cooling and was part of this study, which was used to evaluate the impact of the CFD turbulence models. The standard k - ɛ turbulence model with standard wall function (WF) for y+ values in the range 30 - 45 showed better agreement with the measured data, where all the flow features were predicted correctly. Also enhanced wall treatment approaches (EWT) were used for y+ values from 1 - 5, but there was no significant improvement in the predictions compared with the standard wall function approach. All the turbulence models available in Fluent were investigated for an array of holes in a metal wall, which involves only a computation of one hole that is classic short hole or pipe entry length heat transfer. Many of the models could not predict the flow separation and reattachment within a hole L/D of ~1 and as this was fundamental to both effusion and impingement heat transfer, indicating that these models were all poor at the predictions of impingement and impingement/effusions cooling.
The experimental data base in impingement heat transfer has results that would not normally be expected and the CHT computations enabled the reason for the experimental trends to be explained. This includes the reduction in heat transfer along the impingement gap influenced by cross-flow, which would be expected to increase the heat transfer. The relatively low effect of turbulence enhancing obstacles in the impingement gap was also predicted. The influence of the number of impingement holes, which leads to methodology to choose a particular hole size has been predicted based on thermal gradients in the metal wall, this helps the designer in choosing optimum number of holes. For impingement cooling with single sided coolant exit from the cross-flow duct, it was shown that the deflection of the cross-flow onto the impingement jet wall surface was a major reason for the deterioration in the impingement target surface heat transfer along the gap. The very limited experimental database for heat transfer to the impingement jet wall surface was well predicted, thus showing that both wall surfaces were important in the overall impingement heat transfer.
The design configurations investigated were the hole length, pitch, gap, height and depth to diameter ratios L/D, X/D, Z/D, H/D and E/D respectively. The range of L/D investigated was 0.78 - 4.85, by varying the hole diameter for a fixed metal wall thickness (length) of 6.35 mm. This heat transfer was dominated by thermal and aerodynamic entry length effects including the heat transfer on the hole approach surface. The X/D range investigated was 1.86 - 21.02 by varying D at constant X and also by varying X at constant D, which varies the number of holes per surface area, n. The range of Z/D investigated was from 0.76 - 7.65 at varied and also at a constant Z. The main coolant flow parameter varied was the mass flux G, which is equals to G*/P (kg/sm2bar) in this Ph. D thesis. The requirements for each G with a fixed hole geometry, is a new CHT computation, which is time consuming compared with fairly rapid experimental determinations of the effect of G. The literature survey showed that there were no available detailed flow dynamics investigations of multi-hole impingement cooling. The key experimental measurement that indicates the correctness of the aerodynamic predictions was the pressure loss, which was as a result of the air feed to the impingement gap or effusion hole discharge. The results showed, for the range of geometries, reasonable agreement with the experimental measurements. For heat transfer the experimental measurements were all surface averaged, either for the whole wall or for each row of holes. The predictions were shown to give excellent agreement with surface average heat transfer, which also gave the surface distribution of the heat transfer. It was shown that the surface distribution of heat transfer was directly related to the surface distribution of the turbulence kinetic energy.
The experimental influence of turbulence enhancing obstacles in the impingement gap was well predicted. The experimental data base was for one obstacle per impingement hole using two flow configurations: flow parallel to the obstacles, so that the action was to increase the surface area for heat transfer at low blockage increase and flow across the obstacles, so that the action was to increase turbulence and surface area, but at the expense of higher pressure loss. Two obstacles shapes were investigated experimentally, simple continuous ribs and slotted ribs which gave rectangular pin fins relative to the cross-flow, with both turbulence generation and surface area increased. The predictions agreed with the experiments that showed the main effect of the obstacles, for which the deterioration of heat transfer with distance was reduced, but to only have a relatively small (~ 20%) increase in the surface averaged heat transfer. The validated computational procedures were used to investigate other obstacle geometries for the same impingement configuration: surface dimples, round pin-fins and inclined ribs in a zig-zag of ‘W’ format. The zig-zag design predicted an improvement in overall heat transfer compared with the other designs.
Impingement/effusion internal wall heat transfer was modelled with one effusion hole per impingement hole and a fixed 8 mm gap. It was shown that the key interaction effect was to remove any cross-flow from the gap, provided all the impingement air flow went through the effusion holes. This geometry is then only viable for low coolant mass flow rates and thus the modelling was confined to low G. This limitation of coolant flow was because effusion cooling improves if the hole velocity is low relative to the cross-flow, which occurs at low mass flow rates. Also the proportion of compressor air used for film cooling of combustor walls or turbine blades increases NOx from the combustor as the air used is not available to operate the primary zone leaner with lower NOx. For impingement only cooling, most of the work was carried out at high G, close to 2 kg/sm2bar, as the air in combustor application would be used for regenerative cooling and sent to the combustor low NOx primary zone at the exit from the impingement gap. Impingement/effusion cooling was shown to reduce significantly the reverse flow of the impingement jet back onto the impingement jet wall surface and hence had lower impingement heat transfer. However, the combination of the impingement and effusion wall cooling did lead to more total heat transfer than for impingement wall only cooling at the same G. Also investigated was the used of fewer number of impingement holes and more for effusion holes with a hole number ratio of 1/10. Effusion film cooling improves if the number of holes increases, whereas impingement cooling benefits from a low number of holes due to the reduced influence of the cross-flow. Also it was thought that using 10 effusion holes per impingement hole would act like a near uniform surface suction on the impingement jet leading to enhanced cooling of the effusion wall. The results of the modelling showed little benefit of this technique, which was shown to agree with experimental investigation into this effect.
This research has shown that current CHT CFD software can reliably predict experimental investigations of impingement and impingement/effusion overall wall heat transfer. It is thus considered that it can now be used as an engineering design tool for gas turbine combustor wall and turbine blade cooling optimisation. It is possible that gas turbine development in this area could be mainly using CHT CFD instead of the extensive experimental investigations that have been used to date. This work has also shown that relatively simple turbulence, wall function modelling and grid geometries are very effective and the use of more complex models is not justified.