Measurement of Heat Transfer Distrubution of Cooled Real Geometry Using Infrared Theromography

Over the years, gas turbine industry is continuously exploring different methods to increase the turbine inlet temperature for improving the specific power output and thermal efficiency of gas turbine engines, which leads to thermal loads for the engine components, which are usually managed thanks to the introduction of complex internal cooling systems. For this reason, it is necessary to develop tools able to accurately and quickly estimate thermal loads on turbine components. Typical experimental methods for assessing the heat transfer characteristics of such systems commonly rely upon scaled-up models investigated at nearly ambient conditions, since the size and operating environment of real engine parts make it extremely difficult to perform direct measurements. By doing so, however, many geometric and flow features of the real cooling system get lost, since the studied geometry is ideal and measurement constraints often require a simplification of the system itself. So there is a need for the development of a non-invasive, non-destructive, transient inverse technique which allows testing of real turbine blades temperature measurements. A much more reliable evaluation of cooling performance would thus be obtained by studying the real hardware, which requires the development of a suitable technique. As an additional advantage, a similar method could also be employed for in-line inspection of manufactured parts, as to clearly identify faults and defects before the actual installation. The aim of this work is to present the development and application of a measurement technique that allows to record internal heat transfer features of real components. In order to apply this method, based on similar approaches proposed in previous literature works, the component is initially heated up to a steady temperature, then a thermal transient is induced by injecting cool air in the internal cooling system. During this process, the external temperature evolution is recorded by means of an IR camera. Experimental data are then exploited to run a numerical procedure, based on a series of transient finite-element analyses of the component. Then two different approaches can be followed, which will be refereed as fluid model method and regression method respectively. In fluid model method at the end of transient, finite element output external surface temperature is compared to the one which is obtained with experiment and the convective internal heat transfer coefficient is iterated continually with a root finding algorithm until the convergence between them is achieved. The coolant temperature will be updated during the transient with the help of a fluid model. This approach works very well for simplified geometries, in which convective internal heat transfer coefficient converges specific value but as we move to more complex geometries it may diverge. The reason is the inability of the fluid model to find accurate coolant temperature at certain regions, so a second approach is introduced for more complex geometries, the regression which is a modification version of the first approach. In the regression method, the test duration is divided into an appropriate number of steps and for each of them, the heat flux on internal surfaces is iteratively updated to target the measured external temperature distribution at the end of step. Heat flux and internal temperature data for all the time steps are eventually employed in order to evaluate the convective heat transfer coefficient via linear regression. This technique has been successfully tested on a cooled high-pressure vane of a Baker Hughes heavy-duty gas turbine, which was realised thanks to the development of a dedicated test rig at the University of Florence, Italy. The obtained results provide sufficiently detailed heat transfer distributions in addition to allowing to appreciate the effect of different coolant mass flow rates. The methodology is also capable of identifying defects, which is demonstrated by inducing controlled faults in the component.