Effusion Cooling has a strong potential in protecting hot section components of gas turbines engines such as turbine blades, endwalls and combustor liners. By carefully distributing the cooling holes adaptively to the external thermal load, effusion cooling can dramatically reduce the temperature and thermal stress for the protected components. However, the effectiveness for effusion cooling was not easy to correlate due to the numerous parameters to study and the significant variation from the upstream to the downstream. Conventional equations used in the literature were insufficient to express the complex mechanism for effusion cooling and consequently utilized averaged parameters as variables. This study proposed a convolution method to model the local adiabatic cooling effectiveness for the entire effusion cooled surfaces. The new model treated the cooling hole distribution as an input matrix and applied convolution networks to predict the cooling effectiveness. Compared with conventional correlations, this network based model provided extensive details of the cooling effectiveness distribution while consuming computational time as short as correlations. Training of the proposed model was based on the numerical simulation results of three geometries and the validation was conducted for two additional geometries. Results indicated high accuracy and high robustness of the convolution model. With the aid of this novel model, further designing could adjust hole distribution in a random manner instead of using rows and columns, and generate adaptive effusion cooling based on thermal load.

This content is only available via PDF.
You do not currently have access to this content.