Thermal radiation is typically one of the most important phenomenon to be taken into account in the evaluation of combustor walls thermal loads due to the high temperatures reached into them. A classical approach is based on the so called Zonal Method, originally developed by Hottel and Sarofim (1967), and actually widely employed in the industrial environment. Even if its accuracy has been largely demonstrated, its efficiency is affected by computational costly solution of 4th – 6th fold integrals constituting the DEAs. A direct integration is usually employed, subsequently smoothing the results in order to obey the conservation constraints. The last decades have seen a growing interest on developing new techniques able to simplify these time consuming direct numerical integrations. Among them one of the most promising approaches has been recently introduced by Yuen (2008) which is based on the classic Mean Beam Length concept. The emittance (absorptance) coefficients of the radiating (receiving) gas volume zones of cubic shape are treated as those of a grey gas filled hemisphere. According to Yuen (2008), a correlative expression has been employed for evaluating the MBL corresponding to each single volume zone. In the present work its application has been extended to more complex zone shapes by means of an Artificial Neural Network trained on a properly selected geometry database. In this way the DEA integral folds can be all reduced to the 4th order, and, employing well known geometrical techniques (Walton, 2002), can be further decreased to lower order integrals. The proposed model has been compared with 3D benchmark test cases available in literature: the accuracy has been tested against the results of DTM, FVM and classical Zonal Method. Moreover an industrial application is shown. The geometry of the AE94.2 gas turbine combustion chamber has been considered; the gas mixture has been treated as a non-grey gas using the well known Weighted Sum of Gray Gases (WSGG) model. In comparison with the classical zonal method approach, the efficiency of the proposed one demonstrates the possibility of a zone refinement enabling a more accurate evaluation of radiative thermal loads at the same computational cost. Results are presented and discussed.

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