Gas turbine engineering faces many challenges in the constant strive to increase not only the efficiency of engines but also the various stages of development and design. Development of combustors have primarily consisted of empirical or semi-empirical modelling combined with experimental investigations. Due to the associated costs and development time a need exists for an alternative method of development. Although experimental investigations can never be substituted completely, mathematical models incorporating numerical methods have shown to be an attractive alternative to conventional combustor design methods. The purpose of this study is twofold: firstly, to experimentally investigate the physical properties associated with a research combustor that is geometrically representative of practical combustors; and secondly, to use the experimental measurements for the validation of a computational fluids dynamic model that was developed to simulate the research combustor using a commercial code. The combustor was tested at atmospheric conditions and is representative of practical combustors that are characterized by a turbulent, three-dimensional flow field. The single can combustor is divided into a primary, secondary and dilution zone, incorporating film¬cooling air through stacked rings and an axial swirler centred around the fuel atomizer. Measurements at different air/fuel ratios captured the thermal field during operating conditions and consisted of inside gas, liner wall and exit gas temperatures. An investigation of the different combustion models available, led to the implementation of the presumed-PDF model of unpremixed turbulent reaction. The computational grid included the external and internal flow field with velocity boundary conditions prescribed at the various inlets. Two-phase flow was not accounted for with the assumption made that the liquid fuel is introduced into the combustion chamber in a gas phase. Experimental results showed that incomplete combustion occurs in the primary zone, thereby reducing the overall efficiency. Also evident from the results obtained are the incorrect flow splits at the various inlets. Evaluation of the numerical model showed that gas temperatures inside the combustor are overpredicted. However, the numerical model is capable of capturing the correct distributions of temperatures and trends obtained experimentally. This study is successful in capturing detail temperature measurements that will be used for validation purposes to assist the development of a numerical model that can accurately predict combustion properties.
Dissertation (M Eng (Mechanical Engineering))--University of Pretoria, 2007.