Preliminary gas turbine combustor design using a genetic algorithm

Abstract

A genetic algorithm, coupled with a versatile preliminary design loot, is employed to demonstrate the concept of an autonomous design procedure for gas turbine combustors with user specified performance criteria. The chosen preliminary design program utilises a network based approach which provides considerable geometric flexibility allowing for a wide variety of combustor types to be represented. The physical combustor is represented by a number of independent, though interconnected, semi-empirical sub-flows or elements. A full conjugate heat transfer model allows for convection, conduction and radiative heat transfer to be modelled and a constrained equilibrium calculation simulates the combustion process. The genetic algorithm, whose main advantage lies in its robustness, uses the network solver in order to progress towards the optimum design parameters defined by the user. The capabilities of the genetic program are demonstrated for some simple design requirements, for example zone fuel/air ratio, pressure drop and wall temperatures.