×

Geometrical optimization in gas turbines by applying computational fluid dynamics and genetic algorithms. (English) Zbl 1326.35270

Summary: A methodology that uses genetic algorithms with a computerized vision tool and computational fluid dynamics (CFD) to optimize the geometric parameters of gas turbine components, is presented. The proposed methodology applies the results of the CFD simulation where the temperature and velocity contours are use to create the population of individuals. Each population generated is composed by the geometrical parameters that represent a feasible geometry. The set of parameters corresponding to the individual genotype were decoded and a script for the CFD software (Fluent \(\copyright\)) that describes the geometrical shape was created. The optimization process considers an initial set of individuals and a CFD simulation is created to obtain the temperature and velocity contours according to the desirable properties of the thermal behavior and subsequent populations were generated by applying selection, crossover and mutation genetic operators to the best individuals.
The composition of the new population was created with 2 elite individuals, 6 individuals obtained from the application of the genetic operators and 2 new random individuals. A morphometric analysis computes several geometric properties of the temperature and velocity profiles to provide information to decide the best individuals. For the case of the transition piece of the gas turbine, the temperature and velocity profiles required at the outlet must be uniform because a non-uniformity in the temperature profile affects the useful life of the blades and nozzles of the first stage. However, a diminution in the average value of the turbine inlet temperature produces a reduction of the thermal efficiency and power of the gas turbine. Then, by applying a genetic algorithm and CFD simulation to optimize the geometry of the transition piece it is possible to analyze the geometrical parameters that are not intuitive for a human designer.

MSC:

35Q35 PDEs in connection with fluid mechanics
35Q30 Navier-Stokes equations
76F60 \(k\)-\(\varepsilon\) modeling in turbulence
68T20 Problem solving in the context of artificial intelligence (heuristics, search strategies, etc.)
76M25 Other numerical methods (fluid mechanics) (MSC2010)
90C59 Approximation methods and heuristics in mathematical programming
80A20 Heat and mass transfer, heat flow (MSC2010)

Software:

FLUENT
PDFBibTeX XMLCite