Development of an automated process for turbine blade optimisation

Računalna dinamika fluida (CFD, eng. Computational Fluid Dynamics) postala je neophodna u procesima optimizacije komponenti turbostrojeva. Smanjenje vremena potrebnog za konstrukciju i proračun rezultira nižom cijenom gotovog proizvoda. Brže pronalaženje optimalnog rješenja, a samim time i poboljšanje performansi te smanjenje mase gotovog proizvoda čine optimizaciju privlačnom za industrijsku upotrebu. Osim prethodno navedenih povoljnih svojstava računalne dinamike fluida, izrazito je bitno napomenuti mogućnost vizualizacije kompleksnih pojava u strujanju fluida oko kritičnih komponenti. U ovome je radu razvijen automatizirani proces optimizacije profila turbinskih lopatica. Proces se sastoji od sljedećih koraka: 1. Parametrizacija geometrije turbinskog profila, 2. Deformacija početne proračunske mreže za novu geometriju, 3. Numerički proračun stlačivog strujanja kroz kanal turbinske lopatice, 4. Odabir najboljeg rješenja prema zadanom kriteriju. U prvom koraku se vrši postupak parametrizacije geometrije turbinske lopatice koristeći periodičnu B-spline krivulju 5 stupnja. Njome se aproksimira set kontrolnih točaka (kojima jedefinirana geometrija turbinske lopatice) te se između svake dvije kontrolne točke ekstrapolira 40 točaka parametrizacije. U drugom koraku se početna proračunska mreža deformira upravo prema koordinatama točaka parametrizacije da bi se dobila proračunska mreža oko traženog profila. U trećem koraku se vrši numerički proračun stlačivog strujanja kroz kanal turbinske lopatice te se izračunavaju vrijednosti funkcija cilja kao reprezentativne vrijednosti strujanja. Koristeći vrijednosti funkcija cilja, pomoću višeciljnog optimizacijskog algoritma, izračunava se novi set kontrolnih točaka. Time ponovo započinje petlja optimizacijskog procesa.Computational fluid dynamics (CFD) is a critical part in optimisation of turbomachinery components. Constant commercial pressure to produce parts of highest possible quality both in shortest possible time and with maximal reduction in price, are the reasons why CFD has become a necessity. Utilization of CFD enables faster design cycles, better performance, and reduction in cost and weight of the final product. CFD is of greatest importance in optimisation procedures because much more blade designs as well as blade parameters can be examined than it would be the case in an experimental study. Also CFD enables visualisation of complex fluid flow patterns over critical component geometries. Developed automated optimisation process consists of the following steps: 1. Blade geometry parametrisation using a periodic B-spline curve, 2. Morphing of the computational mesh 3. Numerical calculation of a compressible fluid flow through turbine blade passage, and 4. Obtaining the new set of control polygon vertices. In the parametrisation step, blade geometry, which is defined with the set of control polygon vertices, is approximated with a periodic B-spline curve of 5th degree. 40 parametrisation points are extrapolated between each pair of control polygon vertices. In the mesh morphing step, numerical mesh is deformed in respect with the parametrisation points coordinates. Numerical calculation of compressible fluid flow is conducted in the third step of this optimisation process in which objective function values are obtained. Objective functions are physical values derived from flow variables, which describe the characteristics of each blade geometry. Using this values, new set of control polygon vertices is calculated with Multi-objective evolutionary algorithm. In this way, the optimisation loop is closed and restarted. Due to the possibility of working with very irregular and distorted starting geometries, the developed automated optimisation process has proven to be robust and effective