The aim of this work is to improve an already existing automated tool for the shape optimization of turbomachinery blades based on an evolutionary strategy. This tool has been developed at Politecnico di Milano [Fernandez et al. (2015)]. This optimization scheme will serve to deal with transonic and supersonic blades cascades for applications of Organic Rankine Cycle (ORC) turbines. The blade geometry is defined using parameterization techniques based on B-Spline curves, that allow to have a local control of the shape, but the trailing edge, the region with high gradient, has been interpolated with elliptical shape. The displacements of the control points of the B-Spline curve, from the Baseline positions, along preferable directions (defined by the local gradient of the curve), represent the design variables of the optimization problem. In the present work, the performance of the blade shape is assessed by means of fully-turbulent flow simulations performed with a CFD package, in which the real-gas behaviour is modelled with both the Peng-Robinson equation and with the Lookup Table approach. The solver is set along with the optimization tool to determine the optimal shape of the blade. As only blade-to-blade effects are of interest in this study, quasi-3D calculations are performed, and a single-objective evolutionary strategy is applied to the optimization (the objective function has been defined either as the pressure loss coefficient calculated downstream of the blade, or as the standard deviation of the pressure distribution downstream of the blade). As a result, a non-intrusive tool, with no need for gradient definition of the objective function, is developed. The computational cost is reduced by the use of surrogate models. A Gaussian interpolation scheme (Kriging model) is applied for the estimated n-dimensional function, and a surrogate-based global optimization strategy is proved to yield an accurate way for optimization. In particular, the present scheme has been applied to the re-design of a supersonic stator cascade of an axial-flow turbine and to the case of the first stator and rotor of a centrifugal-flow turbine. Specific codes have been also implemented to deal with the optimization of symmetric profiles for ORC turbines, with the aim of reducing the computational time taking advantage from the special geometrical features. In all the cases a significant efficiency improvement has been achieved. The method shows benefits with respect to gradient-based methods and an optimized blade geometry is easily achieved using the genetic algorithm, with a reduction of computational time with respect to the use of several CFD simulations. The power of this tool is that it is completely automatized: once the settings are indicated, no interventions are required and an optimal geometry can be found.
L’obiettivo di questa tesi è quello di migliorare uno strumento di ottimizzazione geometrica applicato alle turbomacchine, già esistente [Fernandez et al. (2015)]. Lo schema di ottimizzazione già esistente ha mostrato dei limiti e delle potenzialità di miglioramento. Questo "in-house Blade-Shape Optimization Tool" proposto è stato progettato per applicazioni ORC (Organic Rankine Cycle) caratterizzate da flussi supersonici e da effetti di fluido reale non trascurabili. Lo strumento di ottimizzazione non è intrusivo, non richiede la definizione del gradiente della funzione obiettivo, bensì si basa su algoritmi genetici e riduce il costo computazionale attraverso la costruzione di un metamodello che approssima la vera relazione tra geometria del profilo e prestazioni. La geometria palare viene definita attraverso curve B-Splines, eccetto nella regione del trailing edge, che invece viene parametrizzato separatamente con una forma ellittica. Gli spostamenti nel piano dei punti di controllo della B- Spline interpolante, definiti lungo le direzioni delle normali locali, rappresentano le variabili del problema di ottimizzazione. L’ottimizzatore, ricevuti in ingresso un profilo di pala di partenza (Baseline e le variabili del problema, è in grado di ottimizzarne la fluidodinamica Blade-to-Blade secondo una funzione obiettivo scelta dall’utente. L’accuratezza della soluzione della geometria ottima finale è garantita dalla risoluzione del flusso quasi 3D e completamente turbolento operata tramite un codice CFD commerciale (il modello di fluido reale è introdotto sia con l’approccio Lookup Table, che con quello della definizione dell’equazione di stato). Lo strumento di ottimizzazione è stato testato sulla geometria "Biere" di uno statore supersonico di una turbina assiale e sulla geometria del primo statore e del primo rotore di una turbina centrifuga. In tutti i casi un miglioramento nelle prestazioni termodinamiche è evidente. La potenza di questo strumento sta nell’essere completamente automatico: una volta che le condizioni termodinamiche operative sono state specificate, indicate le variabili del problema e la Baseline, si ottiene una geometria finale ottimizzata senza la necessità di un ulteriore intervento dell’utilizzatore.
Development of algorithms for the shape-optimization applied to turbomachinery blade profiles
PAPETTI, VIOLA
2015/2016
Abstract
The aim of this work is to improve an already existing automated tool for the shape optimization of turbomachinery blades based on an evolutionary strategy. This tool has been developed at Politecnico di Milano [Fernandez et al. (2015)]. This optimization scheme will serve to deal with transonic and supersonic blades cascades for applications of Organic Rankine Cycle (ORC) turbines. The blade geometry is defined using parameterization techniques based on B-Spline curves, that allow to have a local control of the shape, but the trailing edge, the region with high gradient, has been interpolated with elliptical shape. The displacements of the control points of the B-Spline curve, from the Baseline positions, along preferable directions (defined by the local gradient of the curve), represent the design variables of the optimization problem. In the present work, the performance of the blade shape is assessed by means of fully-turbulent flow simulations performed with a CFD package, in which the real-gas behaviour is modelled with both the Peng-Robinson equation and with the Lookup Table approach. The solver is set along with the optimization tool to determine the optimal shape of the blade. As only blade-to-blade effects are of interest in this study, quasi-3D calculations are performed, and a single-objective evolutionary strategy is applied to the optimization (the objective function has been defined either as the pressure loss coefficient calculated downstream of the blade, or as the standard deviation of the pressure distribution downstream of the blade). As a result, a non-intrusive tool, with no need for gradient definition of the objective function, is developed. The computational cost is reduced by the use of surrogate models. A Gaussian interpolation scheme (Kriging model) is applied for the estimated n-dimensional function, and a surrogate-based global optimization strategy is proved to yield an accurate way for optimization. In particular, the present scheme has been applied to the re-design of a supersonic stator cascade of an axial-flow turbine and to the case of the first stator and rotor of a centrifugal-flow turbine. Specific codes have been also implemented to deal with the optimization of symmetric profiles for ORC turbines, with the aim of reducing the computational time taking advantage from the special geometrical features. In all the cases a significant efficiency improvement has been achieved. The method shows benefits with respect to gradient-based methods and an optimized blade geometry is easily achieved using the genetic algorithm, with a reduction of computational time with respect to the use of several CFD simulations. The power of this tool is that it is completely automatized: once the settings are indicated, no interventions are required and an optimal geometry can be found.| File | Dimensione | Formato | |
|---|---|---|---|
|
2017_Aprile_Papetti.pdf
solo utenti autorizzati dal 04/04/2018
Descrizione: Testo della tesi
Dimensione
6.24 MB
Formato
Adobe PDF
|
6.24 MB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/133882