The current energy transition is driving changes in many sectors, including the one of centrifugal compressors. Aiming to achieve net-zero carbon growth by 2050 and limit effects of global warming, a rapid reduction of greenhouse gas emissions is necessary. To pursue this goal, the scientific literature shows many possible pathways. However, whether focusing on increased system efficiency, capture technologies, renewable energy sources, or different energy vectors, centrifugal compressors hold a key role in the energy transition. In this context, higher performing centrifugal compressor stages need to be designed within ever shorter timescales. Furthermore, although computing capacities have increased in recent decades, exploiting a design process that requires less computational effort while maintaining good accuracy can offer considerable benefits. To this end, the present PhD thesis provides a methodological framework for the aerodynamic design of centrifugal compressor stages leveraging the know-how of preengineered families. Specifically, the research focused on the structured introduction of low-order models, computational fluid dynamics (CFD), and artificial intelligences (AI) in preliminary and detailed design, performance map generation, and impact evaluation of geometric variations. To address the goal of this study, the following research question was derived and addressed in the present PhD dissertation. RQ) How is it possible to perform the aerodynamic design of centrifugal compressor stages by exploiting the know-how of pre-engineered families and the synergetic use of low-order models, CFD, and artificial intelligence? As a viable answer to the above research question, this PhD thesis proposes an industrialfriendly methodological framework for the aerodynamic design of centrifugal compressor stages based on the synergy of low-order models, CFDs, and AIs, and leveraging the know-how of pre-engineered families. This methodological framework is composed by four phases. In the first phase, a 1D single-zone model was developed and combined with an artificial neural network (ANN). Then, exploiting the achieved surrogate-model new centrifugal compressor stages were derived from pre-engineered ones. In the second phase, CFD analyses and ANN were used to perform a detailed multi-point optimization of the above designed stages. In the third phase, a hybrid loworder/CFD model and a genetic algorithm were developed and combined to reduce the computational effort in generating the performance map of centrifugal compressor stages. Finally, in the fourth phase, a multi-point surrogate approach was defined by training an ANN with CFD results. Overall, the main goal of this research was achieved thanks to the collaboration with Baker Hughes, enabling the proposed methodological framework to be tested and validated during the aerodynamic design of 50 centrifugal compressor stages.

A methodological framework based on artificial intelligence for the aerodynamic design of centrifugal compressor stages / Marco Bicchi. - (2023).

A methodological framework based on artificial intelligence for the aerodynamic design of centrifugal compressor stages

Marco Bicchi
2023

Abstract

The current energy transition is driving changes in many sectors, including the one of centrifugal compressors. Aiming to achieve net-zero carbon growth by 2050 and limit effects of global warming, a rapid reduction of greenhouse gas emissions is necessary. To pursue this goal, the scientific literature shows many possible pathways. However, whether focusing on increased system efficiency, capture technologies, renewable energy sources, or different energy vectors, centrifugal compressors hold a key role in the energy transition. In this context, higher performing centrifugal compressor stages need to be designed within ever shorter timescales. Furthermore, although computing capacities have increased in recent decades, exploiting a design process that requires less computational effort while maintaining good accuracy can offer considerable benefits. To this end, the present PhD thesis provides a methodological framework for the aerodynamic design of centrifugal compressor stages leveraging the know-how of preengineered families. Specifically, the research focused on the structured introduction of low-order models, computational fluid dynamics (CFD), and artificial intelligences (AI) in preliminary and detailed design, performance map generation, and impact evaluation of geometric variations. To address the goal of this study, the following research question was derived and addressed in the present PhD dissertation. RQ) How is it possible to perform the aerodynamic design of centrifugal compressor stages by exploiting the know-how of pre-engineered families and the synergetic use of low-order models, CFD, and artificial intelligence? As a viable answer to the above research question, this PhD thesis proposes an industrialfriendly methodological framework for the aerodynamic design of centrifugal compressor stages based on the synergy of low-order models, CFDs, and AIs, and leveraging the know-how of pre-engineered families. This methodological framework is composed by four phases. In the first phase, a 1D single-zone model was developed and combined with an artificial neural network (ANN). Then, exploiting the achieved surrogate-model new centrifugal compressor stages were derived from pre-engineered ones. In the second phase, CFD analyses and ANN were used to perform a detailed multi-point optimization of the above designed stages. In the third phase, a hybrid loworder/CFD model and a genetic algorithm were developed and combined to reduce the computational effort in generating the performance map of centrifugal compressor stages. Finally, in the fourth phase, a multi-point surrogate approach was defined by training an ANN with CFD results. Overall, the main goal of this research was achieved thanks to the collaboration with Baker Hughes, enabling the proposed methodological framework to be tested and validated during the aerodynamic design of 50 centrifugal compressor stages.
2023
Andrea Arnone
ITALIA
Marco Bicchi
File in questo prodotto:
File Dimensione Formato  
PhD_Thesis_Marco_Bicchi.pdf

embargo fino al 31/12/2028

Tipologia: Tesi di dottorato
Licenza: Creative commons
Dimensione 6.02 MB
Formato Adobe PDF
6.02 MB Adobe PDF   Richiedi una copia

I documenti in FLORE sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificatore per citare o creare un link a questa risorsa: https://hdl.handle.net/2158/1308358
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus ND
  • ???jsp.display-item.citation.isi??? ND
social impact