A DESIGN STUDY OF AN ELASTO-HYDRODYNAMIC SEAL FOR sCO2 POWER CYCLE BY USING PHYSICS INFORMED NEURAL NETWORK

Mohammad Towhidul Islam, Mohammad Fuad Hassan, Karthik Reddy Lyathakula, Sevki Cesmeci, Hanping Xu, Jing Tang

Research output: Contribution to book or proceedingConference articlepeer-review

3 Scopus citations

Abstract

Supercritical carbon dioxide (sCO2) is a promising alternative working fluid in power cycles. It offers several benefits, including improved efficiencies, lower power costs, lower water usage, and reduced equipment footprint. At the subcomponent level, there is a lack of suitable shaft end seals to minimize the sCO2 leakage in the turbomachinery, which would otherwise penalize the efficiencies by up to 0.65 percent, for example, for a 500 MWe utility-scale power plant when existing sealing technologies are used such as labyrinth seals. For a potential solution, we propose a simple sleeve-like Elasto-Hydrodynamic (EHD) seal concept that provides low leakage, minimum wear, and no stress concentration. In this work, a numerical study has been performed to model the structural deformation of the EHD seal and fluid flow by using Lame's equation and simplified Reynold's equation, respectively. However, the analytical solutions for this type of modeling require a set of coupled differential and highly non-linear equations. Conventional approaches for optimizing the solution often end in convergence issues and require selecting specific flow rates. Therefore, the design equations used in this work have been solved by adopting the Physics Informed Neural Network (PINN) to mitigate these issues. Through using PINN analysis, the throttling behavior of the seal was successfully demonstrated. The deformation of the seal was observed to occur at an axial length of approximately 18mm, while a fixed root condition was imposed at 26.5mm. The initial clearance within the seal was 0.2566mm, but it was observed to decrease as the pressure increased within the range of 60MPa to 90MPa. The fluid flow inside the region also decreased from 0.68kg/s to 0.42kg/s at this high-pressure range. The proposed analysis can be used to design EHD seals for specific cases when more comprehensive simulation models are not readily available or are deemed to be costly.

Original languageEnglish
Title of host publicationProceedings of ASME Power Applied R and D 2023, POWER 2023
PublisherAmerican Society of Mechanical Engineers (ASME)
ISBN (Electronic)9780791887172
DOIs
StatePublished - 2023
EventASME Power Applied R and D 2023, POWER 2023 - Long Beach, United States
Duration: Aug 6 2023Aug 8 2023

Publication series

NameAmerican Society of Mechanical Engineers, Power Division (Publication) POWER
Volume2023-August

Conference

ConferenceASME Power Applied R and D 2023, POWER 2023
Country/TerritoryUnited States
CityLong Beach
Period08/6/2308/8/23

Scopus Subject Areas

  • Mechanical Engineering
  • Energy Engineering and Power Technology

Keywords

  • Elasto-Hydrodynamic (EHD)
  • Machine Learning
  • Neural Networks (NNs)
  • Supercritical CO2

Fingerprint

Dive into the research topics of 'A DESIGN STUDY OF AN ELASTO-HYDRODYNAMIC SEAL FOR sCO2 POWER CYCLE BY USING PHYSICS INFORMED NEURAL NETWORK'. Together they form a unique fingerprint.

Cite this