Advanced

Multi-scale modelling of fire in accelerator tunnels: a CERN case study

Schepers, Melchior LU (2018) In LUTVDG/TVBB- VBRM05 20181
Division of Fire Safety Engineering
Abstract (Uncoded languages)
Multi-scale modellering is een nieuwe benadering in het modelleren van rook en brand in systemen waarbij de grootte van het domein een meer conventionele volledige 3D modellering onmogelijk maakt. Het domein wordt opgesplitst in een 1D en 3D gedeelte om zo de simulatietijd te verminderen. In deze thesis wordt de multi-scale
methodologie toegepast op de LHC deeltjesversneller te CERN met behulp van de HVAC module van FDS. Eerst wordt de theoretische basis van multi-scale modellering uitgelegd, alvorens over te gaan tot het modelleren van één sectie van de LHC tunnel. Het gebouwde referentiemodel omvat in totaal 3km aan tunnel, waarvan 640m in 3D, waarin zich een 1MW brand bevindt. Het overige gedeelte bestaat uit een 1D netwerkmodel.... (More)
Multi-scale modellering is een nieuwe benadering in het modelleren van rook en brand in systemen waarbij de grootte van het domein een meer conventionele volledige 3D modellering onmogelijk maakt. Het domein wordt opgesplitst in een 1D en 3D gedeelte om zo de simulatietijd te verminderen. In deze thesis wordt de multi-scale
methodologie toegepast op de LHC deeltjesversneller te CERN met behulp van de HVAC module van FDS. Eerst wordt de theoretische basis van multi-scale modellering uitgelegd, alvorens over te gaan tot het modelleren van één sectie van de LHC tunnel. Het gebouwde referentiemodel omvat in totaal 3km aan tunnel, waarvan 640m in 3D, waarin zich een 1MW brand bevindt. Het overige gedeelte bestaat uit een 1D netwerkmodel. Specifieke aandacht wordt besteed aan de implementatie van de correcte ventilatie-randvoorwaarden en de keuze van het 3D domein. Het geconstrueerde referentiesysteem wordt vervolgens onderworpen aan een reeks gevoeligheidsanalyses, die zowel betrekking hebben op het 1D als 3D gedeelte van het model. De ontwikkeling van de stroming wordt geanalyseerd op basis van zowel het Froudegetal als van het temperatuurs- en snelheidsprofiel van de stroming doorheen de tunnel. Dusdanig wordt aangetoond dat Fring = 3.2 een voldoende maar niet noodzakelijke voorwaarde is voor de bepaling van de lengte van het 3D domein. Suggesties voor mogelijk bijkomend onderzoek vormen het sluitstuk van deze thesis. (Less)
Abstract
Multi-scale modelling is a novel approach to fire modelling in situations where the size of the domain prevents it from being modelled completely in 3D. By splitting the domain in a 1D and 3D portion multi-scale modelling allows for much faster simulations which still adhere to the correct boundary conditions. In this thesis the multi-scale methodology is applied to the LHC accelerator in use at CERN using the HVAC capabilities of FDS. The theoretical foundation of multi-scale modelling is first explored after which a benchmark model representing one section of the LHC tunnel is built. The model contains one 640m long 3D domain, accommodating a 1MW fire, while the remainder of the domain is made up out of 1D ducts. Special attention is... (More)
Multi-scale modelling is a novel approach to fire modelling in situations where the size of the domain prevents it from being modelled completely in 3D. By splitting the domain in a 1D and 3D portion multi-scale modelling allows for much faster simulations which still adhere to the correct boundary conditions. In this thesis the multi-scale methodology is applied to the LHC accelerator in use at CERN using the HVAC capabilities of FDS. The theoretical foundation of multi-scale modelling is first explored after which a benchmark model representing one section of the LHC tunnel is built. The model contains one 640m long 3D domain, accommodating a 1MW fire, while the remainder of the domain is made up out of 1D ducts. Special attention is paid to the correct implementation of the push-pull ventilation strategy and selection of the 3D domain size. Following the construction of the benchmark system, it is subjected to a number of sensitivity analyses focussing on both the 1D and 3D portions of the domain. The development of the flow is investigated based on the Froude number and the temperature and velocity profiles of the flow along the tunnel. As such it is shown that the condition Fring = 3.2 serves as a sufficient but not necessary condition in determining the length of the 3D downstream domain. By suggesting possible improvements to the multi-scale model, the door is opened to future research. (Less)
Popular Abstract
Modern day engineering and architecture push the boundaries of our technical capabilities. As structures become more and more complex a one size fits all approach based on prescriptive regulations is no longer up to the task of ticking all the boxes to ensure a fire safe design. This is where a performance-based approach based on fire modelling comes into the picture. In a performance-based approach the design is not based solely on written regulations, but rather tested by detailed numerical simulations. Running such a simulation however requires huge amounts of computational power, which is costly and often only available at supercomputing centers. To be able to run a large number of different fire scenarios approximating techniques... (More)
Modern day engineering and architecture push the boundaries of our technical capabilities. As structures become more and more complex a one size fits all approach based on prescriptive regulations is no longer up to the task of ticking all the boxes to ensure a fire safe design. This is where a performance-based approach based on fire modelling comes into the picture. In a performance-based approach the design is not based solely on written regulations, but rather tested by detailed numerical simulations. Running such a simulation however requires huge amounts of computational power, which is costly and often only available at supercomputing centers. To be able to run a large number of different fire scenarios approximating techniques therefore have to be used. An example of one such technique is multi-scale modelling. Multi-scale modelling combines detailed 3D computational fluid dynamics with simplified 1D approximations, in order to reduce the computational cost of the simulation and thus significantly enhance its speed.
The European Organization for Nuclear Research, CERN, located at the Franco-Swiss border is a typical facility housing complex infrastructure, which has to be designed using a performance-based approach. This is in part also due to the fact that the special diplomatic status of the CERN site causes it to fall outside of national regulations. At the CERN site near Geneva, huge particle accelerators are used to collide fundamental particles and investigate their properties. These accelerators are housed in large underground tunnels often spanning many kilometers. The largest accelerator in use is the Large Hadron collider or LHC, located in a 27km long circular tunnel, 100m below the surface. Modelling a fire in the 27km long tunnel completely in 3D would lead to a very slow and computationally expensive model, due to the size of the domain. This is where multi-scale modelling comes in, as it allows for the vast majority of the domain to be represented by a simplified 1D model.
By using the Fire Dynamics Simulator software it was possible to build a representative multi-scale model of a three kilometer long section of the LHC tunnel. The domain was split in a 640m long 3D section and a 2360m long 1D part. This division is the strength, but also the weakness of the multi-scale modelling technique as the coupling implies a sudden loss of information when going from 3D to 1D. Not surprisingly the location of the 1D-3D interface is thus a very important parameter in designing the multi-scale model and was extensively researched in the thesis work.
A question that however comes to mind when using approximating techniques such as multi-scale modelling is how exact the results are? Validation of the results either by experiments or by a full 3D simulation is thus the logical step forward for any future research. (Less)
Please use this url to cite or link to this publication:
author
Schepers, Melchior LU
supervisor
organization
course
VBRM05 20181
year
type
H2 - Master's Degree (Two Years)
subject
keywords
Multi-scale modelling, CFD, tunnel fires, FDS, CERN, particle accelerator
publication/series
LUTVDG/TVBB-
report number
5575
other publication id
LUTVDG/TVBB--5575--SE
language
English
additional info
This thesis was made possible due to the extensive support of the fire safety engineering team of CERN.
id
8952161
date added to LUP
2018-06-21 12:09:21
date last changed
2018-06-21 12:09:21
@misc{8952161,
  abstract     = {Multi-scale modelling is a novel approach to fire modelling in situations where the size of the domain prevents it from being modelled completely in 3D. By splitting the domain in a 1D and 3D portion multi-scale modelling allows for much faster simulations which still adhere to the correct boundary conditions. In this thesis the multi-scale methodology is applied to the LHC accelerator in use at CERN using the HVAC capabilities of FDS. The theoretical foundation of multi-scale modelling is first explored after which a benchmark model representing one section of the LHC tunnel is built. The model contains one 640m long 3D domain, accommodating a 1MW fire, while the remainder of the domain is made up out of 1D ducts. Special attention is paid to the correct implementation of the push-pull ventilation strategy and selection of the 3D domain size. Following the construction of the benchmark system, it is subjected to a number of sensitivity analyses focussing on both the 1D and 3D portions of the domain. The development of the flow is investigated based on the Froude number and the temperature and velocity profiles of the flow along the tunnel. As such it is shown that the condition Fring = 3.2 serves as a sufficient but not necessary condition in determining the length of the 3D downstream domain. By suggesting possible improvements to the multi-scale model, the door is opened to future research.},
  author       = {Schepers, Melchior},
  keyword      = {Multi-scale modelling,CFD,tunnel fires,FDS,CERN,particle accelerator},
  language     = {eng},
  note         = {Student Paper},
  series       = {LUTVDG/TVBB-},
  title        = {Multi-scale modelling of fire in accelerator tunnels: a CERN case study},
  year         = {2018},
}