Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.

An Integrated Computational Environment for Simulating Structures in Real Fires

1,194 views

Published on

Presentation made by Prof. Asif Usmani @ University of Porto during the OpenSees Days Portugal 2014 workshop

Published in: Engineering
  • Be the first to comment

An Integrated Computational Environment for Simulating Structures in Real Fires

  1. 1. OpenSees Days, Porto, 3-4 July, 2014 An integrated computational environment for simulating structures in real fires Asif Usmani and Liming Jiang School of Engineering, The University of Edinburgh, UK & Jian Jiang, Guo-Qiang Li and Suwen Chen College of Civil Engineering, Tongji University, China With acknowledgements to (other previous and current PhD students): David Lange, Jian Zhang, Yaqiang Jiang, Panagiotis Kotsovinos, Ahmad Mejbas Al-Remal, Shaun Devaney and Payam Khazaienejad + the IIT Roorkee and Indian Institute of Science teams! & special acknowledgement to Frank McKenna for
  2. 2. Key components of simulating structures in fire modelling fire and smoke 1. 980°C 932,2°C 900 800 850 600 800 750 400°C 750,0°C modelling structural member temperature evolution 2. modelling structural response to the point of collapse 3.
  3. 3. Integrated computational environment for structures in fire Fire models Standard Natural/parametric (short hot/long cool) Localised Travelling CFD (FDS, OpenFOAM, ANSYS CFX) fire – heat transfer middleware Heat transfer OpenSees, ABAQUS, ANSYS heat transfer - thermomechanics middleware Structural response OpenSees, ABAQUS, ANSYS (general) SAFIR, VULCAN (special purpose) Integrity failure structure – fire coupling
  4. 4. Why do we need an “integrated computational enviroment” ? Current widespread practice is “prescriptive” (standard fire + isolated member) Architects are pushing boundaries with larger and taller buildings, internal partitions are disappearing, more unusual architectural shapes are being used, demand for sustainability is bringing in new materials (and hazards), built-environments are getting more complex and dense creating higher risk (consequences of disaster are increasing) => “Just like alternative” or the seismic performance based hazard and engineering (the PBE) approaches structure interact in earthquake engineering (necessitating e.g. soil-structure interaction Even when PBE approaches are used in general uniform compartment fires are assumed (a single compartment temperature at a given instant in time – no spatial variation) for improved fidelity of modelling with But even if one wanted to make a realistic estimate of the fire, there are no tools to simulate the reality) whole – fire process, (if and the commercial vendors structure make them also they would be interact! too expensive – furthermore researchers will have no control over the tools) Yes it is very unlikely that such an environment will be used in routine engineering – but routine engineering can benefit from research to create a better understanding of structural response in real fires – IF ONLY we had such a tool! Currently the only way to do a fully coupled simulation is to “conduct an experiment” How do we learn from our failures (aviation industry is a great example!)
  5. 5. “Performance” in a general structural engineering context Modern structural engineering defines “limit states” d Expressed quantitatively in terms of the fundamental parameters Load: “maximum” load sustainable (ultimate limit state) Displacement: “maximum” allowable displacement (serviceability limit state) Clear and quantifiable link between cause (load) and effect (displacement) Concept easily extended to general structural frames and extreme loadings, inter-storey drift in an earthquake
  6. 6. Performance in the context of struct. fire resistance CONCRETE STEEL Observation Fire heats steel, steel rapidly loses stiffness & strength at temperatures above 400oC with only half the strength remaining at 550oC Solution Protect all steel for a long enough period Issues with this approach 1. How long should a structural member be protected for? 1250 2. Cause (heating) and effect (reduced load capacity 1000 and displacements) works reasonably well for 750 simple determinate structures such as 500 250 but not for large redundant 0 structures 0 30 60 90 120 150 180 time [minutes] Temperature [oC] Fire (BS-476-Part 8) But time on a standard fire curve ? has no physical meaning!
  7. 7. Responses of real structures depend upon the fire Eurocode model of “natural fires” (valid for < 500 m2 & < 4m ceiling height) 1200 1000 800 600 400 200 0 0 2000 4000 6000 8000 10000 12000 14000 Time (sec) Atmosphere Temperature (°C) Well-ventilated OF=0.08 Under-ventilated OF=0.02 Short hot fire Long cool fire Behaviour of a small composite steel frame structure in ‘long-cool’ and ‘short-hot’ fires, Fire Safety Journal, 39:327–357, 2004
  8. 8. e.g. 2x2 generic frame (columns & edge beams protected )
  9. 9. Deflections - 2x2 generic frame max. defl.=310 mm @ 78 mins max. steel tempr.=750 O C Short hot fire Long cool fire OF=0.02 max. defl. = 365 mm @ 23 mins max. steel tempr.=950 O C OF=0.08
  10. 10. Strains (in the horizontal direction) - 2x2 generic frame OF=0.08 OF=0.02 TENSILE TENSILE COMPRESSIVE COMPRESSIVE DifferenSth foirrte sh optr ofidreuce different strLuocntgu rcaol orel sfipreonses
  11. 11. Fires in large compartments Fire tends to travel in large spaces Source: NIST NCSTAR 1-5
  12. 12. Therefore … Permutations are potentially endless! an “integrated computational environment” is required to deal with them Also, we cannot properly address this issue in a deterministic framework (earthquake engineers realised this long ago!) However, there are no tools available to easily implement “proper” Performance Based Structural Engineering for Fire Resistance Our aim is to develop a prototype of such a tool based on the OpenSees framework
  13. 13. Plantation Place (Arup Fire)
  14. 14. Two sub-structure models 10m 9m 9m Model 1 Model 2
  15. 15. Fire scenarios Time long-cool (parametric) fire Temperature short-hot (parametric) fire 1250 1000 750 500 250 0 0 30 60 90 120 150 180 time [minutes] Temperature [oC] Fire (BS-476-Part 8)
  16. 16. Results from Model 1 All beams protected Only secondary beams unprotected 380mm max deflection 470mm max deflection
  17. 17. Unprotected 10m panel (Model 2)
  18. 18. Protected 10m panel (Model 2)
  19. 19. Final Proposal (accepted) Saving of £250K on Plantation Place
  20. 20. Engineers learn from failures Ronan Point,UK (May, 1968) I35W Bridge, Minneapolis, USA (Aug, 2007) “disproportionate” and/or “progressive collapse” => “robustness requirements”
  21. 21. Why some structures collapse and others don’t in large fires ? Have we learnt anything? If so, what has changed?
  22. 22. Delft University Architecture Faculty Building, May 2008 Concrete structures do fail in fire! Gretzenbach, Switzerland (Nov, 2004)
  23. 23. How can we learn from failures? Modelling and Simulation
  24. 24. The “spectrum” of modelling (analytical solutions) L, E, A, I DT, T,z z δz = δcirc(ǫφ) ≈ δsin(ǫφ) P = EA(ǫT − ǫφ) L, E, A, I DT, T,z M = EIφ P P M M x δz δx δx = (ǫT − ǫφ)l L, E, A, I DT, TP = EA(ǫT − ǫφ) ,z P P δz Deflections assuming ǫT > ǫφ δz = δcirc(ǫφ) + (ǫT − ǫφ) Post-buckling deflections 2 πadd to δz above a ǫp) calculation where, ǫp = ǫT − ǫφ − δsin(2 λPre-buckling deflections where, C = 0.3183l2A πI Cδcirc(ǫφ) 1 − C from thermal bowing from P-d moments
  25. 25. The “spectrum” of modelling (computational)
  26. 26. The “spectrum” of modelling (multi-hazard) Is deterministic analysis satisfactory in this context?
  27. 27. The “spectrum” of modelling (multi-hazard probabilistic) How to deal with uncertainty ? Monte Carlo Method Identity random parameters and their pdfs (sensitivity analysis may be necessary) Set pass/fail criteria Determine the acceptable level of “confidence” Run “n” number of deterministic analyses by randomly selecting values of random parameters (reduce “n” using a variance reduction technique) Determine the probability of failure
  28. 28. The “spectrum” of modelling (probabilistic - whole life) Performance Time 100% Extended life & near 100% performance with regular repair/maintenance Normal deterioration (no repair or maintenance) material durability and cyclic load/deformation induced cumulative damage Accelerated deterioration (no repair or maintenance) following a short duration extreme event, e.g. blast, fire, windstorm, earthquake) Likely scenarios in current practice Initial design life Extreme event Repair Maintenance Tolerance threshold for deterioration in performance Life-cycle analysis
  29. 29. Integrated computational environment for structures in fire This is only part of the big picture Fire models Standard Natural/parametric (short hot/long cool) Localised Travelling CFD (FDS, OpenFOAM, ANSYS-CFX) fire – heat transfer middleware Heat transfer OpenSees, ABAQUS, ANSYS heat transfer - thermomechanics middleware Structural response OpenSees, ABAQUS, ANSYS (general) SAFIR, VULCAN (special purpose)
  30. 30. The “spectrum” of modelling Multi-hazard/multi-scale Whole building Uncertainty Life-cycle analysis Real fires (CFD) CURRENT FOCUS Non-uniform fires Real fires (CFD) 3D sub-frames Non-uniform fires Model complexity Computational cost (log scale) Ideal fires (unif.) 2D sub-frames Analytical 3D sub-frames 3D sub-frames Uncertainty 3D sub-frames Uncertainty
  31. 31. Previous modelling work reproduced using OpenSees
  32. 32. Cardington frame 8 Storey steel frame composite structure 2 tests by BRE 4 tests carried out by “British Steel” (Corus), shown on building plan below Restrained beam test Corner test Frame test Demonstration test Download report from: www.mace.manchester.ac.uk/project/research/structures/strucfire/DataBase/References/MultistoreySteelFramedBuildings.pdf
  33. 33. Modelling of Cardington tests using OpenSees Restrained beam test Corner test Experiment OpenSees 0 100 200 300 400 500 600 700 800 900 1000 0 -50 -100 -150 -200 -250 -300 -350 -400 -450 Midspan deflection of joist at gridline 1/2 (mm) Temperature of the joist lower flange at mid span (oC) Experiment OpenSees 0 100 200 300 400 500 600 700 800 900 0 -50 -100 -150 -200 -250 Midspan deflection of joist (mm) Temperature of the joist lower flange at mid span (oC)
  34. 34. WTC Towers Collapse models (3 Floor Fire – no damage)
  35. 35. WTC Towers Collapse models (3 Floor Fire – no damage)
  36. 36. 3D model: Truss deformations
  37. 37. Collapse mechanism from 3D model
  38. 38. Photograph from NIST report
  39. 39. Photograph from NIST report
  40. 40. Are there generic collapse mechanisms? Weak floor Strong floor
  41. 41. Model to test generic mechanisms 10m 6m Universal Beam Universal Column Beam udl (N/mm) Column load (N) Floor span Strong beam 533x210x92 305x305x198 45 6000 10 Weak beam 305x102x28 305x305x198 45 6000 10
  42. 42. Model results Weak floor mechanism Strong floor mechanism
  43. 43. OpenSees models 2D model of WTC collapse
  44. 44. Current status of the work on the “Integrated Computational Environment for Structures in Fire”
  45. 45. Integrated computational environment for structures in fire Fire models Standard Natural/parametric (short hot/long cool) Localised Travelling OpenSees CFD (FDS, OpenFOAM, ANSYS-CFX) & later FDS (NIST) & OpenFOAM Eventually all to be packaged fire – heat transfer middleware Heat transfer OpenSees, ABAQUS, ANSYS heat transfer - thermomechanics middleware Structural response OpenSees, ABAQUS, ANSYS (general) SAFIR, VULCAN (special purpose) within a “wrapper” software capable of carrying out PBE and probabilistic analyses Integrity failure currently Matlab later OpenSees structure – fire coupling OpenSees OpenSees Open source and freely downloadable from UoE and later main OpenSees site at PEER/UC Berkeley
  46. 46. UoE OpenSees Wiki site https://www.wiki.ed.ac.uk/display/opensees o Command manual o Demonstration examples o Downloading executable application o Browsing source code
  47. 47. First version of the “integrated enviroment” in OpenSees ♦ Scheme for Modelling Structure in fire SIFBuilder Fire Heat Transfer Thermo-mechanical User-friendly interface for creating (regular) structural models and enable consideration of realistic fire action Models of fire action (only idealised fires), i.e., Standard fire, Parametric fire, EC1 Localised fire, Travelling fire Heat transfer to the structural members due to fire action Structural response to the elevated temperatures
  48. 48. SiF Builder Developed for creating large models Driven by Tcl Minimum input required Geometry information -XBays,Ybays,Storeys Structural information -Material, Section Loading information -Selfweight, Horizontal loading -Fire action
  49. 49. Floor Building Sub-frame Partitions Heat flux distribution from a localised Fire (at a single instant) Why is this needed? Example showing complexities (currently ignored in practice !) of modelling even a simple real structure Typical beam X-sections Fire exposed surface Typical column X-sections Fire exposed surface
  50. 50. OpenSees heat transfer analysis of RC column X-section Temperature distribution in the RC column after 1 hour of EC1 localised fire Temperature distribution in the RC column after 2 hours of EC1 localised fire
  51. 51. OpenSees heat transfer analysis of RC beam-slab X-section Temperature distribution in the RC beam after 1 hour of EC1 localised fire Temperature distribution in the RC beam after 2 hours of EC1 localised fire
  52. 52. Idealised Fires Uniform (no spatial variation) fires Standard fire: ISO- 834 fire curve Hydro- carbon fire: EC1 Empirical Parametric fire: EC1 Parametric fire model non- uniform (spatial and temporal variation) fires Localised fire Alpert ceiling jet model Travelling fire
  53. 53. Heat Transfer modelling HT materials - CarbonSteelEC3, ConcreteEC2 - Steel ASCE - easy to extend the library, - Entries for conductivity, specific heat HT elements - 1D, 2D, 3D heat transfer elements HT recorders (for structural analyses) Simple Mesh - I Beam, Concrete slab, Composite beam Fire Heat Transfer Heat flux BCs - Convection, radiation, prescribed heat fluxes Tcl interpreter Still under development Tcl commands available Easy to extend
  54. 54. Tcl commands for Heat transfer analysis Fire Initialization of heat transfer module HeatTransfer 2D3D; --To activate Heat Transfer module Definition of Heat Transfer Materials HTMaterial CarbonSteelEC3 1;. HTMaterial ConcreteEC2 2 0.5; Definition of Section or Entity HTEntity Block2D 1 0.25 0.05 $sb 0.10; Meshing the entity #SimpleMesh $MeshTag $HTEntityTag $HTMaterialTag $eleCtrX $eleCtrY; SimpleMesh 1 1 1 10 10; Definition of fire model FireModel Standard 1; …….. Heat Transfer
  55. 55. Strategy for efficient heat transfer modelling Fire Heat Transfer Idealised uniform fires, T(t): Heat flux input is spatially invariant over structural member surfaces; 2D heat transfer analysis for beam section, 1D for concrete slab 3D 2D 1D q convection +radiation q heat flux input
  56. 56. Strategy for efficient heat transfer modelling Idealised non-uniform fires, T(x,y,z,t): oHeat flux input varies with the location ; oComposite beam: a series of 2D sectional analyses oConcrete slab : using localised1D Heat Transfer analyses 2D section with localised BC q convection + radiation q heat flux input Section 1 Section 2 Section 3 1D section with localised BC Localised heat flux
  57. 57. Composite beam - 2D X-sectional versus full 3D model 0.0 0.5 1.0 1.5 2.0 2.5 3.0 800 700 600 500 400 300 200 100 0 Bottom flange of steel I-beam Bottom flange Temperature (oC) Section Location (mm) 3D-300s 2D-300s 3D-600s 2D-600s 3D-900s 2D-900s 3D-1200s 2D-1200s 3D-1500s 2D-1500s 3D-1800s 2D-1800s 0.0 0.5 1.0 1.5 2.0 2.5 3.0 140 120 100 80 60 40 20 0 Mid-depth of Concrete Slab Temperature in concrete slab (oC) Section location (m) 3D-300s 2D-300s 3D-600s 2D-600s 3D-900s 2D-900s 3D-1200s 2D-1200s 3D-1500s 2D-1500s 3D-1800s 2D-1800s Composite beam Length: 3m Steel beam: UB 356 × 171 × 51 Concrete slab: 1.771× 0.1m Material with Thermal properties according to EC2 and EC3 EC localised fire Heat release rate: 3MW Diameter: 1m, Ceiling height:3m Fire origin: under the beam end What we found Exactly the same temperature profile!
  58. 58. Concrete slab – 1D (through-depth only) versus full 3D model Concrete slab: Dimension: 5m×5m× 0.1m Material with Thermal properties according to EC2 EC localised fire Heat release rate: 5MW Diameter: 1m Ceiling height:3m Fire origin: under the slab corner What we found: Localised 1D analysis produces identical temperature profile as 3D analysis Temperature plot for slab bottom 800 Concrete slab exposed to localised fire 0 1 2 3 4 5 600 400 200 0 Diameter:1m Storey height:3m HRR=5MW Concrete slab temperature (oC) Section Location (mm) Bottom_3D Bottom_1D Mid depth_3D Mid depth _1D Top_3D Top _1D
  59. 59. Thermo-mechanical modelling ♦ Thermo-mechanical classees Heat Transfer Thermo-mechanical analysis HT recorders (for structural analyses) Thermomechanical materials - With temperature dependent properties Thermomechanical sections -Beam sections membrane plate section Thermomechanical elements -Disp based beam elements, MITC4 shell elements Loading:Thermal action -2D3D BeamThermalAction, ShellThermalAction - NodalThermalAction
  60. 60. ♦ Tcl commands for material, section, and elements uniaxialMaterial SteelECThermal $matTag EC3 $fy $E0; … section FiberThermal $secTag { Fibre.. Patch.. Layer.. } … element dispBeamColumnThermal $eleID $node1 $node2 $NumIntgers $secTag $GeomTransTag; … block2D $nx $ny $NodeID0 $EleID0 ShellMITC4Thermal $SecTag { …. }
  61. 61. Thermo-meecchhaanniiccaall aannaallyyssiiss ♦ Tcl commands for defining beam thermal actions Uniform along beam length, non-uniform through depth pattern Plain $PatternTag Linear { … eleLoad –ele $eleID -type -beamThermal $T1 $y1 $T2 $y2 $T3 $Y3 … $T9 $Y9 …} pattern Fire $PatternTag $Path $Path $Path $Path $Path $Path $Path $Path $Path { … eleLoad –ele $eleID -type -beamThermal $T1 $y1 $T2 $y2 $T3 $Y3 … $T9 $Y9 …} Using Linear Load pattern Using Fire Load pattern for further non-uniform profile
  62. 62. Thermo-meecchhaanniiccaall aannaallyyssiiss ♦ Tcl commands for defining beam thermal actions Importing external temperature history file pattern Plain $PatternTag Linear { … eleLoad -ele $eleID -type -beamThermal -source $filePath $y1 $y2 $y3…$y9; …} “BeamTA/element1.dat” Time T1 (corresponding to y1)……………………………T9(corresponding to y9) 60 47.2327 46.4181 47.3834 47.5978 47.6355 47.5978 47.3834 46.4181 47.2327 120 75.5848 74.8791 76.5169 77.0164 77.1223 77.0164 76.5169 74.8791 75.5848 180 104.494 103.735 105.762 106.516 106.698 106.516 105.762 103.735 104.494 …. ….
  63. 63. Thermo-meecchhaanniiccaall aannaallyyssiiss ♦ Tcl commands for defining beam thermal actions Non-uniform along beam length, either through depth pattern Plain $PatternTag Linear { … eleLoad -range $eleTag0 $eleTag1 -type -beamThermal -source -node; load $nodeTag -nodalThermal $T1 $Y1 $T2 $Y2; … } Apply Nodal thermal action Source external temperature file … load $nodeTag -nodalThermal –source $filePath $y1 $y2; …
  64. 64. Thermo-meecchhaanniiccaall aannaallyyssiiss ♦ Tcl commands for defining beam thermal actions ThermalAction for 3D I section beams … eleLoad –ele $eleID -type -beamThermal $T1 $y1 …T5 $Y5 $T6 $T7 $Z1 $T8 $T9 $Z2 … $T14 $T15 $Z5 … Lower Flange Temperature T6,8,10,12,14 Web Temperature T1,2,3,4,5 Upper Flange Temperature T7,9,11,12,15
  65. 65. Examples [Available@UoE Wiki]
  66. 66. Examples-Simply ssuuppppoorrtteedd bbeeaam ♦ A simply supported steel beam; ♦ Uniform distribution load q= 8N/mm ♦ Uniform temperature rise ΔT; ♦ Using FireLoadPattern Temperature-time curve defined by FireLoadPattern: Tcl script uniaxialMaterial Steel01Thermal 1 308 2.1e5 0.01; element dispBeamColumnThermal 1 1 2 5 $section 1;
  67. 67. 1) without thermal elongation? 2) UDL removed? Examples-Simply ssuuppppoorrtteedd bbeeaam Deformation shape (without UDL) Deformation shape (with UDL) Beam end displacement Material degradation UDL applied
  68. 68. Examples-Restrained Beam under tthheerrmaall eexxppaannssiioonn ♦ 2D elements, Fixed ends; ♦ Element 1 with ΔT ≠0 , only one free DOF at Node 3 - The effects of Thermal expansion; -stiffness degradation, strength loss; -and restraint effects; set secTag 1; section FiberThermal $secTag { fiber -25 -25 2500 1; fiber -25 25 2500 1; fiber 25 -25 2500 1; fiber 25 25 2500 1; }; … pattern Plain 1 Linear { eleLoad -ele 1 -type -beamThermal 1000 - 50 1000 50 }; set secTag 1; section FiberThermal $secTag { fiber -25 0 5000 1; fiber 25 0 5000 1; }; … pattern Plain 1 Linear { eleLoad -ele 1 -type -beamThermal 1000 -50 1000 50 }; 2D beam element 3D beam element
  69. 69. Examples-Restrained Beam under tthheerrmaall eexxppaannssiioonn ♦ 2D elements, Fixed ends; ♦ Element 1 with ΔT ≠0 , only one free DOF at Node 3 Material softening No strength loss in heated part (stiffness loss considered) Considering strength loss - The effects of Thermal expansion; -stiffness degradation, strength loss; -and restraint effects; Thermal expansion In heated element
  70. 70. Examples-CCoomppoossiittee BBeeaam Composite beams with column connected Deformation shape 1) Column was pushed out by thermal expansion; 2) Being pulled back by Catenary action
  71. 71. Thank you

×