Advanced search
1 file | 812.58 KB Add to list

The MaxEnt method for probabilistic structural fire engineering : performance for multi-modal outputs

Author
Organization
Abstract
Probabilistic Risk Assessment (PRA) methodologies are gaining traction in fire engineering practice as a (necessary) means to demonstrate adequate safety for uncommon buildings. Further, an increasing number of applications of PRA based methodologies in structural fire engineering can be found in the contemporary literature. However, to date, the combination of probabilistic methods and advanced numerical fire engineering tools has been limited due to the absence of a methodology which is both efficient (i.e. requires a limited number of model evaluations) and unbiased (i.e. without prior assumptions regarding the output distribution type). An uncertainty quantification methodology (termed herein as MaxEnt) has recently been presented targeted at an unbiased assessment of the model output probability density function (PDF), using only a limited number of model evaluations. The MaxEnt method has been applied to structural fire engineering problems, with some applications benchmarked against Monte Carlo Simulations (MCS) which showed excellent agreement for single-modal distributions. However, the power of the method is in application for those cases where ‘validation’ is not computationally practical, e.g. uncertainty quantification for problems reliant upon complex modes (such as FEA or CFD). A recent study by Gernay, et al., applied the MaxEnt method to determine the PDF of maximum permissible applied load supportable by a steel-composite slab panel undergoing tensile membrane action (TMA) when subject to realistic (parametric) fire exposures. The study incorporated uncertainties in both the manifestation of the fire and the mechanical material parameters. The output PDF of maximum permissible load was found to be bi-modal, highlighting different failure modes depending upon the combinations of stochastic parameters. Whilst this outcome highlighted the importance of an un-biased approximation of the output PDF, in the absence of a MCS benchmark the study concluded that some additional studies are warranted to give users confidence and guidelines in such situations when applying the MaxEnt method. This paper summarises one further study, building upon Case C as presented in Gernay, et al.
Keywords
fire, probability, uncertainty, MaxEnt, bimodal

Downloads

  • 2019 07 01 Interflam2019 HopkinEa bimodal.pdf
    • full text
    • |
    • open access
    • |
    • PDF
    • |
    • 812.58 KB

Citation

Please use this url to cite or link to this publication:

MLA
Hopkin, Danny et al. “The MaxEnt Method for Probabilistic Structural Fire Engineering : Performance for Multi-modal Outputs.” Proceedings of Interflam 2019. 2019. Print.
APA
Hopkin, D., Fu, I., Gernay, T., Elhami Khorasani, N., & Van Coile, R. (2019). The MaxEnt method for probabilistic structural fire engineering : performance for multi-modal outputs. Proceedings of Interflam 2019. Presented at the Interflam 2019.
Chicago author-date
Hopkin, Danny, Ian Fu, Thomas Gernay, Negar Elhami Khorasani, and Ruben Van Coile. 2019. “The MaxEnt Method for Probabilistic Structural Fire Engineering : Performance for Multi-modal Outputs.” In Proceedings of Interflam 2019.
Chicago author-date (all authors)
Hopkin, Danny, Ian Fu, Thomas Gernay, Negar Elhami Khorasani, and Ruben Van Coile. 2019. “The MaxEnt Method for Probabilistic Structural Fire Engineering : Performance for Multi-modal Outputs.” In Proceedings of Interflam 2019.
Vancouver
1.
Hopkin D, Fu I, Gernay T, Elhami Khorasani N, Van Coile R. The MaxEnt method for probabilistic structural fire engineering : performance for multi-modal outputs. Proceedings of Interflam 2019. 2019.
IEEE
[1]
D. Hopkin, I. Fu, T. Gernay, N. Elhami Khorasani, and R. Van Coile, “The MaxEnt method for probabilistic structural fire engineering : performance for multi-modal outputs,” in Proceedings of Interflam 2019, Egham, United Kingdom, 2019.
@inproceedings{8622466,
  abstract     = {Probabilistic Risk Assessment (PRA) methodologies are gaining traction in fire engineering practice as a (necessary) means to demonstrate adequate safety for uncommon buildings. Further, an increasing number of applications of PRA based methodologies in structural fire engineering can be found in the contemporary literature. However, to date, the combination of probabilistic methods and advanced numerical fire engineering tools has been limited due to the absence of a methodology which is both efficient (i.e. requires a limited number of model evaluations) and unbiased (i.e. without prior assumptions regarding the output distribution type).

An uncertainty quantification methodology (termed herein as MaxEnt) has recently been presented targeted at an unbiased assessment of the model output probability density function (PDF), using only a limited number of model evaluations. 

The MaxEnt method has been applied to structural fire engineering problems, with some applications benchmarked against Monte Carlo Simulations (MCS) which showed excellent agreement for single-modal distributions. However, the power of the method is in application for those cases where ‘validation’ is not computationally practical, e.g. uncertainty quantification for problems reliant upon complex modes (such as FEA or CFD). 

A recent study by Gernay, et al., applied the MaxEnt method to determine the PDF of maximum permissible applied load supportable by a steel-composite slab panel undergoing tensile membrane action (TMA) when subject to realistic (parametric) fire exposures. The study incorporated uncertainties in both the manifestation of the fire and the mechanical material parameters. The output PDF of maximum permissible load was found to be bi-modal, highlighting different failure modes depending upon the combinations of stochastic parameters. Whilst this outcome highlighted the importance of an un-biased approximation of the output PDF, in the absence of a MCS benchmark the study concluded that some additional studies are warranted to give users confidence and guidelines in such situations when applying the MaxEnt method. This paper summarises one further study, building upon Case C as presented in Gernay, et al. },
  author       = {Hopkin, Danny and Fu, Ian and Gernay, Thomas and Elhami Khorasani, Negar and Van Coile, Ruben},
  booktitle    = {Proceedings of Interflam 2019},
  keywords     = {fire,probability,uncertainty,MaxEnt,bimodal},
  language     = {eng},
  location     = {Egham, United Kingdom},
  pages        = {12},
  title        = {The MaxEnt method for probabilistic structural fire engineering : performance for multi-modal outputs},
  year         = {2019},
}