Probabilistic Models for Natural Hazards – MODNAT
Mimic natural hazards in order to better cope with them
High fidelity stochastic models construction for natural hazards and study of their impact on structures.
To take into account non gaussian and non stationnary characters of hazards
The main goal of this project is to develop high fidelity stochastic models of natural events in order to:<br /><br />1.describe their uncertain behaviour <br />2.generate easily synthetic realizations <br />3.evaluate the impacts of natural events on structures<br /><br />High fidelity, meaning here that we want to overcome the classical Gaussian or/and stationary assumptions which are almost always introduced when dealing with stochastic processes or random fields modelling. The construction of such models is much more difficult since the description of a non Gaussian non stationary process requires the knowledge of its non countable, time varying marginal distribution family. Therefore the first step is to overcome this technological bottle-neck. The second objective is to propose user-friendly models which are not CPU-intensive since they have to be used in general in conjunction with Monte Carlo procedures. This represents another challenge. <br />Once a model of a natural event such a gust or seismic signal is constructed, engineers have at their disposal a practical easy-to-use tool to predict the outcomes of the natural event on their structure or process. <br />Manufacturers will gain a greater confidence through risk analysis and will be able to optimize the various safety margins introduced in their models, leading to development and exploitation cost reductions.<br />The use of the proposed mathematical models for industrial applications represents a pioneering advance in the representation of natural events.<br />
The objectives of the task are to construct probabilistic models for random phenomena which are non stationary or non homogeneous. No assumptions on the probability distributions will be introduced (for instance Gaussian assumption). The validation of the models will be demonstrated on academic examples. Numerical models will be given in order to study the statistical properties of system outputs to such inputs. The distribution of maxima will be studied.
work in progress
Expected Scientific, Technological and Economical Outcomes
Numerically efficient models and simulation tools for natural hazard representation.
Best-estimate models accounting for natural variability
More robust and reliable design of structures.
More accurate risk evaluation
Cost reducing for construction and exploitation
Reduction of environmental impact
Fabrice, Poirion, Irmela, Zentner
Non-Gaussian non-stationary models for natural hazard modeling, Applied Mathemetical Modelling, Volume 37, Issue 8, Pages 5938-5950
The Probabilistic Models for Natural Hazards (MODNAT) project has for goal to develop high fidelity numerical probabilistic models in order to reproduce and take faithfully into account the effects of uncertain natural phenomena.
The developed methods will allow to reproduce the most pertinent characteristics of the natural phenomenon involved, either sample paths or statistical quantities such as maxima laws. A key aspect of this project for the model development will be the exploitation of experimental databases. The main applications which will be considered are model construction of turbulence gust, seismic acceleration and sea waves. However the methodologies will be general enough to be applied to other fields such as biology, medicine or finance.
The developed models will be applied to various civil engineering fields dealing with risk and certification:
seismic assessment of dams,
certification of wind turbines to turbulent gusts,
design of sea walls and other maritime structures.
The different approaches will then be compared and benchmarking is performed. Where possible, the models will be implemented in opensource or other reference software used by the respective community.
Another goal of the project is to gather, besides the developed methods in the project, other existing approaches in order to offer a library of numerically efficient methods dealing with probabilistic model construction of natural events to the project partners and the scientific community.
Project coordination
fabrice POIRION (Office National d'Etudes et de Recherches Aéronautiques)
The author of this summary is the project coordinator, who is responsible for the content of this summary. The ANR declines any responsibility as for its contents.
Partner
UBP Université Blaise Pascal Institut Pascal
BU University of Brighton
IFPEN IFP Energies nouvelles
IFREMER Institut francais de recherche pour l'exploitation de la mer
EDF R&D Electricité de France
ONERA Office National d'Etudes et de Recherches Aéronautiques
Help of the ANR 621,257 euros
Beginning and duration of the scientific project:
September 2012
- 39 Months
