The Department of Structural Reliability at TNO (Netherlands Organization for Applied Scientific Research) is organizing a workshop on the computational challenges and aspects in the reliability analysis of engineering structures. The particular focus of the workshop is reliability problems that are: (i) high-dimensional (>40); (ii) have multiple failure modes; (iii) and have small failure probabilities (1e-3–1e-6).
The aim of the workshop is to bring together researchers, practitioners, and software developers from all over the world to discuss challenges related to this topic, share experience, learn from each other, and to jointly find ways of solving these challenges. The application area is intentionally broad, we expect practitioners and researchers from all branches of engineering who face the outlined challenges, e.g. civil, mechanical, aerospace, industrial, and mechatronics engineers.
With the advancements in probability theory and structural reliability, combined with the increase in computational power, it is now possible to conduct probabilistic analysis of engineering models. However, it is still challenging to perform reliability analysis on complex, realistic engineering applications; for example when non-linear finite element analysis is involved.
The reason is that one of the main tasks in reliability analysis is to compute high-dimensional integrals in order to determine the failure probability. This is challenging, because (i) the curse of dimensionality; and (ii) typically we are interested in estimating small probabilities (1e-3–1e-6).
Efficient methods are available to compute failure probabilities if only a single failure mode (SFM) is involved and the number of random variables (≡ dimension of the integral) is moderate (<20). However, problems with multiple failure modes are abundant in practice. For these, many of the efficient methods, which excel in SFM problems, break down and one is often compelled to fall back to more general, but at the same time computationally more demanding methods.
Another direction where the methods, which excel in SFM problems, can break down is where the number of random variables increases. This can easily happen when certain properties are modeled as stochastic processes, for instance as random fields. Again, the challenge is that the generally applicable methods, e.g. crude Monte Carlo or directional sampling, necessitates enormous computational capacity to obtain answers in a reasonable time.
In both cases there are two possible fronts to tackle the challenges: (a) using “smarter” reliability methods that require fewer evaluations of the expensive engineering models; and (b) “smarter” computational techniques that reduce the computational time for the same number of model evaluations. Just to indicate the possible directions and methods, the following examples are given for (a): advanced simulation techniques (e.g. subset simulation, importance sampling, stratified sampling); surrogate models or response surfaces (e.g. artificial neural networks, kriging, polynomial chaos expansion); advanced discretization of stochastic processes. Examples of possible directions for (b) are: parallelization on CPU and/or GPU, and cloud computing.
But how efficient and accurate are these methods and can they be applied for realistic engineering problems? The workshop intends to address these questions by providing a forum to share expertise from diverse disciplines in an open and informal atmosphere.
Location and date
The workshop will take place at the Business Centre Delftech, Delftechpark 29, 2628 Delft, The Netherlands.
The workshop will be held on the 24th of January, 2018.
Lodging and travel
You will need to provide your own transportation to Delft and to take care of your lodging.
We have a limited number of funds to assist those who would not otherwise be able to attend the workshop.
Agnieszka J. Bigaj van Vliet
Nadieh E. Meinen
Arthur T. Slobbe
Raphaël D.J.M. Steenbergen
If you have any questions please contact us using this email address:
We hope you'll join us in Delft!