Stefano Marelli


Last Name

Marelli

First Name

Stefano

ORCID

0000-0002-9268-9014

Organisational unit

03962 - Sudret, Bruno / Sudret, Bruno

Filters

2010 - 2019502020 - 2026110
Reset filters

Search Results

Publications 1 - 10 of 160
  • Reliability analysis of wind turbines using manifold-NARX surrogate models
    Item type: Other Conference Item
    Schär, Styfen; Marelli, Stefano; Sudret, Bruno (2023)
    Modeling the dynamic response of systems is essential for structural health monitoring, reliability analysis or design optimization. Structures such as wind turbines, tall buildings or long bridges are submitted to transient excitation such as ground motions or wave and wind loads, which inherently showcase aleatory uncertainties, however. To address uncertainty quantification questions such as sensitivity or reliability analyses, many runs of the simulator of the system would be needed, which is rather intractable in practice, thus the development of surrogate models in the last decade. The construction of surrogates for models with high-dimensional input and output time-series, is however, a complex topic that is not fully covered in the recent literature. To tackle this class of problems, we propose a novel approach called manifold NARX (mNARX), which leverages prior knowledge of the system's physics to incrementally build an input manifold for efficiently surrogating dynamic system responses. In the case of wind turbines, the input wind box, which is a 10-minute three-dimensional wind field sampled at 20 Hz is first compressed using a 2-D discrete cosine transform, which yields time-series of spectral coefficients that we call features. These features are used to initially predict simple auxiliary quantities such as control system actions using a standard NARX approach. In an incremental process, the spectral features and the previously build auxiliary quantities are used to model increasingly complex variables such as rotor speed and position. In a final step, we use the spectral coefficients and auxiliary quantities to form an exogenous input manifold on which NARX models of the final quantities of interest (e.g., power generation, blade and tower bending moments, etc.) are constructed. The training and evaluation of this surrogate model chain are computationally inexpensive (i.e., a handful of full system simulations is sufficient), up to several orders of magnitude cheaper than the original simulator. It can then be used for reliability analysis using standard simulation methods, as demonstrated in the investigated example.
  • Seismic Fragility Analysis based on Artificial Ground Motions and Surrogate Modeling of Validated Structural Simulators
    Item type: Working Paper
    Abbiati, Giuseppe; Broccardo, Marco; Abdallah, Imad; et al. (2020)
    engrXiv
    This study introduces a computational framework for efficient and accurate seismic fragility analysis based on a combination of artificial ground motion modeling, polynomial-chaos-based global sensitivity analysis, and hierarchical kriging surrogate modeling. The framework follows the philosophy of the Performance-Based Earthquake Engineering PEER approach, where the fragility analysis is decoupled from hazard analysis. This study addresses three criticalities that are present in the current practice. Namely, reduced size of hazard-consistent size-specific ensembles of seismic records, validation of structural simulators against large-scale experiments, high computational cost for accurate fragility estimates. The effectiveness of the proposed framework is demonstrated for the Rio Torto Bridge, recently tested using hybrid simulation within the RETRO project.
  • Multifidelity surrogate modelling with noisy grey-box models
    Item type: Conference Poster
    Giannoukou, Aikaterini; Marelli, Stefano; Sudret, Bruno (2022)
  • Introducing efficient structural reliability methods for stochastic simulators
    Item type: Other Conference Item
    Pires, Anderson V.; Moustapha, Maliki; Marelli, Stefano; et al. (2024)
    SIAM Conference on Uncertainty Quantification (UQ 2024). Searchable Abstract Document
  • An adaptive algorithm based on spectral likelihood expansion for efficient Bayesian calibration
    Item type: Other Conference Item
    Wagner, Paul-Remo; Lataniotis, Christos; Marelli, Stefano; et al. (2019)
  • A global framework for active learning reliability in UQLab
    Item type: Other Conference Item
    Moustapha, Maliki; Marelli, Stefano; Sudret, Bruno (2022)
  • Use of stochastic emulators to establish fragility curves of structures under tornado winds
    Item type: Other Conference Item
    Kroetz, Henrique M.; Costa Macedo, Felipe; Marelli, Stefano; et al. (2025)
    14th International Conference on Structural Safety and Reliability (ICOSSAR’25) - Book of Abstracts
    Design optimization is crucial in ensuring that structures are safe and robust against uncertainties. To account for various uncertainties in the design (e.g., manufacturing tolerances) or the structure’s environment (e.g., loads), reliability-based design optimization (RBDO) trades the design cost against a target reliability level. Solving an RBDO problem is however computationally expensive as it entails performing a reliability analysis for each design considered throughout the optimization procedure. While surrogate models can reduce this cost, they become inefficient in high-dimensional spaces. In this contribution, we present an alternative approach to solving RBDO problems that makes use of stochastic emulators, which are surrogates used to approximate so-called stochastic simulators. Unlike deterministic simulators, stochastic simulators yield different results when evaluated multiple times on the same design. This is due to some latent variability that is beyond the designer’s control.In the proposed approach, we first reformulate the problem by splitting the model inputs into deterministic design parameters and random variables, and aggregate the latter into latent ones. By integrating these latent variables, the response for each design becomes stochastic, enabling approximation by stochastic emulators such as stochastic polynomial chaos expansions (SPCE) [1] or surrogate-based empirical distributions like generalized lambda models (GLaM) [2]. The advantages of introducing stochastic emulators are two-fold. First, they enable efficient handling of high-dimensional problems, as random variables are implicitly represented through the stochastic emulator. Second, they accelerate the optimization process by providing semi-analytical methods to compute quantiles (which can be mapped to exceedance probabilities), thus avoiding the need for time-consuming reliability analysis at each optimization iteration. We demonstrate the effectiveness of this approach through application examples of moderate and high dimensionality. [1] Zhu, X. and B. Sudret (2023). Stochastic polynomial chaos expansions to emulate stochastic simulators. International Journal for Uncertainty Quantification 13 (2), 31-52. [2] Zhu, X. and B. Sudret (2021). Emulation of stochastic simulators using generalized lambda models. SIAM/ASA Journal on Uncertainty Quantification 9 (4), 1345-1380.
  • Active learning for system reliability analysis using PC-Kriging, subset simulation and sensitivity analysis
    Item type: Report
    Parisi, Pietro; Moustapha, Maliki; Marelli, Stefano; et al. (2022)
    RSUQ-Report
    Structural reliability analysis aims at assessing the safety of structures which often operate under uncertain conditions. Approximation-, simulation- or surrogate-based methods are used in this context to estimate the failure probability. Surrogate-based methods are the least computationally intensive and consist in building a cheaper proxy of the original limit-state function, which is calibrated using a limited set of samples known as the experimental design. The latter is sequentially enriched to increase the accuracy of the surrogate in areas of interest, hence allowing for an accurate estimation of the failure probability. A large number of such techniques has been recently developed in the literature (e.g., active Kriging - Monte Carlo simulation). However most of these techniques consider a single limit-state function. These methods lose efficiency when used to solve system reliability problems, where failure is defined by a non-trivial combination of multiple limit-states. This is due to some peculiarities of the system problem such as the presence of disjoint failure domains or the uneven contribution of each limit-state to the overall failure. In this work, we propose an efficient algorithm combining Kriging/PC-Kriging and subset simulation to solve system reliability problems in their most general setting, i.e., with an arbitrary combination of components. We devise a new learning function which first identifies candidate samples and then selects for enrichment the specific limit-state that contributes the most to system failure. The algorithm is validated on a set of analytical functions and compared with existing methods in the literature.
  • Stochastic spectral embedding for Bayesian inverse problems
    Item type: Other Conference Item
    Wagner, Paul-Remo; Marelli, Stefano; Lataniotis, Christos; et al. (2019)
  • Active learning for structural reliability: Survey, general framework and benchmark
    Item type: Journal Article
    Moustapha, Maliki; Marelli, Stefano; Sudret, Bruno (2022)
    Structural Safety
    Active learning methods have recently surged in the literature due to their ability to solve complex structural reliability problems within an affordable computational cost. These methods are designed by adaptively building an inexpensive surrogate of the original limit-state function. Examples of such surrogates include Gaussian process models which have been adopted in many contributions, the most popular ones being the efficient global reliability analysis (EGRA) and the active Kriging Monte Carlo simulation (AK-MCS), two milestone contributions in the field. In this paper, we first conduct a survey of the recent literature, showing that most of the proposed methods actually span from modifying one or more aspects of the two aforementioned methods. We then propose a generalized modular framework to build on-the-fly efficient active learning strategies by combining the following four ingredients or modules: surrogate model, reliability estimation algorithm, learning function and stopping criterion. Using this framework, we devise 39 strategies for the solution of 20 reliability benchmark problems. The results of this extensive benchmark (more than 12,000 reliability problems solved) are analyzed under various criteria leading to a synthesized set of recommendations for practitioners. These may be refined with a priori knowledge about the feature of the problem to solve, i.e. dimensionality and magnitude of the failure probability. This benchmark has eventually highlighted the importance of using surrogates in conjunction with sophisticated reliability estimation algorithms as a way to enhance the efficiency of the latter.
Publications 1 - 10 of 160