arXiv subjects

Pierre Jehel

Publications and source records attributed to Pierre Jehel.

10 recordsLinked to original sources

arXiv2026

An Agentic Interface for End-to-End Probabilistic Seismic Hazard and Risk Analysis

Probabilistic seismic hazard and risk analyses are backbone to building codes, insurance pricing, and disaster management. Yet their open-engine pipelines remain accessible primarily to experts. We present the first agentic interface to the end-to-end probabilistic seismic hazard and risk chain via an open-source server, addressable through the Model Context Protocol (MCP). MCP wraps the OpenQuake engine and the 2020 European Seismic Hazard and Risk Models using twenty-four typed endpoints. Here, an agent is defined as a large language model (LLM) with tools. LLM is confined to the role of an orchestrator so it plans, translates, and explains, while the OpenQuake engine and custom codes compute hazard values, damage probability, and loss. Each response carries source-model, ground-motion model (GMM), and certified data provenance for transparency. Results are benchmarked against ESHM20 at seventy-three cities, the replicated 475-year spectral accelerations match official values within a median of 5 %, and a full hazard-to-loss estimate runs in minutes. The interface additionally accepts a user-supplied empirical or machine-learning GMM on any tectonic region type of the published tree, and adds additional features that existing web services omit: conditional spectra, deterministic scenarios, surface hazard, per-building loss, retrofit comparison, and record selection with waveform retrieval. The proposed design layers transfer to other regional models, and to other hazards.

physics.geo-ph↗

arXiv2025

Towards a collaborative digital platform for railway infrastructure projects

The management of railway infrastructure projects can be supported by collaborative digital platforms. A survey was carried out to identify the needs and expectations of the various stakeholders involved in the design and construction of railway infrastructure projects regarding collaborative platforms. These needs and expectations can then be translated into functional specifications to be included in the digital platforms. A total of 21 interviews were conducted between October and December 2022, during which 35 individuals were interviewed. Key roles were represented across the different project phases: engineers from design and construction firms, project managers, infrastructure managers. And various engineering fields were represented: civil, electrical, telecommunications, tracks, systems. These interviews were carried out by CentraleSup{\'e}lec | Universit{\'e} Paris-Saclay and by SNCF R{\'e}seau using a structured protocol designed to collect the specific needs of the interviewees for collaboration, as well as the guiding principles that shape both individual work practices and collaboration between professions. The resulting material was analyzed and then synthesized into a conceptual model of a collaborative digital platform for supporting the design and construction phases in a railway infrastructure project. Also, from these interviews emerged five core functionalities that the platform must offer: Providing access to existing infrastructure data; Accelerating repetitive tasks; Verifying essential project requirements; Supporting decision-making; Facilitating coordination among stakeholders.

cs.HC↗

arXiv2025

Accelerating the Serviceability-Based Design of Reinforced Concrete Rail Bridges under Geometric Uncertainties induced by unforeseen events: A Surrogate Modeling approach

Reinforced concrete rail bridges are essential components of railway infrastructure, where reliability, durability, and adaptability are key design priorities. However, the design process is often complicated by uncertainties stemming from unforeseen construction constraints, such as the need to reposition piers or alter geometric characteristics. These design adaptations can lead to repeated redesigns, added costs, and project delays if not anticipated in the early design stages, as well as significant computational overhead when using traditional finite element (FE) simulations. To address this and anticipate such unexpected events, this study adopts surrogate modeling as an efficient probabilistic design approach. This methodology integrates key geometric parameters as random variables, capturing the uncertainties that may arise during the design and construction phases and propagating them on the bridge's performance functions. By doing so, we aim to enable the efficient exploration of a large number of design scenarios with minimal reliance on time-consuming finite element (FE) simulations, represent the performance functions of a reinforced concrete bridge as a function of our variable design parameters, and classify the overall design scenarios into failure and safe scenarios In this study, a four-span reinforced concrete bridge deck is modeled using a multi-fiber finite element approach in Cast3M software. This FE model is used to generate the required design of experiments to train the surrogate models. Within this framework, a comparative performance assessment is conducted to evaluate the performance of the Kriging surrogate against alternative methods, including polynomial chaos expansion (implemented in UQLab) and support vector regression (SVR). This methodology supports early-stage uncertainty-informed design, enhancing the robustness and adaptability of reinforced concrete rail bridges in the face of practical constraints and changing site conditions.

math.NA↗

arXiv2025

A surrogate-based approach to accelerate the design and build phases of reinforced concrete bridges

Integrating uncertainties in the design process of reinforced concrete rail bridges, in a fully probabilistic framework, makes their design more complex and challenging. To propagate these uncertainties and convey their influence on the performance of the engineering system, a high-dimensional design space is supposed to be explored. A great challenge to be considered here lies in the computational burden as conducting such an exploration campaign requires substantial calls to computationally expensive finite element simulations. To address this challenge, a surrogate model mapping the design space to the reinforced concrete bridge performance functions is developed in the context of an active learning algorithm. The importance of this model lies in its ability to explore as many design scenarios as possible with minimal computational resources and classify the design scenarios into failure and safe scenarios. This work considers a 4-span reinforced concrete bridge deck. A multi-fiber finite element model of this beam is developed in Cast3m to generate the required design of experiments for the surrogate model. A performance comparison is undertaken to evaluate the Kriging surrogate model effectiveness with and without active learning while the reliability of Kriging predictions is also assessed in comparison to PC-Kriging.

math.NA↗

arXiv2025

Breaking the Black Box: Inherently Interpretable Physics-Constrained Machine Learning With Weighted Mixed-Effects for Imbalanced Seismic Data

Ground motion models (GMMs) are critical for seismic risk mitigation and infrastructure design. Machine learning (ML) is increasingly applied to GMM development due to expanding strong motion databases. However, existing ML-based GMMs operate as 'black boxes,' creating opacity that undermines confidence in engineering decisions. Moreover, seismic datasets exhibit severe imbalance, with scarce large-magnitude near-field records causing systematic underprediction of critical high-hazard ground motions. Despite these limitations, research addressing both interpretability and data imbalance remains limited. This study develops an inherently interpretable neural network employing independent additive pathways with novel HazBinLoss and concurvity regularization. HazBinLoss integrates physics-constrained weighting with inverse bin count scaling to address underfitting in sparse, high-hazard regions. Concurvity regularization enforces pathway orthogonality, reducing inter-pathway correlation. The model achieves robust performance: mean squared error = 0.6235, mean absolute error = 0.6230, and coefficient of determination = 88.48%. Pathway scaling corroborates established seismological behaviors. Weighted hierarchical Student-t mixed-effects analysis demonstrates unbiased residuals with physically consistent variance partitioning: sigma components range from 0.26-0.38 (inter-event), 0.12-0.41 (inter-region), 0.58-0.71 (intra-event), and 0.68-0.89 (total). The lower inter-event and higher intra-event components have implications for non-ergodic hazard analysis. Predictions exhibit strong agreement with NGA-West2 GMMs across diverse conditions. This interpretable framework advances GMMs, establishing a transparent, physics-consistent foundation for seismic hazard and risk assessment.

cs.LG↗

arXiv2025

Graph Transformer-Based Flood Susceptibility Mapping: Application to the French Riviera and Railway Infrastructure Under Climate Change

Increasing flood frequency and severity due to climate change threatens infrastructure and demands improved susceptibility mapping techniques. While traditional machine learning (ML) approaches are widely used, they struggle to capture spatial dependencies and poor boundary delineation between susceptibility classes. This study introduces the first application of a graph transformer (GT) architecture for flood susceptibility mapping to the flood-prone French Riviera (e.g., 2020 Storm Alex) using topography, hydrology, geography, and environmental data. GT incorporates watershed topology using Laplacian positional encoders (PEs) and attention mechanisms. The developed GT model has an AUC-ROC (0.9739), slightly lower than XGBoost (0.9853). However, the GT model demonstrated better clustering and delineation with a higher Moran's I value (0.6119) compared to the random forest (0.5775) and XGBoost (0.5311) with p-value lower than 0.0001. Feature importance revealed a striking consistency across models, with elevation, slope, distance to channel, and convergence index being the critical factors. Dimensionality reduction on Laplacian PEs revealed partial clusters, indicating they could capture spatial information; however, their importance was lower than flood factors. Since climate and land use changes aggravate flood risk, susceptibility maps are developed for the 2050 year under different Representative Concentration Pathways (RCPs) and railway track vulnerability is assessed. All RCP scenarios revealed increased area across susceptibility classes, except for the very low category. RCP 8.5 projections indicate that 17.46% of the watershed area and 54% of railway length fall within very-high susceptible zones, compared to 6.19% and 35.61%, respectively, under current conditions. The developed maps can be integrated into a multi-hazard framework.

eess.SP↗

arXiv2017

Two types of spurious damping forces potentially modeled in numerical seismic nonlinear response history analysis

The purpose of this paper is to provide practitioners with further insight into spurious damping forces that can be generated in nonlinear seismic response history analyses (RHA). The term 'spurious' is used to refer to damping forces that are not present in an elastic system and appear as nonlinearities develop: such damping forces are not necessarily intended and appear as a result of modifications in the structural properties as it yields or damages due to the seismic action. In this paper, two types of spurious damping forces are characterized. Each type has often been treated separately in the literature, but each has been qualified as 'spurious', somehow blurring their differences. Consequently, in an effort to clarify the consequences of choosing a particular viscous damping model for nonlinear RHA, this paper shows that damping models that avoid spurious damping forces of one type do not necessarily avoid damping forces of the other type.

physics.geo-ph↗

arXiv2016

A Stochastic Multi-scale Approach for Numerical Modeling of Complex Materials - Application to Uniaxial Cyclic Response of Concrete

In complex materials, numerous intertwined phenomena underlie the overall response at macroscale. These phenomena can pertain to different engineering fields (mechanical , chemical, electrical), occur at different scales, can appear as uncertain, and are nonlinear. Interacting with complex materials thus calls for developing nonlinear computational approaches where multi-scale techniques that grasp key phenomena at the relevant scale need to be mingled with stochastic methods accounting for uncertainties. In this chapter, we develop such a computational approach for modeling the mechanical response of a representative volume of concrete in uniaxial cyclic loading. A mesoscale is defined such that it represents an equivalent heterogeneous medium: nonlinear local response is modeled in the framework of Thermodynamics with Internal Variables; spatial variability of the local response is represented by correlated random vector fields generated with the Spectral Representation Method. Macroscale response is recovered through standard ho-mogenization procedure from Micromechanics and shows salient features of the uniaxial cyclic response of concrete that are not explicitly modeled at mesoscale.

cond-mat.mtrl-sci↗

arXiv2015

On damping created by heterogeneous yielding in the numerical analysis of nonlinear reinforced concrete frame elements

In the dynamic analysis of structural engineering systems, it is common practice to introduce damping models to reproduce experimentally observed features. These models, for instance Rayleigh damping, account for the damping sources in the system altogether and often lack physical basis. We report on an alternative path for reproducing damping coming from material nonlinear response through the consideration of the heterogeneous character of material mechanical properties. The parameterization of that heterogeneity is performed through a stochastic model. It is shown that such a variability creates the patterns in the concrete cyclic response that are classically regarded as source of damping.

cs.CE↗

arXiv2013

Initial versus tangent stiffness-based Rayleigh damping in inelastic time history seismic analyses

In the inelastic time history analyses of structures in seismic motion, part of the seismic energy that is imparted to the structure is absorbed by the inelastic structural model, and Rayleigh damping is commonly used in practice as an additional energy dissipation source. It has been acknowledged that Rayleigh damping models lack physical consistency and that, in turn, it must be carefully used to avoid encountering unintended consequences as the appearance of artificial damping. There are concerns raised by the mass proportional part of Rayleigh damping, but they are not considered in this paper. As far as the stiffness proportional part of Rayleigh damping is concerned, either the initial structural stiffness or the updated tangent stiffness can be used. The objective of this paper is to provide a comprehensive comparison of these two types of Rayleigh damping models so that a practitioner (i) can objectively choose the type of Rayleigh damping model that best fits her/his needs and (ii) is provided with useful analytical tools to design Rayleigh damping model with good control on the damping ratios throughout inelastic analysis. To that end, a review of the literature dedicated to Rayleigh damping within these last two decades is first presented; then, practical tools to control the modal damping ratios throughout the time history analysis are developed; a simple example is finally used to illustrate the differences resulting from the use of either initial or tangent stiffness-based Rayleigh damping model.

physics.class-ph↗