Interpretable machine learning models for displacement demand prediction in reinforced concrete buildings under pulse-like earthquakes - research article

Aug 2, 2026

This study proposes an interpretable machine-learning workflow for estimating seismic displacement demand in existing reinforced concrete buildings subjected to pulse-like earthquakes. The method is built on two scales: a large-scale nonparametric model trained with Gaussian Process Regression using all candidate building attributes and intensity measures, followed by interpretation via SHAP to select a compact set of intensity measures. A symbolic reduced-scale model is then developed using Genetic Programming, with archetype simplified systems for training efficiency and refined building models for unbiased final evaluation.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages24

Reading Time1 h 0 min

AccessFree to download

Introduction

Importance of seismic demand estimation and PBEE

Role of intensity measures and engineering demand parameters

Limits of scalar intensity measures

Journal of Building Engineering 95 (2024) 110124 | Full length article
Interpretable machine learning models for displacement demand prediction in reinforced concrete buildings under pulse-like earthquakes
Giulia Angelucci a, Giuseppe Quarantab,∗, Fabrizio Mollaioli a, Sashi K. Kunnath ca Department of Structural and Geotechnical Engineering, Sapienza University of Rome, Via Gramsci 53, 00197 Rome, Italy b Department of Structural and Geotechnical Engineering, Sapienza University of Rome, Via Eudossiana 18, 00184 Rome, Italy c Department of Civil & Environmental Engineering, University of California Davis, Davis 95616, CA, USA | | | | --- | --- | --- | | A R T I C L E I N F O | | A B S T R A C T | | Keywords:
Engineering demand parameter Gaussian process regression Genetic programming Machine learning
Pulse-like earthquake Reinforced concrete | | This work proposes a novel procedure to guide the development of machine learning models for estimating the seismic demand in existing reinforced concrete (RC) buildings. The proposed approach is organized across two scales. A large-scale (nonparametric) machine learning model is first obtained by means of Gaussian Process Regression (GPR) using all candidate building attributes and intensity measures. SHapley Additive exPlanations (SHAP) values are utilized to facilitate its interpretation and to assist the rational selection of a small subset of intensity measures, which is finally employed to develop a (symbolic) reduced-scale machine learning model by means of Genetic Programming (GP). Simplified models of archetype buildings are adopted to develop machine learning techniques at both scales, in such a way to alleviate the simulation time for preparing large datasets. Refined models representative of actual buildings are instead considered for the unbiased final assessment.
The proposed approach is applied to develop predictive machine learning models for the maximum inter-storey drift in bare frames, pilotis frames and frames with infills under pulse-like seismic ground motions. Consequently, the critical examination of the SHAP values revealed the most significant intensity measures and unfolded interesting patterns depending on the occupancy rate of the infills. Moreover, the final assessment demonstrates that this approach allows the management of a non-homogeneous building stock consisting of very diverse structural systems (i.e., spanning from existing buildings designed against gravity loads only to buildings that comply with outdated seismic codes) while providing satisfactory predictions of the seismic demand with minimum computational effort. | 1. Introduction The estimation of the seismic demand in structures subject to severe earthquakes is of utmost importance because of its pivotal role in seismic risk assessment and mitigation. Performance-Based Earthquake Engineering (PBEE) provides a valuable methodological framework in this regard: it comprises several steps, of which estimating the structural fragility is especially critical. The structural fragility is the probability of exceeding a predefined Limit State (LS), expressed as Engineering Demand Parameter (EDP), conditioned to different values of a designated ground motion Intensity Measure (IM). In order to quantify the seismic risk, this conditional probability is then convolved with the results of a probabilistic seismic hazard analysis [e.g., 1,2]. Since the ∗ Corresponding author. E-mail address: [giuseppe.quaranta@uniroma1.it](giuseppe.quaranta@uniroma1.it) (G. Quaranta). [https://doi.org/10.1016/j.jobe.2024.110124](https://doi.org/10.1016/j.jobe.2024.110124) Received 2 April 2024; Received in revised form 30 May 2024; Accepted 2 July 2024 Available online 10 July 2024 2352-7102/© 2024 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY license ([http://creativecommons.org/licenses/by/4.0/](http://creativecommons.org/licenses/by/4.0/)).

FAQs

What is the paper’s main objective in displacement demand prediction?

To develop interpretable machine-learning models that estimate seismic displacement demand in existing reinforced concrete buildings under pulse-like earthquakes, supporting rational model development and reliable predictions.

How does the workflow use SHAP in the modeling process?

SHAP values are computed to interpret the large-scale Gaussian Process Regression model and to identify the most influential intensity measures, enabling the selection of a small subset for the reduced-scale model.

Why are two modeling scales and simplified archetype structures used?

A two-scale strategy combines broad candidate learning with a smaller symbolic model, while archetype buildings reduce simulation time for generating large training datasets; refined models are then used for unbiased final assessment.

Aug 2, 2026

Upload by Theodore