Title: Structural condition assessment and prediction using machine learning-based EKF

Authors: Santosh Bisoyi; Amit Kumar Rathi

Addresses: Department of Civil and Infrastructure Engineering, Indian Institute of Technology Jodhpur, Rajasthan – 342030, India ' Department of Civil and Infrastructure Engineering, Indian Institute of Technology Jodhpur, Rajasthan – 342030, India

Abstract: This paper aims to improve the machine learning (ML)-based extended-Kalman filter (EKF) technique for condition assessment of structures. EKF is widely used for estimating structural properties, aiding in the detection and assessment of damage. However, it lacks accuracy in the prediction of aging and remaining service life. This study improves the prediction of structural conditions by incorporating ML algorithms, thus facilitating timely predictive maintenance and disaster prevention strategies. The proposed method employs EKF for the parameter estimation from vibration responses to train ML models for accurate prediction. The framework is validated based on multi-storey framed structures subjected to seismic excitation through numerical simulations and experimental tests. The results illustrate the accuracy and robustness of the proposed method up to 20% Gaussian and non-Gaussian noises. The ML model such as extreme gradient boosting achieves R2 score of 0.99 and RMSE of approx. 0.01 in predicting the stiffness parameters. The proposed approach demonstrates potential in real-time condition monitoring and predictive maintenance.

Keywords: extended-Kalman filter; EKF; machine learning; prediction modelling; structural condition assessment.

DOI: 10.1504/IJSTRUCTE.2025.150122

International Journal of Structural Engineering, 2025 Vol.15 No.4, pp.427 - 457

Received: 10 Aug 2024
Accepted: 11 Jul 2025
Published online: 01 Dec 2025 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article