Short-term damage alarming with limited vibration data in bridge structures - A fully non-parametric machine learning technique

Aug 4, 2026

Upload by Aria

Machine learning-assisted vibration monitoring supports damage alarming for civil structures, but short-term programs face limited vibration data and strong environmental/operational variability, weakening state-of-the-art methods. This paper proposes a fully non-parametric unsupervised technique combining non-parametric data partitioning and density-based anomaly detection to generate damage-alarming indicators. Limited eigenfrequencies from full-scale bridge structures validate effectiveness and practicality for short-term monitoring schemes, enabling reliable alerts despite scarce measurements.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages16

Reading Time40 min

AccessFree to download

Introduction

Vibration-based structural health monitoring

Short-term vs long-term monitoring challenges

Proposed fully non-parametric technique

Non-parametric data partitioning

Density-based anomaly detection

Validation and results

Measurement 235 (2024) 114935 Contents lists available at ScienceDirect Measurement journal [homepage: www.elsevier.com/locate/measurement](homepage: www.elsevier.com/locate/measurement) | Short-term damage alarming with limited vibration data in bridge structures: A fully non-parametric machine learning technique | | | | --- | --- | --- | | Alireza Entezamia, *, Hassan Sarmadib, Bahareh Behkamala
a Department of Civil and Environmental Engineering, Politecnico di Milano, Milan, Italy b Head of Research and Development, IPESFP Company, Mashhad, Iran | | | | A R T I C L E I N F O
Keywords:
Structural health monitoring
Short-term measurement
Environmental and operational variability Unsupervised learning
Anomaly detection | A B S T R A C T | | | | Machine learning-assisted vibration monitoring is an intelligent, automated, and popular strategy for evaluating civil structures and damage alarming. However, implementing this strategy under a short-term monitoring program may encounter challenges such as limited vibration data, profound environmental and operational variations, and the limitations of state-of-the-art solutions under these conditions. The main purpose of this paper is to propose a novel machine learning technique in terms of unsupervised learning for damage alarming with limited vibration data. The crux of this technique lies in two fully non-parametric parts of data partitioning and anomaly detection. Initially, a non-parametric clustering approach with a novel procedure is presented to divide limited vibration data into clusters. Subsequently, a new density-based anomaly detector is developed to prepare indicators for damage alarming. Limited eigenfrequencies of full-scale bridge structures are used to validate the proposed solution. Results can substantiate its effectiveness and practicability in short-term monitoring programs. | | 1. Introduction Civil structures such as buildings, bridges, dams, etc. are critical and expensive assets of every society offering a range of services and benefits including shelter, transportation, and water and energy supply. These structures support the advancement and welfare of human societies by enabling economic development, social integration, environmental protection, and cultural diversity. However, natural and human-induced hazards always threaten safety, functionality, and sustainability of such structures causing unfavorable events and losses. Vibration-based structural health monitoring (SHM) is one of the practical and automated solutions for assessing civil structures. This methodology primarily involves continuous evaluation of a structure over time by measuring vibration and other influential parameters using different sensors [1–4], regular inspections of structural health and safety, and warning of the emergence of any adverse changes [5,6]. The implementation of a vibration-based SHM program needs several initial prerequisites including sensing systems for data acquisition, storage, and communication; engineering software for numerical modeling; feature extractors to derive useful information from measured vibration data; and computational methods for decision-making about the current status of the civil structure. Consequently, model-driven and data-driven techniques are key technical solutions employed in vibration-based SHM. Depending on the strategic importance of the civil structure under monitoring, the main purpose of SHM, the type of structure and its geographical location, and total budgets, an SHM program can be carried out in short-term and long-term schemes. Short-term monitoring is typically implemented for a limited period to achieve specific objectives such as measuring and analyzing structural responses over extreme and sudden events (e.g., earthquakes, hurricanes, floods, fires, etc.), validating sensing systems and their functionalities, ensuring construction protocols align with design specifications, and r

FAQs

Why is damage alarming difficult in short-term bridge monitoring?

Short-term programs often provide limited vibration data and experience large environmental and operational variations, which reduce reliability of existing methods.

What is the core idea of the proposed method?

The method uses unsupervised learning built from two fully non-parametric components: data partitioning and a density-based anomaly detector to produce damage-alarming indicators.

How is the technique validated?

Limited eigenfrequencies measured from full-scale bridge structures are used to validate the proposed solution’s effectiveness and practicality in short-term monitoring programs.

Aug 4, 2026

Upload by Aria