PSI - Issue 84

Available online at www.sciencedirect.com

ScienceDirect

Procedia Structural Integrity 84 (2026) 797–804

III Fabre Conference: Existing Bridges, Viaducts, and Tunnels: Research, Innovation, and Applications Novelty detection of a multi-span PSC bridge using control charts Fulvio Busatta a , Marco Pirrò a , Carmelo Gentile a, *, Lorenzo Lepori b , Paolo Mannella b

a Department of Architecture, Built environment and Construction engineering (ABC), Politecnico di Milano, Milan, Italy b ANAS S.p.A., Rome, Italy

© 2026 The Authors. Published by ELSEVIER B.V. This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0) Peer-review under responsibility of the scientific committee of the Conference Keywords: cointegration; control chart; monitoring; prestressed-concrete bridge; vibration-based novelty detection. Abstract Within Structural Health Monitoring frameworks based on Operational Modal Analysis, novelty analysis aims to detect structural anomalies by inspecting the time evolution of natural frequencies estimated from ambient excitation. Novelty detection algorithms generally require training periods and multivariate control charts are often used to effectively show anomalies that might occur over time. However, in real-world applications, variations of the natural frequencies occur not only because of structural changes, but also due to the environmental and operational variability (EOV). Hence, the removal of EOV effects should precede any novelty analysis. The paper presents and discusses the dynamic monitoring of a continuous 5-span prestressed-concrete bridge managed through a dedicated cloud platform employing an industrial Python-based version of the P3P software. Approximately, 1-year dynamic monitoring was performed, with eleven vibration modes of the bridge being automatically identified and tracked. These modal estimates were exported and used to conduct novelty detection through different techniques as follows: (a) control charts based on the Hotelling T 2 statistics and Mahalanobis distance, with the frequency residuals being evaluated through PCA regression, and (b) control chart based on Cointegration which makes use of a linear combination of non-stationary monitored features without requiring any EOV effect removal. Results show the capability of the control charts based on Mahalanobis distance and Cointegration technique of performing novelty analysis effectively using a short training period than the control chart based on the Hotelling T 2 statistics.

* Corresponding author. Tel.: +39-02-2399-4242; fax: +39-02-2399-2225. E-mail address: carmelo.gentile@polimi.it

2452-3216 © 2026 The Authors. Published by ELSEVIER B.V. This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0) Peer-review under responsibility of the scientific committee of the Conference 10.1016/j.prostr.2026.06.102