Creative Commons source reading · Machine learning for performance
Vibration-Based Machine Learning Model Training for Railway Bridge Health Monitoring †
Alaggio R; Asad M; Cirella R; Costantini S; De Gasperis G
Originally published in Sensors (Basel) · 2026-07-07
From the article
Selected source text. JournalOS removes markup and shortens excerpts where indicated. The full article provides the complete context.
Abstract
AbstractBridge health monitoring and machine learning are increasingly intertwined for civil engineers and artificial intelligence experts. Bridges’ poor health can result in severe outcomes if not addressed in time. Therefore, continuous monitoring is required to detect any anomaly or damage. Sensors, such as accelerometers, inclinometers, thermistors, etc., can help actively monitor these bridges. The signals from these sensors help record physiological activities. Such activities are helpful for anomaly detection, damage localization, and bridge health predictions with the help of machine learning algorithms. The proposed method extracts features from the dynamic response of a bridge to ambient excitation. It focuses on processing the signal received from different accelerometers installed on a steel railway bridge to determine the…
6. Results
ML models (SVM, MLP, RF, KNN, Ensemble (KNN, RF, and SVM), and DNN) were implemented using Pytorch 2.0 and Python 3.9 for programming purposes using the NVIDIA 940MX GPU. As a model optimization kit, a DL framework, the back-end for detecting anomalies and classifications, was used to prepare, train, and optimize the model. All experiments were performed on the Jupyter Notebook, version 7.5.5, platform using an NVIDIA GPU with 16 GB memory.
7. Conclusions
This paper has proposed a practical implementation of the machine learning method to classify various damage scenarios, in their location and intensity, of an in-service steel railway bridge. The framework is based on a deep neural network classifier, trained with features extracted from acceleration data produced by a physics-based reduced-order model of the bridge. In observation, it becomes evident that with smaller slicing, higher accuracy is achieved with both standard and augmented data. Further work has been carried out to introduce four class damage intensities within the standard range and train and test this model in various scenarios using a DNN model.
Editorial reading guide
- What data created and tested the model, and was evaluation independent?
- What is the cost of a false positive, false negative or out-of-distribution decision?
