Anomaly Detection in Structural Health Monitoring with Ensemble Learning and Reinforcement Learning

This research introduces a novel approach for improving the analysis of Structural Health Monitoring (SHM) data in civil engineering. SHM data, essential for assessing the integrity of infrastructures like bridges, often contains inaccuracies because of sensor errors, environmental factors, and transmission glitches. These inaccuracies can severely hinder identifying structural patterns, detecting damages, and evaluating overall conditions. Our method combines advanced techniques from machine learning, including dilated convolutional neural networks (CNNs), an enhanced differential equation (DE) model, and reinforcement learning (RL), to effectively identify and filter out these irregularities in SHM data. At the heart of our approach lies the use of CNNs, which extract key features from the SHM data. These features are then processed to classify the data accurately. We address the challenge of imbalanced datasets, common in SHM, through a RL-driven method that treats the training procedure as a sequence of choices, with the network learning to distinguish between less and more common data patterns. To further refine our method, we integrate a novel mutation operator within the DE framework. This operator identifies key clusters in the data, guiding the backpropagation process for more effective learning. Our approach was rigorously tested on a dataset from a large cable -stayed bridge in China, provided by the IPC-SHM community. The results of our experiments highlight the effectiveness of our approach, demonstrating an Accuracy of 0.8601 and an F -measure of 0.8540, outperforming other methods compared in our study. This underscores the potential of our method in enhancing the accuracy and reliability of SHM data analysis in civil infrastructure monitoring.