Structural Health Monitoring of Port Rail-Mounted Gantry Cranes via Wavelet and Machine Learning-Based Algorithms

Port Rail-Mounted Gantry Cranes are a significant investment for ports and have multiple failure modes. These cranes are complex mechanical and civil systems that run remotely without maintainers or operators present. Determining efficient methods for early fault detection is critical for optimizing that investment. We propose a solution of using commodity IoT acceleration sensors and converting the signals into the Wavelet Domain and making 3D dynamic wavelet fingerprints. This allows neural network methods, specifically autoencoder anomaly detection algorithms, to highlight anomalies within the acceleration signals. Our method is successful, utilizing a single sensor, in a noisy system of systems industrial environment, where standard signal processing and anomaly detection methods are insufficient due to the many confounding signals.