Matched Filtering-Based Sensor Fault Classification for Structural Health Monitoring Systems

Structural health monitoring (SHM) systems rely on sensor networks to assess the condition of engineering structures designed for long-term operation. Over time, structures such as wind turbine towers or bridges may experience degradation-related damage, while the monitoring systems themselves age and become less reliable. This aging can lead to sensor faults that generate plausible yet inaccurate data, potentially causing misinterpretations of structural integrity and even catastrophic failures. Therefore, distinguishing between sensor faults and actual structural damage is crucial for ensuring the long-term reliability of SHM systems. A key challenge is the diversity of sensor fault types, which must first be classified to enable effective compensation. This study introduces a matched filtering approach to address this issue. Sensor fault types, such as bias and drift within a sensor network, are analyzed using matched filtering to classify them effectively. The proposed classification method is validated on a real-world support structure with synthetic measurement data, a 9-meter-high lattice mast exposed to real environmental conditions. The results demonstrate that sensor faults can be accurately classified. A key advantage of this approach is that precise classification enables more accurate sensor fault compensation. Enhancing the robustness of SHM systems will significantly improve the reliability of data-driven structural assessments, a crucial aspect for ensuring the longevity of critical infrastructure.