The importance of predictive maintenance for bridges and roads continues to grow, necessitating efficient and continuous management strategies that can detect early signs of deterioration through systematic inspections. However, the retirement of experienced inspectors, difficulties in training new personnel, and budget constraints leading to outsourced inspections by less experienced subcontractors have compromised the consistency and quality of infrastructure assessments.
To address these challenges, our research employs a multimodal approach to data collection and analysis using a specially equipped inspection vehicle. This vehicle integrates high-resolution optical and infrared cameras, vibration sensors, and acoustic sensors to capture a comprehensive dataset of road and bridge conditions.
Our study investigates how multimodal data analysis can enhance the detection of both surface anomalies and subsurface degradations. We employ system invariant analysis technology to monitor anomalies by analyzing changes in the correlations between different types of time series data. This approach enables us to identify subtle infrastructure changes that may not be apparent through traditional inspection methods.
Furthermore, we explore how this multimodal approach contributes to the analysis of underlying factors causing these issues, potentially leading to more accurate and proactive maintenance strategies.
This presentation will report our findings, discussing the synergistic benefits of multimodal data integration as well as the implications of individual sensor data analysis. We will also address how this approach can improve inspection efficiency and enhance the capabilities of inexperienced inspectors by standardizing the quality of inspections.