Indian Institute of Technology Mandi has partnered with INRIA in France to create advanced Artificial Intelligence (AI) and signal processing techniques for accurately predicting the structural health of bridges and other structures.
The finding of these studies have recently been published in the journals Mechanical Systems and Signal Processing and Neural Computing and Applications. These papers are co-authored by Dr Subhamoy Sen, Associate Professor, School of Civil and Environmental Engineering and his research scholars Dr Smriti Sharma, Mr Eshwar Kuncham, and Ms Neha Aswal from IIT Mandi along with Dr Laurent Mevel from INRIA, Rennes, France.
Bridges play a vital role in India’s infrastructure, with nearly 13500 of them across the country. These structures undergo natural ageing due to environmental factors like temperature fluctuations, and water and air exposure, compounded by heavy road traffic. Traditionally, assessing bridge conditions has relied on visual inspections, a method considered inadequate by experts. It falls short in detecting all structural issues and is subjective and time-consuming, involving manual analysis of numerous photographs.
Recent advances in instrumentation, data analysis, and artificial intelligence (AI) tools like Deep Learning (DL) hold great promise for structural health monitoring (SHM) of bridges and other structures. These technologies make it easier to detect, measure, understand, and even predict the evolution of defects over time. This, in turn, enables more effective planning of renovation or repair work, ultimately reducing maintenance costs and extending the lifespan and availability of bridges.
The research team at IIT Mandi has developed a Deep Learning (DL)-based SHM approach. Their AI algorithms can identify and isolate structural damages by analysing recorded ambient dynamic responses, all without the need for human intervention.
Elaborating on their work, Dr Subhamoy Sen, IIT Mandi, said, “We have employed data-driven methods like Machine Learning, AI, and Bayesian statistical inference to estimate a bridge’s health and predict its remaining usable life. This outcome has the potential to reduce risks to infrastructure under operational and adverse loading conditions.”
Temperature fluctuations can greatly affect a bridge’s dynamic traits, especially prestressed concrete, and cable-stayed bridges. It is therefore important to consider these temperature effects in both real-time and AI-based SHM. IIT Mandi’s algorithm was rigorously validated on a real bridge in a cold region with extreme annual and daily temperature swings.
To assess the algorithm’s damage detection capabilities, the IIT Mandi researchers initially tested it on an undamaged real bridge. Subsequently, they intentionally induced damage in the computer model to evaluate the algorithm’s accuracy in pinpointing the damage’s location. This testing confirmed the algorithm’s efficacy in identifying structural damage.
In another related study, the researchers used advanced filtering techniques to estimate the condition of different structural components without the need for direct measurement of their connections. This technique enables the separate assessment of each component’s health, which can in turn help in evaluating overall structural integrity. The method was validated using computer simulations of a beam exposed to various forces and demonstrated robust performance in handling challenges like background noise and damage severity.
IIT Mandi researchers state that these AI-based algorithms have broad applications, not limited to bridges alone, and can be extended to structures such as ropeways, buildings, aerospace structures, transmission towers, and various infrastructure elements requiring periodic health assessments and protection measures.