Artificial intelligence (AI) and neural networks are transforming how we monitor the structural health of bridges leading to a more proactive approach to infrastructure safety.
A study led by the Australian Catholic University (ACU) has successfully created and tested an AI-driven method to detect minor defects in bridges before they escalate into critical failures.
The research authored by Associate Professor Niusha Shafiabady and her team at the Computational Intelligence and Social Good Lab, focused on four Vietnamese bridges:
The Chumchup, Gocong, Ongdau, and Ongnhieu bridges. The AI system used neural networks and determined the vibration states or ‘loss factor’ from different bridge vibration states to show bridge structural health and integrity results.
“Loss factor is, if you want to think of it in our daily life? If you’ve ever jumped on a trampoline, some of the energy goes into making the trampoline stretch and move and some of that energy doesn’t come back to you,”
Shafiabady stated. She goes onto say:
“That is the lost energy in the bridges, the loss factor is actually that lost energy, so it measures the energy that doesn’t come back.”
The “loss factor” measures the energy lost during a structure’s vibration, often converted to heat or internal friction.
By tracking changes in the loss factor over time, the AI model can signal early signs of structural deterioration. This approach categorises energy signals into three types: structural responses, defect-related indicators, and noise interference.
The study explored three scenarios that have occurred on those Vietnamese bridges:
Heavy Vehicle Loads – Simulating trucks and other large vehicles exceeding standard weight limits.
Light Traffic – Monitoring small vehicles during low-traffic periods.
Heavy Traffic – Assessing mixed vehicle loads during peak congestion.
In each scenario, the AI system monitored vibration patterns, enabling researchers to detect subtle structural changes. Using layered neural networks, the system progressively refined its analysis, improving detection accuracy and allow preventative maintenance to take place.
“Applying these AI methods was primarily for pre-emptive maintenance,” she said. “It’s not necessarily something needed to have immediate attention to that bridge, but just to avoid issues like catastrophic problems that could happen if the maintenance teams didn’t look after the bridge.”
The neural networks operate collaboratively, with one network making initial assessments and another refining the decision. This layered approach ensures reliability and precision in defect detection.
The success of AI in partnership with Neural Networks allows for broader and improved applications in infrastructure monitoring. By integrating advanced machine learning algorithms, governments and engineers can build safer and resilient transport networks allowing for greater longevity of existing assets as well as a reduction in overall cost.
Comments