EngineeringIndustry 4.0Manufacturing

Researchers at Drexel University develop AI-enhanced system for automated infrastructure inspection

Researchers at Drexel University have developed an system that harnesses the power of AI and computer vision to address the task of ascertaining infrastructure integrity. This new approach, aimed at detecting fissures in concrete, paves the way for robots to undertake the scanning, modelling, and monitoring of ageing infrastructure, marking a significant leap forward in the field of engineering.

A recent spate of building collapses, evident by the Raac scandal in the UK, and the evident wear and tear on roads and bridges signal a dire need for more efficient inspection methods. Traditional techniques are struggling to differentiate between benign ageing effects and critical structural flaws, a challenge that the Drexel team’s system seeks to overcome.

By augmenting visual inspections with AI-driven machine learning algorithms, the initiative promises to revolutionise how engineers assess and manage structural health. Published in the journal Automation in Construction, the system integrates computer vision with deep learning algorithms to accurately identify problem areas. These are then meticulously mapped out using laser scanning to create a comprehensive digital model, facilitating closer monitoring and evaluation.

Arvin Ebrahimkhanlou, PhD, Assistant Professor, and Ali Ghadimzadeh Alamdari, Research Assistant, both from Drexel’s College of Engineering, emphasise the importance of early detection. They draw a parallel between identifying cracks in structures and diagnosing medical conditions early to prevent further complications. Their work underscores the pressing need for an efficient system to prioritise and address infrastructure repairs, especially in light of the American Society of Civil Engineers’ report highlighting a $786 billion backlog in road and bridge maintenance.

The project utilises a sophisticated stereo-depth camera system that feeds data into a convolutional neural network. This setup is adept at recognising patterns indicative of structural damage, thus enabling targeted inspections. Once a potential issue is flagged, robotic arms equipped with lasers perform detailed scans, generating three-dimensional images that are crucial for tracking damage progression and planning repairs.

Central to the Drexel University researchers’ methodology is the creation of digital twins for the inspected structures. This innovative aspect of their system involves the detailed mapping of detected fissures through laser scanning to generate precise digital replicas of the real-world infrastructure. These digital twins serve as dynamic models that can be analysed and monitored over time, offering a revolutionary approach to understanding and predicting structural behaviour under various conditions. By employing digital twins, the team not only enhances the accuracy of current inspections but also provides a forward-looking tool for simulating potential future issues, thereby enabling preemptive maintenance and repair strategies. This approach represents a significant step forward in the field of structural engineering, merging the latest in AI technology with practical applications to safeguard infrastructure integrity.

Preliminary tests in laboratory settings have demonstrated the system’s remarkable ability to detect and measure cracks with unparalleled precision, significantly outperforming existing technologies.

Despite the advanced capabilities of the AI and robotic systems, the Drexel researchers envision human inspectors playing a crucial role in the decision-making process for repairs. The primary aim of this technological advancement is to reduce the manual workload, enhance the accuracy of inspections, and minimise the potential for human error.

Looking to the future, the team is exploring ways to integrate this cutting-edge system into a larger, autonomous monitoring framework. This would likely include drones and unmanned vehicles, creating a comprehensive solution for maintaining and monitoring infrastructure. Successful real-world application and collaboration with industry partners are key to refining and expanding the use of this technology, promising a new era of efficiency and reliability in infrastructure management.

There’s plenty of other editorial on our sister site, Electronic Specifier! Or you can always join in the conversation by commenting below or visiting our LinkedIn page.