In a groundbreaking development for the construction and energy sectors, researchers have combined ultrasonic tomography with deep learning to revolutionize the way we inspect concrete structures. This innovative approach, detailed in a recent study published in the journal *Developments in the Built Environment* (translated from Dutch as “Advances in the Built Environment”), promises to make non-destructive testing (NDT) faster, more accurate, and more scalable than ever before.
At the heart of this research is Inad Alqurashi, a researcher from the Department of Civil, Environmental, and Construction Engineering at the University of Central Florida. Alqurashi and his team have developed a method that uses a nanoscale object detection model to automatically identify and quantify internal defects and embedded components in concrete, such as reinforcement bars and ducts. This is a significant leap forward from traditional methods, which often rely on manual interpretation and can be time-consuming and subjective.
The process begins with controlled concrete samples containing artificial defects of varying shapes and depths, along with embedded rebars and ducts. Ultrasonic signals are collected using a MIRA A1040 tomograph and reconstructed into 3D volumes via Synthetic Aperture Focusing Technique (SAFT). These volumes are then converted into 2D slices and segmented using Chan-Vese segmentation and morphological post-processing. A partial histogram matching procedure unifies color scales across segmented slices, minimizing color-related biases before model training.
“Our goal was to create a robust, automated system that could reduce the need for manual interpretation and minimize subjective variability,” Alqurashi explains. The team achieved this by using segmentation-assisted labeling to provide robust ground truth annotations, resulting in 7220 labeled images. The trained AI model accurately detected delaminations, rebars, and ducts (both grouted and ungrouted), achieving a mean Average Precision ([email protected]) of 0.73 and an Average Intersection-over-Union (IoU) of 0.80.
The implications for the energy sector are substantial. Concrete structures are integral to energy infrastructure, from power plants to wind turbines. Ensuring the integrity of these structures is crucial for safety and efficiency. This new method could enable rapid assessment and monitoring, reducing downtime and maintenance costs.
Testing on real-world bridge data demonstrated the model’s generalization to unseen conditions, highlighting its potential for widespread application. “This integrated approach has the potential to provide an effective, scalable NDT/E solution for rapid assessment and monitoring of concrete infrastructure,” Alqurashi says.
The key innovations include automated segmentation-based labeling, robust color standardization via histogram matching, and a lightweight deep learning model optimized for real-time deployment on resource-constrained devices. These advancements could pave the way for more efficient and accurate inspections across various industries, including energy, transportation, and construction.
As the world continues to invest in infrastructure, the need for advanced inspection techniques becomes ever more critical. This research not only addresses that need but also sets a new standard for the future of non-destructive evaluation. By combining the power of ultrasonic tomography with the precision of deep learning, Alqurashi and his team have opened up new possibilities for maintaining and monitoring the integrity of our built environment.
In the words of Alqurashi, “This is just the beginning. The potential for this technology to transform the way we inspect and maintain our infrastructure is immense.” As the energy sector continues to evolve, so too will the tools and techniques we use to ensure the safety and efficiency of our critical infrastructure. This research is a testament to the power of innovation and the potential for technology to drive progress in the built environment.