Ensuring that the materials used in vehicles are structurally sound is vital for public safety. Sina Negarandeh, a master’s student in computer science with a concentration in applied artificial intelligence, aims to support this with his AI model that significantly accelerates analysis of X-ray data on structural damage. This model is as accurate as standard techniques but takes a fraction of the time.
A federally funded, collaborative effort
Looking back on the history of engineering motivated Negarandeh to pursue this research. “I often think about how much of our daily lives are shaped by the engineers and scientists who came before us,” he says. “Being able to contribute to that collective effort, even in a small way, is what drives me.” He also had a keen interest in integrating AI into physical science.
This research was part of a larger project funded by the Natural Sciences and Engineering Research Council of Canada to reduce waste and lower carbon emissions. Negarandeh worked with the University of Waterloo, specifically its Fatigue and Stress Analysis Laboratory, and with several automotive industry partners. He credits Professor Vasiliki (Verena) Kantere as his primary supervisor for the research. Professor Hamid Jahed and PhD candidate Sepehr Ghazimorady from the University of Waterloo were also collaborators.
Training AI to quickly detect structural damage
Negarandeh used AI to analyze data on different materials to inspect structural integrity. First, he collected data by applying stress to an aluminum alloy. This simulated the general wear and tear on metals used in vehicles. He then viewed the damage with a non-destructive technique known as X-ray diffraction. This method looks inside materials without breaking them, detecting damage that can’t be seen from the outside.
Negarandeh’s goal was to analyze this data with a new, fast, efficient model. “I want to combine the precision of traditional modelling with the speed of modern AI networks,” he says. “This way, materials can be screened in real time rather than after hours of analysis.”
He developed his AI model using a “teacher–student” framework. It begins with the original, time-consuming model (known as the teacher), which then feeds information to a “student” AI model. As the student AI model learns from the teacher, it becomes highly efficient, producing precise and accurate results.
Secure, efficient, sustainable results
Negarandeh’s approach was successful. The student AI model learned from the teacher model without losing accuracy. When using the models to test aluminum samples, both presented the same results, but the student model did it more quickly. “The AI model reduced the computational time required to analyze a material scan from nearly two hours down to just a fraction of a second,” he says.
The model is also capable of following proper engineering guidelines and meeting safety standards.
This new AI model solves the issue of engineers having to choose between efficiency and accuracy. “Our framework shows you no longer have to make that choice,” Negarandeh says. “By having a slower yet accurate model act as a teacher for a time-efficient AI student, we get the accuracy of the old approach in a fraction of the time.”
This is especially important in the automotive and aerospace industries, where both time and accuracy are crucial to avoid catastrophic failures. It also offers environmental benefits. Since experts can detect internal damage early, it extends the usable life of vehicles and reduces the resources required to replace worn or broken parts.
The next goal for this research is to expand the framework into other research areas, such as biomedical diagnostics.
Negarandeh hopes to continue his career developing meaningful, impactful projects that improve society and the environment. “Engineering and AI are becoming increasingly interconnected, and I want to be part of the community that ensures this integration leads to real, positive change,” he says.
Transformative technology a uOttawa Engineering priority
Technology for the digital transformation of society is one of the five primary research areas at the uOttawa Faculty of Engineering. The University is dedicated to being part of the technological advancement of society.
Negarandeh earned first place in the Technology for the digital transformation of society category at the 2026 Engineering and Computer Science Graduate Poster Competition, held during Engineering Research Celebration Day. His poster, Neural Posterior Estimation via Hierarchical Bayesian Distillation, captivated judges with its innovative approach.
Discover other winning projects from the graduate poster competition.
