CU Denver Engineers Develop AI Tool to Automate Finding Defects in Sewer Lines

With more than 1,500 miles of sanitary sewer lines in Denver, the job of ensuring these pipes are in a state of good repair is one that requires continuous analysis and review. Breaks can cause disruptions to daily lives by shutting down businesses (including the state capitol for a day in August 2024) and causing a literal stink.  

Enter Brian Zavareh, a graduate student in CU Denver’s College of Engineering, Design, and Computing. While interning with Denver’s wastewater IT department in 2019, he noticed that crews spent days and days reviewing film from the inside of pipelines across the city. Like most U.S. cities, Denver records sewer video, which people then review to detect defects. This is a time-consuming process, especially when you consider the size of the system and the fact that there are more than 250 different types of pipe failures—from small cracks to complete breaks—to look for.  

Zavareh quickly realized he could apply what he’d learned during his doctoral studies at CU Denver to help the city spend less time looking at videos and more time making repairs. If his idea worked, he hoped he might even be able to create a model that would predict the locations of future breaks, which would help the city get ahead of problems. “I thought I could help by automating some of the process and offered to develop an ML [machine learning] model to help the operators,” he said. “It worked.” 

Zavareh took his idea and early work to Farnoush Banaei-Kashani, one of his favorite CU Denver computer science and engineering professors and his mentor. Banaei-Kashini’s research is primarily focused on fundamental and applied data science and machine/deep learning, with a special interest in intelligent data-driven systems (systems that automate the process of decision-making based on real data).  

The professor listened to Zavareh and provided feedback and support. “Brian’s internship gave him the opportunity to identify a real-life problem, and he came up with an idea to solve it,” Banaei-Kashani said. “That’s what we engineers do.” Zavareh’s model could process five hours of film and identify defects with more than 85% accuracy in less than half an hour.  

Faculty Student Mentorship Key to Success  

When Zavareh first enrolled at CU Denver for his master’s degree, his goal was to become a software developer. Then, he took a class from Banaei-Kashani, and his whole life changed.  ”I knew right then, that’s what I wanted to do … to use data to address problems that are big, like a data scientist,” he said.  

Over time, Zavareh decided to also get his PhD and now works directly with Banaei-Kashani. “Your relationship with your faculty is instrumental to success,” Zavareh said. “He’s been nothing but enthusiastic and supportive of my ideas. That’s why we are here today.”  

Zavareh, who estimates he’ll finish his PhD in 2026, also has a full-time job as a data scientist for a global company that develops medical devices. And while his experience is unique, it is common for CU Denver students to find ways to apply their classroom experiences in jobs and careers.  

The Launch of a Company  

In addition to studies and teaching, the professor-student team has launched the start-up company InfraSmart Solutions through CU Denver’s Smart Futures Lab, which is home to an incubator and accelerator program. “Right now, we are at a base model,” Zavareh said. “I’m really excited about the potential once implemented. I can’t wait to see it grow.”   

They have begun talks with the city of Denver about a possible pilot program and with two other municipalities about potential projects. What they’ve learned through Smart Futures Lab is that it takes time to launch a business and to get a product to market.   

The duo continues to fine-tune their AI tool while they explore ways it can be enhanced to manage the logistics of things like the dispatch of work crews and managing multiple projects simultaneously. This approach can potentially be applied to many other types of infrastructure, including bridges and airports. “AI is an incredible technology that keeps evolving,” Banaei-Kashani said. “We’re working hard at CU Denver to innovate and apply what we are seeing in the lab to industry and business. We’re training our future scientists and engineers. It is one of the most exciting times to be in this field.”