Carleton co-op student uses fresh approach to glean data from legacy equipment at the NRC
- Boucherville, Quebec
Artificial intelligence (AI) is becoming part and parcel of new products designed for consumers as well as industrial processes and applications. But the power of AI to make predictions, recommendations, and draw conclusions requires well‑structured digital data – meaning that valuable information from analog or legacy systems, which can be expensive and disruptive to upgrade or rewire, is often left out in the cold.
"I took one look at the analog consoles on legacy equipment at the NRC's large‑scale plastic processing facility in Boucherville and thought, if my eye can see the numbers on the gauges, then a different kind of eye could too," says Fatima when describing how she got the idea for the project. She set up a raspberry pi camera in front of various consoles to capture images of the different types of data they indicate: temperature, pressure, rotation speed, torque, and amperage, among others. She converted those values into a digital dataset suitable for analysis in an Industry 4.0 environment using algorithms to extract the numerical values from the images with 96% accuracy.
Fatima worked closely with her supervisors at the NRC, Data Analytics Team Leader Stéphane Tremblay and Senior Data and Solutions Analyst Patrick Paul, to bring the accuracy of image processing to a human level, close to 100%. "The model became so precise that it is able to detect human errors in the picture labelling step," says Stéphane.
On the factory floor in Boucherville, researchers now understand exactly what is happening during the production process of plastic components such as fibre‑reinforced compounds for the automotive sector, or bio‑based polymers for packaging, and how different conditions generate variations in end products.
"Fatima's project was not only beneficial to the NRC's operations in Boucherville – the ability to digitalize analog data in a non‑intrusive way is very appealing to Canadian polymer producers. These SMEs employ about 90,000 people in Canada and welcome technologies that help them stay competitive in a global economy."
Solving real world problems
This hot new technology is not the only discovery that Fatima made during her co‑op term at the NRC. "My time at the NRC showed me how my studies at Carleton translate into solving problems in the real world. Seeing the impact and applications of my project in industrial and academic settings has been very motivating," says Fatima.
Fatima plans to complete her Bachelor's degree in Statistics in 2020, then pursue a career in data science and machine learning. She is also considering a Master's degree to expand her knowledge and contribute to research in the field.
"Fatima's experience and success illustrates exactly why we strive to provide co‑op, internship, and other experiential learning opportunities for our students at Carleton University. The opportunity for the student to solve real‑world problems in association with partners such as the NRC is incredibly valuable and beneficial for our students, our partner organizations, and for our university."
Exciting student work opportunities at the NRC
Each year, the Digital Technologies Research Centre welcomes about 40 students to work alongside leading experts in advanced analytics, computer vision, human‑computer interaction, natural language processing, bioinformatics, and artificial intelligence. Students bring fresh ideas and tremendous energy for project execution, helping NRC experts advance cutting‑edge research; while NRC experts provide coaching, advice, experience, mentorship, direction, and access to research infrastructure in industrial settings, helping students reach their scientific and professional goals more rapidly and in a supportive environment.
"Coaching promising students like Fatima and giving them opportunities to try out their ideas in new settings is something I really enjoy about my job, and one of the reasons why I love working at the NRC," says Patrick Paul, Fatima's supervisor.