April 05, 2019KITCHENER, ONT. — A Canadian company is using artificial intelligence to predict flaws in vehicles before they roll off the assembly line, a potentially game-changing technology that could cut costs in the supply chain, improve overall vehicle quality and increase industry competitiveness.“We apply AI to the data [that] manufacturers are able to collect during various stages of production and use it to predict earlier in the process whether the system is going to fail,” said 28-yearold Greta Cutulenco, co-found-er and CEO of Kitchener-based Acerta Analytics Solutions Inc.“And we can do this within the production process itself to eliminate scrap and rework earlier on, identifying that something might fail.
“But also we do that to predict how likely any of the units are to fail in the field once the system is shifted out of the plant. That saves on manufacturers’ warranties.
“What we’re doing is allowing [automakers] to find these issues much earlier in the process and then act much faster,” Cutulenco said.
‘WE MUST STAY AHEAD’
Companies such as Acerta that apply machine learning and data to traditional manufacturing will be critical cogs in the supply chain if Canada wants to remain globally competitive, said Colin Dhillon, chief technical officer for the Automotive Parts Manufacturers’ Association (APMA).
“You have to look at countries in the east, such as China and India, which are fast-growing and are potentially leaders in some of these general areas surrounding the factories of the future,” Dhillon said.
“So, we must stay ahead and promote individuals like Greta and companies like Acerta and build more and make it a more competitive playing field.”
Outside of Silicon Valley, Ontario has the largest tech hub in North America, according to a 2018 report by Ray Tanguay, the federal and provincial governments’ automotive adviser. Emerging technologies, such as AI, cybersecurity and robotics are key to “heralding the era of connected, autonomous and electric vehicles.”
But Canada’s auto sector must be aggressive in harnessing the automation needed in the vehicles of the future.
“If we are to do that,” Dhillon said, “we will need the backing of companies like Acerta and their technology.”
HUNDREDS OF SENSORS
Manufacturers are having more difficulty zeroing in on problems because of the complexity of vehicles, said Cutulenco, noting that one vehicle can contain hundreds of sensors that generate hundreds of thousands of rows of data.
Acerta’s website references case studies, including one of a leading automotive manufacturer that asked her company to analyze data collected during a road test of one of its vehicles. The client provided 250 megabytes of data recorded from 350 sensors during 80 hours of driving. Acerta devised an algorithm that scanned it and then pinpointed an anomaly in data generated by eight of the sensors.
Using the information, a combustion-engine expert identified the root cause of the problem within 60 minutes.
Cutulenco, who earned a bachelor’s degree in software engineering from the University of Waterloo, Ont., started Acerta with Jean-Christophe Petkovich, the company’s chief technical officer, and Sebastian Fischmeister, a University of Waterloo professor of electrical engineering and head of WatCAR, the university’s autonomous-driving research group.
In its first three years, Acerta has developed a growing list of customers, including General Motors, Fiat Chrysler Automobiles, Daimler AG, Nissan Motor Co. and Germany-based auto parts maker ZF Group.
The company recently started working with GM on the automaker’s self-driving-vehicle program. GM applied Acerta’s technology to traction control, allowing the car to react faster to changing road conditions such as snow and ice.
“For example, if the vehicle is starting to lose traction, we can warn the autonomous system so that it doesn’t fly out of control,” Cutulenco said.
“In general, even for vehicles today, we can apply that to monitor whether your transmission is going to break down or if your motor is going to be OK. So we can start to monitor quality and reliability of all the various systems in the car.”
PREDILECTION FOR PREDICTION
As automakers race to develop self-driving vehicles, they look to companies such as Acerta to learn how to use high volumes of data to predict software glitches in vehicles, Cutulenco said.
“Manufacturers are spending a lot of their AI resources on autonomous driving. Autonomous driving is still an unsolved, huge problem.”