Startup finds glitches in auto parts
Anomaly detection is lucrative field for downtown Kitchener company
KITCHENER — A local startup has built technology that quickly finds software problems long before vehicles are assembled, avoiding warranty claims, recalls and reduced profits.
Acerta, which employs 16 people and raised US$2 million last year, specializes in detecting anomalies among the reams of data generated by the sensors and software embedded in car parts. It is among Waterloo Region startups that are applying cutting-edge data science, artificial intelligence and cloud computing to traditional manufacturing.
“We can get very, very specific about what area of the car is experiencing the issue,” said Greta Cutulenco, Acerta’s co-founder and chief executive officer.
She cites the example of a company that makes transmissions. When a new transmission is tested at the factory, the manufacturer collects a lot of data from the sensors inside it. The data is uploaded to the cloud, where Acerta’s platform analyzes it in less than a second and sends the results back to the manufacturer.
“We take data from a variety of sensors within a vehicle and analyze it using machine-learning technology to both detect and predict failures,” said Cutulenco.
Another automotive company noticed that a rear-view camera kept shutting off. When the vehicle’s data was put through the Acerta platform, the problem was quickly identified. Every time someone changed gears, it set off a software glitch that shut down that camera.
After graduating with a degree in software engineering from the University of Waterloo in 2014, Cutulenco worked for Magna, one of North America’s biggest auto parts manufacturers. While working on camera-based navigation systems in the company’s research and development office in Brampton, she saw increasing numbers of recalls and warranty claims because of software in new car parts.
At the time, auto parts manufacturers had software developers go through code, line by line, looking for the errors that caused the parts to malfunction. As Cutulenco returned to UW to do a master’s degree, she was sure there had to be a way to use machine learning and analytics to quickly pinpoint problems.
“When I started doing my master’s at the University of Waterloo, my co-founders had been working around the idea of how to use large volumes of data produced by modern technologies to detect issues,” she said.
Sebastian Fischmeister, a professor of electrical and computer engineering at UW and head of the group working on experimental vehicle technology called WatCar, and Jean-Christophe Petkovich, the firm’s
chief technology officer, are the other founders.
Cutulenco put her studies on hold when Acerta was founded 2 ½ years ago.
“We took a lot of the research that has been done in this area and decided to commercialize it,” said Cutulenco. “And we started Acerta and started targeting automotive companies to get them to use more automated analysis.”
Vehicle assembly and auto parts manufacturing remain vital to southern Ontario’s economy, even after numerous plant closings. There are five assembly plants supported by a network of more than 700 parts makers, 500 tool-and-die shops and mould makers, says the provincial agency Invest Ontario.
Acerta, which is based in the Velocity Garage in the Tannery building in downtown Kitchener, quickly gained traction. It went through the Techstars accelerator in Detroit, and Startup Autobahn, a tech accelerator in Stuttgart, Germany.
With each passing year, more computer code and sensors are deployed in vehicles. Cars can have more than 100 million lines of code. And sensor counts range from 300 to 5,000, depending on the vehicle make and model. They are found in camera systems, electric power steering, infotainment systems, transmissions, engines, batteries, powertrains, and hybrid power controllers that switch the vehicle from electric to internal combustion, and back again. All of this technology generates huge amounts of data.
“Automated analysis is definitely needed here to deal with all that volume,” said Cutulenco. “It all comes down to volume. It’s just so much data, so many things that happen in a car. You just can’t deal with that complexity as easily anymore.”
Among the companies using Acerta’s software and services are Chrysler, and ZF, a big German manufacturer of gearboxes and transmissions.
“Every single new system that they produce, they can upload the data to us, and then we run the analysis,” said Cutulenco.
“It is done in real time so they can get the analysis within a second.”
Auto parts makers need a lot of computing power to move the huge amounts of data into the cloud where Acerta’s platform analyzes it.
“So a lot of times we are actually bottlenecked by upload speed, rather than actual time to analyze the data,” said Cutulenco.