Musashi Seimitsu Discusses AI Manufacturing At GTC 2019

Mar 19, 2019
7Shares
Submit

NVIDIA: A robot arm glides to scoop up high-precision automotive gears from a glowing cube used for inspections as a nearby autonomous guided vehicle carts heavy parts across the manufacturing floor, yielding to an employee.

It’s a glimpse at the future at Japanese auto-parts giant Musashi Seimitsu, the world’s largest manufacturer of transmission gears. But it’s also a view of how the company plans to stay ahead of the country’s population decline.

Musashi Seimitsu discussed its next-generation autonomous robots at our GPU Technology Conference this week.

With Japan’s population forecast to decline by nearly a third in the next 40 years from a peak of 128 million in 2010, Musashi is betting that AI will be transformative for more than just a labor-shortage crunch.

Founded in 1938, the company is no stranger to change. It started by manufacturing carburetor parts for aircraft, changing course to sewing machine parts after War War II, moving next into motorcycle parts by the mid-1950s and jumping into the automobile industry for its anticipated growth.

To be sure, CEO Hiroshi Otsuka embodies the samurai spirit in his efforts at driving innovation in the manufacturing industry.

The grandson of Musashi’s founder, Otsuka has been a big proponent of technology for market advantages, particularly AI. Now, facing a difficult labor market, he’s developing an AI roadmap that extends well beyond high-precision auto parts manufacturing.

Otsuka’s Musashi is taking a role typically reserved for software startups but that the company’s leader says is the rightful provenance of a massive operation with vast data: It’s developing AI manufacturing with its sights on other industries. And it’s establishing its presence in the premier research centers of the world for AI.

“Artificial intelligence for manufacturing inspections and logistics running on the NVIDIA Jetson platform represents the next chapter for performance in manufacturing everywhere,” Otsuka said. “This is just the start of industrial transformation.”

On Monday, Musashi announced its NVIDIA Jetson-based AOI box, dubbed Neural Cube. Neural Cube, the brains of its parts inspection, just might be the start of something bigger.

Inspects Dimples, Spatters

Parts inspection has become more critical in recent years. That’s because it’s not just mechanical failure that’s being addressed. Slight imperfections in gears and other parts can make noise that is now more than ever audible in automotive cabins because of the proliferation of electrical vehicles.

Musashi’s AI inspection system consists of a robotic arm and its Neural Cube that sports an NVIDIA Jetson TX2 and a camera. It can inspect differential bevel gears by tapping into image classification from Inception-v2 models from Keras or TensorFlow libraries, scoring inference in 5 seconds to spot 1 millimeter dimples.

Neural Cube can also spot welding spatters that can occur from joining metal with a weld. Using MobileNets from Keras or TensorFlow on Jetson, it can see spatters as small as 0.5 millimeter within 7.5 seconds.

Its accuracy is on par with humans, according to Musashi.

AI Aids Labor

With a labor pool that is retiring, autonomous operations are growing more important for Japan.

Musashi’s AGV self-driving robot uses lidar and a stereo camera to autonomously detect people, equipment and floor lines. It taps into AI for path planning to navigate parts about the production line, which is connected with the Jetson AOI system, and can haul over 1 ton of weight.

The moves at Musashi will redistribute 20 percent of its workforce that focuses on inspections, freeing workers for higher level work while potentially boosting inspection effectiveness — and is just the start for a company turning to AI.

Musashi is working on its next-generation Neural Cube with Jetson AGX Xavier, building more complex neural networks to identify the smallest of imperfections.

It’s part of a bigger vision — think AI manufacturing as a service — to take what it has developed from its massive manufacturing datasets to bring to other industries, Otsuka said.

Submit