As part of my role at Verne Global I head-up our business development work within manufacturing and as such I’m often looking at how the industries within the “global manufacturing family” are implementing the latest tech trends. Top of these is definitely machine learning and not a week goes by without me hearing about another design, production or testing process being given the “machine learning makeover”, even in highly complex manufacturing environments.
I’ve picked out a few examples in this blog that demonstrate how machine learning is being used to reach greater efficiency and reliability at every step of the manufacturing process. Firstly I’ve heard about Solido Design Automation, an electronic design automation (EDA) company working within the semiconductor industry, based in Saskatoon, Canada. They’re developing a suite of machine learning tools aimed at helping semiconductor manufacturers improve the speed and reliability of semiconductor design. Solidio, which has received both significant investment, uses machine learning to help semiconductor manufacturers reduce the resource-intensive cell, memory, and I/O characterisation processes, aspects of IC design that become more difficult as integrated circuits shrink in size.
As a keen driving enthusiast I’m especially interested in how the automotive industry - which I know well due to our ongoing and well publicised work with BMW and VW - is transforming technically. It appears machine learning is touching all aspects of the production life cycle. For example, Continental AG, a world leader in automotive components parts, has started to use machine learning to help design car tyres that provide better traction in inclement weather, while American tyre manufacturer Goodyear has proposed a very high-concept tire design called the “Eagle 360” that integrates – somewhat amazingly - machine learning into the tyre itself, allowing it to adapt to road conditions by making adjustments to an external membranes of sensors. Quite literally machine learning on the move!
Assisted by its technology partner IBM, Schaeffler, the manufacturer of bearings for the automotive and aerospace industry, has embarked upon a wholesale “digital transformation” to integrate machine learning into its production. In particular Schaeffler has begun to explore machine learning application in bearing maintenance, and how machine learning can help produce greater accuracy and reliability in the bearing manufacturing process.
Interestingly larger companies are forgoing the partnership route to build machine learning capacities of their own. For example, I saw that General Electric has made a very public pivot toward machine learning, hiring a large team of machine learning researchers to help develop its Predix software, a platform designed to help manufacturers monitor, record, and analyse each stage of their manufacturing process. In pursuit of its ambition to become a global top-ten software company, GE has been gaining machine learning expertise through acquisition too, like data integration platform Bit Stew and machine learning company Wise.io, both of which it purchased in 2016. So far, GE has attracted over 270 companies to employ its Predix software suite, including oil giant BP, which has outfitted 650 of its oil wells with sensors to feed data into the platform.
In Germany, Siemens has developed its own solution, which it calls Mindsphere. The Mindsphere platform is an open IoT operating system that offers product lifecycle management, digital twin capability, and a robust machine learning capability. In the case of Siemens, Mindsphere is just the latest incarnation of a long-standing commitment to machine learning, which stretches back over twenty years. According to Siemens, Mindsphere has already been able to reduce the emissions of their gas turbines an additional 10% to 15% beyond what their human engineers had previously thought possible.
Looking further afield, Australian software company 1Ansah has developed a natural language processing platform to help consolidate equipment manuals and maintenance logs into a single, intelligent source of knowledge. The company has had early success helping Airbus engineers maintain their helicopters. Because Airbus has equipped the latest generation of A350s with over 10,000 sensors in each wing to capture data, 1Ansah and machine learning companies like them are in a great position to provide on-going, synergistic support to Airbus, and possibly disrupt the field of aerospace maintenance in the process.
This is just a sampling of the many early use cases for machine learning in the field of manufacturing, and there are plenty of others. We’re still at the very early stages of this new data-driven era of industry. As the number of industrial robots continues to grow, to nearly 2.6 million in the next couple years, the desire for more intelligence that enables to them accomplish more complex tasks will increase along with it.
This creates a very exciting situation for companies in the field of industrial machine learning, and specialised, industrial-scale data center providers like ourselves at Verne Global that are committed to helping them innovate, develop and power their technologies.