Industry 4.0: bringing AI to manufacturing

In the second of a series of four articles on the matter of Industry 4.0, Mike James from ATS Global looks at developments around artificial intelligence

Guest
By Guest on 16 August 2017
Industry 4.0: bringing AI to manufacturing

This article first appeared in the Summer 2017 issue of The Record.

Current Industry 4.0 and smart manufacturing developments clearly point to the adoption of artificial intelligence (AI). And yet few AI developments have led to the creation of smart manufacturing applications. 

Head to Wikipedia and a basic definition of AI reveals five steps to achieving real intelligence – from simple reflex agents leading up to the more advanced type of AI: learning agents. I think most readers would agree that current systems are at the lowest level of AI and we have a long way to go to get to level five – fully contextualised decision making.

People have the wonderful ability to make decisions by taking account of a massive range of variables, which can be instinctive, learned and factual. A simple example: before going out we might check the weather forecast before deciding to take an umbrella with us. When we think this through, we not only consider factors such as how heavy the rain might be and how hard the wind might blow, but also how long we might be outside, and whether we trust the forecast or care if we get wet.

Only time will tell if computers can match the way we think. Google AI just won a GO games series against the world champion by replaying and learning from old games. Yet not one of Google’s stated aims for AI relates to industrial applications. Why? Is it just too hard?

On the shop floor, most processes are close to unique, or have few similarities worldwide (at the plant cell level). At the minute, this could mean that an investment in AI is hard to justify. But there are signs of AI being tested in shop floor environments. 

Gartner recently published a report about “Cool Vendors” and one of them is Bennit.AI. Their software follows the same principles as Google AI: it learns best practices by following the decisions made by operators. It runs in the background, unobtrusively, and correlates the operator actions to product quality and productivity. It’s not only cool, but I think it’s the way forward to capture valuable knowledge easily and effectively in times when we lose shop floor expertise when skilled operators retire or leave.   

 

Mike James is president of the Manufacturing Operation Management Institute

 

Number of views (1020)/Comments (-)

Comments are only visible to subscribers.