At the recent Africa Automation Fair, SA Instrumentation and Control had the opportunity to catch up with Benjamin Bruns, Beckhoff Automation’s business manager for process industries, who was out from Germany for the exhibition. Along with Kenneth McPherson, managing director of the company’s sub-Saharan Africa office, he brought us up to speed with the advances in artificial intelligence (AI) happening as part of Beckhoff’s ongoing investment in R&D.
“AI can mean many different things depending on the area in which you are working,” begins Bruns. “For us at Beckhoff, by AI we mean machine learning, and we have incorporated this as a suite of tools in the TwinCAT automation software.”
“This is a platform our customers can just use,” adds McPherson. “They need to understand their application and apply the tools in that context, but they do not need any specialised AI expertise to get started.”
The approach at Beckhoff is an integrated one. Machine learning is not something the company felt it should just add to its portfolio as a spinoff of the hype around Industry 4.0. Rather, it is a logical extension to its fast control technology, underpinned by the speed and deterministic capabilities of the EtherCAT fieldbus.
“In packaging applications, for instance, our control platform is the foundation on which we build a machine learning capability,” explains Bruns. “And, of course, since vision systems are an indispensable element of modern packaging lines, the new TwinCAT Vision capability is an important enhancement.”
“Exactly,” adds McPherson. “Our machine learning modules are not simply an add-on to the TwinCAT suite, they are an extension of the entire Beckhoff philosophy and are based on the deep understanding we have acquired over decades of solving complex control and positioning problems. With the digital technologies available today, we are now able to code that expertise in ways that allow us to share the benefits with our customers. It’s not something that happened because suddenly we were caught unawares by a change in market requirements – it is simply the next evolutionary step in the development of our offering.”
“Let me describe it using an example,” says Bruns, “it’s my favourite application at the moment.
“We have a customer in the business of filling household gas bottles with LPG. Each of the bottles need to be filled accurately, which sounds easy enough, but they have a problem because in order to refill the bottle with the correct amount of gas, they need to know the empty weight of that bottle – and there is about a ten percent variation in this across all the bottles in the system.
“In the early days, the bottles were all marked with their weight, but over time these markings have become scratched and damaged to the point where it is now very difficult for a machine to read the labels using a traditional vision system like a camera. So they approached us to see if the problem could be solved using machine learning techniques.”
After an analysis of the problem, the Beckhoff TwinCAT development team figured they could solve the problem by building an extensive image database and then ‘teaching’ the machines to recognise a specific bottle based on correlation. Bruns describes it as a machine learning path in TwinCAT, i.e. creating a master database and then using that information to optimise operational performance. Initially it required human operators to interact with the machine and correct any mistakes in identification, but gradually the need for human intervention was reduced as the system advanced from an accuracy of around thirty percent, to the current operating level of around ninety eight percent correct identifications.
“What made it possible was our capability to handle the enormous amounts of data, thanks to the speed of the EtherCAT fieldbus,” explains Bruns, “coupled of course with the intelligent algorithms in the TwinCAT software.”
“What’s even more impressive is how easy we have made it to add this type of machine learning capability to one of our existing control platforms,” adds McPherson. “The basic weighing and transfer of the bottles is all handled by Beckhoff controllers and I/O terminals interconnected via the EtherCAT bus system. All we needed to do was activate the intrinsic TwinCAT machine learning functionality.”
The implication is that very little hardware – other than cameras – had to be added to an existing operation to achieve a quantum leap in plant efficiency. This is the essence of Industry 4.0 and the digital technologies that underpin it, which Beckhoff has mastered as a logical extension of the PC-based control architecture it pioneered in the late 1980s.
For more information contact Michelle Murphy, Beckhoff Automation, +27 11 795 2898, michellem@beckhoff.com, www.beckhoff.co.za
| Tel: | +27 11 795 2898 |
| Email: | info@beckhoff.co.za |
| www: | www.beckhoff.com |
| Articles: | More information and articles about Beckhoff Automation |
© Technews Publishing (Pty) Ltd | All Rights Reserved
printer friendly version