Last year I looked at the growing trend for digitalized factories that can be controlled by mobile devices. One of the main use cases for such technology is to spot problems before they get so big as to become very costly to fix.
It’s a challenge that researchers at the EU ProaSense project are grappling with. The project, which features research partners from six countries aims to test the analysis of data from the Slovenian company HELLA and MHWirth to develop more proactive management.
Modern machinery generates a huge amount of data, with sensors capable of monitoring the status of those machines on the production line.
“Operators want status reports and advance notification”, the team say. “They’re concerned about why the machine in question has stopped, and what they have to do to get it running again. It is possible from the massive flow of data to identify patterns of interest that can tell us something about how a machine’s status will change in the future.”
The data will hopefully provide some detailed insights into the functioning of each piece of machinery, and indeed the plant as a whole. Which machines are prone to breaking, for instance, or the general trend for machinery up-time.
Understanding Complex Manufacturing
“Our aim is to obtain a better overview of machine status, and adopt a precautionary approach. The sensors pick up trends and tendencies”, the team say.
The challenge is particular strong at HELLA, who make headlamps for cars. As these lamps have become larger and more complex in recent years, it has placed a strain on the manufacturing process.
Being able to gain greater insight into how manufacturing is going therefore, and why things break could be invaluable to them. The authors liken it to having a weather forecast for their machinery.
“You can compare it to the forecasts provided by the Met Office”, they say. “We’re also using sensors to help make predictions. Meteorologists refer to a multitude of variables in order to come to the single forecast they deliver to their clients, such as — ‘tomorrow it will be 13 degrees and overcast’.”
The initial results are mixed. The sensors attached to the machines have undoubtedly provided a wealth of data, but not all of that data has proved useful in understanding the machinery.
Despite these mixed results, the team are confident that as the approach evolves it will provide ever better levels of insight, with an eventual goal of reducing downtime by 50%.
It will be initially tested at a wider range of small and medium-sized firms in Norway before hopefully then expanding it to a wider range of scenarios.
It will certainly be an interesting project to keep tabs on.