The pharmaceutical industry is warming to the concept of smart manufacturing — and predictive maintenance in particular.
The reason for that trend is partly financial. A single hour of downtime can potentially cost millions of dollars in lost productivity, and predictive maintenance promises to make maintenance more efficient and effective.
One traditional barrier to predictive maintenance adoption has been the industry’s resistance to decentralized computing. Pharmaceutical manufacturers, like many other industrial sectors, initially resisted cloud computing. “Several years ago, there were some headwinds in terms of making a connection that was going out of the building to the cloud,” said Bob Neagle, business unit manager at Danaher’s Videojet Technologies.
But now, manufacturing professionals – even in highly regulated industries – are “more willing to embrace the cloud because they understand the benefits they can get,” Neagle said.
A potential gateway to cloud computing — and predictive maintenance — is pharmaceutical companies’ traceability initiatives, Neagle explained. Printers in manufacturing facilities “end up talking to the cloud because the individual serial numbers [they produce] come down from the cloud,” Neagle said. And after serial numbers have been printed onto a pharmaceutical product, many printers can back up that information to the cloud.
Neagle said many pharmaceutical companies have had cloud-enabled traceability initiatives in place for the better part of a decade.
But now, pharmaceutical companies have become interested in more sophisticated functionality such as predictive maintenance.
One of its main benefits is using information from sensors integrated into manufacturing systems to schedule maintenance rather than waiting for the system to break.
But predictive maintenance can reduce traditional scheduled maintenance as well. “Manufacturers don’t want their maintenance people having to run around and touch things that don’t need to be touched,” Neagle said. Machine learning systems that draw data from sensors within machines can advise when technicians should perform maintenance tasks. Predictive maintenance supplants scheduled maintenance.
“There are two advantages — you have increased device availability but also more efficient use of your maintenance teams,” Neagle said.
Yet predictive maintenance is not always easy to deploy. A 2019 report from Bain & Company concluded that such IoT-enabled functionality was often difficult to deploy.
Vendors have responded by building predictive maintenance functionality into products.
Videojet, for instance, has a service known as Maximize that uses an ink build-up sensor within its industrial 1880 continuous inkjet printer to inform operators when the system needs maintenance. If, for instance, the printer detects a potential problem related to ink viscosity, it can inform an operator that maintenance is required. The system doesn’t require cloud connectivity but supports it.
Videojet also has a cloud-based machine-learning-enabled offering known as Rapid Recover that helps operators respond if a printer fails. The company estimates that the system can speed recovery from a problem by 60%.
If the problem is relatively simple, the system could instruct an operator to follow the steps outlined in a video and be up and running in 20 minutes. In more complicated situations, the Rapid Recover service might tell an operator to switch to a spare printer while also notifying the company’s technicians of the problem. Once the spare printer is online, the system can restore prior configurations from the previous printer in use. “You don’t have to manually set it or use USB sticks to transfer settings,” Neagle said. “It’s just automated through the cloud.”
Videojet is currently testing more advanced predictive maintenance functionality to assess that a given printer is likely to fail in a specific time interval.
Pharmaceutical companies that wish to have a holistic approach to predictive maintenance spanning their facilities will likely need help from a large vendor or system integrator to help pull together data streams from across different device types in a factory.
The rewards of such a comprehensive strategy can be significant. “It goes beyond just predictive maintenance,” Neagle said. There’s a lot of interesting information that devices that are on the packaging and production lines can provide.”