Process monitoring technologies that gather data in real-time give drug firms greater control over product quality. They could also help identify problems with equipment more effectively than current approaches according to a new study.
The study—by researchers at the University of Tokyo in collaboration with Roche—sets out a “data-driven” method for detecting the deterioration of processing equipment on biopharmaceutical production lines.
The approach uses exiting process monitoring technologies, which is its major advantage, says lead author Philipp Zuercher, a PhD student from the department of chemical system engineering, at the University of Tokyo.
“The implementation of my approach enables a fast and cost-efficient introduction of data-driven monitoring for equipment reliability assessment,” he says. “Commercially available solutions are based on additional sensors which are difficult to install in an operating facility due to re-validation requirements and additional downtime in tight production schedules.
“Furthermore, the custom-made approach based on Python [a programming language widely used in industry] allows for continuous improvements without interfering with licensing issues that can be made in-house.”
Co-opting in-situ process sensors for equipment monitoring could help biopharmaceutical companies significantly reduce production downtime by accelerating maintenance programs, according to Zuercher.
“With a potential application of the approach in an operating sterile filling line, several days and up to weeks of unexpected downtime could be prevented especially through the localization step that allows for identification of the problem source,” he continues. “Tailored maintenance actions due to the knowledge of the location reduce maintenance time and therefore overall downtime further.”
The researchers used data from an established sterilization unit on an aseptic filling line at a Roche-owned drug manufacturing facility in Kaiseraugst, Switzerland, to develop and test the approach.
This minimal data requirement is another advantage, points out Zuercher.
“Our approach only required sensor data from technologies already in place for univariate process monitoring—such as pressure, temperature, spray time, etc,” he explains. “In contrast, typical measurements for predictive maintenance applications would also include vibrational or IR measurements.”
The reduced data requirement means most biopharmaceutical companies would have access to the process information needed to implement the approach with little need for additional equipment or infrastructure, says Zuercher.
“Minimal investment would be required due to the fact that the implementation of the method could be achieved in parallel to commercial manufacturing—with no additional downtime—and the fact that no additional sensors are required except for any blind spots.”