THE OLD MANUAL PROCESS CONTROL SYSTEM
In 1990 the Process Science group at this company had a typical fermentation lab, with an eclectic collection of instrumentation.
Scientists collected data and managed recipe development for new drugs in the tried-and-true, old-fashioned way. To monitor
the development process, they walked around the lab recording data on a clipboard from each instrument, and then manually
entered all the data into a spreadsheet. Based on their analysis of this aggregated data, they returned to each controller
to adjust the recipe process by hand, in order to optimize the recipe for maximum effectiveness and best yield.
Figure 2. Fermentation lab
A few years ago, the company acquired the pharmaceutical assets of another company. With this acquisition they got a mixed
set of expensive instrumentation. Adding these instruments to the lab resulted in a wide variety of equipment of assorted
age and from various vendors. They had reactor vessels connected to either B. Braun Biostats or New Brunswick Scientific BioFlos.
They had probes, pumps, valves, meters, gas mixers, and scales. Just about every make and model was represented. Each vendor's
instrument had its own proprietary interface and few of the instruments talked to the others. Each biocontroller operated
in isolation. The data aggregation, integration, and analysis became even more cumbersome.
This scenario of manual methods for instrument monitoring, control, and data management is still the status quo in many R&D,
process development labs, and pilot plants today. They are "tried and true," and many businesses depend on established, "trusted"
procedures rather than risk introducing change. On the other hand, plenty of lab directors would welcome automation, as would
the scientists and technicians who do the daily tasks of managing cell culture or fermentation experiments. Often, they believe
that manual process control and data collection are the only alternatives available, given the equipment they have to use.
Most comprehensive automation systems are large-scale, expensive systems with required hardware (such as PLCs) and fully customized
software. These systems are just not practical in the process development lab or pilot plant, where automation rightfully
takes a back seat to experimentation, innovation and cost-savings.
This status quo of manual process control and data management is partly dictated by the instrumentation. Instrument manufacturers
focus on excellence in function, accuracy, and speed of their devices. They are typically less concerned with acting as part
of a larger whole, sharing and communicating with other instruments in "the big picture." The result is islands of data: isolated
measurements and analyses, and isolated subsystems like scattered puzzle pieces. Instrument manufacturers have specific expertise,
and concentrate more on perfecting their own island, and less on building bridges. A particular instrument could be a piece
in many different puzzles, serving a variety of larger processes. It is a challenge to be a team player in those many different
DRIVERS FOR A CHANGE
The Lab Director of Process Science at the company recognized that the costs of managing the isolated islands of data and
trying to bridge the gaps between them were hurting his biopharmaceutical business. He foresaw opportunity costs growing steeply,
because his products and processes were becoming increasingly complex, patent competition more intense, and time-to-market
more critical. For example, Table 1 shows three examples of drugs and their average daily revenue. You can see a large lost
opportunity cost for every additional day that was spent in its development.
Table 1. Opportunity Costs
The Process Science Lab Director was not going to replace or upgrade all the instrumentation in order to achieve a unified
system. After all, that would also be a waste of good equipment. His strategy was to bring old devices, new devices, and future
unknown devices under one umbrella.
The Process Science group had been taking small steps toward automation, contracting for custom software that did crude control
and data collection for individual reactors, and trying software applications supplied by the biocontroller vendors. The limitations
of these led to the decision to put a comprehensive, automated data management and process control software system into place.