To understand why batch manufacturers increasingly embrace open control standards, consider how to make a cake. The recipe might say to mix together the ingredients and bake at 350°F for an hour. Recipes always include materials and processing parameters (add, mix, pour) but rarely specify what brand of mixer or oven to use. Their universality is what makes them so easy to share.No one would mistake baking a cake with synthesizing complex and highly purified pharmaceuticals. Making drugs requires pinpoint control, which is in turn highly dependent on the equipment processing them. Yet the cake analogy provides crucial insight into why pharmaceutical manufacturers are embracing the ISA S88 batch control standard: S88 separates the recipethe materials and how to process themfrom production equipment capability.By adhering to the standard, users can write recipes without knowing anything about how their manufacturing equipment actually implements the recipe. Equally important, process equipment and control system vendors can integrate factories without knowing which specific recipes they will run.S88 promises a quick, efficient and consistent way to run multiple recipes on a single batch line, or to run individual recipes on batch lines around the world. That, of course, is the theory. When it comes to any control technology, mastery of details determines how close anyone comes to achieving that goal. Last month, at the World Batch Forum annual meeting in Atlantic City, N.J., leading manufacturers shared insights and best practices regarding S88. This article discusses how Bristol-Myers Squibb and Pfizer are using open batch control to improve plant efficiencies.
BMS Pilot Plant Uses S88 for PATIt was easy for the WBF attendees to sympathize with Paul F. McKenzie, executive director of process R&D at Bristol-Myers Squibb Co.s New Brunswick, N.J., facility.McKenzies scaleup facilities house 60 different reactors of various sizes and capabilities. They are used primarily to develop, study and optimize process routes for large-scale manufacturing. To do that effectively, McKenzie would like to use the S88 batch control standard to simplify the transfer of processing instructions to bulk manufacturing plants. He would also like to use plug-and-play process analytical technologies (PAT) to provide real-time feedback on process behavior.Both goals, he told the audience at WBF, are works in progress.Not just developmentBoth goals are worth achieving because the changing pharmaceutical regulatory environment has created a huge opportunity for companies that understand and master the science behind their manufacturing operations.Until now, says McKenzie, Federal regulators have encouraged an industry manufacturing philosophy that could be summed up in two words: Dont change. More recently, however, FDA has begun to encourage continuous improvement and risk-based assessment.According to McKenzie, FDA sees two risk factors in pharmaceutical manufacturing. The first is a plants implementation of systems-based quality assurance and its regulatory track record. The second is its ability to fundamentally understand, monitor and control its manufacturing processes. FDA plans to reward companies that score well in these areas with less scrutiny, fewer inspections and reduced regulatory burden.By using his scaleup facility as a test bed, McKenzie hopes to show how S88 recipe-based batch execution and plug-and-play and can improve flexibility, compliance and process understanding from the laboratory through the factory floor.BMS has a roadmap to integrate relevant technologies into its operations. In the laboratory, it has embraced high-throughput process optimization, electronic notebook number crunching, and spectroscopic analysis. New Brunswick has put a full-court press on quality systems for materials, equipment, facilities and procedures. It bases its approach on International Conference on Harmonization (ICH) Q7A, which he calls "GMP in a reasonable format.Process control remains more problematic. BMS is already using Emersons PlantWeb to manage assets and S88 batch control and Fieldbus-based plug-and-play devices. It has begun to implement broader use of in-line PAT (which McKenzie redefines as process assurance technology). It plans to use the output to develop better processes and drive real-time process and predictive models. Neither of those applications are there yet, he concedes.Harmonization savesThe attraction of S88, as mentioned, is that it enables manufacturersat least theoreticallyto run different recipes on the same equipment or clone existing recipes for use in new plants. S88 also provides a consistent way of thinking about processing. The advantage of S88 is that it gives us a common vocabulary from the lab all the way up to manufacturing, McKenzie explains. The unit operations are really the sameheat, charge, distilland we use S88 to leverage that commonality.McKenzie is putting S88 to the test by attempting to harmonize the New Brunswick pilot plant with a BMS manufacturing plant in Ireland. Any application of S88 has to start with good equipment standards, says McKenzie. Its very difficult between multiple plants, but an absolute necessity.At this point, he says, both facilities use consistent equipment PIDs and more than 125 standardized, validated S88 phases (devices grouped to perform a common function). We have full electronic execution, McKenzie says, including recipe execution on Emersons DeltaV, materials management on SAP, a common data historian, and real-time quality review reports delivered to the desktop.We did a back-of-the-envelope calculation, he adds. We estimated that harmonization saved us 7.5 man-years, 1,300 documents, and 7,800 signatures simply by reusing documents.Mixed successMcKenzie reports mixed successes in harmonization of operator interfaces, control modules, phases, recipes, and data historian/reporting.BMS, for example, standardized its operator messages and prompts and reuses 90% of existing graphics on new projects. One of our biggest anti-S88 colleagues went to our plant in Ireland and within two hours he had found the coffee and was comfortable in front of its screens because standards make a big difference, McKenzie recalls. Yet BMS still needs to do a better job of avoiding customization and integrating portable equipment into the interface.By standardizing equipment layouts, BMS has been able to recycle roughly 80% of its existing control modules for projects. The challenge is to incorporate bus-based portable equipment into control architectures that are not designed for flexible facilities. Plug-and-play has been a big success, but it has been difficult to manage, he says.Probably two-thirds of the phases used in New Brunswick are common to the phases used in Ireland, McKenzie says. Yet their commonality is limited to activities and not code. This is because subtle differences in equipment and engineers introduce slight idiosyncrasies that force changes in code parameters. Harmonization often fails to strike the proper mean: Phases are often too complicated for a pilot plant or not complicated enough for Irelands automated facility.A common vocabulary for recipes forces better pre-batch planning, which for R&D scientists is often very difficult to accept, says McKenzie. Recipes have replaced 100-page logsheets and eliminated documentation errors.The biggest challenge has been the training curve on how to write a recipe; he adds. I'm not talking about pulling blocks together into a flowchart, but about the fundamental knowledge you need to write a recipe well and to translate what you do in the lab to that recipe setting.BMSs quality group already releases batches based on paperless DeltaV review, but they have to move from plant to plant to do it because the data historians are relegated to their own facilities and do not feed data directly to a centralized DeltaV system. As a result, McKenzie says, our QA people are very fit. We havent got to where I want to be, which is to have a supply chain view of all these operations.The data historian is also able to funnel continuous process data to researchers. The scientists love getting it at the desktop and are really starting to use the tools to look at it in a format that makes sense, McKenzie reports.Real-time batch data, on the other hand, is a long-awaited almost success. BMS is currently validating its system, but had significant issues with the event-file interface. It was not as easy as I would have hoped 18 months ago, says McKenzie.Plug-and-play puts Ph.D.s in the labPlug-and-play is really extrapolating the advantages of Fieldbus to our system, McKenzie says. Foundation Fieldbus, in particular, enables BMS to power and communicate to numerous instruments along one cable. This reduces installation costs. Because Fieldbus automatically senses devices, tags and configurations, it simplifies shifting instruments from one reactor to another.Thats a must, especially when using such high-end process analytical technologies as Fourier Transform infrared (FT-IR) or near infrared (NIR) spectroscopy systems. Such laboratory instruments could cost about $100,000 plus another $100,000 for validation. We have rows of reactors and cannot afford to put every reactor on life support, says McKenzie. We have to move them in when we need them and replace them with something else after weve utilized them.Unfortunately, todays plug-and-play reality still resembles the computer industrys early plug-and-pray days. Configuration remains difficult when moving from system to system and downloading on Fieldbus segments can affect the equipment on it. Nor is Fieldbus system architecture aligned for flexible facilities.This has not kept McKenzie's team from putting charge, extraction, filtration and other units on skids. If anyone's ever seen a batch run in an R&D environment, theres one poor operator and eight Ph.D.s looking over his shoulder, he relates. What weve been able to do by using portable equipment and measuring instruments is get the Ph.D.s back into the lab where they can watch the data in real time on their desktops.BMS already uses skid-mounted field instruments. Instead of using a flashlight and sightglass to locate a phase split during liquid extraction, a portable mass flowmeter measures density and closes a control valve when it detects the interface. Rather than grab slurry samples from a cake wash for lab analysis, the in-line conductivity detector automatically stops the wash when it reaches its target.Analytical instruments are further behind, though McKenzie credits vendors with making them more rugged for at-line service. Analytical methods such as gas chromatography, Raman spectroscopy, Lasentec particle analysis, FTIR, and NIR are all moving to the plant floor. They enable operators to accomplish such complex tasks as monitoring a polymorph transition during crystallization.Yet unlike field instruments, PAT instruments have traditionally operated as stand-alone entities. They are not easily incorporated into industrial control and data acquisition systems.Still, as McKenzie notes, the heat is on to generate process knowledge that pharmaceutical manufacturers can share with regulators and use to drive continuous process improvement. Pharmaceutical manufacturers and their vendors must build the links that make that possible, the connections between processing equipment, control systems, data historian, and analytical and field instruments. But the carrotreduced regulatory burdencertainly makes the task worthwhile.
Standards and Code AutomationDemand for Listerine mouthwash was soaring, and the two high-volume continuous production lines at Pfizer Global Manufacturing's Lititz, Pa. plant could scarcely keep up. The facility was already using its batch system to make a single product. It needed a new, more flexible batch line for smaller runs of new and specialty products.The company decided to go with an S88-compliant batch line because it promises outstanding flexibility, Pfizer automation team leader Rick Brill told a WBF audience. He chose to work with system integrator EZSoft, Inc. (Downington, Pa.), because its bid came in nearly 30% lower than its competitors."I called them to make sure they understood the specifications," says Brill. EZSoft did. It simply believed its proprietary ControlBuilder software would save time by automating the design, programming, simulation and documentation of the Allen-Bradley PLC-based control system.Model by designControlBuilder's object-oriented design model is organized along the same lines as S88. It starts with devices, such as valves, pumps and sensors. Each device, or object, is associated with specific properties, methods and events. Because it starts with a library of objects, EZSoft is able to pull projects together quickly.A group of devices that perform a specific function is called a phase. Phases are one of the most critical parts of any S88 system. When properly defined, they are easy to rearrange into new recipes and take advantage of S88s ability to manage recipes on the fly.Its always a struggle to find the appropriate phase size, says Brill. Larger phases tend to specify a particular application or process step, so they're not really generic. You have to create new phases every time you want a new recipe. The key to creating reusable phases is to break them into small components. If you go with smaller phases, you can simply rearrange them and you're off and running.Once the phases are defined, they are combined into such unit operations as charge and batch tanks. Unit operations are combined into process cells, such as manufacturing and storage areas. Everything we do, right down to visualization, follows this model, says EZSoft president Gregory Skovira.Manual still makes senseIn addition to automating batch production, Pfizer wanted the new system to produce two formulations at a time while reducing labor.Simultaneously batching two recipes starts with a flexible layout. Pfizer's two charge tanks can connect with any of its five batch tanks. Each batch tank, in turn, can feed into either of two storage tanks.Transfer panels determine the systems configuration. To redirect the flow, the operator disconnects the triclamp fittings and manually reroutes the piping. Some facilities use a matrix valve arrangement, but we had cleaning issues with the valves, says Brill. A little bit of manual labor outweighed the amount of time wed have to spend cleaning. Like most automated batch lines, manual operations still make sense in some areas. GE Fanucs iWorkInstruction manages the manual work instructions, which it stores as files. They pop up when the recipe calls for a phase linked to a work instruction.EZSoft satisfied our systems flexibility requirements by letting us modify the order of addition and grouping of manually added ingredients, Brill says. We simply make changes in the recipe file parameters rather than create new work instructions.Pfizer also found a way to eliminate one of the two operators needed to run the system. GMP requires one operator to handle the charge and a second to check everything, Brill explains. Now weve linked the batch control system to the weigh/dispense system. We weigh and dispense each batchs manually-added ingredients automatically, then barcode it.When the operator receives a batch kit, he scans the barcode, Brill adds. The system then compares the code with what its expecting. This ensures we have the right material at the right time and eliminates the need for a manual double-check.On time, on budgetFactory acceptance testing (FAT) took only about one month, thanks to careful preparation. ControlBuilder spit out 1,000 pages of FAT documentation, says Brill. We had our quality people approve the protocols and final reports so we didnt have to redo the functional simulation tests as part of the validation.All units were approved prior to site acceptance testing (SAT), which took about two months. SAT ensured proper control system installation and communications with existing plant systems. All we had to do then was validate the systems GMP batch and recipe execution functionalities, explains Brill.The project came in on time and on budget, even though Pfizer made last-minute recipe changes during the final changes of the project.Brill is confident of the new system. As long as there are no equipment changes, we could create a new recipe in a matter of one day, he says. Its the documentation, change control and process validation requirements that add a whole new layer to the schedule.