Anyone searching the web for discussions on smart manufacturing and Industry 4.0 will find thousands of articles and countless blog posts on the topic. Many begin with an observation on how consumer tastes are evolving to where shoppers today demand products exactly the way they want.
When Sally Smith, for example, sets out to buy a new car, she doesn’t choose one off the dealer’s lot. She goes to the company’s website, and using the build-it feature, checks all the boxes to design the vehicle with exactly the options she wants. Then the assembly plant builds it just for her. This works because the whole production system is automated, and creating a one-off car is no trouble at all.
Most smart manufacturing discussions focus on the needs and operations of discrete industries — plants building things, which can be cars, smartphones or electric guitars. This is much different than pharmaceutical manufacturing, for obvious reasons. Mercifully, manufacturers don’t receive orders from individual consumers asking for tablets with the exact formulation and dosage wanted. Given these facts, does smart manufacturing even apply to the world of life sciences?
What Makes Manufacturing Smart?
All the elements of smart manufacturing do not apply the same way in every manufacturing environment, but there are some universal underlying concepts. It helps to think about what makes smart manufacturing smart, and how the concepts extend to life sciences.
Smart equipment: Taking the temperature of a product in a reactor can be done manually with a thermometer, or with an electronic instrument equipped with multiple redundant sensors and self-diagnostic capabilities. The latter, used in a smart manufacturing environment, can provide much more data about the product, about itself and about what’s happening inside the reactor.
Flexible production equipment: Getting to market quickly with new products combined with easily scalable production are central to smart
manufacturing. Products showing promise need fast ramp-up to accommodate demand, so the notion of building a production line to make only one type of product with a fixed production rate is fading fast. Whether building cars or active ingredients, new lines must be configurable and adaptable to a longer list of possible products with quick and easy changeover. Plant facilities are smaller so the equipment footprint needs to be minimized.
Plug-and-play connectivity: Any electronic devices used in a production unit — such as field instruments, valves, PLCs and so on — must be easy to connect and configure, or better yet, self-configuring. Designing more elaborate systems should be easy using a unified programming environment able to communicate with the individual devices. With this type of drag-and-drop programming, it’s easy to change the process software.
More sophisticated production equipment: Basic process instruments are giving way to more sophisticated devices. For example, determining the moisture content of granules tumbling in a dryer can be inferred by measuring the humidity of air coming out. But the same measurement can be done more quickly, directly and accurately by using a near-infrared spectrometer, which may also be able to measure and communicate other useful attributes at the same time.
Data gathering, analysis and distribution: The glue that binds smart manufacturing implementations together is data. While the realities of data wrangling are largely new to many kinds of manufacturers, life science companies are old pros. There is probably no other industry that has been creating and using big data as long. It is available in countless forms in a pharmaceutical manufacturing environment due to the range of regulations watching over every element, from cradle to grave.
The sheer volume of data generated has kept life science companies at the forefront of developing and deploying data retention and distribution methods. There are still technological backwaters to be found in some small areas where manual methods still exist, but these are increasingly rare.
Coordinated, not Isolated
Looking at these elements together in a group illustrates how different they are. For example, selecting instrumentation is not the same as designing a network, and developing them individually to achieve a common goal requires different approaches. The concepts necessary to make a CIP skid more flexible don’t look anything like those needed to reach to the cloud to support better data retrieval. Nonetheless, improvements done in any areas will deliver pockets of better performance to some extent. The real payoff comes with a more coordinated effort to connect the improvements.
Setting up the structure for data exchange is a basic element of smart manufacturing, because getting the right information to the right people at the right time is critical for success. Consequently, the mechanisms to support data flow connect all the elements, making manufacturing into a cohesive process rather than disconnected islands.
Making this work effectively depends on three strategies:
1. Effective process control and instrumentation
Understanding what is going on in a process depends on having the right kinds of instruments in the right places. A temperature sensor mounted in the wrong part of a reactor will not provide the desired information about what is happening in the process. It will create data, but nothing very useful. Processes must be evaluated to ensure the best technologies are being used and applied optimally.
Manufacturing in life sciences evolves slowly due to the conservative nature of the industry and regulatory considerations. Process Analytical Technology (PAT) practices have made improvements easier to implement, but once a process is working and validated, the steps involved with changes are complex. Improvements must promise a significant payoff to be implemented, but they do happen. The industry does evolve, and for the most part it is moving in smart manufacturing directions.
For example, newer production equipment tends to be more modular and built on wheels so it can be moved easily to where it is needed. Connections to join these modules are more standardized in size and position, so assembly of complex trains is easier. Sub-units, such as CIP and SIP skids, need to be easy-to-move and designed to work with a wide variety of process equipment.
Control and automation equipment, including instrumentation, valves and so on, is also more modular to avoid the need for special programming code and awkward connections, both physical and electrical.
One of the biggest developments is the growing use of single-use equipment and processes. Since this approach requires a combination of matching permanent and consumable elements, these are typically handled as pre-configured modular units sold by the equipment vendors. Often this includes predetermined instrumentation placement, and even selection following general GMP and typical process concepts.
In most cases, instrumentation for single-use equipment is adequate for basic needs, but these units tend to be under-instrumented for complex and difficult processes. If deeper process data is required, additions and modifications will likely be needed.
2. Getting data where it’s needed
Data in a pharmaceutical manufacturing facility takes many forms. Those responsible for actual production don’t necessarily pay attention to results from clinical trials or other phases prior to manufacturing, as they need to zero in on what’s happening in the plant, within individual reactors, tablet presses, etc.
Data must make its way from individual instruments and field devices through low-level PLCs to the DCS and process historians. This is a normal situation and happens in any number of process industries. The thing that makes life science environments different is the volume of data and its many forms. The level of detail necessary just to support traceability goes far beyond what most other conventional industries deal with. This kind of data retention and distribution requires sophisticated networking capabilities so all the people that need to see specific bits of information can access the necessary areas.
For example, a lab examines a given batch of product and finds an attribute has drifted out of spec. Lab personnel may then have to dig into the detailed information on the history of all the ingredients, along with all the information related to the specific batch and perhaps any intermediaries produced in the plant. This can involve an enormous amount of data, all of which must be easy to identify and retrieve.
Cloud storage certainly addresses the volume requirements since its capacity is effectively infinite, but it does not ensure easy retrieveability. Good tracking mechanisms must be set up to help guide users to the right data and to keep them from other areas where they have no reason to venture.
3. Understanding what data is saying
As has been said a thousand times: Data is not information. Data becomes information when the right people can apply the right tools to it, extract the secrets buried within, make decisions and take actions based on it. This is a point where many companies fall down. Well-designed networks supporting effective data collection can deliver pages of numbers, but individual users are left to their own devices to make sense of it.
An engineer trying to determine the optimum reaction temperature by comparing temperature curves from multiple batches against critical attributes can end up with a spreadsheet with tens of thousands of cells. With some effort, he or she can probably generate some useful graphs, but it will be tedious and time consuming. Data analytics software is available with capabilities to perform this kind of task far more easily, helping support decision making and pointing out where processes can be improved.
The ability to improve processes, even in a regulated and validated environment, is key to smart manufacturing. All the new features must work together to point at problems and inefficiencies waiting to be corrected, or at least mitigated.
Creating Return on a Smart Manufacturing Investment
It all sounds good, but what’s the financial payoff? How does this help a company improve production and reduce costs? The answers come in many forms and should not be difficult to see. Being able to draw on enough historical batch manufacturing data with sufficient detail to characterize critical product attributes can help optimize a process and increase yields. Equipment utilization can be improved when it’s easier to setup and requires less configuration.
The challenge is creating the systems necessary to make these kinds of changes possible. While the tools to create such systems have gotten better, the systems have grown more complex. Few companies within the world of life sciences have the inclination and internal bandwidth to implement complex networking solutions without engaging external help.
Many companies, such as a DCS vendor, can help with specific parts of the project. But trying to subdivide something this large and complex so individual sections can be given to different service providers ends up creating gaps and awkward hand-off points. Other service providers may have the know-how to create a comprehensive solution, but don’t understand the specific needs of life sciences.
A truly helpful service provider needs to bring four attributes to the table:
Thorough understanding of networking mechanics at all levels, from an individual field device to the c-suite
Experience within the world of life sciences, including all the intricacies of this heavily regulated business
Process manufacturing expertise to help unravel the complexities of plant-floor equipment and activities, and
Business sense to know how all these elements affect the bottom line in the complex world of life sciences.
When all these elements come together in close partnership between the client company and the service provider, smart manufacturing becomes an effective tool to deliver the desired ROI.
