Intelligence-based Manufacturing

  • June 18, 2012
  • Feature
Moving to preventative & proactive manufacturing Exclusive – Part 6 from Pharmaceutical Automation Roundtable (PAR) 2011
June 2012
By Bill Lydon – Editor
This article about using data, modeling and other automation technology for Intelligence-based Manufacturing, is the sixth and final article in a series covering the recent annual Pharmaceutical Automation Roundtable (PAR).
About PAR
I had the privilege of attending the Pharmaceutical Automation Roundtable as an observer in November 2011. This PAR was hosted by Johnson & Johnson in Spring House, PA, with Dave Stauffer, Terry Murphy, and Joel Hanson of Johnson & Johnson participating.
Lead automation engineers from various parts of the world attended the invitation-only, two-day event. This is the most knowledgeable group of automation professionals gathered in one place at any one time focused on discussing automation issues.  A range of companies participated including Abbott, Amgen, Biogen Idec, BMS, Genentech, Genzyme, Glaxo, Imclone, Johnson & Johnson, Eli Lilly, Lonza, NNE Pharmaplan, Novo Nordisk, Pfizer, and Sanofi-Aventis.
The PAR was founded about 15 years ago by Dave Adler and John Krenzke, both with Eli Lilly and Company at the time, as a means of benchmarking and sharing best practices for automation groups among peer pharmaceutical companies. The group specifically does not discuss confidential or proprietary information, cost or price of products, price or other terms of supply contracts, or plans to do business or not do business with specific suppliers, contractors, or other companies.
The individual PAR group members have a wealth of practical knowledge and knowhow to share with other participants, truly learning from each other.
Topics are agreed upon prior to the meeting and a member with make a presentation on their organizations views and approach to the topic. After this presentation others comment on their organizations situation.
Intelligence-based Manufacturing
The presentation started with a discussion of intelligence based manufacturing concepts focused on harnessing the complementary power of data, modeling, engineering and IT infrastructure in order to create a game changing paradigm by transforming data into knowledge and ultimately intelligence.   The goal is to move from responsive and reactive actions to preventative and proactive manufacturing strategies. Ideally this will lead to a holistic system shifting from stand-alone and isolated unit operations towards integrated e-manufacturing infrastructure at process, plant, and enterprise level. The PAR member presenter defined intelligence as the ability to accommodate uncertainty in data and the adaptability to cope with prevailing conditions and risks. Intelligence-based manufacturing is based on a combination of a number of elements including process measurements (inputs and outputs), soft sensors (derived values), process models (simple to AI models), process simulation (dynamic or static), and process optimization algorithms.
The presenter offered a roadmap for implementation:
Data & Process Models
Generation & Accessibility of Critical Operational Data
Transformation of Data into Process Intelligence
Manufacturing Intelligence
Integration of Intelligence from Different Sources
Predictive, Adaptive, Multi-Scale, and Multi-Unit Control Strategies
Integrated Enterprise
Integrated Intelligence-based Planning, Scheduling & Operations
Skilled, Engaged, & Enabled Workforce Making Technology Savvy Decisions
It was also observed that intelligence-based manufacturing is the technology backbone enabling high process capability including Lean and Agile manufacturing.
The presenter pointed out that data from all sources needs to be used productively. This encompasses supply chain to production for multiple sites. Modeling capabilities at different levels are needed as well as integration of intelligence from different sources leveraging supporting technologies including predictive, adaptive, and advanced control. Another challenge is getting the right information to the workforce and having them make decisions based on real-time data.   “We have been talking about it for a long time but are just now getting to the point of predictive and proactive type activities.”  
There is interest in analytics and optimization at a high level of management and they are looking for something that is available today that can help reduce costs, increase yields, and profits. This PAR member’s company has a process analytics sciences group leading an optimization initiative with people that are research oriented and don’t get intimately involved with plant sites as much as they should. The philosophy behind it may be in the right direction but they haven’t yet included and asked the right people. We look at this optimization system and many in the automation group feel much of this should be in the automation controller. Also, the quality group has concerns about advanced software changing setpoints dynamically. Maybe in a few years they will get to the point that manufacturing will be based on a model developed under QbD (Quality by Design). “I just don’t see it right now,” said the PAR member.
He also described local efforts, “Every automation engineer in every site is doing something novel, something innovative using historical data and SCADA systems to make life easier for the operators.” Historical data is primarily used for predictive maintenance or operational improvements. It is important to find further ways to encourage sharing and collaboration between site engineers.
The following comments came out of the discussion by PAR members on this topic:
The following thought was echoed by many in the group. “Over the last years we saw automation engineers coming up with good ideas for improvements but in the current business cycle with patents running out and people reductions, so many engineers do not have the time to focus on looking at improvements and refinements.   This is happening at the same time that the company wants to get the most out of their assets, desire higher productivity, and operate with lower costs. Reduced automation headcount goes squarely against this goal.”
“The available time experienced engineers have to mentor younger engineers is now almost non-existent. Experienced engineers are so busy they don’t have the time to do this mentoring anymore. We are creating a less efficient, capable workforce now than we had ten years ago.”
“Early retirements are depleting companies of a great deal of expertise, knowhow, and activity-based knowledge that is not being transferred to the new engineers.”
“Often finding young engineers reinventing things we did 15 years ago…they are back in the mode of learning by error.”
“We have been using historical data with the maintenance system to do more predictive maintenance vs. time based maintenance.”
“In the last 5 years our people have had access to historical data and they are starting to use it, seeing the possibilities, and learning the value of this information.   We have made it our priority to get as much data into the historian as possible knowing that it is ultimately valuable and the need is not always obvious until the information is required or requested.” 
“We have a site that has doubled their production in one year by adding a historian and applying RCA (Root Cause Analysis) and OEE (Overall Equipment Effectiveness) to fully leverage the data.”
“We use golden batch curves to benchmark and monitor production for change or issues.”
“We do multivariant analysis and use it to manually tweak a batch.   We do not yet use it to change processes in real-time.”
“We have a lot of process modeling going on in the R&D organization and we have an operational feasibility team to determine if implementation in plants is feasible and business value justified. This has been an effective way to sort these out.”
“Management has a growing interest in data driven manufacturing.”
“We also have an analytical modeling group at our central engineering team and they are generally called in on site specific problems rather than looking at processes used at many sites. It is a fairly strong group of 7 people with the right tools.” 
“An enterprise system performs planning and scheduling, defining production orders at the sites. At the site operation level we don’t yet have dynamic scheduling.”
Data Management
“Data management is seen as the next step in our business model and it is very complex with such a wide range of equipment and systems to extract the data.”
“Sometimes we wait for the ideal, perfect model before trying some things to learn about what is really required. An iterative prototype process can be much more productive.”
“About 18 months ago we started a heavy focus on data business intelligence for manufacturing and dedicated a group of people to it. This was championed by manufacturing to understand what is happening within and across manufacturing sites. We have been fortunate that we have been able to deploy most of the S95 and S88 blueprint at most sites. The basic building blocks are there including historians, standardized MES, and LIMS. We started small by prototyping plant data warehousing and we are able to get data now across all four levels of S95 with a common frontend and data context. When we expose that to a higher level of manufacturing management, they can drill down and compare various things at sites. We have seen success with this.  In contrast that same group sponsored a new program that challenges existing manufacturing looking for cheaper ways to produce in various parts of the world.”
“The plants view PAT (Process Analytical Technology) as a very expensive endeavor and are quite resistant.”
“The main area we have advanced in is using metrics for fill finish lines with database driven morning meetings about the lines that explore information including how many times equipment stopped, for what reason, and how long so maintenance and designers can review to make good improvements.  People are asking for more of this based on early adopter success.”
Intelligent-based Manufacturing in many ways appears to be the extension of “front office” business intelligence into manufacturing and with a site-based focused. In the context of business intelligence (BI), it refers to computer-based techniques used in identifying, extracting, and analyzing business data. Business intelligence technologies provide historical, current and predictive views of business operations with common functions including reporting, online analytical processing, analytics, data mining, process mining, complex event processing, business performance management, benchmarking, and predictive analytics. Marrying business intelligence methods with automation systems appears to be the next logical step in the evolution manufacturing landscape.
I was reminded of the thoughts about “big data” being important when the comment was made about making it a priority to get as much data into the historian as possible knowing that it is valuable and the need is not always obvious until the information is required and requested. In this case, they had a site that doubled their production in one year that leveraged their historical data once they had it in place. They also got a team together to leverage that data and drive to root cause on issues and improve performance bottle-necks.   Google is another great example of a company collecting data without knowing its immediate value and later using it. This is further emphasized by a well noted technologist in a recent article, Creating Value in a Hyper‚Äêconnected World
Your thoughts and comments are welcomed.
Links to other articles in this series:
Part 6: Intelligence-based Manufacturing (You are currently reading this article)

Did you enjoy this great article?

Check out our free e-newsletters to read more great articles..