MES integration with sensor-based data in life sciences--8 years later
![MES integration with sensor-based data in life sciences--8 years later MES integration with sensor-based data in life sciences--8 years later](/pic/mes_integration_with_sensor-based_data_in_life_sciences--8_years_later/Ykk2OGZvYWg1WlE_.jpeg)
Good morning, everybody. Welcome to conversations on critical operations. And today we're talking to Barry Higgins. He is the APC test lead at Johnson & Johnson. Good morning, Barry. Good morning, Nick. Thanks for coming. And we're also being joined by Petter
Moree. He is the industry principal for pharmaceuticals at OSIsoft. Good morning, Peter. Good morning, Nick. And good morning Barry. It's great to see you both here. Morning. Good. All right. So the reason I wanted to interview you today,
Barry is because here it is, it's almost 10 years after you presented at our user conference back in 2012, in San Francisco in 2013, in Paris, and your video is being viewed by hundreds of people a year. It's one of the best, the most popular videos we have. So can you remind us again, what was it? What was your your topic that year? My topic on those presentations was a an MES integration project that we had done across the our pharma sites in Europe and North America. And and I mean, why would you guess that it's so popular still after all these years? I think the the utilization of any MES type systems across industry and not just pharma, and has increased over the last number of years. And I think the the work that we did with integrating it with our existing data infrastructure, and some of the approaches we took and must still be striking a chord with with companies today.
Okay, well, can you give us a quick recap, I mean, if you can have a 40 minute presentation, but do what you can. And it as I said, it was a it was a major project in our former division. And in starting off in Europe, with with three sites, where we wanted to have a an implemented MES system in our large molecule API sites. And we had a quite an existing and mature data infrastructure in place, and we were using a lot of paper records, copying of test results and putting that into our paperback book for release at the end of when when the battle is completed. So the business challenged and challenge was to streamline or our manufacturing process, move towards a paperless and manufacturing floor. And, and also get rid of visibility and the end to end of our of our process, from warehouse to plant and back to the warehouse.
If I remember correctly, that's one of the what's one of the goals you accomplish, you did reduce a lot of paperwork, if I remember correctly. Yeah, we reduced a large number of documents and forms and record batch record books, and we move to ebrt. If I remember correctly, we removed over 1000 individual documents from from from our manufacturing floor.
Would it be fair to say that, that that was one of the biggest outcomes of this is it you were able to look at your data going from, you know, go all the way through the process. And that was a view you just didn't have before? Yeah. Prior to the Emmys and being implemented, the process could only be viewed in individual steps or stages within within the within the process and that meant looking at paper records and going to our historian to look at a particular bioreactor or filter or autoclave at a time and but with integrating and bringing meds and integrating or existing earpiece systems, they're historians. It was the business that and develop their product, great, greatest value and greatest visibility because they could see when when the order was released Murray RP system and where it was going, what stage it was at at any stage by going to our me system and looking at that. And it was only because of our existing data structure that allowed and that to take place. I think the presentation that Barry did here a few years ago has has really gained a lot of attention because I think that we see that it Johnson and Johnson Johnson had been ahead of the industry, they have started to implement MEMS and among other things of the year. So coming into the mspc. Now, many, many of the
pharmaceutical companies are just starting to implement them as systems in a similar type of fashion. And I think that's one of the reasons they would like to learn from from the experience and learn from the experts like you, Barry and and and your team here at Johnson and Johnson. So a few common question I often get when we are interacting with customers could be what what is the order of implementation? For example? Is it better to start to implement an mes system before implement implementing a data historian or data infrastructure? or What can you say about that? I can, I can only speak to, to the situation we that we had in Johnson and Johnson, where we had in on those farmers sites, we had an existing data infrastructure in place, it enabled us to was more quickly implemented an MES solution. And
because that allowed us to connect to one source for our process data, and we didn't, we weren't going going between level four, to level three to level two to level one, we were able to go directly only go as far as level two. So we weren't, I suppose crossing the firewall between our internal networks. And which gave us from a risk based approach allowed us to, to not put our software systems at risk from a level four system. And and it also gave, because it wasn't the existing install base that we had. It gave our our colleagues in, in regulatory and quality at the Ria conference, because they we weren't, it was the same source of information that we were processing that we were connecting to. So those are the main benefits that the historian or data protection would give to me a system to be like the single source for the operations data. Yes. And also it allowed us to, I think I mentioned it in
my presentation to bring instrumentation and instruments that we were using in this on the shop floor and tour EMEA system quite quickly, by connecting it directly to our tour or their historian that we had, we're not creating new and data pipelines, from the announcements into we could upload we could we could deploy a standard so that pi would be seen then as the as the as a gateway for all the all the data, not just all the details Qaeda or DCS or control data, but also from other type of load sensor building management systems or whatever that could be of relevance. Yeah, I PC benchtop device that we use in in the in the plant. And I believe in in my presentation that I referenced the pH meter that we we've got that they're bringing it directly into me as bringing via pipe.
Right. So so the journey of implementing MES. Is that mostly through the business advantages? And what are the type of clients and insights where you're implementing MES at first. And as I said, it was our family. There's multiple APA sites that we were focused on. And as that was sent, that was
it was into delivered that would deliver the greatest business value and visit had the greatest business need. And so we have roll little has been rolled out across a number of farmers sites in both in Europe and North America. And and also, we've seen expos rodeos and to serve our consumer sector as well. And from a technical point, of course, the end date I can be contextualized a lot of information and valued what's the second EMEA system that can also be sent into let's say, a data story and to understand better when the batch dot and and to have a better traceability. Also when you
would like to do other type of applications? Yes, having a stream of data is is one level of information. But if you can add more context, it gives you more, it brings more understanding. And to the data mix. It makes us a richer data set. The Richard david package. You know, that's great. And I've seen a number of other presentations from from from your organization. So you're not yesterday.
pioneer or a leader here under MES implementation, but also in the adaptation of the use of, let's say, advanced analytics to do different type of process modeling and golden batch technologies, for example. And of course, that is also very important to understand that the batch start and batch and phase transitions in, for example, in the large molecule operations. So, I would assume that data has has really become much more valuable within your organization due to that the tremendous value they, your colleagues and you have really demonstrated within within both AP application of MES. But also
with with the usage of machine learning and advanced analytics. Yeah, and I think what having the the me SNR data structure there has brought the data to a wider audience, it's not just the process, and quality engineers and departments looking at the data for, for release purposes, we know have a rich, historical and contextualized data set that we can we can give to our, our data science teams and globally and with easy access, and to allow them to build in the type models to give us a better understanding of how our processes are performing. And as you said, to develop goals and batch where we can monitor our processes in real time and monitor how it is performing against that golden batch. So
we're widening the, the use and the audience and of using that data set. And again, I think it's critical what you said there, having a context having the Event Frames or batch or having a context arise data set is, is crucial. In a previous role, I was I was working in our one of our r&d sites. And we
were talking about tech transfer of a product from the r&d side to commercials site and I showed him a trend from PI Processbook of one of our, one of the reactors in in another site, another country. And while I was while I was there, and just off the the profile, the temperature profile, there is identify the product, and that's the product that they were worked on. And they felt they were able to say, Okay, now we can monitor, I suppose have a view of the products that leave the research and the pilot plans. And as they transfer to commercial, so it enables them to monitor the how the process is performing. And also if if if there are issues that need to be addressed on the commercial sites, and they can quickly access that data and give support.
And that's great. And it's a great, great example here, because one thing we do see is that many many customers now, they rather would like to implement a data story on the data infrastructure, also in the in the process or end and in in a pilot and before the pilot plant and through the pilot plans to build up the knowledge management to reduce time to market that's one thing but also to build up a much better understanding like knowledge management about the project and the process. So that can be brought with the process to through the commercial, clinical and later commercial manufacturing. So that seems to have a be have a tremendous
value in organizations today. So so you see now the usage of of also data infrastructure in in your process development areas. Yes, we had Johnson Johnson undertook a program a number of years ago to bring the same and I suppose applications and data structures and data structures that we have in our commercial sites, and move that and bring that to the to the pilot plants and the development labs and as well. And so we would we look to apply the SAP structure for batch processing, and right down to our development labs. So there's an allows us to track
the process throughout the choice lifecycle. Excellent. So what why don't we have your former colleague colleagues here, Paul McKenzie, present that presentation A while ago, on exactly that type of topic where he spoke about the importance of interacting with r&d all the way to to the patient.
To real address those things come into that how important do you think it is to have leadership and management to support this type of methodology and implement them in in an organization like this, without a doubt, having senior leadership support for these, these types of strategies, and not enough just a strategy on paper, but to, to have that strategy executed across the organization, as I think Paul's mantra at the time was led to patient to do that. And we continue to have that strong leadership within within j&j to, to look at and to innovate, and to look at developing the next and the next technology the next and look for the opportunities to to, to accelerate, or time to market and accelerate the time, or reduce time with our products spent in our warehouse so they get quicker to our patients. Yeah, that is excellent. I think that if you look at the technology, and what is available in the market today, with all this type of machine learning and artificial intelligence, and data, infrastructure, and all those things have been around for for quite a few decades, that's nothing really new. What I think is new is that leadership are seeing the benefits and the business value of doing these things. And also that you have, you have better knowledge within
the workforce. So you have you can hire people with better understanding about data science, statistics, data modeling, Process Management, automation, and other things. And when you bring all those different parts together, that's when you really can start to be to take the leverage of being digital. So that brings me at least into a topic of this industry. 4.0 that is a buzzword, you hear about that everywhere. And everyone is is probably talking about that. But
I would say that you have been doing that, not just for a few years, you have been doing that for quite a few, quite many years, actually. So So can you kind of share some insight of what you're doing with regards to industry 4.0 and maybe even for my 4.0 as just an adaptation of that to the life sciences
markets. Yeah, and I suppose, as you said, they're we're we're we're bringing, there's a new skill set a new generation of, of employees coming in that are, are more data science aware. And I suppose the one thing that you as a data scientist needs his data. And so being able to supply them with the data from
multiple sources, and it's not just from a historian, we it's our MDS or europei, or maintenance systems, we need to be able to give that information to our data science team in our digital teams to to give them a robust data set to allow them to to develop those those models that will, that will deliver value back back to the business. And in information, in particular, where we're moving ahead with a number of initiatives in the industry 4.0, or area, and around the predictive and adaptive process control, and the area and digital assistant and automated vehicles. And so again, what all these
initiatives need as a basis is data. And in particular, in pharma. We're looking at trying to reduce our release times. And take as I said before, to get the products to our customers and our patients quicker, to do a lot more of our end of line testing in line to do it while the process is being is progressing. rather than waiting for a result
at the at the end when the battery is finished, and modeling our our equipment and our steps to again replace some of those those tests. And we're also looking at how we manage our assets, how we do our maintenance and scheduled maintenance and moving more to condition based predictive base maintenance. But we're all what all those strategies needs. And all this program needs is data. Yeah, no, it's great to hear that you spend the whole reister of applications here on really consuming and using the value of the data. And I would assume that having having data accessible and in a trustworthy way is of course super important for for the majority of the data consumers today within Your organization here. Absolutely, that the policies
and procedures that that j&j have across their other, their data systems from a data integrity viewpoint, and is a very strong and robust, but having the having the data digitized, and having a digital infrastructure. It also increases the strength around that. There are some challenges that it brings as well. But it also allows you to have one version of the truth. Yeah. Excellent. Well, one of the things I noticed, you had mentioned back in 2012, is that you were thinking of scaling this up to other sister plants, the same system you'd set up? Did you ever do that? And what kind of results did you get? yet? We took the the the learnings and the, the, I suppose challenges that we had in the first install, and roll that make sure they were incorporated into our and just to say, that follows on both in Europe and North America. Okay, great. I know that our acid structure the AF
was a critical part of a lot of what you did, what were you able to to do any duplication or do able to make any use of the structures you had created? Or? or What did you? Was it easier? The second third time around? I guess, is what I'm asking. Yeah, the use of M unit class templates that would that we had developed in on the first slide. And for our process equipment, in particular, we were able to, to share those with the with with our sister sites and allowed them to, to build on them. And so they weren't starting from from scratch, as
to how to, to to build a reactor, for instance, they we were able to give them okay, this is a template and allow them to extend or if they didn't want to use all the attributes, and that wasn't that wasn't an issue, but it gave them and I suppose because we had done the qualification and the testing of that those templates upfront, it reduced the time for implementation on the other side. Okay, and you were using a package live point from marabu. I think you were using that for a visualization or for analysis and visualization? I'm not sure. Is that right back at that time? Yeah, that's right. And it's still being used across a number
of our sites, both in Europe and North America, and I believe in Asia Pacific. And it's not just the process data that that that light point tool allows the visuals, it has connectors to, to other data sources as well. And so we we can view have a greater view of the of our process, but also some business dashboarding capabilities and connected to some of our businesses as well. So, is it used in about eight sites, I
think across the 10 seconds, okay. And you would also mention a lot of non pi data, SAP data quality control data LIMS system data, have you added any other non non PI System or non control system data sources since then. And I suppose, shortly after this project, we would have looked, we looked at independence on the maintenance systems as well, or pm systems. And so we would we would, we had connections to our SAP Pm our Maximo Systems Incorporated as well. But it was at a time, Nick, this is back in eighth 10 years ago. Now. We're moving I suppose to Amman, I spoke to the pettibone earlier about it essentially 4.0
we're, we're moving our I suppose, our analytics and our data platform to to the cloud. And so the connections for I suppose non pi systems are being done there. But the PI data still forms a crucial part of that. That data layer. Right, okay. Okay. So that means, yeah, I guess one, two, this is great, because I think they also demonstrate they are a head of head of the industry even here. So the industry is often Of course talking about cloud and have been doing that for a while but in my opinion, rather few companies have really taken the step to the clouds. It's great to hear so I would
assume then you are your leverage. The say the day integrator another type of oversight of technology to bring the PI data into into these cloud platforms for analytics for visualization and data access, yet they At advanced, the PI Integrator has been an integral tool in a lot of the, I suppose that the proof of concepts that we've been doing it on a number of our projects, it allows us to to get data quickly and into our cloud based platforms. And no, there have been some challenges there. And I know we're working with OSI to resolve some of those challenges. And but I It has been an invaluable tool. And maybe you already answered this
in answering Petters questions, but do you have any plans for the future? Yes, as I document what we're better, we're looking to bring our pts bring PT instruments from the lab, bring them onto the shop floor, integrate those into our, into our processes, and to do more inline testing of our, our processes to, again to to reduce time to to America. Right. So, you know, you it's interesting that you mentioned that this Ph. D. and above. I was I was part of this Ph. D. Journey backward reading back in early 2000 2002 1003. And then the guidance PT guidance came out in 2004. So it's like 15
plus years ago. And right now, if I just do a snapshot over the industry, I think that right now, many companies have really started to really implement PHP and then take that into value when in development to speed up time to market and have much better understanding about their complex products they are doing, for example, in a large scale operation, you see a lot of usage of brahmam, to listen to evaluate the metabolites and other important components in the in the cell culture, for example. And we have seen a great use case examples, I think from from some of your colleagues as well together with Martin Gadsby from optimal on 20 q. So I think they have been around for 15 plus years doing this type of PT, and we have promo sites of yours, of course, partnering with those companies closely to be able to really add, let's say operations state and PT data in combination as that would definitely add much more value to your, your processes and your product development. Yeah, the PTS mutation that we're using gives us another data source another with an a more more complex data source coming from the ram as a proposal in particular, and but allows us to bring that and from the lab, and all the way through to through to commercial. And with the work that we did with
with optimal in the past, we will utilize it again, going back to work toward data infrastructure two or PI System we're so we have the connector, our PI System tour Cinta que system, or Syracuse's and protect the RAM. And so we're bringing an app running our nvm models on that platform. And to give to loaded the allow people to allow personnel to see how the product process is performing. And with with large molecule processes. It's quite
complex. And as you said, there's a lot of a lot of data that comes in and when you're when you're dealing with a large molecule. Well, thanks so much. Thanks for joining us, both of you. Hey, like we normally do. How about if we finished with a quick
lightning round of things that, you know folks in manufacturing know about, but nobody else does? I love these questions. So, Barry, have you ever worked a compressed workweek, you know, 12 hours on 12 hours off? That kind of thing? Yes, I am. I, in my early days in the industry, I worked a lot of shift work. Okay, dislike it. And there was some pros and cons and time off but long hours as well. So yeah, so long breaks. It's very, very interesting. Hey, what's the first computer you ever used?
first computer I have is a Commodore 64. Okay, cool. Cool. So, you know, we work with data? What's the coolest calculation or code that you've ever written? code and put me on the spot. Yeah, that's what this is going back to my automation days working on the an Emerson RS three system. And so I would have done a lot of coding on
that. So I can't I have too many examples to call Oh, no problem. So now Is there anything working in pharma? That's the kind of take to your personal life For example, I used to work in a chemicals plant, right, whose history was first gunpowder. So safety was everything. I mean, man, you know You just had to be Mr. Safety there. And so you kind of took that home with you when you work in pharmaceuticals or anything like that. Yeah. And one thing that I I take from not just
Federal from j&j is his or his or credo, and that j&j have had for a long years. And I suppose it's also a good life message. And a reminder of just do good and be good and be kind. Very cool. Very cool. So do you, like many engineers have a piece
of broken gear, or like a fried motherboard or something you keep on your desk as a memento. And I, it's, it's back. It's not from my engineering time. But when I was, again, in manufacturing, and I inadvertently left a piece of equipment inside another piece of equipment, let's say and there was a there was a and a low noise in it, and didn't cause any damage to the equipment, but someone wanted a piece of equipment, for me presented to me as a trophy. That's, that's pretty, it's pretty sweet. mounted it for you. Well, that's great. Hey, is there anything that we, you know, we forgot to talk to you about, you know, sometimes we just don't ask the right questions, anything you want to tell people? Are you trying to make better use of real time data? I suppose the one thing I would say is contextualized. And it's
not just not just there are certain benefits to having your data in time series. And, but bringing contractualization context to your data will allow you to, to share it more share more easy with a wider audience and give you greater benefits. And so that's my one takeaway. Right now.
Okay. Well, thank you so much. Gentlemen, thank you so much for joining me. No problem. Always good to talk to you guys. Good to talk to Patrick. Good talk, Nick. great talking to you, Barry. And next time, I hope that we can see each other face to face at the PI World Meeting in a few months time. So thanks a lot, Nick, for for hosting this and
putting it all together. I find it very, very exciting. Thanks a lot. Well, thank you. Thanks for joining us. And thank you all for joining us again. I'm Nick Dorazio. See you next time.
2021-01-06 23:02