COMPANY CONFIDENTIAL. DO NOT DISCLOSE.
In a smart production environment, a little process understanding goes a long way
At ADM, multivariate statistical capabilities are key to generating process optimization and improvements that scale to production sites around the world
Challenge: With the adoption of fully automated production systems at ADM sites around the world, technology engineers tasked with driving proactive improvement and analytics enablement are challenged to replicate successes at scale.
Solution: ADM standardized around JMP® Pro, a statistical discovery software that provides wide-ranging tools for validation, process optimization, multivariate analysis and customization. “JMP Pro really enables people to do work faster,” says Senior Analytics Engineer Isaac Himanga, adding that in his estimation, what used to take 15 minutes now takes just one. “You multiply that by 100 users and all the analyses they do in a year, and it has a huge impact.”
Results: In addition to reducing time spent on data access and preparation, JMP Pro has helped engineers across ADM tighten process control and resolve costly process variation and validation issues. Furthermore, the team has found ways to leverage process knowledge gained from one optimization project to improve SPC at other sites via scripted JMP add-ins and dashboards. Says Himanga: “We put it online – that’s a great way to replicate [the impact] many times over.”
_____________________________________________________________________________________
Archer-Daniels-Midland Company (ADM) is a major player in the global food industry with an extensive portfolio of ingredients and flavors for everything from breakfast cereals to livestock multivitamins. By leveraging its capabilities across the supply chain, ADM has revolutionized the way natural raw materials are transformed into dietary staples for hundreds of millions of consumers worldwide. Moreover, the company is now using its influence to drive sustainable solutions in agricultural technology and food processing.
Much of ADM’s capacity for innovation in recent years can be traced directly to its analytics transformation. The company’s widespread adoption of state-of-the-art IT and data systems has catalyzed a sea change in the way process efficiency and effectiveness are optimized at every stage of the production pipeline.
“We’ve seen escalating momentum for analytics initiatives across ADM,” says Senior Analytics Engineer Isaac Himanga. That groundswell in support for analytics enablement, he explains, came partly as ADM reorganized around centers of excellence – one each for improvement and analytics – that established more interdisciplinary roles for technology engineers who, like their process engineering counterparts, are subject matter experts for each individual process. The other motivator, he adds, was simply demonstrating value through a number of high-profile case studies: “It takes some time to bring people up to speed, but you just have to get them over the initial hump. It helps to take an opportunity where we understand the process well to do a lot of work in a minimal amount of time in order to demonstrate improvement,” he says.
As the analytics capability of ADM was gradually elevated through trainings and data infrastructure updates, Himanga explains, teams came to view Excel as no longer fit for purpose. “We were trying to take these analytics tools and use our data to make better decisions,” he says.
The demands of more advanced process control systems and automation led them to JMP® Pro, a statistical discovery software that has been a key partner to ADM not only in culture transformation but in many of the analytics team’s early proofs-of-concept.
Moving away from reactive problem-solving
Like any multinational corporation operating between food processing and commodities markets, ADM has many millions of time series data and equally as many tags. Moreover, ADM often acquires new facilities – many of which, at the time of acquisition, rely on outdated analog systems and must be retrofitted with digital sensing technology. Accounting for the shortcomings in antiquated systems, Himanga explains, has presented an opportunity to scale improvements and process knowledge from ADM’s state-of-the-art flagship sites.
Across ADM’s many global production facilities, he says, Industry 4.0 has proven there were gaps in existing instrumentation and shown the business value of resolving data systems challenges. “There are things we’ve learned in one facility that we realized we could replicate in others,” he explains. “[This scalability] is now enabling us to move faster and focus on specific improvements.”
Today, Himanga’s remit lies in both one-off troubleshooting projects – reducing steam, increasing yield or solving a quality issue, for example – and projects where he and his colleagues seek out opportunities to increase revenue and drive down costs. “We’re putting a lot of effort into proactive analytics,” he says. “Take anomaly detection for example. We can call out variables or parts of a process that are working abnormally and then use our platform to help find problems before they really manifest themselves as effecting quality or yield.”
With its straightforward analytics workflow and platforms that enable advanced approaches like multivariate modeling, JMP Pro has broken wide open the field of possibilities when it comes to improvement. Not only has Himanga uncovered areas where data was being collected but not utilized, he has also identified countless opportunities to move from a traditional reactive approach to a more proactive model.
“In a lot of cases, we’ll do a big analytics project and it’ll turn out that even the set point was enough to get started and pointed in the right direction!” he says. “We have to balance not spending too much time on any one optimization opportunity because there are just so many opportunities out there. We’ll start optimizing something, make recommendations, and when the facility later gets better instrumentation, we can go back and do a version two.”
Standardizing around JMP Pro helps scale successes across ADM’s global facilities
In addition to its large presence in North America, ADM now has significant manufacturing operations in Europe, Brazil and the Asia Pacific region. By taking a global approach to analytics excellence, Himanga and his colleagues have acted on unique opportunities to scale the business value created through engineering improvements.
“Instead of having someone focus only on their own geographic region, we intentionally mix and match so that each [team member brings a fresh perspective] to plants in other parts of the world,” he says. Creative, data-driven approaches to problem-solving transcend national borders, and common tools help individuals overcome cultural and language barriers that might otherwise impede collaboration.
“Our vision is to have every engineer, maybe even every manager, go through a standard set of training so that we all speak the same language around data analytics,” Himanga says. “But to do that, we needed to standardize on a tool.” JMP Pro provided a mechanism for time and value-adding activities to scale, particularly as the tool’s scripting feature allowed analytics experts to quickly roll out customizations.
The “JMP model engine” reduces data access time by 15 times
One such add-in created by Himanga and his team pulls data into JMP Pro directly from ADM’s data historian. The add-in – what Himanga calls a “JMP model engine”— can pull information from a source system to a desktop PC. By then running the script against new data and sending the results back to the historian, much of the data access process is fully automated, he explains; users can go from offline analysis to scoring a model online relatively quickly.
“Accessing data used to be a lot of the work, especially for small projects… and you could spend quite a lot of time doing it,” he explains. “But our custom add-in now takes care of all that work. You can type a tag into JMP and it’ll go get the data and give you a script. And if tomorrow you want to see the same graph using more recent data, all you have to do is then copy and paste the graph, put the scripts together and click ‘run.’”
Himanga estimates that automating data access processes in JMP Pro has reduced what used to take 15 minutes into just 1. “And if you multiply that by 100 users and all the analyses they do in a year, it really adds up! JMP really enables people to do work faster,” he says. “We’ve got enough positive use cases now that people are talking. People who were once among the skeptics are starting to say ‘well, the other plants have done it already. I want that benefit too.’”
Automated dashboards advance tighter process control
Enthusiasm aside, Himanga is clear that analytics transformation is not something that happens overnight, even with the right tools; rather, it’s a long-term project and a journey. Improvement processes may take 6 months, but the process knowledge they generate is worth the investment.
To illustrate, Himanga shares examples from two process-related initiatives where multivariate statistics in JMP Pro solved an otherwise unsolvable problem. In one case, production issues had plagued one of ADM’s corn wet mills where corn gluten feed pellets are produced. At times, the liquid corn solution would not run through the mill as fast as was needed, but operators could not identify the source of variation.
“The cause was really hard to identify as a single mill can have millions of points you could potentially look at, along with a bunch of quality parameters and more inputs than you could possibly graph in Excel,” he explains. “We’d had a lot of people working on this problem for years. But it was really exacerbated two years ago to the point where the plant as a whole was slowing down.”
“So we used JMP to look at the thousands upon thousands of tags with different lag times. It was a really big data set and we ended up using all of the different platforms in JMP. In the end, we did a bootstrapping of the data set to identify the most likely candidates. Then we could use the multivariate tools in JMP to identify the real benefit and in six months, we found our answer. We figured out that we had to run the process a certain way. A few months after that, we were seeing improvements.”
In another case study, Himanga and his colleagues identified a pattern in how and when operators were cooling equipment used to create a soybean meal product often sold to customers as an additive refined to custom specifications. Prior to looking at the data in JMP Pro, the team had been unable to understand why equipment maintenance and cooling were having such a variable impact. By using analytics to develop process understanding, however, Himanga was able to define a set of guidelines to help plant operators know precisely when to cool the production line. Then they took it one step further.
“We realized we didn’t need to give guidance to them at all – we could have the computer do it for us,” he explains. Now operators can simply look at an automatically updated dashboard to see when the production line’s steam valves should be closed.
“[This dashboard] gives consistent response across all of the operators, across all of the different shifts, and results in tighter control. And JMP helped us to do that at not just that one facility, but others as well,” Himanga says. “Sometimes there are different scenarios that cause moisture differences. But using JMP, we’re able to quickly identify the key factors, build a model and generate a recommendation that can mitigate those fluctuations. Then we put it online. It’s a great way to replicate [the impact] many times over.”
