THIS STORY IS AWAITING CUSTOMER APPROVAL. CONSIDER IT COMPANY CONFIDENTIAL. DO NOT DISCLOSE.
Driving towards a better tire
At Michelin, multi-factor experimentation is creating key efficiencies in R&D. That innovation just might unlock the future chemical processes needed for a 100% sustainable tire
_____________________________________________________________________________________
Challenge: As Michelin strives to achieve ambitious environmental sustainability goals for its tire portfolio, it has become clear that classical design of experiments (DOE) is no longer sufficient for the imperative pace of innovation. To accelerate chemical process development and reduce the costs associated with experimentation, the company sought to overcome inefficiencies created by their dependence on an inadequate toolkit.
Solution: Recognizing the need to create economies of scale, technical leadership at Michelin greenlit a proposal put forward by Florent Mathieu, process engineer at the company’s site in Auvergne-Rhône-Alpes, France, to license JMP® statistical discovery software. “The DOE capabilities afforded by JMP are much more efficient than other tools,” he explains, adding that in addition to streamlining data management and exploration, the software now enables engineers and scientists to optimize a multitude of experimental variables with complex interactions.
Results: Since its implementation, JMP has helped engineering teams to achieve multi-objective optimization; for example, reaching the longest possible tire lifespan while also maximizing grip. Says Mathieu: “Since we’ve begun using JMP, the number of experiments we need has been drastically reduced, as has the related cost of those experiments.”
_____________________________________________________________________________________
In addition to its ambitious targets for reducing emissions from logistics operations and manufacturing, the Michelin Group announced in 2021 it will reduce raw material related CO2 emissions by 70%, increase the energy efficiency of its product portfolio (10% by 2030) and manufacture a fully sustainable tire by 2050. Such targets, while attainable, will require unprecedented innovation – a reality the company has not shied away from.
As Michelin seeks to replace all fossil materials with the biomaterial substitutes now available through emerging supply chains, the company has also embraced creative solutions to increase efficiency and accelerate development. In R&D, those tasked with creating the future chemical processes required to produce a 100% sustainable tire see the company’s commitment to environmental sustainability as going hand-in-hand with a step change in analytics culture.
Innovating chemical processes for a 100% sustainable tire
When Florent Mathieu, PhD, first joined Michelin in 2021, he understood well the brief – to establish chemical processes to produce entirely biosourced raw materials – and what would be required to achieve it. An expert in building process intensification strategies based on non-linear experimental designs, Mathieu came to the tire industry from the life sciences where he had designed a process for the synthesis of artificial red blood cells for biotechnology startup EryPharm.
Brought on to a chemical process engineering team at Michelin’s site in Clermont-Ferrand, Auvergne-Rhône-Alpes, France, Mathieu began by taking stock of existing data workflows to identify strategic improvement opportunities that could introduce a new level of efficiency to the R&D organization. “Statistical analysis allows us to be more objective about everything we do,” he says.
“I strongly believe that all the data we capture – be it historical process data or otherwise – is very important for understanding [the science] and the phenomena we’re observing. It is now possible to access more and more data, so we absolutely need more efficient tools with which to manage that data – as well as to analyze, explore and model it.”
A change agent advocates for enhanced statistical capabilities
While Michelin’s process engineering team already had a set of statistical best practices in place, Mathieu saw an opportunity to create transformative efficiencies by moving from classical design of experiments (DOE) to modern DOE and implementing automation to create economies of scale. Raising the statistical capability of the entire organization on the face of things seemed an ambitious goal, he says, but much of that challenge could be achieved simply by introducing a new tool.
Reflecting on his own first foray into an industry role at EryPharm, Mathieu says his initial impulse was to use the programming tools he had relied on in academia. Mathieu soon learned, however, that the pace of development required in industry would not allow sufficient time for coding.
“You have to not only be so much quicker, you have to also share your data and tools with colleagues,” he says. “I had to find another way to work more visually and be more efficient, so I started looking for a statistical tool and came across JMP®.”
Having used the statistical discovery software previously, Mathieu quickly observed that the key rate limiter at Michelin was not inadequate knowledge, but inadequate tools. Without access to the right software, scientists and engineers were limited to much slower experimentation approaches.
“Classical design of experiments is just not as efficient as what can be done in JMP,” Mathieu explains. “So one of the first things I did at Michelin was to propose we buy JMP.” Once the company’s technical leadership saw what the tool could do, he says, “it was not very difficult to prove to them that JMP was really powerful and efficient.” That power soon translated into fewer experiments, lower costs of experimentation and shorter development cycles.
“At Michelin, experiments are very expensive, so if you can reduce the number of experiments you need to arrive at an answer, you can save a lot of money and time,” he says. “Since we’ve begun using JMP, the number of experiments we need has been drastically reduced… In fact, I did an estimation of how many experiments we can save – and the costs associated with that – and found that in less than one month, with just the economy of experiments we didn’t have to do thanks to JMP, the savings more than pays for a license.”
Custom design of experiments provides more complete process understanding
Despite its ubiquity in everyday life, the tire is, in actuality, a very high-tech object. The product of complex chemical formulations, the average tire comprises more than 200 unique raw materials including natural and synthetic rubbers, carbon black and silica, resins, oils, sulfur, cables, bead wire and textile reinforcements like rayon and polyester.
One of the most important ingredients in any tire formulation is elastomer, a polymer characterized by both viscosity and elasticity. The properties of a constituent elastomer frequently dictate the composition of formulation outputs, and elastomers themselves vary to a great extent. In other words, Mathieu explains, manipulating elastomers within a formulation – from small tweaks to major changes like replacing nonrenewable materials with biosourced components – affects the other factors via complex interactions.
“This is why it’s so important to explore the entire design space,” Mathieu explains, adding that when applied to chemical reactions, DOE provides insight into variables like temperature, pressure, reactivity and the ratio of catalyst to raw material. With more than 200 components in any formulation, conditions must be screened either through one-factor-at-a-time experimentation or through DOE.
“There are many people who believe that to understand something, they have to take a one-factor-at-a-time approach,” Mathieu explains. “We think linearly and often forget that we don’t understand interactions intuitively. That’s why we have to do design of experiments to get more information.”
With JMP, Michelin’s engineers and scientists now have access to a suite of tools for definitive screening designs and custom DOE that can be used in almost any experimental situation to provide highly customized solutions. Entire designs can be run within resource constraints. Only have the resources to run 12 experiments? No problem. Have a partial data set and don’t want to start over from scratch? It’s easy to augment new experiments with data you’ve already collected. “That means you’re not running an experiment all over again,” Mathieu says. “Utilizing data you’ve already collected is a very important efficiency.”
On top of its cost and time efficiencies, he says, one of the strongest capabilities of JMP is its ability to arrive at the optimal compromise between conflicting objectives. A tire designed to last a long time, for example, would need to be very hard, while a tire designed to have a lot of grip would need to be very soft. Achieving the correct balance between the two is essential to producing a high-quality product and, he says, multi-objective optimization is where JMP really shines.
“The best way to prove to [skeptics] that DOE is important is to show it to them with their own data,” Mathieu explains. “When I show them a map of all the data they’ve gathered and all the experiments they’ve run, they understand that [without DOE] they weren’t exploring the whole design space.”
From basic data visualization to advanced DOE and Functional Data Explorer all within one tool
With his colleagues now taking full advantage of increased capacity for state-of-the-art DOE, Mathieu has been freed up to investigate other ways of leveraging statistics to introduce efficiencies into process design. A JMP Pro user, he has explored ways of using the tool’s Functional Data Explorer (FDE) platform to predict certain properties of elastomers using data from mass spectroscopy. By measuring key parameters like pressure, temperature and conversion with batch reactions, Mathieu can understand how a batch profile influences the properties of a product – and adjust parameters over time to optimize desirable properties.
“The idea of FDE is to transform functional data into a vector, create a model and do statistical analysis on that vector,” he explains. “It’s so powerful because with a spectrum, you can predict some very important properties for an elastomer and imagine how best to modify the reaction and chemical components. It’s a very predictive tool.”
While not everyone in the organization has the need to look at functional data, the fact that users of all skill levels can work within the same tool is a major asset. JMP, Mathieu says, has introduced efficiencies in data management and access as well as data visualization and exploration – capabilities that affect everyone. Visualization is what he sees as one of the primary features of the tool and encourages all users to first visualize their data as a mandatory step in any statistical workflow. Moreover with JMP, all these steps can be standardized around a single set of best practices, automated and rolled out across the organization.
“Management has very quickly seen the advantages of using such a tool because it helps to formulate the problem clearly, identify your objectives, and have a good methodology,” he says. “The community [of users at Michelin] is growing more and more every day. And when we present the reasons that we use JMP, everybody becomes interested and curious to know more and get the software.”
