Leveraging JMP for Enhanced Data-Driven Decision Making in BASF Antwerp’s Chemical Production Plants

Hello everyone, I'm David Autrique, working in the data analytics team at BASF Antwerp, and today I'll be sharing how we leverage JMP to enhance data-driven decision-making in our chemical production plants. I would suggest to dive in.

Now, before diving into the details, I would like to present BASF Antwerp in a nutshell. Let's first look at the broader context of BASF event group, or production sites, which is located in the Port of Antwerp, as you can see. Is the second most important BASF Verbund site worldwide, consisting of a surface of more than 6 square kilometers containing more than 50 world-scale installations, and welcoming on a daily basis more than 3000 employees. With these 50 production installations, we manufacture a wide range of products including base and specialty chemicals, synthetics, refining products, inorganic chemicals and so on. Now, as you can imagine, these this extensive production footprint lays the foundation for a data driven approach when it comes down to optimizing operations.

Now what does that imply? World-scale manufacturing process of core chemicals. Let us take a closer look at the challenges that we, tend to face on a daily basis with operations producing, over more than 1000 tons a day, of product, we face several issues. This is more or less depicted on the block schema that you find at the bottom of the slide, including unwanted byproducts, variable yields, and so on. The idea is that we tend to produce, inspect final product for end customers while reducing variable costs to the minimal extent. With minimal raw material input, minimum energy input and so on.

What we believe in is that we can reduce variable costs, employing data analytics. By applying data-driven strategies, we presume that we can predict outcomes more accurately, reduce waste, improve efficiency, and so on. What we simply tend to do is that we want to valorize the vast amount of production data that we produce on a daily basis.

On a daily basis, we have multiple sensors, on our facilities that measure pressure, temperatures and so on. We simply want to consolidate them and get the maximum of information out of all the sensor data. As failing to do so would simply imply that we would build up a data graveyard, which is exactly not the idea. Now JMP comes here into the picture, and you could state that it definitely acts as an enabler that helps us to unlock valuable insights out of our processes.

Now, to reduce these variable costs: raw material costs, energy costs, and so on. Site management. Launched an operational excellence program. The key component of that program is Applied Data Analytics for Operations, a training program for process managers. By combining Utah State process knowledge with data analytics, we believe that we can unlock synergies where really one plus one equals three. Leading to, let's say, deeper insights. Put it in this way, measurable and more sustained improvements as well as better user acceptance because we combine and process knowledge and data analytics.

Now the data analytics team in Antwerp simply serves you state as some kind of catalyst in the whole process. Practically, and this is depicted here on the slide. Process managers bring a kind of problem statement to the table tied to the OpEx Target Picture of their plant. Then we simply start to dig into the problem. We quantify the problem, and we convert it, in fact, in an analytics, program.

This is stated on the fourth bullet we transform the operational issues into data exploration quests. Then, the way that we do that is by applying or leveraging simply point and click tooling. We don't want to make developers of our process managers, but we simply want that they learn to get the maximum out of the data.

In the final step, of course, we hope to suggest and validate improvement measures that we then can implement, and then the respective measures have to be locked in an OpEx database. As you can imagine. Now, as stated earlier, then the Applied Data Analytics for operations program is specifically tailored to the process managers. You can find a picture here on the right side of a classroom training that we're organizing, showing, the functionalities of JMP to a core group of process managers, 20 process managers that were selected for this specific training, and we try to explain them how to use JMP, and how to use this data analytics tool that we have on site in a practical, graphical and analytical way.

The delivery format that we have in mind combines in fact three main blocks. We combine self-paced learning because they are all very busy. They have to be able to work with the tooling, study the tooling, study the newest features on their own, at their own rhythm. We provide also consultancy on the job. We enter the plant, help them to solve problems they're facing with, at a given stage in their OpEx projects and so on.

Last but not least, we have this classroom training. Now, basically, the workflow that we tend to follow, is that participants are in fact taught to first, let's say, explore and visualize data, and identify in fact, the operational opportunities. What they have to do is that they, in a first step, analyze cost variability. Typically, they do this by using dashboards that they have in the plant, where they tend to follow the variability in their daily KPIs. Then they start to dig deeper in it. What they then do is that they start to delve, in fact, in the underlying time series or process data, sense temperature readings, pressure readings, whatever that they have in mind. In order to do that, they typically use conventional time series workhorses, such as seek, such as trend miners, and so on.

In a final step they convert, or they get in fact data out of this, first data exploration. They consolidate or convert, their quest in a kind of data table that they start to analyze using JMP. There you could state that the final step involves using JMP to conduct a really advanced analysis, where they typically do sensitivity analysis. They try to correlate key process factors with the cost function of interest. What we now found and observed is that this approach really starts to drive actionable improvements. We see that it gets traction. We also see that they get interested in using the tool that they start to get interested in using JMP. That they also see the value of the features that it brings to the table and so on.

Finally, we really finally set to empower our teams in operations to make data-driven decisions that enhance efficiency and reduce costs for BASF Antwerp. The next step that we have to do is to get the whole thing operational, and simply to do it. That's what we will have on our plate for the coming year. Thank you very much.