Manufacturing Intelligence

The “Golden Batch” is the name given by chemists to a perfect result, a batch in which everything comes together, from the specifications to the energy consumption. But hitting such a bull’s-eye can be a difficult task. A team of experts from Bayer Technology Services and Crop Science is making it easier – with the intelligent analysis of process data.

A confusing array of possibilities: In chemical and pharmaceutical production, many variables can be modified in the search for gold.

Hobby chefs know it well: sometimes a meal works brilliantly – sometimes it’s a flop. There can be many reasons why the results are not always identical: the size or type of the potatoes, for example, or the temperature of the water, the amount of salt. Each of those variables can change, and each has its corresponding effect on the outcome. Of course, you could keep a tally of all of the factors involved, and that, of course, would lead to a perfect result every time. But how many amateur chefs do that?

In the chemical and pharmaceutical industries, a perfect result is called a “Golden Batch.” Needless to say, all of the other batches that are delivered to the customer fulfill all of the necessary specifications. Purity, melting point, moisture content – everything complies precisely with the specifications. But it is still entirely possible to differentiate batches from one another. In terms of manufacturing time or energy consumption, for example. These are only tiny differences, to be sure, but in the long run, even tiny differences can make themselves felt.

But how do you turn a product into a “Golden Batch”? Does it come down to stirring for an extra minute before heating, so that the source material is better mixed? Or does opening valve X only at point Y, thus maintaining the high pressure just a little longer, make the difference?

There are countless possible adjustments in a chemical process. Plant operators with training and a lot of experience are well aware of that. And it is precisely their expertise – and also their gut feeling – that is so valuable to any operation. And, of course, it would be wonderful to be able to ensure optimal process management based on objective criteria. For this purpose, Crop Science is currently putting together a central process data archive, the Production Information Center (PIC). The project team, under the leadership of Dr. Karsten-Ulrich Klatt, is supported by experts from the Manufacturing IT (MIT) area at Bayer Technology Services.

One of these is Martin Schmitz. He works in a field within MIT known as Manufacturing Execution Systems (MES). For Schmitz, it is not particularly important to know which chemicals are present in a reactor or how they react with each other. Schmitz is a computer scientist. His interest lies in data. “Data is a treasure trove,” Schmitz knows. Assuming that one can analyze it properly and glean information from it that is not immediately obvious. And this is exactly what the PIC aims to make possible in future for the process experts who work in the production areas of Crop Science. Karsten-Ulrich Klatt is currently one of Schmitz’s most important customers.

“We want optimal production at all times. Together with Bayer Technology Services, we are constantly getting closer to that goal.”

Dr. Karsten-Ulrich Klatt

Manufacturing Systems Technology, Crop Science

As in other areas of life, a tremendous amount of data is produced in the chemical production field. Every pump, every valve, every temperature sensor, every pressure gauge – all of them supply a continuous stream of information. Normally, the control system that regulates the particular process works with this data. In future, the data will be stored at Crop Science in the PIC, where the experts in the project team aim to conflate it with data from batch systems and quality control. “In this way, bit by bit, we want to get closer to the process conditions for the production of Golden Batches.”

But that’s not all: the experts also have the process conditions that could lead to noncompliant batches in their sights. In practice, it happens again and again that, for example, a reactor temperature briefly exceeds the normal range. “If luck isn’t on your side, the formulation has had it, and has to be thrown away,” explains Schmitz. To avoid this, one looks for indicators that precede an undesired increase in temperature. A pattern in the temperature profile, for example. Or it might be a particular mixing setting, a striking raw material characteristic, a certain external temperature: that is, correspondences that only become apparent when one looks more deeply into the data.

Mastering such a complex evaluation requires intelligent methods. “So we provide IT-based solutions which we can use to discover characteristic patterns even in enormous quantities of data,” says Schmitz. If such a pattern is known, then in future one can intervene in the process sequence and, for example, prevent a temperature increase in a reactor – ideally fully automatically via the process control system.

Martin Schmitz (left) from Bayer Technology Services and his project partner Dr. Carsten Welz from Crop Science
An experienced team: Martin Schmitz (left) from Bayer Technology Services and his project partner Dr. Carsten Welz from Crop Science.

Experts like Klatt and Schmitz call this approach “Manufacturing Intelligence” – information is harvested from data and, in turn, is used to come up with concrete instructions for appropriate action. They talk cheerfully about “Smart Data,” meaning a further development of “Big Data.” The “smart” tag points to the fact that the data is not merely gathered, but also intelligently (or smartly) evaluated. In production optimization, this approach is still very new, although it has already proven its worth in subdisciplines such as the life cycle management of plants or individual components. Its capabilities can be shown with one example from a Bayer operation in which pressure sensors were breaking down with unusual frequency, but for no easily identifiable reason. It was only when the plant data was subjected to exhaustive analysis that the cause was discovered: pressure fluctuations in a pipe system, triggered by frequent opening and closing of valves, were responsible.

The project team, which also includes Dr. Carsten Welz from Crop Science, is thinking far beyond the applications for individual plants. The Manufacturing Intelligence approach at Crop Science is intended for global use from the very start. In the PIC, process and other data is gathered from around 30 sites worldwide. This is no trivial matter. For one thing, the local site systems from which the PIC draws its data are very different. For another, the global data transfer has to work securely at all times. In a year, all of the data in the PIC should be available and analyzable with the tools available there.

Then the process experts in every connected production operation of Crop Science will be in a position to optimize the process conditions step by step toward the “Golden Batch.” “Thanks to smart data, Bayer will definitely hit a few bull’s-eyes,” says Schmitz. “And if our treasure trove of data really translates into Golden Batches in the end, then we’re talking about real money.” Karsten-Ulrich Klatt formulates his expectations more soberly: “We’re aiming for optimal production at all times. Together with Bayer Technology Services, we are constantly getting closer to achieving that goal.” 

(*) Required
Your opinion? Comment now