Digital Process Twin cuts J&J production time & costs

Cooking with pharmaceutical ingredients

How does a medication come into being at J&J Innovative Medicine? The lab develops a stepwise chemical plan (a process) to prepare a new active ingredient. Once the lab phase is done, fabrication is ramped up in several phases, from one liter to tens of thousands of liters in the case of commercial production. Development takes place at two Belgian facilities, the Chemical Development Mini Plant (the CDMP, in Beerse) and the Chemical Development Pilot Plant (CDPP, in Geel).

“Making chemical products is a little like cooking,” Niels explains. “You mix ingredients in a reaction vessel, and out comes something new. In the ‘cooking’ process, you have to concentrate a lot on important parameters like temperature, pressure and mixing speed to make sure you always get the right products safely and with reliable quality. We watch those parameters constantly.”

“If we optimize the solvent switch, we save a lot of time and increase efficiency across the board. Any business wants to produce as efficiently as it can, but things get even more urgent where human lives are involved. An optimized switch also helps us use less chemical products. Which is important because at J&J, we try to minimize the environmental impact of our production processes. That’s why we set up a pilot project with the Digital Process Twin from Siemens — optimizing one solvent switch at first, with the ultimate goal of optimizing all switches.”

“After we built the model, we calibrated it with data derived from the process, meaning from real life. With that data, we built a digital application with the gPROMS Digital Applications Platform, or gDAP. That procedure takes place in an open loop. Then we looked at how the model responded to the input from the process control system, and we wound up with a closed loop. The gDAP constantly makes predictions along the way, which it then compares with the actual data. That lets us constantly improve the process.”

“I often compare it with a GPS,” Niels adds. “We’re traveling from a certain composition of solvents — location A, to another composition — location B. The model will guide us from A to B in real time by the shortest route, or the fastest one or the most ecological one. It outputs the ideal route and keeps optimizing it as a function of actual conditions — is there a detour or an accident?”

“On top of that, there are also substantial advantages when you’re setting up new processes. Just optimizing existing solvent switches is very profitable all by itself. And if we can apply the model before we expand production — meaning in the lab — we save even more time and resources. After all, experimenting in the laboratory is much less expensive than experimenting on an industrial scale. So now we’re looking at how we can use lab data to make models that we can apply later for larger-scale production.”

Broad availability of models

“Just recently we also used the Digital Process Twin for a new process development involving freeze-drying. This was a reaction that caused the temperature inside the reaction vessel to rise, yet the temperature also had to be kept below a certain value or the reaction would fail. The process worked in the lab, but expanding to industrial scale makes the parameters change. In that case, normally we would set up a test configuration and experiment until we get the right parameters. But this time we could run the configuration virtually and the model showed that the process just wouldn’t work at all at that scale. So that saved us a lot of time and resources.”