One of the absolute biggest challenges in drug filling lines is to ensure consistency—not only across batches but also within batches. Issues such as foaming of the product within the container, dripping through the needle, and wicking—especially with hydrophobic and low surface tension products—make it hard to maintain a consistent fill level across products. Product characteristics like density, polarity, and surface tension all greatly impact the ease of filling a product and, when using the same machinery to fill more than one product type, it can be difficult to correctly set up the equipment for every changeover.

Underfilled products obviously are lead to higher rejection rates while overfills outside of normal tolerance ranges could have been used to fill other units and thereby increase the yield of the production. Thus, both scenarios are rather undesirable and, as a result, mitigating these issues should be mitigated in order to maintain high product quality, patient safety, and economic gain.


Similar to filling, producing tablets and capsules of high quality at consistent rates is increasingly challenging. Many of the most troublesome problems that manufacturers encounter, incl. insufficient tablet hardness, inconsistent tablet weight, incorrect flow of formulation through the feed system, and tool damage, can often be corrected by making adjustments to the tablet press or its systems.1 Though, figuring out what adjustments to make and at what times is hard and even very experienced operators often do not get it right.

Source: Wikipedia

In both the case of filling and the case of tableting, making use of AI models to continuously monitor and analyze parameters produced by the machinery can help improve quality levels, lower rejection rates, increase patient safety, and boost gross margins.

AI models can suggest recommendations for how to improve the operation in real time and either present the recommendations to operators who can take action or directly implement them in a closed-loop system. AI models trained in Criterion AI can be exported to run in on-premise environments and, by making use of industry standard protocols such as OPC UA, models can easily be integrated with existing equipment.

With Criterion AI you can

  • Train AI models to control filling and tableting processes
  • Mitigate under- and overfilling to improve product quality
  • Reduce variations in tablet weight, friability, and hardness
  • Empower operators to make better decisions in real time
  • Integrate models directly with existing equipment

There is a lot of potential in improving filling and tableting processes with AI, so please reach out to us today to start the dialog on how Criterion AI can help you boost your operations.

This website collects usage data to help us improve your experience of the site. Click here to accept or here to opt out.