Direct compression modeling

We aid in solving formulation and process development challenges for a direct compression process to upgrade in efficiency, workflow and effort to bring products to market faster.

Download our poster contributions to the 6th APV Continuous Manufacturing Conference below:

Machine learning in direct compression: supercharging process and formulation design with quantitative tools

Facilitating scale-up in Direct Compression: from a small-scale single-punch to a large-scale rotative tablet press via transfer learning

Overview, from data to predictions

Jump straight to more information on our modelling

Case study

The Virtual Process Modeling Suite of Elegent has been created to bring the practical use of high-quality models to your fingertips!

Current functionalities include:

Evaluation of formulation powder flow in a tablet press
In function of a chosen formulation, the probability of achieving good flow in the tablet press is predicted. Good flow is denoted when the predicted relative standard deviation on tablet weight for this formulation is smaller than 2%. Through a ternary diagram, this flow quality can immediately be assessed in function of varying concentration of the formulation components.
Tabletability assessment
In function of a chosen formulation, tablet tensile strength is predicted in function of various critical process parameters such as compression force and tableting speed.
Ejection stress
Similar to the tablet's tensile strength, the ejection stress is predicted in function of critical process parameters such as compression force and tableting speed to minimize the wear of the tableting tooling.

If you are interested in trying out the Suite, do not hesitate to get in touch with us for a free trial.

You can also check out our pre-release version in the video below.

Behind the predictions: Q&A on models

What does the dataset look like?

Combining several data sources proprietary to Ghent University, data of which Elegent is the sole commercial exhibitor, we have created an exhaustive dataset that serves as the base for modelling of flow and tableting CQA prediction.

More than 50 powders were characterized in terms of density, morphology, rheology and other flow- and tabletability-related powder properties.

More than 350 different formulations were composed out of these individual powders, varying in API type, API concentration (up to 60%), filler combinations and ratios, as well as lubricant type and content.

How were the formulations processed?

Formulations were processed using a STYL’One Evolution compaction simulator (Medelpharm, Beynost, France) equipped with cylindrical, flat-faced Euro B punches. A single-paddle feeder with six fingers was used for die filling. The tablets were produced at varying compression forces and speeds.

How are blend properties derived from the raw material properties?

Several strategies are used and continuously evaluated on being the best representation of a mixed powder blend property. This is studied per property that is important for every different CQA that is the subject of modelling. The combination of the best mean type to be used and the format of the fraction for weighting (volume, mass, volume-in-die, ...) are always investigated for their optimal representation of a mixed powder property.

Which modeling technique is used?

Boosted tree classification and regression are currently used to model CQAs at Elegent.

Any other questions?

Please do not hesitate to contact us if you would have specific questions!