Most products are first produced at small scale, such as lab or pilot scale. Subsequently scaling up to full scale production can be time consuming, frustrating and costly.
An effective alternative is to apply numerical optimization on multivariate models that span both scales. This provides an efficient and systematic method to determine the optimal operating conditions at full-scale.
A Fortune 500 pharmaceutical company wanted to investigate the ProSensus approach to scale-up a lab-scale roller compactor/granulator to full-scale. By contrast, the traditional approach would involve the use of similarity scaling factors followed by some degree of trial and error to search for suitable operating conditions at full-scale.
The scaled-up operating conditions were calculated in one step using the ProSensus approach. Unfortunately, time and budgetary constraints did not allow the calculated solution to be run. However, the recommended conditions were a close match to actual operating conditions of a very similar full-scale roller compactor running similar product, which served to verify the metholodology.
The ProSensus Approach
The ProSensus method for scale-up involves a joint multivariate model to predict the effect of operating conditions at both scales, even if the set of measured variables at each scale is different. Numerical optimization is then applied to this model to translate the known operating conditions at one scale to the equivalent conditions at the other scale that will produce product with the same quality characteristics.
In this particular example, a roller compactor is used to convert the pharmaceutical powder blend to granulates that are then stamped into tablets downstream in a tablet press. Powder is fed through a roller and exits as a ribbon that is milled into granules, as shown in the diagram below. The data available is also shown.
The multivariate model structure is a joint-Y-PLS (JYPLS) model. Only the final product properties (Y variables) have to be the same between the lab- and full-scale process. Although the measured process variables may be similar at both scales, they may not be exactly equivalent due to differences in sensor location and scale effects. As a result, the process conditions (Xa and Xb) are treated separately in the JYPLS formulation. The schematic of a scale-up problem is shown below, as well as the results of one of the two predicted ribbon variables. The joint model is able to predict the quality well for both machines. Using this model, optimization methods are employed to determine the operating conditions of the scaled-up compaction process that will result in the desired ribbon properties.
As mentioned earlier, the calculated operating conditions closely matched the known operating conditions for a very similar machine running very similar product. This demonstrated that the ProSensus method was effective in calculating appropriate operating conditions for full-scale in one step.
1. Z. Liu, M.J. Bruwer,J.F. MacGregor,S.S.S Rathore, D.E. Reed, M.J. Champagne, “Scale-up of a Pharmaceutical Roller Compaction Process Using a Joint-Y Partial Least Squares Model”,Ind. Eng. Chem. Res,50,10696-10706,2011