(745f) Modelling Liquid Drug Product Manufacturing: Method and Case Study

Authors: 
Eberle, L., Swiss Federal Institute of Technology, Zurich (ETHZ)
Capon, E., Universitat Politecnica de Catalunya
Sugiyama, H., The University of Tokyo
Tien, G., ETH Zurich
Graser, A., F.Hoffmann-La Roche
Schmidt, R., F.Hoffmann-La Roche
Hungerbuehler, K., Swiss Federal Institute of Technology, Zurich (ETHZ)

Modelling drug product manufacturing processes both for optimizing production schedules as well as for simulating process enhancements is an increasingly important topic in times of rising economic pressure on drug manufacturers. In real-life, production schedules resulting from the execution of drug production manufacturing models can only be consulted reasonably, if a consolidation of all relevant production constraints is achieved in one model. Such constraints include demand, shelf-life, due date, workforce availability as well as utilities and equipment limited resources. On top of these constraints, production schedules are often rescheduled due to product reprioritizations, raw material scarcities and holding time limitations, e.g., sterile holding times. Hence, optimal scheduling on the shop floor level is a very complex task, which can be a cornerstone for increasing production efficiency.

In this work, the aforementioned challenges have been solved and a slot-based formulation is presented that was inspired by, and was applied to a case study of liquid drug product production line. While considering demand, shelf-life, due date, product-dependent processing rates and sterile holding times, the formulation allows finding the optimal batch order with respect to makespan minimization. This goal is achieved by the introduction of two operation modes, similar to the concept described by Moniz et al. (2014). The model includes batches produced individually, and the more time-efficient variant, where multiple batches are produced within a campaign, i.e., cleaning and sterilization activities are reduced. A second key feature of this model is the complexity reduction of modelling the filling processes for liquid products. Such fillings are performed in several hundred working steps; however, grouping the filling tasks into process elements that need to be performed without major interruptions, e.g., weekends, allowed an efficient modelling with only six process elements.

As an industrial case study, the proposed model was applied for scheduling a filling line for liquid drug products of Roche in Switzerland. Results were used to validate the model and optimize a three-months production schedule; including orders of various biopharmaceutical drug products and respecting phases of technical services as well as media-fills. Production plans resulting from the application of the scheduling model withstood an expert evaluation on executability with real life constraints, which provided the successful validation of the scheduling algorithm. The makespan for batches as scheduled in the sampled production schedule by company experts and the schedule provided by the presented algorithm differ by about 2 % of the overall planning horizon duration. Such a difference of few percentages on the line entails important benefits to the processing plant.

References: Moniz, Samuel; Barbosa-Póvoa, Ana Paula; de Sousa, Jorge Pinho (2014): Simultaneous regular and non-regular production scheduling of multipurpose batch plants: A real chemical–pharmaceutical case study. In Computers & Chemical Engineering 67, pp. 83–102. DOI: 10.1016/j.compchemeng.2014.03.017.