(423c) Sequential, Simulation-Based Manufacturing Execution System for the Planning of Pharmaceutical Production

Fast development timelines for new medicines call for accelerated engineering workflows that enable evaluation of manufacturing options while ensuring product quality, patient safety and uninterrupted supply. This evaluation can involve resource planning and operation forecasting (equipment, personnel, materials) to fulfill the demand of a product in a timely manner.

Different approaches exist to study production and operations planning. Scheduling and resource allocation optimization approaches based on mixed-integer problem formulations are a wide area of research [1]. A common approach to modeling the unit operations in these contributions is that of using coarse models of low computational cost that neglect the phenomena internal to the process. Therefore, forcing the analyst to make assumptions about the distribution of disturbances that may affect the planning and scheduling, even when these variations can have a physical cause associated with them.

Another less common approach is the use of process simulators coupled with planning tools as an upper decision layer. In this case, task durations as well as the internal behavior of the processing units are retrieved after solving material and energy balances. This not only helps with operation planning, but also to process understanding as information of the physics and chemistry of the process becomes available. An important feature of the said simulators is the ability to deal with a combination of continuous and batch/semi-batch operations, as is often encountered in the pharmaceutical industry. To this end, we leverage the numerical library PharmaPy [2] to model the operation (resource consumption, schedule, timings) of a chemical manufacturing plant.

In this work, we propose a methodology to simulate the operation and scheduling of an entire chemical plant given a series of models for each unit operation involved, and the information about the resources available and those necessary for each unit operation to be able to run. In our approach, a Manufacturing Execution System (MES) is simulated using a sequential-oriented engine to manage the execution of subsets of unit operation models involved in the flowsheet, as opposed to an equation-oriented approach that solves the entire flowsheet simultaneously. To this end, the transfer of material across the different material reservoirs and operations in the flowsheet is managed by the simulated MES by following pre-established logical rules about the operation. This approach enables us to study the effect of the variations of process consumables (e.g. quality of raw materials) and available resources (number of intermediate storage units) onto the overall operating schedule of a facility.

Our methodology is demonstrated by analyzing multiple process configurations of a hypothetical flowsheet for the manufacture of an active pharmaceutical ingredient (API), where a continuous synthesis step is followed by batch separation and purification. The fictitious process includes multiple batch processing units and storage tanks, for which the MES decides task assignment and process timing/sequencing. Of particular interest, scenarios of increased throughput for a fixed plant configuration, as well as alternatives with less or more processing or storage units are analyzed and compared based on makespan and mean throughput. While in this contribution we utilize a concocted scenario to illustrate the features of the methodology, the authors have applied this technique to aid the study and optimization of a real manufacturing train.

[1] Harjunkoski, I., et al., Computers & Chemical Engineering, 2014. 62: p. 161-193.

[2] Casas-Orozco, D., et al., Computers & Chemical Engineering, 2021. 153.