Production scheduling of multipurpose batch plants. Application to a case of production plant for active pharmaceutical ingredients (API’s)

Abstract

Scheduling is a key topic on production. On manufacturing and engineering, scheduling is the planning of the production or operation in a factory. It is of special importance on multipurpose and multiproduct plants, where scheduling can have a major impact on the productivity of the process. In this work, our main objectives are, in one hand, the analysis of the different mathematical models developed to find the optimal production scheduling in order to give selection criteria depending on specific problem, and, in the other hand, to use selected models on a real-world case of a multipurpose plant in order to show the magnitude of the problem. In order to classify the optimization models, we used as criteria the event representation of them, as it’s the aspect that defines the approach of the model. For each event representation mentioned on this work, an example extracted from the bibliography is shown and the main characteristics of the event representation are explained. After that, a real-world problem of a multipurpose plant producing both rifaximin and fosfomycin tromethanol is solved using a self-developed model using as event representation the global time points, which is the more appropriate for the problem as it’s explained later. Using computational software (Wolfram mathematica) we try to solve the problem, but it’s so large that for computer limitations it can’t be solved. A simplified version of the model is proposed and solved with the same software. Results show that optimized schedules can produce an extra batch of fosfomycin when two batches of rifaximin are being produced, while on the other instances there is no difference between optimized and non-optimized schedules. As the horizon time is only of two weeks, this little difference is enough to show that the optimization have a profit. Also a huge gap is observed between the number of possible batches on simplified and non-simplified models when production lines are being used, which suggest that the simplification leads to worse results.