Optimization of Production and Packaging Schedules in a Mixed Discrete/Continuous Manufacturing Environment

This research was driven by the need for a more efficient production scheduling system in a consumable liquid product division of a large consumer products company. The manufacturing process under inspection consists of continuous and discrete elements, on both production and packaging lines. The production lines are split into continuous production lines and batch production lines which produce the product in fixed batch amounts. Then there are several bottling lines, some of which package a particular bottle size and others that can package multiple bottle sizes. The main objective of this research was to reduce the amount of time it takes for the client to create production and bottling schedules. An optimization model was developed to automate this process and provide the client with the best possible schedule. The objective of the model is to minimize cost by minimizing the number of switches across the production and bottling lines, as well as minimizing the amount of overproduction. Inputs into the model include model parameters, like the number of shifts to schedule, and monthly demand numbers for each stock keeping unit (SKU). The variables being solved for are the amount of each flavor to be produced across the production lines during each shift, and the number of bottles of each SKU to be bottled across the bottling lines during each shift. Due to the unique constraints and resources of the client, a custom formulation using mixed integer programming was necessary to achieve these objectives. Overall, our model fell short in some areas but succeeded in others. Our analysis showed that the model had a 13% average decrease in production switches but an 87% average increase in bottling switches compared to the current manual scheduling system. However, the ability of our system to create a good enough initial schedule reduces the time it takes expert human schedulers to develop a final schedule by up to 85%. Runtime and computational constraints barred us from creating an optimal, cost-minimized solution for our client, and future work can be directed toward solving these issues.