Food & Beverage Plant Background
In Planning and Scheduling, the “devil is in the details” – unless all relevant details of the production environment are accurately modeled and solved, the resulting plans and schedules will be sub-optimal, or even unbuildable. A case in point is a Food and Beverage (F&B) plant with a complex production environment that Eyelit Technologies can model and solve in necessary detail. In contrast to simple heuristic scheduling systems, the schedules generated by Eyelit Technologies achieve significantly higher utilizations and reduced makespans.
The organization of production in the F&B plant is shown in the schematic above. Among the many real-world details that need to be considered in scheduling this plant:
- The plant produces over 150 SKUs with each product requiring 4 distinct operations.
- Mixed batch and continuous flow production: The first stage (Operation10) is batch production that feeds semi-finished goods (SFGs) into storage tanks. The remaining three stages (Operations 20, 30, 40) are continuous flow.
- Because of continuous flow production, the resources required for Operations 20, 30, and 40 must be available simultaneously, and are held for the duration of production.
- There are 28 resources in all, and a given operation may require 1 or 2 resources.
- Another consequence of continuous flow production is that the effective production rate is throttled down to the rate of the bottleneck.
- The bottleneck can vary from SKU to SKU.
- Sequence dependent changeover times, that can also vary by resource and SKU pairs. The business goals for the scheduling process are to reduce the makespan required to produce a given set of orders and maximize resource utilizations.
Eyelit Technology vs. Legacy Production Schedule
All relevant aspects of the production environment can be captured within Eyelit Technologies and further, the resulting model can be solved to generate high-quality schedules using Eyelit Technologies’ patented AI and optimization based technology. In contrast, legacy systems lack descriptive power, and details such as throttling, or sequence dependent changeovers are simply skipped. Legacy systems also lack sophisticated solvers, and instead rely on simple heuristics to generate schedules.
The contrast between the schedules generated by simple heuristics and Eyelit Technologies could not be clearer.
Legacy Production Schedule – Gantt Chart
The figure above is a Gantt chart of a schedule generated by a sorting heuristic that seeks to minimize changeovers. As can be seen, the schedule has a large amount of forced idle time and a makespan that exceeds 4 weeks. This is because the legacy system did not consider bottleneck resources, alternate routings and wait times for all required resources to become available. Most of the inefficiencies arise not due to changeovers but due to starvation of resources waiting for other required resources to become available.
Eyelit Technology Production Schedule – Gantt Chart
The figure above is the Gantt chart of the schedule generated by the Eyelit Technologies solver after all relevant details of the production environment were accurately modeled. Eyelit Technologies reduced the makespan from 4 weeks to less than 2 weeks.
Secondly, there is far less wait time for resources to become available, and many of the resources have utilizations in excess of 90%. This shows the importance of considering all relevant detail in a scheduling model.