Daniel Sherwood - Optimising a factory using discrete event simulation

🏠 Home

Optimising a factory using discrete event simulation

In this study, I lead a team of students in investigating the possibility of optimising a factory producing scissor jacks. We chose two conflicting KPIs: production cost and delivery reliability. This choice was based on a predetermined production scenario: supplying jacks to several compact car manufacturers in France. The production consisted of several operations illustrated below:

We simulated the factory using two discrete event simulations in Anylogic. One simulated the production of jacks over the course of a day, accounting for employee absences.

The second simulated the delivery of jacks to customers on a weekly basis, accounting for machine downtime.

We discovered that the the quantity and allocation of workers influenced both KPIs, resulting in a multi-objective optimisation problem. We observed that the delivery reliability was highly sensitive to the mean daily production of jacks and variability in the order quantity. In contrast, the variability of daily production is not a significant factor in delivery reliability.

We performed a heuristic parameter variation experiment to quantify the effect of altering the allocation of workers on both KPIs. From this we identified a non-dominated solution. For a given value of reliability, we achieved a cost saving of approximately C0.12 per jack, compared with other possible solutions for the allocation of workers.