Industrial engineering juniors hit home run with Unilever project
No one expected stunning results when five industrial engineering juniors took on a project for Unilever Foodsolutions of Milwaukee in 2009. The project seemed simple enough: try to improve a scheduling process that already appeared to be running efficiently. It was not clear that any improvement was even possible. So when the final results showed an opportunity to eliminate more than 16 hours of non-productive changeover time per month, there were quite a few surprised expressions in the room.Unilever Foodsolutions produces a variety of dressings and sauces, packaged in one-gallon plastic containers for institutional use. The facility processes and packages as many as 108 different products per month. Ingredients are mixed in 2,000-gallon tanks, and the liquid product is pumped to a bottling line where the plastic containers are filled.
Each of the 108 different products is manufactured about once per month. After each product is made and bottled, it is necessary to shut down the line and clean the equipment before the next product can be manufactured. This down time, or changeover time, is non-productive and totaled about 110 hours per month.
The goal of the students’ project was to reduce these changeover times, if possible. Jesse Broaddrick, Alex Gerdmann, Mason Josie, Theresa Wenszell and Ben Wheeldon recognized this as an optimization problem with multiple constraints. Four different types of cleaning can be done, depending on how similar or different the next product is from the preceding product. The actual inter-process cleaning times are 38, 48, 69 or 77 minutes. The students’ major challenge was to find the optimal sequence of products to minimize total changeover times while still producing all required products. Actual changeover times for all combinations of processes were well defined.
Simply stated, the goal was to use as many short-time changeovers as possible, while avoiding as many high-time changeovers as possible.
The students chose to use mathematical programming since it would seek a minimum-time solution while recognizing existing real-world constraints. In addition to the four different cleaning times, the students also needed to decompose the month’s schedule into weekly blocks and recognize product groupings by shipper. These additional constraints modified the problem into nine individual sub-problems which ranged from four to 31 products to be sequenced in each group.
Their first attempt to solve this problem led to initial success when considering only four or five products, but the technique proved unworkable on large problems. After further investigation, they found that the factory’s scheduling could be classified as a “traveling salesman problem,” which has a known solution technique. Josie found shareware software on the Internet, downloaded it and demonstrated how it could address the scheduling issues. However, implementation required a great deal of manual effort to input data from Unilever’s system to the solution software. The team then found and recruited another student to write Java code which demonstrated how this could be done.
The traveling salesman solution approach provided a substantial improvement in the facility’s scheduling—more than 16 hours of non-productive time per month were eliminated. This represented a significant cost savings, or increased production, for Unilever Foodsolutions. It also demonstrates the students’ ability to apply the operations research knowledge gained during their industrial engineering coursework at MSOE.
