Determination of Batch-Sizes
A case-study in the process industry
Viktor Farbäck & Simon Gottfridsson Lund University, Faculty of Engineering (LTH)
Picture the scenario; you work at a company which produce around 400 types of drinks, in various bottles and selling units. Your responsibility is to plan which beverage to be produced. Today’s assignment is to plan the production sequence for 10 different kinds of beers. How should this be done in a cost-efficient way?
You know approximately, but not exactly, how much of these beers that are demanded every week. Every time you want to change the kind of beer the machines are producing, there is a cost
associated to it. For each day you store the beer that you have produced but not yet sold, there is also a cost. Add to this scenario that the products have an expiry date which is different for each beverage. Products that not are sold by this date, must be thrown away. The question now becomes, how much of each beer should you produce every time you have the machine set up for a product? You don’t want to produce each type too often, because of the cost when changing products. But you don’t want to produce it too seldom either, as the amount you must carry in stock each day is costly, and there is a risk that you produce goods that perish.
But what if the buyer wants to buy more than expected of a certain beer? What if the yield from the production does not meet the expectations? There need to some sort of mechanism in place to protect us from these uncertainties, to minimize the risk that we cannot satisfy the customer demand from our stock.
During our thesis, we have developed a quantitative model that will help the planner to make these decisions. The model will not only tell you how much of each beer to make every time, in a more cost-efficient way, the suggestion is also based on operational constraints and the risk for products to perish.
Since you don’t know exactly how much beer the customer will buy, the model also helps you to decide how much safety stock you will need, in case the quantity is not enough until the next
production time. The model contains a way of calculating this safety stock that incorporates both the variations in yield and customer demand, while also recognizing the reductions in risk that a greater batch size provides.
When we have tested the created model, we have been able to generally increase the batch sizes, thus increasing the performance of the production facility. Since our new way of calculating safety stock incorporates the positives of greater batch sizes, the result doesn’t drastically increase the average stock levels.