This graduation project analyses the fully automated (bulk) pallet hall of EDCO Eindhoven B.V. by means of simulation.
EDCO Eindhoven B.V., founded in 1978 as an import/export company, has since that time earned itself a unique position as a global trading corporation. Starting as a small convenience store at a petrol station, it has grown into a global trading company with sales offices and showrooms spread over Europe and Asia. EDCO is known for its particular wide assortment, consisting of more than 20,000 products, including household goods, electronics, garden-, bicycle-, sports- and seasonal accessories, toys, tools, car- and truck accessories, as well as licensed articles.
One of EDCO's focal points is its stock availability: nearly all articles are always on hand and may thus be immediately delivered to customers, leading to a particularly high customer service level. However, the large amount of stock also implies a high demand for storage space.
Over the past years, EDCO's sales increased, and the storage capacity in its main warehouse in Eindhoven became insufficient. This led to the use of external warehouses. In order to save on expenses related to this storage allocation and in order to facilitate inbound, outbound and picking operations, management decided to build a new warehouse. As this warehouse is capable of storing more articles, the majority of the storage agreements with external warehouses could be terminated.
In EDCO's new warehouse, half of the bulk stock (boiling down to approximately 75,000 pallet positions) is stored in a fully automated environment, controlled by an Automated Storage and Retrieval System (AS/RS). Except for the reduction in labour costs, the introduction of an AS/RS is supposed to lead to other advantages, such as increased reliability and lower error rates. EDCO's AS/RS is particularly interesting, as it makes use of stationary pallet racks in combination with APMs and aisle-changing high rack stackers.
In order to achieve a high performance, EDCO wants to keep the number of aisle changes as low as possible. Additionally, they want to exploit the possibility of creating dual command cycles (cycles which include both a storage as well as a retrieval operation).
However, in order to create dual command cycles, the picking operations of some loads need to be postponed if there is currently no opportunity for a dual command in the associated aisle. This postponement could cause problems for orders waiting in the shipping area or replenishments to the case picking area. EDCO's desire to reduce the number of aisle changes to a minimum only amplifies these issues.
Hence, a careful balance is to be found between efficient order picking and time-wise acceptable pallet arrivals in the shipping and case picking area.
Taking the aforementioned arguments into consideration, the problem statement is defined as follows:
How could EDCO arrange the pallet picking process in its fully automated hall as efficient as possible, taking into account the requested pallet arrival times in the shipping and case picking area?
The objectives of this graduation project include advising EDCO about operating and storage policies, as well as discovering relationships or trends between the several input factors and performance indicators of the AS/RS.
There are two main reasons why it has been decided to model the problem by means of simulation. First of all, EDCO's AS/RS is highly complex. Building an analytical model or solving a model analytically would imply too many (over)simplifying assumptions, making the model unreliable. Secondly, it is desired to use EDCO's actual data. In most warehouses, and so in EDCO's, no day is the same and hardly any day can be described by the "average day". Moreover, a policy which works well in the average situation might be disastrous on a real working day. An analytical model is not able to deal with this, since it needs fixed data, such as fixed numbers or a fixed probability distribution. On the other hand, in simulation models, a company's actual data may be used and no concessions are necessary to be made.
After the data was collected, the simulation model was written. Test runs were performed in order to validate and fine-tune the model. Eventually, the replication size was determined and the scenarios could be simulated. Some interesting results of this graduation project include:
- Allowing high rack stackers to work ahead by not only picking pallets belonging to the current wave, but also to later waves, strongly increases their efficiency. However, this efficiency gain comes with a price, as the number of required buffer locations increases as well.
- An increase in the hall's fill rate slightly decreases the high rack stackers' efficiency. However, this is a minor decrease, confirming the robustness of the hall against fill rate changes.
- The spread of (identical) inbound pallets over the storage racks and aisles appears to be relatively stable. However, it slightly increases if the hall's fill rate increases or if the number of trucks decreases.
- There exist multiple possibilities for dealing with busy waves or large pallet flows. Except for purchasing additional material handling equipment, it is also an option to increase the number of buffer locations, such that vehicles are allowed to work "ahead" (which subsequently increases their efficiency). Naturally, one could also simply accept the fact that (some) pallets are delayed.
Overall, the simulation model is able to give predictions about efficiency, idle rates, required number of buffer locations, order delays and spread of pallets. Adapting the model in order to simulate (forecast) the situation of 2015, revealed a potential bottleneck (as illustrated).
Katinka van de Velde, MSc in Operations Research & Management Science - Tilburg University.
