Within a logistic chain, a considerable amount of goods are physically moved from manufacturers to end users. During this process, these goods may be stored in warehouses for a certain period of time in order to satisfy customer service level. In many local distribution centers (DCs), products stored in slot of racks are picked and distributed in less than unit load quantities. These are the so called order picking systems (OPS): order picking is the process of retrieving products from a storage area in response to a specific customer request.
It has long been identified a very labor intensive operation in manual systems, and very capital intensive activity in automated system. In particular it may consume as much as 60% of all labour activities in a warehousing system. And for a typical warehouse, the cost of OP is estimated to be as much as 55% of the total warehouse operating expense.
In last decades, warehousing systems mainly focused on managing raw materials, in process products and finished goods in storage areas. In particular they were generally considered to be disconnected with the operating context. With growing importance of supply chain management, these systems have changed their role to strategically achieving the logistic goal of shorter order cycle times, lower inventory levels, lower costs and better customer service. The order processing activities in higher modern warehousing systems have to be faster than in the recent past. As a consequence products may stay in warehouses for just a few days or even a few hours. For these reasons, the main objective of an OPS is to minimize the total distance travelled by pickers and the system throughput (the travel time is an increasing function of the travel distance).
The aim of this study is to develop, test and compare a set of different storage allocation rules based on the application of original similarity coefficients and clustering techniques.
The proposed systematic approach gives managers and practitioners an effective method to solve the Order Picking Problem (OPP). The approach can be outlined in two consecutive processes: the first named family grouping and the following one storage allocation. Both processes concern two steps:
· correlation analysis and clustering for the family grouping process;
· priority list and storage position for the storage allocation process.
Each phase can be achieved by a suited management tool, that is a set of strategies, parameters and techniques. The approach gets information about customer orders in input, and using the proposed management tools for each phase, generates the storage locations for products in output. These positions can be used to simulate OP orders and to evaluate the key performance indicators target.
Family grouping process
The first phase deals with the determination of the correlation, also called similarity, between the products of the product mix. There are many measures to quantify the similarity between pairs of objects. This step influences the shape (i.e. the configuration) of the clusters, as some products may be close to one another according to one measure and further away according to another measure. The measures of correlation between products is generally obtained by some similarity indexes. In particular there are two main kind of indexes:
· general purpose index
This kind of index is uniquely based on the information about the belonging of different products to common picking orders (so called belonging frequency information - BFI). As a rule, this information are formalized on the incidence matrix, a Boolean representation of the presence of the products in the different orders.
· problem oriented index
This type of index evaluate the similarity of two products in addition to the incidence matrix information using some peculiar measurement of suitability for the problem (e.g. production and logistics information). An original problem oriented similarity index is illustrated in next sections. The second phase of family grouping process is about the clustering analysis. Clustering is the organization of products into different clusters (or family), so that the products in each cluster have high values of similarity/correlation. The clustering algorithms are the specific management tools for executing this phase. These algorithms can be hierarchical or partitional. Hierarchical algorithm group successive clusters using previously established clusters, whereas partitional algorithm determine all clusters at once. The traditional representation of this hierarchy is a diagram called dendrogram (similar to a tree diagram), with individual products at one end and a single cluster containing every products at the other.
Aditya Vairagkar
SIBM Bangalore
Operations Batch of '11
SIBM Bangalore
Operations Batch of '11
No comments:
Post a Comment