A neural network clustering model for miscellaneous components production planning

Abstract

The growing importance of providing service to customers, e.g. post-sale assistance, supplying of spare parts, upgrading and integration of new elements in installed systems, enhances the importance of planning and management of upgrading parts in most manufacturing industries. These parts are generally characterized by high technical heterogeneity and have a highly variable and difficult to forecast demand. In some areas (especially the most dynamic, e.g. high-tech products), these kinds of components are quite common, and represent a very strong relation between the manufacturing firm and the market. These parts are generally too many to be efectively supported on a planning database system with individual records and too heterogeneous ( and sometimes with a too high value) to be supported all together in a single record. In this paper, we want to study the application of adaptive techniques for the clustering of these components in classes based on the similarities in their market behaviour in order to build an optimal database for planning production and supplying of these components.