Research on Optimization of Raw Material Inventory Decision Based on Clustering of Part Dimensions

With the increasing trend towards large-scale, multi-category customized production of mold steel products, the dimensional differences between products have made it difficult for single-specification raw material cutting methods to meet the needs of large factories. Therefore, a solution is needed to address the problem of multi-specification raw material cutting in mold steel processing enterprises, particularly focusing on how inventory decisions based on raw material specifications can improve plate utilization under specific three-dimensional nesting constraints. This paper aims to address an inventory decision problem within a production cycle. Starting from the part dimensions and cutting processes, and considering the dimensional differences of parts, a clustering approach based on the Self-Organizing Map (SOM) and K-means algorithm is employed. The K-means algorithm is used for secondary clustering of SOM neuron weights to identify parts with similar features. The cluster centers are used as the specifications for raw materials, effectively solving the problem of low cutting utilization and high variability encountered when making inventory decisions based on manual experience. Verification using actual factory data demonstrates that the inventory decision method based on part dimension clustering can significantly improve the utilization rate of block raw materials.