Production Lead-Time Estimation Of Variant Parts In OKP Enterprises Based On RBF Neural Network

OKP( One-of-a-Kind Production ) can quickly satisfy needs from continuous increasing consumer orders and market competing, and stands for a direction in Mechanical manufacturing field. Application of OKP needs production management and control system of enterprises keeps up with changes of markets outside, and quickly reflects those changes to satisfy order needs. But in our OKP enterprises, ERP,MRP can not get enough data supports because of inefficient data management in production, and their reflection to market can not effectively work. So it is one of important questions to our OKP enterprises that how to make our ERP, MRPII get enough data support to improve enterprise abilities of quickly reflection to market.This research bases on a typical OKP Mechanical manufacturing enterprise, and tries to resolve the question of estimating production lead-time, which is a core question effecting its changing time and quantity of MRP in ERP system. The study selected RBF Neural Network, which has drew more and more attention in AI field, used questions about estimation of Production Lead Time in variant parts as a material for research, integrated theory clustering analysis and MATLAB tool, and pre-processed original data to select those representative parts as the material for disposal. The RBF network's input variants were factors affecting Production Lead Time, and output variants were PLT data. The RBF training method used improved K-means clustering method to select Gauss centers, and used Gradient-descent algorithms to train the position and radius of these centers and output weights.Practice result shows the RBF Neural network based on the improved k-means clustering algorithms can estimate PLT data of variant parts relatively exactly, brings scientific data guarantee for MRP making in enterprises under OKP, and develops the application range of RBF theory.