关键词: 订单完工期；预测建模；神经网络集成；仿真模型

Abstract:

An empirical regression predictive method was proposed to address the order due date prediction problem in the discrete production system with multiple manufacturing resources and multiple product types. First, two main classes of factors affecting the order due date, the real-time state of workshop and the components of order, were determined and quantitatively described. Based on the platform of ExSpect, a high-level Petri net simulation model for production process was constructed. Through stochastic simulation and simulation execution, a lot of sample data sets were collected by which many individual neural networks were trained. By using the improved Bagging method based on cluster analysis of prediction errors, the ensembled neural network was set up. Finally, the whole process of order due date prediction was discussed through a case study. The comparison of the predictive values with the simulation results shows that by using this method we can obtain perfect predictive results.

Key words: order due date, predictive modeling, neural network ensemble, simulation model