中文摘要：

为准确地进行定量预测，提出了一种将仿真分析和集成径向基网络模型结合起来的制造系统性能指标预测方法。在定义和量化制造系统各类性能指标的基础上，分析了影响这些指标的静态和动态因素，并建立起径向基集成网络预测模型。通过基于Simul 8平台的仿真分析来收集样本数据，最终利用Bagging方法训练出集成神经网络，实现对工件平均完工时间和设备利用率等关键性能指标值的预测。试验结果表明，采用该方法输人动态影响因素的取值后，能快速获得比较理想的性能指标预测结果，并且其预测精度明显高于其他的神经网络方法。

英文摘要：

In order to predict exactly, an approach for performance index prediction of manufacturing system, which combined simulation analysis and radial basis network model was proposed. After defining and quantizing several performance index of manufacturing system, the influencing factors which were either static or dynamic were analyzed, and the radial basis network prediction model was constructed. The sample data were collected by the simulation process on the Simul 8 platform. Finally, the ensemble neural network was trained by the Bagging approach, by which some key performance index such as the mean throughout time of parts and the utilization ratio of equipments were predicted. The experiment showed that by this method, the satisfactory prediction results of performance index could be obtained rapidly by inputting the values of the dynamic influencing factors. Furthermore, the prediction precision was obviously superior to that of other neural network methods.