【摘要】 利用贝叶斯网对印刷线路板(PCB)微小孔钻孔工艺的孔壁粗糙度进行建模。首先通过鱼骨图对钻孔质量的影响因素进行整理和筛选,确定可用于建模的因素;然后采用灰色关联法对各因素与孔壁粗糙度的关系进行分析,并以此为依据精简次要因素和建立贝叶斯网模型结构;最后在实验数据和生产线采样数据的基础上对模型的条件概率表进行学习与进化。经过检验,模型在进化过程中精度能不断提高,并且当数据量较少的情况下,贝叶斯网能够获得比线性回归及BP神经网络模型更高的精度。更多还原

【Abstract】 Bayesian-Network was used to modeling the hole-wall roughness occurred in PCB micro-drilling.First,the influence factors of drilling quality was arranged and selected with the help of fishbone diagram.Second,Grey-Relation method was employed to analyze the relationships between hole-wall roughness and factors.Then the minor factors were removed and the model structure was built accordingly.Last the Conditional Probability Table was learnt and evolved with the data obtained from organized experiment and product line sampling.It was verified that the prediction accuracy increased with the model evolving.The accuracy of Bayesian Network was higher than that of linear regression and BP neural network in conditions of small data amount.更多还原