摘要
对模糊自适应共振理论(FART)神经网络进行改进,使用改进的FART神经网络对半导体生产线晶圆合格率进行在线检测,对晶圆合格率特征向量进行聚类分析,将合格率损失中拥有相类似特征的晶圆分为一类,一旦检测到生产线发生异常,便可找出故障设备并及时维护,从而使生产线处于高生产率状态。
The fuzzy adaptive resonance theory neural network （FARTNN） was improved to monitor the semiconductor manufacture line on-line. The wafers having similar yield loss characteristic were classified into the same class via clustering analysis of wafer yield characteristic values. Once manufacture line abnormity is detected, the failure machine will be repaired timely, and so the manufacture line can keep high productivity.
出处
《实验室研究与探索》
CAS
2008年第11期6-9,共4页
Research and Exploration In Laboratory
基金
上海高校选拔培养优秀青年教师科研专项基金(科07-43)
校博士启动基金
关键词
模糊自适应共振理论神经网络
半导体生产线
聚类分析
fuzzy adaptive resonance theory neural network
semiconductor manufacturing line
clustering analysis
