• 主题
高级搜索  >
历史搜索 清空历史
首页> 外文会议>World multi-conference on systemics, cybernetics and informatics >A Hybrid Data Driven Approach for Cycle-Time Forecasting in Semiconductor Wafer Fabrication System
【24h】

A Hybrid Data Driven Approach for Cycle-Time Forecasting in Semiconductor Wafer Fabrication System

机译：半导体晶圆制造系统中周期时间预测的混合数据驱动方法

获取原文
页面导航
  • 摘要
  • 著录项
  • 相似文献
  • 相关主题

摘要

Estimating the cycle time of each job in a wafer fabrication factory is critical to help semiconductor manufacturers keep the promises of a good delivery-time and thus improve the level of customer service. Since massive data can be captured by millions of networked sensors, which are embedded in the Semiconductor Wafer Fabrication System (SWFS), this paper describes a hybrid data-driven approach for analyzing large amounts of manufacturing data to forecast the orders' CT in the SWFS. Firstly, a regression-based feature selection model which takes account of massive parameters is proposed to obtain the correlation between order related variables, system condition variables and CT. A Fisher Z-transformation is then applied to combine different correlations into a single measure of the candidate variable's performance. Variables with highest mean Z-transformed correlations are deemed to be 'CT-related'. Secondly, a case-based reasoning method is designed to evaluate the order's dissimilarity with historical orders to search the best fit cases. At last, the numerical experiments show the proposed approach has higher accuracy than BPN in CT forecasting.
机译：估算晶圆制造厂中每个作业的周期时间对于帮助半导体制造商兑现良好交货时间的承诺，从而提高客户服务水平至关重要。由于海量数据可以被嵌入半导体晶圆制造系统（SWFS）的数百万个网络传感器捕获，因此本文描述了一种混合数据驱动的方法，用于分析大量制造数据以预测SWFS中的订单CT 。首先，提出了一种基于回归参数的特征选择模型，该模型考虑了大量的参数，以获得阶序相关变量，系统条件变量与CT之间的相关性。然后应用Fisher Z变换将不同的相关性组合为候选变量性能的单个度量。具有最高Z转换平均相关性的变量被视为与“ CT相关”。其次，设计了一种基于案例的推理方法来评估订单与历史订单的不相似性，以搜索最适合的案例。最后，数值实验表明，该方法在CT预测中具有比BPN更高的精度。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱：kefu@zhangqiaokeyan.com

京公网安备：11010802029741号 ICP备案号：京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有

  • 客服微信

  • 服务号