NSF Young Investigator: Toward Neural Network Based Modeling, Control and Diagnosis of Semiconductor Manufacturing Processes

NSF 青年研究员：面向半导体制造过程的基于神经网络的建模、控制和诊断

• 批准号: 9358163
• 负责人: Gary May
• 金额: $ 27.5万
• 依托单位: Georgia Tech Research Corporation
• 依托单位国家: 美国
• 项目类别: Continuing Grant
• 财政年份: 1993
• 资助国家: 美国
• 起止时间: 1993-08-15 至 2000-01-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=9358163&HistoricalAwards=false
• 关键词: NSF; Young; Investigator; Toward; Neural

9358163 May The research involves applying implementation of computer- integrated manufacturing (IC-CIM) techniques to two critical fabrication processes used in the production of high performance multichip modules and other integrated circuits: plasma-enhanced chemical vapor deposition (PECVD) and reactive ion etching (RIE). These precesses are used in the process sequence to deposit and etch silicon dioxide dielectric layers. Several aspects of computer-integrated manufacturing successfully applied to other etching and deposition processes. For these applications, traditional statistically-based approaches to process modeling, control and diagnosis represent the predominant methods to achieve IC-CIM objectives. However, more recent efforts have shown that neural networks offer great promise in process modeling. In fact, neural process models have been shown to be both more accurate and more robust than their statistical counterparts. This is due to the high degree of parallelism and connectivity in these networks, which crudely mimics the architecture of the human brian. It is anticipated that the innovative use of neural networks will also have a significant impact in process monitoring, control, and equipment diagnosis as well. This program of research will investigate and implement several novel neural network based approaches to the monitoring, modeling, control, and diagnosis of RIE and PECVD. These techniques will be applied specifically to a high performance multichip module program, but the methodology developed will be generic and readily applicable to other electronics manufacturing processes and equipment.

9358163可能的研究涉及将计算机集成制造（IC-CIM）技术应用于用于生产高性能多芯片模块和其他集成电路的两个关键制造过程：等离子体增强化学蒸气沉积（PECVD）和电力离子蚀（RIE）。 这些进攻在过程序列中用于沉积和蚀刻二氧化硅介电层。 计算机集成制造的几个方面成功地应用于其他蚀刻和沉积过程。 对于这些应用，传统的基于统计的方法来处理建模，控制和诊断代表实现IC-CIM目标的主要方法。 但是，最近的努力表明，神经网络在过程建模方面提供了巨大的希望。 实际上，已经证明神经过程模型比其统计学对应物更准确，更健壮。 这是由于这些网络中的高度并行性和连通性，这些网络粗略地模仿了人类Brian的建筑。 预计神经网络的创新使用也将对过程监测，控制和设备诊断产生重大影响。 该研究计划将调查并实施几种基于神经网络的新方法，以进行RIE和PECVD的监视，建模，控制和诊断。 这些技术将专门用于高性能多芯片模块程序，但是开发的方法将是通用的，并且很容易适用于其他电子制造工艺和设备。