Skip to Main Content
 

Global Search Box

 
 
 
 

Files

File List


29063.pdf (7.08 MB)

ETD Abstract Container

Abstract Header

An Adaptive Recipe Compensation Approach for Enhanced Health Prediction in Semiconductor Manufacturing

Shelly, Aaron
http://orcid.org/0000-0002-6171-6454
http://rave.ohiolink.edu/etdc/view?acc_num=ucin1511793532937998

Abstract Details

2017, MS, University of Cincinnati, Engineering and Applied Science: Mechanical Engineering.
Process monitoring is an essential technique in prognostics and health management, to identify system degradation and potential faults. It provides predictive information that helps prevent unexpected breakdowns, ensure healthy conditions for process equipment, and maintain product quality. As a result, manufacturing industries have been increasingly moving away from traditional reactive and time based preventative maintenance practices, towards more predictive condition-based maintenance realized through process monitoring. However, high-tech fabrication industries, like semiconductor manufacturing, continue to face difficult challenges which have inhibited the deployment of condition-based maintenance strategies. Frequent changes in production recipe and prominent levels of chamber to chamber variability, have limited the performance of traditional process monitoring methods. Although significant research has already been completed to solve the multimodal issue characteristic of semiconductor manufacturing data, these methods are often case specific and lack the versatility for wide-range process monitoring of semiconductor systems. To engineer an effective data-driven process monitoring model for semiconductor manufacturing, the developed model must be adaptable and robust to dynamically changing recipe conditions. In this thesis, an adaptive recipe compensation (ARC) approach is proposed for process monitoring to normalize the effect of dynamically changing production recipes in semiconductor manufacturing. The ARC methodology effectively introduces a feature normalization routine into the standard single global model process monitoring approach. An ANOVA based input selection procedure is used to construct similarity based normalization models to offset the multimodal effect of different recipes. The tool health is calculated using a self-organizing map model and health prediction is performed using a regression based curve-fit. The developed approach is demonstrated in two case studies; both, individual real-world datasets from semiconductor etching processes. The result of the proposed method is benchmarked with a global self-organizing map model.
Jay Lee, Ph.D. (Committee Chair)
Manish Kumar, Ph.D. (Committee Member)
David Siegel, Ph.D. (Committee Member)
128 p.
Mechanics
Predictive Maintenance; Recipe Compensation; Semiconductor; Multi-regime; process monitoring; normalization

Recommended Citations

Citations

  • Shelly, A. (2017). An Adaptive Recipe Compensation Approach for Enhanced Health Prediction in Semiconductor Manufacturing [Master's thesis, University of Cincinnati]. OhioLINK Electronic Theses and Dissertations Center. http://rave.ohiolink.edu/etdc/view?acc_num=ucin1511793532937998

    APA Style (7th edition)

  • Shelly, Aaron. An Adaptive Recipe Compensation Approach for Enhanced Health Prediction in Semiconductor Manufacturing. 2017. University of Cincinnati, Master's thesis. OhioLINK Electronic Theses and Dissertations Center, http://rave.ohiolink.edu/etdc/view?acc_num=ucin1511793532937998.

    MLA Style (8th edition)

  • Shelly, Aaron. "An Adaptive Recipe Compensation Approach for Enhanced Health Prediction in Semiconductor Manufacturing." Master's thesis, University of Cincinnati, 2017. http://rave.ohiolink.edu/etdc/view?acc_num=ucin1511793532937998

    Chicago Manual of Style (17th edition)

ucin1511793532937998
542
© 2017, all rights reserved.
This open access ETD is published by University of Cincinnati and OhioLINK.