repository
Special Graduate Schools
Graduate School of Science and Technology
Department of Industrial Engineering
4. Theses(Ph.D)
Fault Detection with Index Regression Correction and Variational Smoothing for Semiconductor Equipment Sensor Signal
| DC Field | Value | Language |
|---|---|---|
| dc.contributor.advisor | 신현정 | - |
| dc.contributor.author | 하태경 | - |
| dc.date.accessioned | 2022-11-29T02:33:06Z | - |
| dc.date.available | 2022-11-29T02:33:06Z | - |
| dc.date.issued | 2022-02 | - |
| dc.identifier.other | 31707 | - |
| dc.identifier.uri | https://dspace.ajou.ac.kr/handle/2018.oak/20580 | - |
| dc.description | 학위논문(박사)--아주대학교 일반대학원 :산업공학과,2022. 2 | - |
| dc.description.tableofcontents | 1. Introduction 1 2. Fundamentals: Time Series Model 7 2.1. ARIMA 7 2.2. EWMA 12 2.3. Kalman Filter 16 3. Fundamentals: Machine Learning Model 20 3.1. Autoencoder 20 3.2. Convolutional Neural Network 23 3.3. Support Vector Machine 29 4. Data Acquisition 33 4.1. Fault Data Generation 33 4.2. Introduce to Facility 34 4.3. Sensor Description of Facility 35 4.4. Data Set Description 38 4.5. Feature Selection 41 5. Fault Detection 46 5.1. Index Regression and Correction 46 5.2. Variational Smoothing Autoencoder 59 5.3. Hybrid Fault Detection with Index Regression Correction and Variational Smoothing Autoencoder 73 6. Conclusion 76 | - |
| dc.language.iso | eng | - |
| dc.publisher | The Graduate School, Ajou University | - |
| dc.rights | 아주대학교 논문은 저작권에 의해 보호받습니다. | - |
| dc.title | Fault Detection with Index Regression Correction and Variational Smoothing for Semiconductor Equipment Sensor Signal | - |
| dc.type | Thesis | - |
| dc.contributor.affiliation | 아주대학교 일반대학원 | - |
| dc.contributor.alternativeName | Ha Tae Kyung | - |
| dc.contributor.department | 일반대학원 산업공학과 | - |
| dc.date.awarded | 2022. 2 | - |
| dc.description.degree | Doctoral | - |
| dc.identifier.localId | 1244968 | - |
| dc.identifier.uci | I804:41038-000000031707 | - |
| dc.identifier.url | https://dcoll.ajou.ac.kr/dcollection/common/orgView/000000031707 | - |
| dc.subject.keyword | Fault detection | - |
| dc.subject.keyword | Index regression | - |
| dc.subject.keyword | Semiconductor | - |
| dc.subject.keyword | Sensor signal | - |
| dc.subject.keyword | Variational smoothing | - |
| dc.description.alternativeAbstract | Fault detection is an important method in semiconductor manufacturing for monitoring equipment condition and examine the potential causes of the fault. The vacuum leakage is considered one of major faults in semiconductor processing. Unnecessary O2, N2 mixture, major components of atmosphere, creates unexpected process results hence drops yield. Currently available vacuum leak detection systems in vacuum industry are based on helium mass spectrometers. It is used for detecting the vacuum leakage at sole isolation condition where chamber is fully pumped, but unable to be used at in-situ detection while the process is ongoing in the chamber. In this study, a chamber vacuum leak detection method named Index Regression & Correction (IRC) and variational smoothing autoencoder has been presented, utilizing common data which gathered during normal chamber operation. This method was developed by analyzing a simple list of data, such as temperature of the chamber wall and the position of auto pressure control (APC) to detect any change of leakages in the vacuum chamber. First approach is index regression & correction method. We defined vacuum leak modeling based on the engineer's experience. Based on this modeling, index regression was proposed. A prediction method was developed to reduce the error by applying the prediction update loop idea of the Kalman filter. These two loops each other and make predictions by reflecting the previous prediction error. The update term has an error value between the measured value and the predicted value. The best condition result of each parameter was extracted and used from the final results sequence. The best gain of the IRC method was a=0.5, b=0.6. As a result of the accuracy comparison evaluation, the excellent performance was confirmed with an accuracy of 0.70 and an F1 score of 0.78. Second approach is variational smoothing autoencoder. The weakest point of data smoothing is the loss of information. To improve this problem, a variational smoothing method was developed. The length of the process log data is slightly different for each process due to the limit of command processing according to the sequence of the computer. To improve this problem, we partition the time series data and extract the segment information. The extracted segment information is strongly related. So, the autoencoder model was applied to train well on highly relevant data. The proposed method, variational smoothing autoencoder model, showed the best performance, AUC by 0.84 and accuracy by 0.76. The IRC method and variational smoothing autoencoder were effective in classifying abnormalities by predicting time series data of semiconductor facility sensors. | - |
- Appears in Collections:
- Special Graduate Schools > Graduate School of Science and Technology > Department of Industrial Engineering > 4. Theses(Ph.D)
- Files in This Item:
- There are no files associated with this item.
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.
-
- License
- STATISTICS
- Total Visit :3,726,081
- Total Download :1,818
- Today View :3,739
