近年，電動車因環保意識提高以及各國政府政策上的支持，再加上技術不斷進步與成本下降，使需求量大幅增加。預估到2035年，電動車的累計銷售量將超過5.25億萬輛，占全球新車銷量中的50%以上。隨著電動車市場迅速擴大，碳化矽高功率元件，如車載電力轉換元件、逆變器和充電樁等設備的需求也同步增加。預計到2033年，碳化矽元件的市場將高達到112.9億美元左右。
然而，在碳化矽晶圓需求量不斷增長的情況下，相繼吸引許多廠商進入此市場。面對層出不窮的新進者以及原先的競爭者，必須持續提升品質、降低生產成本做出產品差異化才能從中脫穎而出。
因此，為提升製程品質，本研究主要分為兩大部分：第一部分為提升晶圓平坦度與降低厚度標準差。首先，針對個案的磨平製程進行現況分析，透過統計檢定找出需要改善的平坦度項目為總厚度差(Total Thickness Variation , TTV)。接著，使用田口方法進行實驗，找到主要影響TTV與厚度標準差的因子為中心齒輪與外環齒輪轉速。為了維持個案公司的產能，再透過第二次田口方法以及限縮機台壓力的範圍，找到TTV、厚度標準差與產能三者兼具的最佳參數組合。最後進行確認實驗，驗證在實際生產中有達到預期的效果；第二部分則透過隨機森林及其他三種方法預測磨平製程的移除率，讓個案公司的產線不再由人工自行判斷，而是透過移除率預測的結果來決定下一盤的加工時間，以減少多磨與少磨所造成的物料成本浪費。
透過上述的研究過程，個案公司實現以下兩個主要結論與成果：(1)磨平製程加工後，碳化矽晶圓的TTV平坦度可由2.2µm降低至0.84µm、厚度標準差可由1.48µm降低至0.98µm；(2)藉由隨機森林法預測下一盤的移除率後，平均多磨的厚度由5.35µm降低至4.5µm，約減少15.9%；平均少磨的厚度則由4.39µm降低至2.94µm，約減少33.0%；整體每月平均物料成本約可減少1,356,000(NTD)。

In recent years, electric vehicles (EVs) have seen a significant increase in demand due to heightened environmental awareness, supportive government policies, continuous technological advancements, and cost reductions. It is estimated that by 2035, the cumulative sales of EVs will exceed 525 million units, accounting for more than 50% of global new car sales. As the EV market rapidly expands, the demand for silicon carbide (SiC) high-power components, such as onboard power conversion devices, inverters, and charging piles, is also increasing. It is expected that by 2033, the SiC component market will reach approximately 11.29 billion USD.
However, with the growing demand for SiC wafers, many manufacturers are entering this market. To stand out amid the influx of new entrants and existing competitors, it is essential to continuously improve quality, reduce production costs, and create product differentiation.
Therefore, to enhance process quality, this study is divided into two main parts. The first part focuses on improving wafer flatness and reducing thickness standard deviation. Initially, a current situation analysis of the lapping process for the case was conducted, identifying Total Thickness Variation (TTV) as the flatness item needing improvement through statistical testing. Next, the Taguchi method was used to conduct experiments, identifying the primary factors affecting TTV and thickness standard deviation as the rotational speeds of the central gear and the outer ring gear. To maintain the production capacity of the case company, a second Taguchi method was employed, narrowing the machine pressure range to find the optimal parameter combination that balances TTV, thickness standard deviation, and production capacity. Finally, confirmation experiments were conducted to verify the expected effects in actual production.
The second part involves predicting the removal rate in the lapping process using Random Forest and three other methods, enabling the case company's production line to determine the next batch's processing time based on the predicted removal rate rather than manual judgment, thus reducing material cost waste caused by over-lapping and under-lapping.
Through the research process described above, the case company achieved the following two main conclusions and results: (1) After processing, the TTV flatness of SiC wafers can be reduced from 2.2µm to 0.84µm, and the thickness standard deviation can be reduced from 1.48µm to 0.98µm; (2) By predicting the removal rate of the next batch using the Random Forest method, the average thickness due to over-lapping decreased from 5.35µm to 4.5µm, a reduction of approximately 15.9%; the average thickness due to under-lapping decreased from 4.39µm to 2.94µm, a reduction of approximately 33.0%; and the overall average monthly material cost could be reduced by about 1,356,000 NTD.

摘要...I
ABSTRACT...II
誌謝...IV
目錄...V
圖目錄...VII
表目錄...IX
第一章 緒論...1
1-1 研究背景與動機...1
1-1-1 碳化矽晶圓簡介與應用...1
1-1-2 碳化矽晶圓主要應用-電動車市場需求...1
1-1-3 碳化矽晶圓主要應用-電動車碳化矽高功率元件市場需求...2
1-1-4 碳化矽晶圓產量與遭遇的挑戰...4
1-2 研究目的...5
1-3 研究架構...7
第二章 文獻探討...9
2-1 晶圓研磨與磨平加工製程...9
2-2 碳化矽晶圓品質改善...11
2-3 碳化矽晶圓移除率提升...14
2-4 晶圓移除率預測...15
第三章 研究方法...18
3-1 研究流程...18
3-2 提升晶圓平坦度的方法-田口方法...20
3-3 磨平移除率預測...20
3-3-1 決策樹...21
3-3-2 隨機森林...21
3-3-3 極限梯度提升(XGBoost)...22
3-3-4 多元線性回歸...22
第四章 個案分析...25
4-1 田口實驗...25
4-1-1 現況分析...25
4-1-2 確定反應值...32
4-1-3 選擇控制因子與其水準...32
4-1-4 選定直交表與配置因子於直交表中...34
4-1-5 平坦度TTV與厚度標準差實驗結果...35
4-1-6 平坦度TTV與厚度標準差結果分析...43
4-1-7 第二次田口實驗選擇控制因子與其水準...46
4-1-8 第二次田口實驗選定直交表與配置因子於直交表中...47
4-1-9 第二次田口實驗平坦度TTV與厚度標準差實驗結果...49
4-1-10 第二次田口實驗確認實驗...56
4-2 移除率預測...59
4-2-1 現況分析...59
4-2-2 找尋磨平站移除率不穩定的原因...61
4-2-3 資料預處理...62
4-2-4 關聯度分析...63
4-2-5 選擇預測模型...64
4-2-6 驗證集移除率預測...70
4-3 改善效益分析...71
第五章 結論...73
5-1 結論...73
5-2 未來研究建議與方向...73
參考文獻...75

英文文獻
Breiman, L. (2001). Random forests. Machine learning, 45, 5-32.
Chen, X., Li, j., Ma, D., Hu, X., Xu, X., Jiang, M. (2006). Fine Machining of Large-Diameter 6H-SiC Wafers. Materials Science and Technology, 22(05), 681.
Chen, T., & Guestrin, C. (2016). Xgboost: A scalable tree boosting system. In Proceedings of the 22nd acm sigkdd international conference on knowledge discovery and data mining (pp. 785-794).
Deng, J., Lu, J., Zeng, S., Yan, Q., & Pan, J. (2021). Processing properties for the Si-face of the 4H-SiC substrates using the magnetically-controlled abrasive solidification orientation–solid-phase Fenton reaction for the fabrication of the lapping–polishing plate. Diamond and Related Materials, 120, 108652.
Deng, J., Yan, Q., Lu, J., Xiong, Q., & Pan, J. (2021). Optimisation of lapping process parameters for single-crystal 4H–SiC using orthogonal experiments and grey relational analysis. Micromachines, 12(8), 910.
Gao, S., Li, H., Huang, H., & Kang, R. (2022). Grinding and lapping induced surface integrity of silicon wafers and its effect on chemical mechanical polishing. Applied Surface Science, 599, 153982.
Guo, F., Shao, C., Chen, X., Xie, X., Yang, X., Hu, X., & Xu, X. (2022). Shape modulation due to sub-surface damage difference on N-type 4H–SiC wafer during lapping and polishing. Materials Science in Semiconductor Processing, 152, 107124.
Guo, X., Yao, W., Zhu, X., Li, Y., & Kang, R. (2022). Research on the shape of ground wafer in Back Grinding of Wafer with Outer Rim. Materials Science in Semiconductor Processing, 139, 106344.
Hu, Y., Shi, D., Hu, Y., Zhao, H., Sun, X., & Wang, M. (2019). Experimental investigation on the ultrasonically assisted single-sided lapping of monocrystalline SiC substrate. Journal of Manufacturing Processes, 44, 299-308.
Li, Z. C., Pei, Z. J., & Fisher, G. R. (2006). Simultaneous double side grinding of silicon wafers: a literature review. International Journal of Machine Tools and Manufacture, 46(12-13), 1449-1458.
Li, W., Yan, Q. S., Lu, J. B., & Pan, J. S. (2013). Effect of abrasives on the lapping performance of 6H-SiC single crystal wafer. Advanced Materials Research, 690, 2179-2184.
Li, X., Wang, C., Zhang, L., Mo, X., Zhao, D., & Li, C. (2018, March). Assessment of physics-based and data-driven models for material removal rate prediction in chemical mechanical polishing. In 2018 2nd International Conference on Electrical Engineering and Automation (ICEEA 2018) (pp. 116-121). Atlantis Press.
Li, Z., Wu, D., & Yu, T. (2019). Prediction of material removal rate for chemical mechanical planarization using decision tree-based ensemble learning. Journal of Manufacturing Science and Engineering, 141(3), 031003.
Luo, Q. F., Lu, J., & Xu, X. P. (2016). A comparative study on the material removal mechanisms of 6H-SiC polished by semi-fixed and fixed diamond abrasive tools. Wear, 350, 99-106.
Luo, Q., Lu, J., & Xu, X. (2016). Study on the processing characteristics of SiC and sapphire substrates polished by semi-fixed and fixed abrasive tools. Tribology International, 104, 191-203.
Moeggenborg, K., Kegg, T., Parfeniuk, C., Stoney, T., & Quast, J. (2015). Effect of surface damage on SiC wafer shape. In Materials Science Forum (Vol. 821, pp. 545-548). Trans Tech Publications Ltd.
Moeggenborg, K., Manning, I., Searson, J., & Chung, G. Y. (2019). Impact of Subsurface Damage on SiC Wafer Shape. In Materials Science Forum (Vol. 963, pp. 530-533). Trans Tech Publications Ltd.
Montgomery, D. C., Peck, E. A., & Vining, G. G. (2021). Introduction to linear regression analysis. John Wiley & Sons.
Neslen, C. L., Mitchel, W. C., & Hengehold, R. L. (2001). Effects of process parameter variations on the removal rate in chemical mechanical polishing of 4H-SiC. Journal of electronic materials, 30, 1271-1275.
Phadke, M. S. (1995). Quality engineering using robust design. Prentice Hall PTR.
Quinlan, J. R. (1986). Induction of decision trees. Machine learning, 1, 81-106.
Su, C. T. (2013). Quality engineering: off-line methods and applications. CRC press.
Sun, C., Lu, Y., Xiu, S., Li, Q., & Zhang, P. (2021). Analysis on the removal mechanism of disc grinding based on dynamic thermal–mechanical coupling. International Journal of Mechanics and Materials in Design, 1-23.
Tönshoff, H. K., Schmieden, W. V., Inasaki, I., König, W., & Spur, G. (1990). Abrasive machining of silicon. CIRP annals, 39(2), 621-635.
Tsai, M. Y., Li, K. Y., & Ji, S. Y. (2021). Novel abrasive-impregnated pads and diamond plates for the grinding and lapping of single-crystal silicon carbide wafers. Applied Sciences, 11(4), 1783.
Wang, P., Gao, R. X., & Yan, R. (2017). A deep learning-based approach to material removal rate prediction in polishing. CIRP annals, 66(1), 429-432.
Xi, Z. H. A. N. G., Ruzhong, Z. H. U., Xuqing, Z. H. A. N. G., Minghua, W. A. N. G., Yu, G. A. O., Rong, W. A. N. G., & Xiaodong, P. I. (2023). Influence of Abrasive Morphology and Dispersion Medium on Lapping Quality of 4H-SiC Wafers. Journal of Synthetic Crystals, 52(1).
Xia, L., Zheng, P., Huang, X., & Liu, C. (2022). A novel hypergraph convolution network-based approach for predicting the material removal rate in chemical mechanical planarization. Journal of Intelligent Manufacturing, 33(8), 2295-2306.
Yiqing, Y., Zhongwei, H., Wenshan, W., Huan, Z., Jing, L., & Xipeng, X. (2020). The double-side lapping of SiC wafers with semifixed abrasives and resin–combined plates. The International Journal of Advanced Manufacturing Technology, 108, 997-1006.
Yu, T., Li, Z., & Wu, D. (2019). Predictive modeling of material removal rate in chemical mechanical planarization with physics-informed machine learning. Wear, 426, 1430-1438.
Zhang, J., Jiang, Y., Luo, H., & Yin, S. (2021). Prediction of material removal rate in chemical mechanical polishing via residual convolutional neural network. Control Engineering Practice, 107, 104673.
Zhang, X., Liu, X., Wang, Y., Zhu, R., Zhang, X., Zhang, Y., & Pi, X. (2023). Optimizing the flatness of 4H-silicon carbide wafers by tuning the sequence of lapping. Semiconductor Science and Technology, 38(3), 034001.
Zhou, L., Audurier, V., Pirouz, P., & Powell, J. A. (1997). Chemomechanical polishing of silicon carbide. Journal of the Electrochemical Society, 144(6), L161.
Zhong, J., Chen, J., Wang, H., Chen, H., Gu, Y., Shen, J., & Sun, T. (2023). Influences of Nonaqueous Slurry Components on Polishing 4H-SiC Substrate with a Fixed Abrasive Pad. Crystals, 13(6), 869.

中文文獻
黃韋翰(2003)，新式超精密拋光機之矽晶圓拋光特性研究，國立中山大學機械與機電工程學系研究所碩士輪文。

網站文獻
Digitimes (2024)。SiC又成一片紅海？中國擴產潮下價格戰、產能去化成隱憂。https://www.digitimes.com.tw/tech/dt/n/shwnws.asp?id=0000688537_26J4E8V942XW276YO9BWE
E+H Metrology GmbH (2024). MX20 Series. https://www.ehmetrology.com/en/products/manual-tools/mx20-series.
IEA (2024). Global EV Outlook 2024-Outlook for electric mobility. https://www.iea.org/reports/global-ev-outlook-2024/outlook-for-electric-mobility
Pecedence Research (2023). Silicon Carbide Market Size and Manufacturers. https://www.precedenceresearch.com/silicon-carbide-market
TECHCET (2023). Silicon Carbide (SiC) Wafer Supply Gets Squeezed. https://techcet.com/silicon-carbide-sic-wafer-supply-gets-squeezed/
Wafer Technology Ltd (2024). Wafer Flatness. http://www.wafertech.co.uk/technology/wafer-flatness/
Yole Group (2023). 2022-2028 YOLE Silicon Carbide (SiC) Power Semiconductor Market Repor https://www.yolegroup.com/