聚合物為人類經濟生活非常重要的材料，其又可細分為熱固性塑料、彈性體和熱塑性塑料。其中，熱固性塑料在航空、汽車交通、電子封裝、醫療等產業使用廣泛。譬如，碳纖維強化聚合物具備高強度、抗腐蝕老化能力強、輕盈、便宜等特性，有效取代了合金和天然資源達到更高性能、減少成本和減少對環境破壞。熱固性樹脂原料也運用於IC晶圓代工程序中後端封裝，將元件固定和灌封保護元件的完整性，
不過，在熱固性聚合物成型的生產過程中，難免會有缺陷的產生導致產品最終品質把關出現挑戰，也使得產品的信賴性檢測不合格影響其功能及可能造成安全問題。例如，一個碳纖維強化聚合物的表面底下，會隱藏許多大小不一的缺陷，作業員通過肉眼是觀察不到的。為克服這些困難，工業上存在許多非破壞性檢測方法如主動熱成像，針對昂貴的複合材料進行大面積的缺陷偵測。除了透過實驗增加缺陷可見度，過去學者也提出許多熱成像數據後處理的方法，能夠考量眾多變數和最終補抓的溫度訊號的複雜關係，進行更有效的特徵提取達到效果更良好的缺陷偵測。所以，在本文第一部分，我們提出了一個新穎的方法，基於神經網路進行熱成像數據處理。
機器學習過去實現了擬合輸入變數及輸出結果的目標，加上計算機速度近年來大幅提升，該領域成為當今學術研究炙手可熱的議題。不過，因為使用深度學習的類神經網路過去存在著黑箱的角色，與熱成像背後蘊含的物理意義仍不明確，也導致不受熱成像數據處理看好。在本文，我們使用一套具備物理導向功能的類神經網路，並由多組熱傳輸送統御方程式引導，將其運用到主動熱成像的數據後處理。此方法在學術界稱之為Physics-informed Neural Network(PINN)。此方法借用已知的物理訊息引導模型最終預測，有望提高預測的準確性。我們使用多個案例，展現其成功描述熱成像圖像的背景訊息。然後，從原圖像刪減背景所獲得的殘差圖像能獲得更好的缺陷偵測及降噪功能，解決不均勻樣品表面和不均勻脈衝加熱的問題。同時，我們也引入Principal component thermography(PCT)的方法，進行特徵提取達到數據壓縮降維的作用，能獲得更優良的缺陷偵測效果。我們使用兩個分別為碳纖維強化聚合物和古物鑲嵌的樣本，進行非破壞性檢測的脈衝熱成像方法後，獲得該樣本相關缺陷位置、形狀的訊息。此外，我們也生成模擬數據並使用PINN進行模型訓練，可獲得準確的樣本材料熱參數的數值。
在射出成型及IC封裝的領域中，短射和包封是非常常見的成型問題。近期有許多針對數值模擬方法，先是仿真真實模流流動和缺陷位置，然後進行優化作業及在前期通過設計降低前述的成型問題。此方法有望取代過去比較耗時耗成本的試誤法。因此，本文的第二個部分，我們針對當今興起的電子元件灌膠的製程工序進行相關研究。第一個部分，我們探討使用於灌膠的聚氨酯熱固性聚合物的特性分析，以及使用多個模型進行特性擬合。過去針對這種高分子材料的分析文獻非常少。因此，我們先是使用微示差掃描熱卡分析儀（DSC）的穩態和動態分析進行固化速率的比較。黏度隨溫度、時間、剪切速率和固化程度則可以使用流變儀，並以Cross-castro Macosko Model進行擬合。在固化膠點以後，模數的影響更大，所以我們也使用動態機械分析（DMA）方法探討材料的固態黏彈性，並使用Generalized Maxwell Model進行擬合。材料的表面分析和熱參數也可以通過動態接觸量測儀和DSC進行分析。獲得正確的材料參數後，我們就能在第二個部分進行非牛頓流體的灌膠模流模擬的數值分析。通過比較實驗和模擬結果，可以同時驗證第一部分的材料分析手法可靠性和模型參數的準確性，也同時可以達到仿真的效果，以便後續能有進行優化分析和可靠度分析。

Polymer is one of the most important materials in our economic activities, which can be subdivided into elastomers, thermoplastics and thermosets. Thermoset matrix is widely applied in aerospace, electronics packaging, automotive and medical industry. For example, prepregs have been used in carbon fiber reinforced polymer (CFRP) to substitute other materials such as alloys and natural materials due to its excellent tensile strength, high resistance, low density and low costs. Thermoset resins have been applied as a molding compound in electronics packaging and encapsulation at the end of the production line.
Faults during polymer processing and in end products inevitably arise and efforts have been made to reduce reliability issues. In a CFRP matrix, defects may be hidden in the subsurface layers which undermine the structural integrity and expected performance of the material. To achieve better quality control and safety, non-destructive testing methods have been employed to facilitate defect detection on expensive and precious products. Many thermographic data analysis methods have been proposed to refine and enhance defect detection in the work of active thermography where complex relationships between variables and outputs exist. In the first part of our work, we aim to improve the effectiveness of thermographic data analysis using a novel machine learning method.
Machine learning has shown promises in being able to universally approximate complex functions of inputs to outputs, and its computational speed has improved exponentially in the recent century. However, its limitation as a purely data-driven “white-box” has long been a subject of research. We introduce to the field of thermography the application of a deep learning method that is constrained by known heat transfer phenomena described by a series of governing equations, also known in the literature as physics-informed neural network (PINN). This family of neural network is able to incorporate a priori physical knowledge, guiding the predictions within the boundaries of physical phenomena in chemical engineering. This is demonstrated by its ability to accurately reconstruct the background information from the thermograms. As a result, residual thermograms formed from removing background from the original images facilitate identification of subsurface defects and reduction in noises caused by uneven background and heating. Principal component thermography (PCT) is incorporated and performed on residual data to further enhance the contrast and saliency of subsurface defects. We illustrate the method’s feasibility through experimental results obtained after performing pulsed thermography (PT) on a carbon fiber reinforced polymer (CFRP) specimen and an ancient marquetry sample. We also identify important quantitative material parameter using PINN trained on a simulated dataset.
Aside from defect formation, short shot and air traps in molded plastic products is also a common engineering challenge in the molding and packaging industry. Numerical simulation of melt front development in mold-filling process has seen increased use in recent years due to advancement in simulation technology, where optimization and realization of design via traditional trial-and-error experimental studies can potentially be replaced by three-dimensional numerical simulation of the full process. In the second part of our work, the molding process of a potted electronic is studied. This is done in two parts. In the first part, we investigated the methodologies for characterizations and modeling of properties of polyurethane thermosets. Experimental study on the properties of this material has previously been limited. Isothermal and dynamic analysis of differential scanning calorimetry (DSC) were compared, indicating different curing kinetics. The reactive viscosity was studied via temperature sweep and modeled after Cross-Castro Macosko Model. Viscosity becomes irrelevant after gel point, in which case modulus becomes prominent. We performed experiments on structural viscoelasticity described by Generalized Maxwell Model which is parametrized by modulus data obtained via dynamic mechanical analysis (DMA). In the second part, numerical simulation of non-Newtonian flow in a potting process is established. By studying the melt front development of the experimental and simulated results, the chemorheological behaviors of the materials were verified, confirming the suitability of our experimental procedures.

Abstract II
摘要 V
Table of content VII
List of figures X
List of tables XIV
First Part 1
Chapter 1 Introduction 2
1.1 Preface 2
1.2 Literature review 4
1.3 Motivation 7
1.4 Thesis structure 9
Chapter 2 Thermographic Data Analysis 10
2.1 Physics-informed Neural Network 10
2.2 Physics-informed Gated Recurrent Unit and Long Short-term Memory 17
2.3 Principal Component Thermography 22
Chapter 3 Experimental Setups 26
3.1 Experiment 27
3.1.1 Specimen 1 (CFRP) 28
3.1.2 Specimen 2 (Ancient Marquetry) 29
3.2 Simulation setup 31
3.2.1 Specimen 3 (Simulated CFRP) 31
Chapter 4 Results and Discussions 34
4.1 Specimen 1 37
4.2 Specimen 2 43
4.3 Specimen 3 47
4.4 SNR comparisons 49
Chapter 5 Conclusion and Future Work 62
Second Part 63
Chapter 1 Introduction 64
1.1 Preface 64
1.2 Literature review 66
1.3 Motivation 68
1.4 Thesis structure 69
Chapter 2 Theories and Principles 70
2.1 Overview of potting process 70
2.2 Curing kinetics 72
2.3 Viscosity 74
2.4 Surface tension and contact angle 76
2.5 Viscoelasticity 77
Chapter 3 Simulation Analysis 84
3.1 Principles 84
3.2 Governing Equations 85
3.3 Meshing 89
Chapter 4 Experimental Setups 91
4.1 Experiment 91
4.1.1 Curing kinetics 91
4.1.2 Viscosity 92
4.1.3 Surface tension 93
4.1.4 Contact angle 93
4.1.5 Specific heat capacity and thermal conductivity 93
4.1.6 Viscoelasticity 93
4.2 Simulation setup 96
Chapter 5 Results and Discussions 98
5.1 Material characterization results 98
5.1.1 Curing kinetics 98
5.1.2 Viscosity 101
5.1.3 Contact angle 105
5.1.4 Surface tension 106
5.1.5 Specific heat capacity and thermal conductivity 107
5.1.6 Viscoelasticity 109
5.2 Simulated mold-filling results 114
5.2.1 Case study 1 114
5.2.2 Case study 2 118
Chapter 6 Conclusion and Future Work 124
Bibliography 126

1.Gholizadeh, S., A review of non-destructive testing methods of composite materials. Procedia Structural Integrity, 2016. 1: p. 50-57.
2.Wen, J., Z. Xia, and F. Choy, Damage detection of carbon fiber reinforced polymer composites via electrical resistance measurement. Composites Part B: Engineering, 2011. 42(1): p. 77-86.
3.Lai, W.L., et al., Characterization of flaws embedded in externally bonded CFRP on concrete beams by infrared thermography and shearography. Journal of Nondestructive Evaluation, 2009. 28(1): p. 27-35.
4.Ladva, N., Numerical Solutions of Simplifications of Navier-Stokes Equations. 2017.
5.Zeng, Z., et al., Specified value based defect depth prediction using pulsed thermography. Journal of Applied Physics, 2012. 112(2): p. 023112.
6.Zeng, Z., et al., Depth prediction of non-air interface defect using pulsed thermography. NDT & E International, 2012. 48: p. 39-45.
7.Pantano, A. and D. Cerniglia, Simulation of laser generated ultrasound with application to defect detection. Applied Physics A, 2008. 91(3): p. 521-528.
8.Foudazi, A., et al., Active microwave thermography for defect detection of CFRP-strengthened cement-based materials. IEEE Transactions on Instrumentation and Measurement, 2016. 65(11): p. 2612-2620.
9.Tian, G.Y., et al., Eddy current pulsed thermography with different excitation configurations for metallic material and defect characterization. Sensors, 2016. 16(6): p. 843.
10.Cannas, B., et al. Modeling of active infrared thermography for defect detection in concrete structures. in Proceedings of the 2012 COMSOL Conference. 2012.
11.Balageas, D.L., et al., The thermographic signal reconstruction method: a powerful tool for the enhancement of transient thermographic images. Biocybernetics and biomedical engineering, 2015. 35(1): p. 1-9.
12.Benítez, H., et al. Modified differential absolute contrast using thermal quadrupoles for the nondestructive testing of finite thickness specimens by infrared thermography. in 2006 Canadian Conference on Electrical and Computer Engineering. 2006. IEEE.
13.Madruga, F.J., et al., Infrared thermography processing based on higher-order statistics. NDT & E International, 2010. 43(8): p. 661-666.
14.Milovanović, B., M. Gaši, and S. Gumbarević, Principal component thermography for defect detection in concrete. Sensors, 2020. 20(14): p. 3891.
15.Rajic, N., Principal component thermography for flaw contrast enhancement and flaw depth characterisation in composite structures. Composite structures, 2002. 58(4): p. 521-528.
16.Wu, J.-Y., S. Sfarra, and Y. Yao, Sparse principal component thermography for subsurface defect detection in composite products. IEEE transactions on industrial informatics, 2018. 14(12): p. 5594-5600.
17.Maldague, X.P., Introduction to NDT by active infrared thermography. Materials Evaluation, 2002. 60(9): p. 1060-1073.
18.Van Drunen, G. and V. Cecco, Recognizing limitations in eddy-current testing. NDT international, 1984. 17(1): p. 9-17.
19.Li, Y. Research and application of deep learning in image recognition. in 2022 IEEE 2nd International Conference on Power, Electronics and Computer Applications (ICPECA). 2022. IEEE.
20.Yang, C.-H.H., et al. Decentralizing feature extraction with quantum convolutional neural network for automatic speech recognition. in ICASSP 2021-2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). 2021. IEEE.
21.Chen, Z., et al., A deep learning method for bearing fault diagnosis based on cyclic spectral coherence and convolutional neural networks. Mechanical Systems and Signal Processing, 2020. 140: p. 106683.
22.Qiu, X., et al., Pre-trained models for natural language processing: A survey. Science China Technological Sciences, 2020. 63(10): p. 1872-1897.
23.Saeed, N., et al., Automatic defects detection in CFRP thermograms, using convolutional neural networks and transfer learning. Infrared Physics & Technology, 2019. 102: p. 103048.
24.Duan, Y., et al., Automated defect classification in infrared thermography based on a neural network. NDT & E International, 2019. 107: p. 102147.
25.D'Orazio, T., et al., Defect detection in aircraft composites by using a neural approach in the analysis of thermographic images. NDT & E International, 2005. 38(8): p. 665-673.
26.Liu, K., et al., Generative principal component thermography for enhanced defect detection and analysis. IEEE Transactions on Instrumentation and Measurement, 2020. 69(10): p. 8261-8269.
27.Raissi, M., P. Perdikaris, and G.E. Karniadakis, Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations. Journal of Computational Physics, 2019. 378: p. 686-707.
28.Sun, L., et al., Surrogate modeling for fluid flows based on physics-constrained deep learning without simulation data. Computer Methods in Applied Mechanics and Engineering, 2020. 361: p. 112732.
29.Goswami, S., et al., Transfer learning enhanced physics informed neural network for phase-field modeling of fracture. Theoretical and Applied Fracture Mechanics, 2020. 106: p. 102447.
30.Yang, Y. and P. Perdikaris, Adversarial uncertainty quantification in physics-informed neural networks. Journal of Computational Physics, 2019. 394: p. 136-152.
31.Zobeiry, N. and K.D. Humfeld, A physics-informed machine learning approach for solving heat transfer equation in advanced manufacturing and engineering applications. Engineering Applications of Artificial Intelligence, 2021. 101: p. 104232.
32.Chen, T.Y. and M.-H. Kuo. Converting the infrared thermal image into temperature field for detection the defects inside materials. in Eighth International Symposium on Precision Engineering Measurement and Instrumentation. 2013. SPIE.
33.Carslaw, H.S. and J.C. Jaeger, Conduction of heat in solids. Conduction of heat in solids, 1947.
34.Baydin, A.G., et al., Automatic differentiation in machine learning: a survey. Journal of machine learning research, 2018. 18.
35.Hochreiter, S. and J. Schmidhuber, Long short-term memory. Neural computation, 1997. 9(8): p. 1735-1780.
36.Chung, J., et al., Empirical evaluation of gated recurrent neural networks on sequence modeling. arXiv preprint arXiv:1412.3555, 2014.
37.Nascimento, R.G. and F.A. Viana, Fleet prognosis with physics-informed recurrent neural networks. arXiv preprint arXiv:1901.05512, 2019.
38.Richardson, M., Principal component analysis. URL: http://people. maths. ox. ac. uk/richardsonm/SignalProcPCA. pdf (last access: 3.5. 2013). Aleš Hladnik Dr., Ass. Prof., Chair of Information and Graphic Arts Technology, Faculty of Natural Sciences and Engineering, University of Ljubljana, Slovenia ales. hladnik@ ntf. uni-lj. si, 2009. 6: p. 16.
39.Wall, M.E., A. Rechtsteiner, and L.M. Rocha, Singular value decomposition and principal component analysis, in A practical approach to microarray data analysis. 2003, Springer. p. 91-109.
40.Liu, K., et al., Enhanced defect detection in carbon fiber reinforced polymer composites via generative kernel principal component thermography. Polymers, 2021. 13(5): p. 825.
41.Wen, C.-M., et al., Thermographic data analysis for defect detection by imposing spatial connectivity and sparsity constraints in principal component thermography. IEEE Transactions on Industrial Informatics, 2020. 17(6): p. 3901-3909.
42.Liu, K., et al., A thermographic data augmentation and signal separation method for defect detection. Measurement Science and Technology, 2021. 32(4): p. 045401.
43.FLIR, A., The Ultimate Infrared Handbook for R&D Professionals: A Resource Guide for Using Infrared in the Research and Development Industry. FLIR AB. URL http://www. globalspec. com/FeaturedProducts/Detail/FLIR/The_Ultimate_Infrared_ Handbook_for_RD/112511/0. Cited on, 2013: p. 5.
44.Gruber, J., et al. Thermal diffusivity measurements on porous carbon fiber reinforced polymer tubes. in AIP Conference Proceedings. 2017. AIP Publishing LLC.
45.Lu, L., et al., DeepXDE: A deep learning library for solving differential equations. SIAM Review, 2021. 63(1): p. 208-228.
46.Liu, D.C. and J. Nocedal, On the limited memory BFGS method for large scale optimization. Mathematical programming, 1989. 45(1): p. 503-528.
47.Csáji, B.C., Approximation with artificial neural networks. Faculty of Sciences, Etvs Lornd University, Hungary, 2001. 24(48): p. 7.
48.Liu, Y., et al., Independent component thermography for non-destructive testing of defects in polymer composites. Measurement Science and Technology, 2019. 30(4): p. 044006.
49.Baylakoglu, I., C. Hillman, and M. Pecht. Characterization of some commercial thermally-cured potting materials. in International IEEE Conference on the Business of Electronic Product Reliability and Liability, Hong Kong, January. 2003.
50.Carlsson, L.A., D.F. Adams, and R.B. Pipes, Experimental characterization of advanced composite materials. 2014: CRC press.
51.Phansalkar, S.P., C. Kim, and B. Han, Effect of critical properties of epoxy molding compound on warpage prediction: A critical review. Microelectronics Reliability, 2022. 130: p. 114480.
52.Dimier, F., et al., Curing kinetics and chemorheological analysis of polyurethane formation. Polymer Engineering & Science, 2004. 44(3): p. 518-527.
53.Chao, N.-H., et al., The use of potting materials for electronic-packaging survivability in smart munitions. 2011.
54.Gan, Z., et al., Comparative study of pressurized and capillary underfill flow using lattice Boltzmann method. Arabian Journal for Science and Engineering, 2019. 44(9): p. 7627-7652.
55.Chang, R.-Y., et al., Three-dimensional modeling of mold filling in microelectronics encapsulation process. IEEE Transactions on Components and Packaging Technologies, 2004. 27(1): p. 200-209.
56.Ardebili, H., J. Zhang, and M.G. Pecht, Encapsulation technologies for electronic applications. 2018: William Andrew.
57.Lin, Y.-J., et al., Modeling of viscoelastic behavior of an epoxy molding compound during and after curing. IEEE Transactions on Components, Packaging and Manufacturing Technology, 2011. 1(11): p. 1755-1760.
58.Liang, G. and K. Chandrashekhara, Cure kinetics and rheology characterization of soy‐based epoxy resin system. Journal of applied polymer science, 2006. 102(4): p. 3168-3180.
59.Garschke, C., et al., Cure kinetics and viscosity modelling of a high-performance epoxy resin film. Polymer Testing, 2013. 32(1): p. 150-157.
60.Kinjo, N., et al., Epoxy molding compounds as encapsulation materials for microelectronic devices. Speciality Polymers/Polymer Physics, 1989: p. 1-48.
61.Domínguez, J., J.-C. Grivel, and B. Madsen, Study on the non-isothermal curing kinetics of a polyfurfuryl alcohol bioresin by DSC using different amounts of catalyst. Thermochimica acta, 2012. 529: p. 29-35.
62.Halley, P.J. and M.E. Mackay, Chemorheology of thermosets—an overview. Polymer Engineering & Science, 1996. 36(5): p. 593-609.
63.Hieber, C. and H.H. Chiang, Some correlations involving the shear viscosity of polystyrene melts. Rheologica Acta, 1989. 28(4): p. 321-332.
64.Arashiro, E.Y. and N.R. Demarquette, Use of the pendant drop method to measure interfacial tension between molten polymers. Materials Research, 1999. 2: p. 23-32.
65.Zobeiry, N., R. Vaziri, and A. Poursartip, Computationally efficient pseudo-viscoelastic models for evaluation of residual stresses in thermoset polymer composites during cure. Composites Part A: Applied Science and Manufacturing, 2010. 41(2): p. 247-256.
66.Li, R., Time-temperature superposition method for glass transition temperature of plastic materials. Materials Science and Engineering: A, 2000. 278(1-2): p. 36-45.
67.Zhang, X., et al. On the moduli of viscoelastic materials. in 4th Electronics Packaging Technology Conference, 2002. 2002. IEEE.
68.Del Nobile, M.A., et al., Use of the generalized Maxwell model for describing the stress relaxation behavior of solid-like foods. Journal of Food Engineering, 2007. 78(3): p. 978-983.
69.Gopala, V.R. and B.G. van Wachem, Volume of fluid methods for immiscible-fluid and free-surface flows. Chemical Engineering Journal, 2008. 141(1-3): p. 204-221.
70.Wang, H., et al., Three-dimensional simulation of underfill process in flip-chip encapsulation. Computers & fluids, 2011. 44(1): p. 187-201.
71.Brackbill, J.U., D.B. Kothe, and C. Zemach, A continuum method for modeling surface tension. Journal of computational physics, 1992. 100(2): p. 335-354.
72.Bonab, V.S. and I. Manas-Zloczower, Chemorheology of thermoplastic polyurethane and thermoplastic polyurethane/carbon nanotube composite systems. Polymer, 2016. 99: p. 513-520.
73.Chiacchiarelli, L.M., J.M. Kenny, and L. Torre, Kinetic and chemorheological modeling of the vitrification effect of highly reactive poly (urethane-isocyanurate) thermosets. Thermochimica acta, 2013. 574: p. 88-97.
74.Song, J.-W. and L.-W. Fan, Temperature dependence of the contact angle of water: a review of research progress, theoretical understanding, and implications for boiling heat transfer. Advances in Colloid and Interface Science, 2021. 288: p. 102339.
75.Song, J.-W., M.-C. Ma, and L.-W. Fan, Understanding the temperature dependence of contact angles of water on a smooth hydrophobic surface under pressurized conditions: An experimental study. Langmuir, 2020. 36(32): p. 9586-9595.
76.Khor, C., et al., Finite volume based CFD simulation of pressurized flip-chip underfill encapsulation process. Microelectronics Reliability, 2010. 50(1): p. 98-105.