熱遮罩的發展從轉換熱力學的建立延伸至多層理論再到雙層理論，從使用複雜的非均值且各向異性的熱學超材料慢慢演變成均值且等向性的自然材料，而雙層理論更是僅使用兩種自然材料即可達到隱身的效果。現今雙層理論僅有圓形和橢圓形的解析解，然而內層必須是完美絕熱材料(熱導係數等於零)的條件下才能達到隱身效果，而這在實際工程上是難以實現的。因此，本文提出了僅以雙層自然均值材料作為遮罩之建構進而實現隱形的效果，並不局限於雙層理論內層需為完美絕熱條件的需求，同時也嘗試了不同的幾何結構和於非線性背景溫度場下之情況並探討其他可行性。此外本文使用機器學習中深度學習的方式設計遮罩外層所需熱導係數，除了在計算上減少了時間成本且能以更智慧的方式設計遮罩外層熱導係數，為應用層面上增進其實用性。以數值模擬的方式驗證此方法的正確性。本文通過幾個案例闡述此方法的性能，可適用於多種不同幾何形狀且在無解析解的情況下也能準確地預測出所需熱導係數，未來在工程應用上可以利用此方法快速且準確的計算所需參數達到遮罩隱形效果。

The invisible cloak always gives people a sense of inaccessibility and magic. With the efforts of scholars, good results have finally been achieved in the field of optics, and other fields are also competing for imitation and application, such as the field of thermal science has also been deeply inspired. The development of thermal cloaks extends from the establishment of translational thermodynamics to multi-layer theory and then to double-layer theory, from the use of complex non-homogeneous and anisotropic thermal metamaterials to homogeneous and isotropic natural materials. The double-layer theory can achieve the effect of invisibility by using only two natural materials. Nowadays, the dual-layer theory only has the analytical solutions of circular and elliptical shapes, but the inner layer must be a perfectly insulating material (thermal conductivity equal to zero) to achieve the stealth effect, which is difficult to achieve in practical engineering. Therefore, in this paper, we propose to achieve the stealth effect by using only two layers of natural homogeneous material as the structure of the cloak, which is not limited to the requirement of perfect insulation for the inner layer of the two-layer theory, and also try different geometrical structures and the case of non-linear background temperature field and explore other feasibility. In addition, this paper uses a deep learning approach in machine learning to design the required thermal conductivity of the outer layer of the cloak, which not only reduces the computational time cost but also enables a smarter way to design the thermal conductivity of the outer layer of the cloak to improve the practicality of the application level. In this paper, we illustrate the performance of this method through several cases, which can be applied to three different geometries and can accurately predict the required thermal conductivity without analytical solutions.

[1] Rumelhart, D. E., Hinton, G. E., & Williams, R. J. ,Learning representations by back-propagating errors. Nature, 323(6088): p.533-536, (1986).
[2] Rosenblatt, F., The perceptron: a probabilistic model for information storage and organization in the brain. Psychological Review, 65(6): p.386, (1958).
[3] Fan, C.Z., Y. Gao, and J.P. Huang, Shaped graded materials with an apparent negative thermal conductivity. Applied Physics Letters, 92: p. 1-4, (2008).
[4] Leonhardt, U., Optical conformal mapping. Science, 312: p. 1777-1780, (2006).
[5] Pendry, J. B., Schurig, D., & Smith, D. R., Controlling electromagnetic fields. Science, 312(5781): p1780-1782, (2006).
[6] Chen, T., Weng, C.-N., and Chen, J.-S., Cloak for curvilinearly anisotropic media in conduction. Applied Physics Letter., 93(11): p. 114103, (2008).
[7] Guenneau, S., Amra, C., and Veynante, D., Transformation thermodynamics: cloaking and concentrating heat flux. Optics Express, 20(7): p. 8207-8218, (2012).
[8] Han, T., X. Bai, D. Gao, J.T.L. Thong, B. Li, and C.W. Qiu, Experimental demonstration of a bilayer thermal cloak. Physical Review Letters, 112: p. 1-5, (2014).
[9] Alekseev, G.V. and D.A. Tereshko, Particle swarm optimization-based algorithms for solving inverse problems of designing thermal cloaking and shielding devices. International Journal of Heat and Mass Transfer, 135: p. 1269-1277, (2019).
[10] Yang, F.-Y., Hung, F.-S., Yeung, W.-S., and Yang, R.-J., Optimization Method for Practical Design of Planar Arbitrary-Geometry Thermal Cloaks Using Natural Materials. Physics Review Applied., 15(2): p. 024010, (2021).
[11] Liu, B., Xu, L., & Huang, J., Thermal transparency with periodic particle distribution: A machine learning approach. Journal of Applied Physics, 129(6): p.065101, (2021).
[12] Liu, D., Tan, Y., Khoram, E., & Yu, Z., Training deep neural networks for the inverse design of nanophotonic structures. ACS Photonics, 5(4):p.1365-1369, (2018).
[13] Han, T., P. Yang, Y. Li, D. Lei, B. Li, K. Hippalgaonkar, and C.W.J.A.M. Qiu., Full‐Parameter Omnidirectional Thermal Metadevices of Anisotropic Geometry. Advanced Materials, 30(49): p. 1804019, (2018).
[14] Yeung, W.-S., and Yang, R.-J., Introduction to Thermal Cloaking-Thermal and Analysis in conduction and convection, Springer Nature Singapore, (2022).
[15] Hinton, Geoffrey E., Simon Osindero, and Yee-Whye Teh., A fast learning algorithm for deep belief nets. Neural Computation ,18(7): p.1527-1554, (2006).
[16] Kingma, Diederik P., and Jimmy Ba., Adam: A method for stochastic optimization. arXiv preprint arXiv:1412.6980 (2014).
[17] Qian, N.,On the momentum term in gradient descent learning algorithms. Neural Networks, 12(1): p.145-151, (1999).
[18] Duchi, J., Hazan, E., & Singer, Y., Adaptive subgradient methods for online learning and stochastic optimization. Journal of Machine Learning Research.,12(7), (2011).
[19] Tieleman, T., & Hinton, G., Lecture 6.5-rmsprop: Divide the gradient by a running average of its recent magnitude. COURSERA: Neural Networks for Machine Learning, 4(2): p.26-31, (2012).
[20] Dede, E. M., Zhou, F., Schmalenberg, P., & Nomura, T., Thermal metamaterials for heat flow control in electronics. Journal of Electronic Packaging, 140(1), (2018).