Date of Award
Doctor of Philosophy (PhD)
Computational Analysis and Modeling
Layered structures have appeared in many engineering systems such as biological tissues, micro-electronic devices, thin lms, thermal coating, metal oxide semiconductors, and DNA origami. In particular, the multi-layered metal thin lms, gold-coated metal mirrors for example, are often used in high-powered infrared-laser systems to avoid thermal damage at the front surface of a single layer lm caused by the high-power laser energy. With the development of new materials, functionally graded materials are becoming of more paramount importance than materials having uniform structures. For instance, in semiconductor engineering, structures can be synthesized from di erent polymers, which result in various values of conductivity. Analyzing heat transfer in layered structure is crucial for the optimization of thermal processing of such multi-layered materials.
There are many numerical methods dealing with heat conduction in layered structures such as the Immersed Interface Method, the Matched Interface Method, and the Boundary Method. However, development of higher-order accurate stable nite di erence schemes using three grid points across the interface between layers for variable coe cient case is mathematically challenging. Having three grid points ensures that the nite di erence scheme leads to a tridiagonal matrix that can be solved easily using the Thomas Algorithm. But extension of such methods to higher dimensions is very tedious. Recently there have been some solution to such complex systems with the use of neural networks, that can be easily extended to higher dimensions. For the above purposes, in this dissertation, we rst develop a gradient preserved method for solving heat conduction equations with variable coe cients in double layers. To this end, higher-order compact nite di erence schemes based on three grid points are developed. The rst-order spatial derivative is preserved across the interface. Unconditional stability and convergence with O( 2 + h4) are analyzed using the discrete energy method, where and h are the time step and grid size, respectively. Numerical error and convergence rates are tested in an example. We then present an arti cial neural network (ANN) method for solving the parabolic two-step heat conduction equations in double-layered thin lms exposed to ultrashort-pulsed lasers. Convergence of the ANN solution to the analytical solution is theoretically analyzed using the energy method. Finally, both developed methods are applied for predicting electron and lattice temperature of a solid thin lm padding on a chromium lm exposed to the ultrashort-pulsed lasers. Compared with the existing results, both methods provide accurate solutions that are promising.
Bora, Aniruddha, "" (2021). Dissertation. 910.