Dimensionally reduced modeling and gradient-based design of microchannel cooling networks

Tan, Marcus Hwai Yik

Permalink

https://hdl.handle.net/2142/97540 Copy

Description

Title

Dimensionally reduced modeling and gradient-based design of microchannel cooling networks

Author(s)

Tan, Marcus Hwai Yik

Issue Date

2017-03-20

Director of Research (if dissertation) or Advisor (if thesis)

Geubelle, Philippe H.

Doctoral Committee Chair(s)

Jacobi, Anthony M.

Committee Member(s)

• Duarte, Carlos Armando
• White, Scott R.

Department of Study

Mechanical Sci & Engineering

Discipline

Theoretical & Applied Mechans

Degree Granting Institution

University of Illinois at Urbana-Champaign

Degree Name

Ph.D.

Degree Level

Dissertation

Keyword(s)

• Microvascular composite
• Dimensionally reduced model
• Thermal model
• Hydraulic model
• Microchannel
• Interface-enriched generalized finite element
• Gradient-based
• Optimization
• Multi-objective
• Pareto front
• Blockage
• Battery cooling
• Nanosatellite radiator panel

Abstract

"Microvascular composites constitute a novel class of biomemetic materials with the ability to perform multiple functions such as dynamic tuning of electromagnetic properties, self-healing and thermal management depending on the fluid circulated in the embedded microchannels. Recent breakthroughs in the vaporization of sacrificial component (VaSC) manufacturing technique have allowed for the creation of intricate microchannel networks and large scale production of these composites. As the design of these networks is key to the performance of the composites and designer's intuition is insufficient to achieve optimal performance, the development of ``automated"" design tools is of paramount importance. The primary goal of this work is to fulfill that need in the specific application of thermal management. To that end, we develop three ingredients: dimensionally reduced thermal and hydraulic models, a numerical solver and a shape optimization scheme. Another goal of this project is to verify and validate the dimensionally reduced models against a commercial computational fluid dynamics software package and experiments. The final goal is to apply the design tool to various 2D and 3D problems. In the dimensionally reduced thermal model, the microchannels are collapsed into lines/curves to simplify mesh generation and their thermal impacts are added to the heat equation. Two versions of the thermal model are considered: (i) a linear model that does not involve radiative heat exchange or linearizes the Stefan-Boltzman radiation equation and (ii) a nonlinear model that incorporates the original radiation equation. The hydraulic model uses the Hagen-Poiseuille law to describe the flow rates and pressure drops in the microchannel networks. To capture the gradient discontinuity in the temperature field due to the microchannels, we employ the interface-enriched generalized finite element method (IGFEM) as the numerical solver, which greatly simplifies mesh generation by allowing for the use of meshes that do not conform to the microchannel network. While previous IGFEM works are based on polynomial enrichment functions, we demonstrate the flexibility of the IGFEM by developing non-uniform rational B-splines (NURBS) enrichment functions for branched network of curved microchannels. We then develop a method to address the convergence issue due to the singularity associated with the thermal model in 3D and combine that method with polynomial IGFEM. The thermal fields obtained from the resulting modified IGFEM agree with those of the significantly more complex and costly ANSYS FLUENT conjugate heat transfer simulations. The final ingredient involves the development of analytical IGFEM-based shape sensitivity analyses for both linear and nonlinear models. These analyses allow the design tool to efficiently exploit existing powerful gradient-based optimization algorithms, especially for large number of design parameters. We then apply the gradient-based shape optimization scheme to solve a diverse range of problems, which demonstrate two key advantages of the scheme due to the use of stationary non-conforming meshes by (i) eliminating the cumulative mesh generation cost and (ii) avoiding severe mesh distortion issues as the microchannel geometry evolves during the optimization process. The first problem involves parallel networks of microchannels for 2D microvascular composite battery cooling panels. Using a differentiable alternative to the maximum temperature (the $p$-norm of the temperature field) of a cooling panel as an objective function, we obtain optimized designs superior to the reference designs in terms of cooling performance. We also extensively validate the IGFEM solutions associated with the designs against ANSYS FLUENT simulations and experiments. We further extend the uses of the tool to include multi-objective optimization, pressure drop as objective function, channel diameters as design parameters and localized heat sources. In the multi-objective optimization, the Pareto fronts of the maximum temperatures and the pressure drops across the networks are generated using the normalized normal constraint method. Next, we apply the optimization scheme to design blockage-tolerant cooling networks embedded in 2D PDMS panels. In this novel application, a minmax problem that minimizes the worst case of a set of predetermined blockage scenarios is formulated and converted to a simpler single-objective optimization problem. In the worst blockage scenario, the designs optimized in this manner exhibit substantial reduction of cooling performance loss compared with designs optimized without considering blockages, with greater reduction as the redundancy of the network decreases. The designs are also validated against experiments. Another novel application of the optimization scheme is related to the design of 2D microvascular panels for nanosatellite. In this application, the sensitivity analysis based on the nonlinear thermal model is used since the nonlinear effect of radiation cannot be neglected. Taking advantage of the optimization tool, two formulations are proposed to satisfy the design constraints. We perform extensive benchmarking of the results obtained from the dimensionally reduced models against those from ANSYS FLUENT, and provide analytical estimates of the thermal performance of optimized designs. In the final application, we design multiple parallel microchannels embedded in 3D microvascular panels using the modified-IGFEM-based optimization scheme. Due to the importance of the straight microchannel design, we propose a semi-analytical model of the maximum temperature in a panel with multiple straight channels."

Graduation Semester

2017-05

Type of Resource

text

Permalink

http://hdl.handle.net/2142/97540

Copyright and License Information

Copyright 2017 Marcus Hwai Yik Tan

Owning Collections