Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Rapid Flow Behavior Modeling of Thermal Interface Materials Using Deep Neural Networks

Simon Baeuerle, Marius Gebhardt, Jonas Barth, Andreas Steimer, Ralf Mikut

TL;DR

A lightweight heuristic is proposed to model the spreading behavior of TIM and an Artificial Neural Network (ANN) is trained on data from this model to speed up the calculation.

Abstract

Thermal Interface Materials (TIMs) are widely used in electronic packaging. Increasing power density and limited assembly space pose high demands on thermal management. Large cooling surfaces need to be covered efficiently. When joining the heatsink, previously dispensed TIM spreads over the cooling surface. Recommendations on the dispensing pattern exist only for simple surface geometries such as rectangles. For more complex geometries, Computational Fluid Dynamics (CFD) simulations are used in combination with manual experiments. While CFD simulations offer a high accuracy, they involve simulation experts and are rather expensive to set up. We propose a lightweight heuristic to model the spreading behavior of TIM. We further speed up the calculation by training an Artificial Neural Network (ANN) on data from this model. This offers rapid computation times and further supplies gradient information. This ANN can not only be used to aid manual pattern design of TIM, but also enables an automated pattern optimization. We compare this approach against the state-of-the-art and use real product samples for validation.

Rapid Flow Behavior Modeling of Thermal Interface Materials Using Deep Neural Networks

TL;DR

A lightweight heuristic is proposed to model the spreading behavior of TIM and an Artificial Neural Network (ANN) is trained on data from this model to speed up the calculation.

Abstract

Thermal Interface Materials (TIMs) are widely used in electronic packaging. Increasing power density and limited assembly space pose high demands on thermal management. Large cooling surfaces need to be covered efficiently. When joining the heatsink, previously dispensed TIM spreads over the cooling surface. Recommendations on the dispensing pattern exist only for simple surface geometries such as rectangles. For more complex geometries, Computational Fluid Dynamics (CFD) simulations are used in combination with manual experiments. While CFD simulations offer a high accuracy, they involve simulation experts and are rather expensive to set up. We propose a lightweight heuristic to model the spreading behavior of TIM. We further speed up the calculation by training an Artificial Neural Network (ANN) on data from this model. This offers rapid computation times and further supplies gradient information. This ANN can not only be used to aid manual pattern design of TIM, but also enables an automated pattern optimization. We compare this approach against the state-of-the-art and use real product samples for validation.
Paper Structure (13 sections, 6 equations, 12 figures, 1 table, 1 algorithm)

This paper contains 13 sections, 6 equations, 12 figures, 1 table, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related Work
  3. Simulation methods
  4. Data representation and 2D discretization
  5. Heuristic
  6. Artificial Neural Network
  7. Experimental setup
  8. Training the ANN
  9. Physical experiments
  10. Benchmarking
  11. Results
  12. Discussion
  13. Conclusion

Figures (12)

  • Figure 1: Material flow of Thermal Interface Material (TIM) during joining the heatsink of an Electronic Control Unit (ECU). Left: state before joining, right: state after joining.
  • Figure 2: Overview of our approach for a single line of TIM. Inputs are the start point coordinate, feed rate and end point coordinate. The TIM distribution is spatially discretized before and after the compression step.
  • Figure 3: Visual representation of the 2D discretization for a line segment
  • Figure 4: Visual representation of an exemplary iteration of the heuristic. Left: top view; right: sectional view A-A.
  • Figure 5: Architecture of the Artificial Neural Network (ANN). Hyperparameters such as the number of optional layers (yellow) are optimized. Mandatory layers (blue) are always included.
  • ...and 7 more figures