Improved Bayesian Optimization Framework for Inverse Thermal Conductivity Based on Transient Plane Source Method.

Area of Science:

• Heat Transfer
• Computational Modeling

Background:

• Conventional analytical models in transient planar source (TPS) methods introduce errors due to idealizations.
• Accurate heat transfer analysis is crucial for material characterization and thermal property determination.

Purpose of the Study:

• To develop a more accurate finite element model for TPS analysis, reducing errors compared to analytical models.
• To propose and evaluate an improved Bayesian optimization algorithm for solving inverse heat transfer problems.

Main Methods:

• Construction of a finite element model to simulate heat transfer processes more realistically.
• Development of an optimization model for inverse heat transfer problems using probabilistic and heuristic algorithms.
• Introduction of a Bayesian optimization algorithm with an adaptive initial population (BOAAIP) for enhanced inversion.

Main Results:

• The finite element model achieved an average error below 1%, a significant improvement over the analytical model's ~5% error.
• The proposed BOAAIP demonstrated better adaptability and stability, unaffected by initial population parameters.
• BOAAIP showed comparable inversion accuracy (~3%) to the genetic algorithm for thermal conductivity below 100 Wm-1K-1, but was 3-4 times faster.

Conclusions:

• The finite element model offers superior accuracy for TPS analysis.
• The improved Bayesian optimization algorithm provides a more efficient and stable solution for inverse heat transfer problems.
• This research contributes to more reliable material thermal property determination through advanced computational methods.