A Stepwise Bayesian Optimization Algorithm Combining Analytical Solution Fitting of Transient Plane Source Method

doi:10.1002/htj.70034

A Stepwise Bayesian Optimization Algorithm Combining Analytical Solution Fitting of Transient Plane Source Method

2025-08-04

0

PRE

AI

DOI：10.1002/htj.70034

Original

Original request for help

Save

Abstract

En 中文

The differences between ideal and actual models of the transient plane source (TPS) method result in measurement errors that need to be eliminated through correction. To enhance measurement accuracy and inversion efficiency, a numerical algorithm incorporating the actual structure was developed. This led to the proposal of the stepwise Bayesian optimization-analytic solution (SBO-AS) algorithm. SBO-AS builds upon traditional Bayesian optimization (BO) by integrating analytic solution (AS) fitting and single-factor loop inversion. The traditional BO algorithm was validated using benchmark functions. Compared with the traditional BO algorithm, the iteration number and time for the inversion of the SBO-AS algorithm were reduced by 81.50% and 89.31%, respectively. The absolute value of the fitting errors for the thermal conductivity of materials, such as glass wool (GW), polystyrene foam (PSF), and extruded polystyrene (XPS), was greater than 20% for the AS algorithm. The algorithm without introducing AS had a smaller inversion error than the one without single-factor loops. The increase in the number of single-factor loop inversions could have improved the inversion accuracy of the thermal contact resistance. The absolute inversion errors of the thermal conductivity and specific heat capacity of different materials were less than 5% for the SBO-AS algorithm, and the number of single-factor loops was three. The inversion error of the thermal contact resistance was less than 20%. With an increase in the initial population number, the actual iteration number of the single-factor inversion showed a decreasing trend; however, the inversion time showed an increasing trend, and the inversion error of each parameter fluctuated. The traditional BO algorithm exhibited a maximum repeatability of 23.4%. The repeatability of the inversion results of the SBO-AS algorithm was less than 0.6%. The robustness of the SBO-AS algorithm was better than that of the traditional BO algorithm.

Keywords:

Bayesian optimization

body-fitted grid

inversion efficiency

transient plane source method

A Stepwise Bayesian Optimization Algorithm Combining Analytical Solution Fitting of Transient Plane Source Method

Abstract

Journal

Organization

Cited Papers