Page 15 - IJAMD-2-1

International Journal of AI for
Materials and Design
Predicting thermal conductivity of sintered Ag

However, the mechanical behavior and heat conduction machine learning-assisted methods has also garnered
ability of sintered nano-Ag generally vary under different significant attention in electronic packaging. Machine
power cycling conditions, possibly due to damage learning has been applied to different aspects of electronic
accumulation and crack formation within its structure. packaging, such as materials extraction, 17-19 solder fatigue
1
These variations in thermal performance could reduce lifetime prediction, 20-22 defect detection, and mechanical
23
24
the reliability of the modules and even lead to their response prediction. In the field of electronic packaging,
failure. Thus, accurate evaluation and prediction of heat Long et al. proposed a convolution neural network (CNN)-
25
conductivity of sintered nano-Ag are essential. assisted nanoindentation method to rapidly determine
the mechanical properties of thin-film elastoplastic
To calibrate the effective thermal conductivity of
sintered nano-Ag, different methods have been presented to materials and predict the constitutive parameters with
26,27
understand the variations of effective thermal conductivity. high accuracy. Recently, Du and coworkers adopted
Ordonez-Miranda et al. measured and calculated the CNN for support vector regression models to predict the
2
thermal conductivity of sintered nano-Ag for a sample with thermal conductivity of sintered nano-Ag, demonstrating
22% porosity, demonstrating that pancake-shaped pores have the potential applications of machine learning in predicting
the thermal conductivity of sintered nano-Ag. However,
a more prominent effect on thermal conductivity compared machine learning requires large datasets. In addition, neural
to sphere pores through numerical comparisons. Signor et al. networks have demonstrated great potential in predicting
3
computed the thermal conductivity of sintered nano-Ag the mechanical and thermal behavior of porous materials.
with finite element analysis using actual 3D microstructures. For example, Wei et al. used three different machine
28
Recently, Sghuri et al. tested the thermal conductivity of learning approaches to quickly and accurately predict the
4
sintered nano-Ag under aging conditions to obtain the equivalent thermal conductivity of composite materials.
thermal conductivity of sintered nano-Ag more directly. Similar studies have been conducted in recent years. 18,29,30
Meanwhile, Hu et al. explored the process-microstructure-
5
thermal relation using focused ion beam scanning electron In this study, we focus on the research status and existing
microscopy. Although these studies have reported various problems of physical and mechanical parameter evaluation
behavior of sintered nano-Ag, predicting its thermal methods for sintered nano-Ag nanomaterials. An image
conductivity remains a major challenge due to the complexity dataset, based on the Gaussian random reconstruction
of the process and flaw-dependent thermal behavior. of sintered nano-Ag nanostructures, was proposed, and
the equivalent thermal conductivity of sintered nano-Ag
Various analytical and numerical methods have been nanostructures was efficiently predicted based on a machine
developed to predict the effective thermal conductivity of learning model. Image processing and model simulation
sintered nano-Ag. Zhao et al. indicated that the existing were performed using MATLAB and Ansys software from
6
models are not suitable for predicting the thermal scanning electron microscopy (SEM) images of sintered
conductivity of sintered nano-Ag due to its high porosity nano-Ag nanostructures; batch calculation of the thermal
and complex microstructure. Qin et al. have presented a conductivity of 2D images of sintered nano-Ag nanostructures
7
semi-analytical formulation to predict the effective thermal was conducted thereafter. According to the SEM image
conductivity of sintered nano-Ag by considering the characteristics of actual sintered nano-Ag nanostructures,
modification in the microstructure. To investigate changes the images of the nanostructures at different sintering
in the effective thermal conductivity of sintered nano-Ag temperatures were generated by the Gaussian random model.
due to crack formation, they also proposed a semi-analytical Based on the artificial neural network (ANN) model of
formulation to predict the effect of cracks on the heat Bayesian optimization, the equivalent thermal conductivity
conductivity of sintered nano-Ag. Lately, the effect of mud corresponding to different sintered nano-Ag microstructures
8
cracking on the heat conductivity of sintered nano-Ag was was accurately predicted using the microstructure image and
also studied, revealing variations in heat transfer behaviors characteristic parameters of sintered nano-Ag, with minimal
9
of sintered nano-Ag in the entire SiC module. Kim et al. loss and a high determination coefficient (0.96).
10
also studied the effect of pore shape and porosity on the
effective thermal conductivity of sintered nano-Ag. 2. Methods
With the development of artificial intelligence, the 2.1. Effective thermal conductivity computation
prediction of thermal conductivity for different materials scheme
with machine learning methods has garnered significant
attention due to advantages, such as high accuracy, 2.1.1. Finite element model of sintered nano-Ag
efficiency, and potential for physics-based interpretation. 11-16 Image-to-parameter automated programming can be
Predicting mechanical and material properties using used to improve the efficiency of analytical calculations.

Volume 2 Issue 1 (2025) 9 doi: 10.36922/ijamd.5744