.png){kind=icon}
.png){kind=icon}
Return
Data-driven predicting and optimizing electro-thermal properties for epoxy resin-based composites via an XGBoost hybrid model
.png){kind=icon}
PRE
AI
W
S
C
Y
K
H
S
DOI：10.1088/1361-6463/ae4a39.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Epoxy resin offers outstanding electrical properties and mechanical strength, making it a widely adopted insulating material in high-voltage equipment. However, its limited thermal conductivity restricts its broader utilization in ultra-high-voltage applications. The development of epoxy resins with both high thermal conductivity and electrical insulation has traditionally relied on empirical trial-and-error approaches involving filler formulation and complex characterization. Hence, this study proposes a method for designing and predicting the electro-thermal performance of epoxy resin composite materials utilizing particle swarm optimization (PSO) and eXtreme Gradient Boosting (XGBoost). Finite element simulation is utilized to calculate the electro-thermal properties of epoxy resin composite samples, thereby generating a dataset for model training. The XGBoost model trained on this dataset delivers the highest predictive accuracy and robustness, with its hyperparameters optimized by the PSO algorithm. Additionally, SHapley Additive exPlanations (SHAPs) analysis is employed to quantify the contribution of input features to the electro-thermal performance. The experimental validation of the SHAP-derived trends is conducted using AlN-doped epoxy resin samples. Overall, this study improves the efficiency of the optimization for epoxy resin composites and offers a method for developing advanced composite materials.
Keywords:
epoxy resin
thermal conductivity
electrical insulation
particle swarm optimization
XGBoost
Journal
J
IF:
3.2
Papers:
726
Citations:
0
Organization
Cited Papers