A Hybrid Stacking-Bayesian Model for Defect Detection in the Lithium-Ion Battery Industry

A defect Lithium-ion battery defect detection is critical due to the widespread use of these batteries. Ensuring their safety and performance is essential, as defects can lead to serious issues, such as overheating and explosions. Early defect detection enhances battery reliability, extends lifespan, and reduces manufacturing costs by mitigating warranty claims and recalls. This paper presents a machine learning-based framework for detecting defects in lithium-ion batteries used in neon signs. Our approach combines various ensemble classification algorithms as base estimators for final stacking meta-learners. A Genetic Algorithm (GA) is used for feature selection, followed by model optimization using Bayesian hyperparameter tuning. Stacking methods with Multi-Layer Perceptron (MLP) and Support Vector Machine (SVM) as meta-learners are employed to enhance classification accuracy for defect detection in lithium-ion batteries used in neon panels. Our Hybrid Stacking-Bayesian (HSB) approach demonstrates the effectiveness of these models in accurately identifying defective batteries, contributing to improved manufacturing quality control and sustainability by minimizing waste and optimizing resource utilization. Implementing our model on real lithium-ion battery data showcases its potential for practical applications in the industry.