AI-driven quality control techniques in manufacturing processes to enhance Six Sigma approach

Abstract

Six Sigma is a business strategy focused on reducing defects and variations in products and processes, thereby enhancing quality. The rise of Industry 4.0 technologies has increased data volumes in manufacturing, creating challenges for integrating Six Sigma methodologies. To adapt, we propose an enhanced Six Sigma approach incorporating AI technologies to optimise performance in this new landscape. This paper proposes a new methodology, the 5I method (identify, inspect, investigate, implement, improve) combines statistical methods and predictive analysis using machine learning, aiming to achieve data-driven predictive quality. To develop a robust model for detecting all types of steel plate defects, we propose a hybrid statistical sampling algorithm that combines the synthetic minority over-sampling technique (SMOTE) and edited nearest neighbour (ENN). Additionally, we applied a universal deep neural network (DNN) as a classifier of defects, achieving an impressive prediction accuracy of 99.05%, surpassing other machine learning algorithms.