A study on the application of the latent dirichlet allocation model in production optimization

• The application of LDA topic models can effectively identify issues in lean production. • Machine learning methods can be effectively integrated with the Lean Six Sigma framework. • Practical cases have significantly improved the overall equipment efficiency in production engineering. In recent years, the global manufacturing industry has continuously undergone digital transformation amidst intense competition. The traditional experience-based lean production model is inadequate to resolve issues like declining equipment efficiency and frequent failures. This study proposes a fault diagnosis and improvement strategy that integrates the LDA topic model with the Six Sigma DMAIC methodology, carrying out in-depth analysis of failure factors in key processes on the production line and implementing a 12-week lean improvement initiative in a case enterprise. The results show that after these improvements, the production line's overall equipment effectiveness (OEE) increased significantly from 71 % to 84 %, with a marked reduction in equipment failure frequency, thereby ensuring enhanced production stability and reliability. This study overcomes the limitations of traditional lean methodologies' reliance on structured data by integrating data-driven decision-making into the conventional Lean Six Sigma framework, not only providing a quantitative foundation for equipment fault diagnosis but also offering new theoretical perspectives and practical pathways for continuous improvement and intelligent transformation in manufacturing.