Keeping CRISP-DM in the (Human) Loop: Insights for Complex Manufacturing Systems

Modern high-precision manufacturing demands quality management systems that integrate advanced analytics with human expertise. Traditional, purely data-driven methodologies, including the Cross Industry Standard Process for Data Mining (CRISP-DM), often fail to capture the operational complexities of today's manufacturing, leading to inefficient defect detection and limited process adaptability. This study proposes an extended human-in-the-loop (HITL)-CRISP-DM framework that synergizes domain-centric insights with data-driven analytics to address these challenges. The framework incorporates structured root cause analysis, iterative feedback loops, and dedicated systems engineering to bridge the gap between domain experts and data scientists. Application of the framework in hairpin welding processes demonstrated significant improvements in defect detection and operational efficiency. The integrated approach enhances quality management and lays the groundwork for scalable applications across various industries. Future adoption of such hybrid intelligence frameworks will be critical for evolving quality control methodologies in increasingly complex manufacturing environments.