Data-Driven Root-Cause Analysis in the Scope of Continuous Improvement Projects

Lean Manufacturing is widely recognized as a prominent methodology for implementing Continuous Improvement (CI) in industrial settings. Root Cause Analysis (RCA) plays a vital role in problem-solving projects, serving as a key component within CI methods like the Plan-Do-Check-Act cycle. However, the RCA process can be time-consuming, relying heavily on the expertise of technicians to manually analyze substantial amounts of data. To address these challenges Industry 4.0 technologies, namely Machine Learning (ML), are being implemented into RCA bringing new challenges such as the need for ML expertise in the lean field. Aiming to contribute to overcoming these challenges, this work presents the development of an Assistance System (AS) that integrates descriptive analysis with ML techniques to support lean technicians in the root cause identification phase. This AS encompasses feature selection, hyperparameter tuning, data balancing and feature importance. Logistic Regression, Decision Tree, Random Forest and XGBoost are the models employed, with F1-Score being the evaluation metric. Besides the ML analysis, the AS also allows for descriptive analysis without ML and data profiling. The results of the ML analysis are presented and compared with the standard descriptive analysis to highlight the effectiveness of the AS in the root cause identification process. The successful integration of descriptive analysis and ML techniques enables a systematic approach to problem-solving, leading to improved overall efficiency and competitiveness in industrial settings.