Transparent machine learning for crimp force monitoring using phase-based SHAP explanations

Abstract Ensuring consistent product quality in modern manufacturing is crucial, particularly in safety-critical applications. Conventional quality control approaches rely on manually defined features and lack adaptability to the complexity and variability inherent in production data. Conversely, data-driven methods, such as machine learning, demonstrate high detection performance but typically function as black-box models, thereby limiting their acceptance in industrial environments. This paper introduces a methodology for industrial fault detection in the domain of crimping, a safety-critical joining technique, which is both data-driven and transparent. The approach integrates a supervised machine learning model for multi-class fault classification, Shapley Additive Explanations for post-hoc interpretability and a domain-specific visualization technique that maps model explanations to interpretable features. The model explanations are assessed with a quantitative perturbation analysis and the visualization technique is evaluated qualitatively by domain experts. The approach achieves a fault detection accuracy of 95.9 %, and both quantitative selectivity analysis and qualitative expert evaluations confirmed the relevance of the generated explanations. This case study contributes to data-driven and interpretable quality control systems in manufacturing.