Explainable Artificial Intelligence Based Fault Diagnosis and Insight Harvesting for Steel Plates Manufacturing

XAI makes AI more accessible to industry experts: The paper demonstrates that explainable AI tools (Ceteris Peribus profiles, Partial Dependence, and Breakdown profiles) can extract human-understandable insights from complex models, lowering the barrier for domain experts in manufacturing to understand and trust AI decisions without requiring deep technical expertise.

Interpretability improves model development: By harvesting insights from XAI methods and converting them into IF-THEN rules, the researchers incorporated explainability directly into the classifier development process, achieving 94% accuracy on fault diagnosis—suggesting that understanding why a model makes decisions can lead to better overall performance.

XAI insights have practical business value: The paper emphasizes that extracted insights "benefit the human experts of steel manufacturing industry," indicating that explainable AI isn't just an academic exercise but can provide actionable knowledge that helps domain specialists make better decisions in real-world industrial settings.

Multiple explanation methods provide complementary perspectives: The use of several XAI techniques (medoids, SMOTE, multiple profiling methods, and association rule mining) together suggests that no single explanation method tells the complete story—combining different approaches to interpretability provides more comprehensive understanding of model behavior.