Explainable mechanism for production process anomalies based on digital twin

Area of Science:

• Manufacturing Engineering
• Industrial Engineering
• Data Science

Background:

• Abnormal production in manufacturing leads to significant economic and reputational damage.
• Existing methods often lack clarity in identifying root causes of production anomalies.

Purpose of the Study:

• To present an explainable mechanism for production process anomalies (EM2PA).
• To clarify complex relationships, identify influencing factors, and provide causal explanations for abnormal production.
• To enable trace-back analysis for manufacturing anomalies.

Main Methods:

• EM2PA comprises three modules: data augmentation for small sample abnormal data, influence factor recognition to decouple relationships, and causal interpretation for explanations.
• Utilized a case study from a discrete manufacturing workshop.

Main Results:

• The EM2PA effectively identified root causes of production anomalies.
• Demonstrated the capability to analyze the impact of various factors on production processes.
• Validated the importance of explainability and causal analysis in anomaly detection.

Conclusions:

• The developed EM2PA provides a robust solution for identifying and explaining production anomalies in manufacturing.
• Explainable AI and causal analysis are crucial for enhancing manufacturing process control and reliability.