Interpreting Data in the Absence of Production Time Consistency and Source Tracking

Data analytics is providing data features to a large range of industrial processes that provides insights that extend well beyond traditional process control. However, data analytics has its own limitations particularly when confronted with challenges such as incomplete source traceability and random process-time window variation. In this context, robust data preparation emerges as a critical prerequisite for the successful deployment of AI-driven methodologies. This paper presents a structured approach to data preparation, following logic strategies applied to normally distributed datasets. The methodology is adapted to account for constraints such as limited data variability and process gradual drift, both presented in the glass bottles manufacturing.