Two-Dimensional Statistical Pattern Analysis for Effective Modeling and Monitoring of Limited Batch Process Data in Food Industry

This study presents a novel two-dimensional statistical pattern analysis (2D-SPA) method, specially designed to address the challenges inherent in monitoring and modeling processes within low-volume, high-quality food manufacturing settings. By ingeniously integrating robust statistical techniques within a two-dimensional sliding window framework, this approach significantly expands the analytical dataset by operating across both batch and time dimensions, effectively overcoming the constraints posed by limited batch sizes. This innovative strategy not only augments the detection and diagnosis capabilities for subtle process deviations and emerging trends but also capitalizes on the dynamic statistical variances observed across batches and over time, thus offering a comprehensive view of the process variability. The application of 2D-SPA is meticulously validated through an industrial case study focused on lactic acid bacteria fermentation, where it demonstrates marked improvements in accuracy and sensitivity for process monitoring. The results underscore the method's potential to significantly enhance quality assurance measures in food batch production.