Hybrid Unsupervised Learning Strategy for Monitoring Industrial Batch Processes

TL;DR

This paper presents a hybrid unsupervised learning strategy (HULS) for monitoring complex industrial processes that addresses the limitations of traditional Self-Organizing Maps (SOMs) in scenarios with unbalanced data sets and highly correlated process variables.

Abstract

Industrial production processes, especially in the pharmaceutical industry, are complex systems that require continuous monitoring to ensure efficiency, product quality, and safety. This paper presents a hybrid unsupervised learning strategy (HULS) for monitoring complex industrial processes. Addressing the limitations of traditional Self-Organizing Maps (SOMs), especially in scenarios with unbalanced data sets and highly correlated process variables, HULS combines existing unsupervised learning techniques to address these challenges. To evaluate the performance of the HULS concept, comparative experiments are performed based on a laboratory batch

Figures (4)

• Figure 1: Flow chart of the simplified laboratory batch process with measured values flow (F), pressure (P), level (L), and control values pump speed (D), and valve position (Z)
• Figure 2: (a) Recorded training and verification dataset. (b) Resulting quantification error $e_{i\mathbf{v^*}}$ for standard SOM model and (c) for proposed HULS procedure.
• Figure 3: 3D/2D visualization of UM and WT with corresponding WT clusters of the trained standard SOM (a) in comparison to the proposed HULS concept (b).
• Figure 4: (a) Recorded Batch sequence for evaluating anomaly detection performance. (b) Resulting quantification error $e_{i\mathbf{v^*}}$ and class assignment $c_i$ for standard SOM model and (c) for proposed HULS procedure.