A DI-SRM Model for Production Bottleneck Prediction in Flexible Production Line

Abstract

In the context of Industrial Internet of Things (IIoT) flexible production lines, accurately predicting production bottlenecks is crucial for optimizing efficiency and resource allocation. However, the dynamic and uncertain nature of these processes poses significant challenges. This study introduces a novel bottleneck prediction method by integrating the Drift Index (DI) with a Stacked Regression Model (SRM). This study marks the first application of stacked ensemble learning techniques in predicting bottlenecks within IIoT-enabled flexible production lines, leading to notable improvements in prediction accuracy and model robustness. The proposed method utilizes both real-time and historical data collected from IoT devices, encompassing three core steps: bottleneck data analysis, quantification of the drift index, and construction of the stacked regression model. By incorporating multiple production parameters such as equipment utilization and queue length, the method employs advanced time series analysis to forecast potential bottleneck drifts. Experimental results confirm that the DI-SRM model achieves high prediction accuracy and real-time responsiveness, effectively addressing the challenges of dynamic production environments. This approach provides reliable decision support for production scheduling and resource allocation, thereby optimizing production efficiency and enhancing market competitiveness.

Keywords

• production line
• drift index
• stacked model
• production bottleneck prediction