Intelligent Optimization of Resin Level in Industrial Silos Using Radar Sensors and Random Forest Algorithms

Abstract

The implementation of Siemens SITRANS LR560 radar sensors in industrial silos, combined with a predictive system based on machine learning, has optimized the control of polypropylene resin levels. A 75% reduction in maintenance costs and an 87.5% reduction in downtime were achieved, improving operational efficiency. A Random Forest model was used to predict failures, validated with class balancing techniques and K-Fold cross-validation, reaching an accuracy of over 95%. The use of TIA Portal to integrate the SCADA system enables real-time monitoring and the generation of alerts for critical events. The results were compared with previous studies, demonstrating that the use of artificial intelligence and IoT in Industry 4.0 improves the reliability of granular material storage and distribution. Future improvements are recommended, including optimizing communication infrastructure in industrial environments, real-time data processing, and enhancing the decision-making process.