Automated Testing and Validation of Digital Twins Using Real-Time Production Data : A Case Study of an Industrial Heating Furnace

Sammanfattning

The increasing adoption of digital twins in industrial settings has revolutionized manufacturing processes by enabling real-time monitoring and control. However, ensuring the accuracy and reliability of digital twins remains a significant challenge, particularly in dynamic production environments where conditions constantly change and traditional validation approaches rely heavily on offline data and manualtuning.This thesis presents a systematic approach for automated testing and validation of digital twins using real-time production data from an industrial heating furnace at Bharat Forge Kilsta AB. The methodology introduces a novel snapshot-based testing framework that enables continuous monitoring of the digital twin’s performance by creating and comparing data snapshots from the production environment with the digital twin’s output.The implementation leverages a comprehensive testing architecture that integrates OPC-UA servers, Kafka message brokers, and a custom-developed Faust agent for real-time data processing. The framework employs probability density functions (PDF) and cumulative density functions (CDF) to analyze temperature and position deviations across multiple sensors, providing quantitative measures of the digital twin’s accuracy.Results demonstrate the effectiveness of the automated testing approach in both normal production and warm holding modes. The analysis shows that FilteredTemp, which represents temperature readings processed to remove noise and inconsistencies, exhibits improved precision compared to ActualTemp, which reflects the raw, unfiltered sensor measurements. This research contributes to the field by providinga robust methodology for continuous digital twin validation, essential for maintaining accuracy in AI-driven industrial process control systems. The findings have significant implications for improving the reliability of digital twins in manufacturing environments.