Leveraging AI and MLOps for IoT-Edge-Cloud Industrial Digital Twins: a Practical Case Study

Artificial Intelligence (AI) and Machine Learning Operations (MLOps) are crucial for the practical application of Digital Twins (DT) in industrial environments. DT facilitate enhanced modelling, forecasting, and optimisation of assets and processes, while MLOps ensure resilient, reliable, and continuously updated models. This paper proposes an industrial-grade DT example for the manufacturing industry, utilising cloud/edge-based AI technologies from the Microsoft Azure ecosystem. Data handling and ML model training are performed at the cloud layer, while models are deployed and maintained at the edge. To enhance interoperability and portability, trained ML models are converted into lightweight and portable models using the Open Neural Network Exchange format (ONNX) standard. Continuous Integration and Continuous Deployment (CI/CD) are performed on an industrial edge device using GitHub Actions. REST APIs are implemented using standard HTTP methods to manage the IoT Central edge manifest. In-the-field experimental results demonstrate the effectiveness of our DT prototype in a specific industrial use case.