Belt rotation in pipe conveyors: Development of an overlap monitoring system using digital twins, industrial Internet of things, and autoregressive language models

Belt rotation is a critical failure mode in pipe conveyors. However, overlap position monitoring remains unexplored. Existing monitoring solutions are expensive and lack scalability, primarily relying on detecting angle variations beyond predefined thresholds. We developed a digital twin of a low-cost test rig for measuring belt rotation using 3D prototyping, infrared distance sensors, and a 3-axis accelerometer. We implemented a controller-responder industrial Internet of Things network using ESP-NOW protocol. Applying a minimum viable prompt, we fine-tuned an autoregressive language model for predicting the overlap position. We then developed a software application for cloud computing and local storage of measurements. Additionally, we developed a human–machine interface for future real-case applications. The bias and precision for the predicted angle (0.2294° ± 11.8822°), angular speed (0.1630 ± 0.7344 RPM), number of laps (0.0001336 ± 0.0448 laps), diameter (–0.1984 ± 6.3743 mm), and outer-edge side (100 % agreement) were suitable for the intended use.