Hybrid model monitoring approach for screw tightening processes

In order to ensure the quality and sufficient clamping force of a bolted joint, real-time monitoring using measurement data is essential. Traditionally, expert systems or control windows analyze process signals such as torque and rotation angle. However, due to the large volume of data, some anomalies or faulty screw connections may go undetected. This article introduces a new hybrid method that combines machine learning with control windows. The method works by scaling both torque and rotation angle, making it applicable to all types of bolted connections. The algorithm requires only 20 training curves, significantly reducing the time needed for setup while still maintaining high accuracy.