Abstract
Micro screws with sizes M2.0 or smaller are widely used to assemble smartphones, tablets, and wearable devices. Due to the delicate nature of the assembled parts, the screws often require tightening torques less than 100 mNm with an accuracy in the range of ±5%. In recent years, high-performance electric drivers have been developed to meet these requirements. However, optimizing screw tightening conditions for micro screws of M2.0 or smaller remains challenging because the axial force during screw tightening cannot be directly measured using conventional methods. Consequently, the optimization process relies on time-consuming and costly trial-and-error methods. To address this issue, this study proposes a system that leverages IoT sensing technology and a machine-learning regression prediction model to accurately estimate the axial force during micro screw tightening. In the data collection phase, a system was designed to sense and capture multivariate time-series data, such as tightening torque, pushing force, and vibration of the electric screwdriver during screw fastening. Subsequently, a machine-learning regression prediction model was developed to accurately estimate the axial force of the screw based on the collected data. Experimental validation using M2 cross-recessed pan head screws demonstrated that the proposed system achieves axial force estimation accuracy within 10%. This study offers a potentially effective approach to optimize screw tightening conditions for micro screws, enabling efficient and reliable assembly processes in the production of electronic devices.
