CNN‐Based Determination of Bolt Preload in High‐Strength Bolted Connections

Abstract This publication presents the first part of a comprehensive study on a novel methodology for real‐time monitoring of preload in high‐strength bolted connections, applied to a standardised beam‐to‐column frame corner. The approach combines electromechanical impedance (EMI) measurements with convolutional neural networks (CNNs) to enable a scalable, robust, and industrially applicable solution for structural health monitoring (SHM). This first part focuses on demonstrating the applicability of the EMI‐CNN concept, explaining the underlying physical principles, and developing an automated evaluation environment. Two experimental series were conducted on a standardised frame corner specimen with M16 HV bolts. EMI spectra were acquired using surface‐mounted piezoelectric transducers and processed with trained CNN models to quantify the preload. In the conducted experiments, the model achieved a mean absolute error (MAE) of 1.07 % with respect to the applied preload. An interactive software tool was developed to support practical application, including visualisation of the bolted connection, model import, and automated real‐time evaluation of new EMI measurements. The preload level is displayed as a percentage, while an integrated traffic light system enables intuitive real‐time condition assessment. The results confirm the feasibility and reliability of the EMI‐CNN framework for preload determination and provide a solid basis for scalable industrial implementation using pre‐calibrated, sensor‐integrated bolts.