볼트의 소리 신호를 이용한 합성곱 신경망 기반 체결력 측정 방법

This paper presents a novel method for measuring the clamping force using sound that occurs during bolt fastening. The resonance frequency of the bolt increases with the progress of the fastening process. This characteristic change is utilized as the feature analyzed by a convolutional neural network (CNN). The clamping force is measured using a load cell, and is then used during labeling for classification. To measure the radiated noise, a microphone is installed near the fastening part. In addition, a signal-processing method is proposed to apply the measurement to deep-learning classification and perform data augmentation. The CNN architecture was modeled, and the fastening force was determined using the classification method. The estimated value was compared with the actual load cell measurements.