http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
Deep Learning-based Planetary Gear Fault Diagnosis using Frequency Domain Averaging
Jeongsan Kim(김정산),Iljeok Kim(김일적),Jungchan Kim(김정찬),Taegyu Choi(최태규),Seungchul Lee(이승철) 대한기계학회 2021 대한기계학회 춘추학술대회 Vol.2021 No.4
Planetary gearboxes are widely used throughout industries because it is easy to deal with large gear ratios. The conventional diagnostic method for the planetary gearboxes is to measure the signal with an accelerometer to check the corresponding frequencies to faults types. However, it is difficult to find fault frequencies of mixed planet gears, sun gears, and bearings, since multiple gears are operated simultaneously. In addition, much noise is inherent in the signal. While time synchronous averaging (TSA) is a widely used noise reduction technique that averages multiple vibration signals within one revolution of the gear, tach signal cannot be synchronized in case of planet gears. To overcome this issue, we apply frequency domain averaging, an averaging technique performed in the frequency domain, to planet gear signals to eliminate the unnecessary noise components. In addition, we designed a convolutional neural network (CNN) to distinguish 12 classes of defect signals which consist of planet gear faults, sun gear faults, bearing faults, and combinations of these generated under steady, acceleration, and deceleration conditions. The result shows over 98% accuracy for test data.