A Study on Gear Defect Detection via Frequency Analysis Based on DNN

Abstract

In this paper, we introduce a gear defect detection system using frequency analysis based on deep learning. The existing defect diagnosis systems using acoustic analysis use spectrogram, scalogram, and MFCC (Mel-Frequency Cepstral Coefficient) images as inputs to the convolutional neural network (CNN) model to diagnose defects. However, using visualized acoustic data as input to the CNN models requires a lot of computation time. Although computing power has improved, there is a situation in which a processor with low performance is used for reasons such as cost-effectiveness. In this paper, only the sums of frequency bands are used as input to the deep neural network (DNN) model to diagnose the gear fault. This system diagnoses the defects using only a few specific frequency bands, so it ignores unnecessary data and does not require high performance when diagnosing defects because it uses a relatively simple deep learning model for classification. We evaluate the performance of the proposed system through experiments and verify that real-time diagnosis of gears is possible compared to the CNN model. The result showed 95.5% accuracy for 1000 test data, and it took 18.48 ms, so that verified the capability of real-time diagnosis in a low-spec environment. The proposed system is expected to be effectively used to diagnose defects in various sound-based facilities at a low cost.

Detecting Helical Gearbox Defects from Raw Vibration Signal Using Convolutional Neural Networks

“…Vibration data can be accompanied by acoustic signals acquired from microphones. In [12], a pure acoustic analysis based on the gears' sound is mentioned. The spectral data of a few specific frequency bands of the recorded sound are used as input for a deep-learning model in order to diagnose the defects in gears.…”

Mechanical varying non-stationary signal separation method based on instantaneous frequency estimation

“…Considering that the actual mechanical equipment often run in large speed fluctuation condition, its IF often presents curvilinear varying trend. As a result, we simulate a mechanical varying non-stationary signal as shown in formula (12). It contains two sub-signals s 1 (t) and s 2 (t), they reflect modulation property and their instantaneous frequencies are overlapping as shown in formula ( 13) and ( 14), s t ð Þ represents gaussian white noise, sampling frequency is 1024 Hz, the corresponding time waveform is shown in Figure 8.…”

“…As instantaneous frequency reflects the dynamic state of mechanical equipment, so improving the estimation accuracy of IF becomes the core of TLOT. Furthermore, considering the structure of most mechanical parts has spatial symmetry, its fault vibration signal presents essentially cyclostationary in the angle domain, by using order tracking and synchronous averaging, we can convert the non-stationary signal in time domain into cyclostationary signal in angle domain again and establish order cyclostationarity analysis 10 ultimately which is convenient for processing mechanical varying non-stationary signal, such as order spectrum, 11 order cepstrum, 12 envelope order spectrum, 13 order bispectrum, 14 and high-order spectrum. 15 As for time-frequency analysis, considering only from the time domain or frequency domain, we can not obtain the instantaneous time-frequency property which is the core of non-stationary signal processing, 16 the time-frequency analysis provides the joint distribution information in the time domain and frequency domain which can effectively give attention to both time resolution and frequency resolution, it is very suitable for estimating instantaneous frequency and separating mechanical varying non-stationary signal.…”