Fault Diagnosis for Rotating Machinery Using Vibration Measurement Deep Statistical Feature Learning

Abstract

Fault diagnosis is important for the maintenance of rotating machinery. The detection of faults and fault patterns is a challenging part of machinery fault diagnosis. To tackle this problem, a model for deep statistical feature learning from vibration measurements of rotating machinery is presented in this paper. Vibration sensor signals collected from rotating mechanical systems are represented in the time, frequency, and time-frequency domains, each of which is then used to produce a statistical feature set. For learning statistical features, real-value Gaussian-Bernoulli restricted Boltzmann machines (GRBMs) are stacked to develop a Gaussian-Bernoulli deep Boltzmann machine (GDBM). The suggested approach is applied as a deep statistical feature learning tool for both gearbox and bearing systems. The fault classification performances in experiments using this approach are 95.17% for the gearbox, and 91.75% for the bearing system. The proposed approach is compared to such standard methods as a support vector machine, GRBM and a combination model. In experiments, the best fault classification rate was detected using the proposed model. The results show that deep learning with statistical feature extraction has an essential improvement potential for diagnosing rotating machinery faults.

Keywords: deep learning; fault diagnosis; rotating machinery; statistical feature; vibration sensor.

Figure 1

Illustration of the network connections with a GRBM. Note the GRBM exhibits same structure compared to its RBM counterpart.

Figure 2

Schematic of the three-layer GDBM: (a) network structure; and (b) pretraining and composition of the GDBM.

Figure 3

Flowchart of the deep statistical feature learning technique for the fault diagnosis of the rotating machinery.

Figure 4

Gearbox fault diagnosis configurations: (a) experimental set-up; and (b) three different faulty gears and five different faulty pinions.

Figure 5

Fault diagnosis configurations for the rolling element bearings: (a) experimental set-up; and (b) 3 different faulty bearings with an inner race fault (left), an outer race fault (middle) and a ball fault (right), respectively.

Figure 6

Time domain features for the gearbox fault diagnosis: (a) time domain waveform of the first signal; (b) time domain statistical features of the first signal; (c) time domain waveforms of the 3600 collected signals; and (d) time domain statistical features of the 3600 collected signals.

Figure 6

Time domain features for the gearbox fault diagnosis: (a) time domain waveform of the first signal; (b) time domain statistical features of the first signal; (c) time domain waveforms of the 3600 collected signals; and (d) time domain statistical features of the 3600 collected signals.

Figure 7

Frequency domain features for the gearbox fault diagnosis: (a) frequency domain representation of the first signal; (b) frequency domain statistical features of the first signal; (c) frequency domain representations of all the collected 3600 signals; and (d) frequency domain statistical features of all the collected 3600 signals.

Figure 8

Time-frequency domain features for the gearbox fault diagnosis: (a) WPT representation of the first signal; (b) time-frequency domain statistical features of the first signal; (c) time-frequency domain representations of all the collected 3600 signals; and (d) time-frequency domain statistical features of all the collected 3600 signals.

Figure 9

Bearing fault diagnosis experiments: (a) the time domain signals; (b) the time domain statistical features; (c) the frequency domain representations; (d) the frequency domain statistical features; (e) the WPT results; and (f) the time-frequency domain statistical features.

Figure 9

Bearing fault diagnosis experiments: (a) the time domain signals; (b) the time domain statistical features; (c) the frequency domain representations; (d) the frequency domain statistical features; (e) the WPT results; and (f) the time-frequency domain statistical features.

Figure 10

Relationship between the classification rate and the number of the modeling epochs: (a) classification rates v.s. pretraining epochs; and (b) classification rates vs. fine-tuning epochs of the time-frequency domain GDBM models.

Figure 11

Bearing fault diagnosis results in different domains: (a) the time domain; (b) the frequency domain; and (c) the time-frequency domain.