Instrumentation & Measurement Magazine 26-2 - 45

Fig. 4. Confusion matrix corresponding to the prediction results of the RGB-TDResNet model .
rate dropped from 71.78%
to 44%. The detection accuracy
of the 2D-Resnet
dropped from 94% to
55.78% and the RGB-Resnet
dropped from 98.67% to
59.55%. However, the accuracy
of the RGB-Resnet
using the RGB image input
is generally about 5%
higher than that of the 2DResnet,
which reflects the
robustness of the RGB image
in describing the fault
properties. Comparatively,
Table 1 - Prediction results of the
RGB-TDResNet model
Fault type
BF007
BF014
IR007
IR014
Normal
01
87
86
90
90
90
02
85
85
90
87
90
03
90
80
90
88
90
sum will approach a Gaussian distribution as the number of
noise sources increases. During the validation experiments,
the Gaussian white noise was selected and superposed on the
CWRU dataset to simulate the noise interference in reality.
All the diagnosis models were trained by the original signal
of 0 HP without noise and tested on the samples which were
generated through superposing the Gaussian white noise on
the original signals with
different signal-to-noise ratios
(SNRs), i.e., SNR = 10
dB, SNR = 5 dB, SNR = 2
dB and SNR = 0 dB. Among
100 samples in the target
domain under the noise
interference, 10 labeled
images were randomly selected
and employed to
fine-tune the model while
the other 90 unlabeled
samples were used for the
testing. Fig. 5 illustrates
the diagnosis accuracies of
all the six models on the
test samples involved with
the noises. We can see that
noise interference led to a
negative effect on diagnosis
accuracy. Method1 is
very sensitive to changes
in noise, and the success
April 2023
the improved RGB-ResNet, i.e., the RGB-DResnet and TRGBResnet,
still provides satisfactory diagnosis results with
accuracies of 74.44% and 96%, respectively, due to their usage
of the dense connection and the transfer learning mechanism
respectively. RGB-DResnet yields a 25% accuracy drop, while
TRGB-Resnet yields a 3.33% accuracy drop with the increasing
noise interference. Moreover, the RGB-TDResnet keeps a consistently
high level of fault-diagnosis performance, yielding
only a 2.44% accuracy drop from 100% to 97.56%. The RGBTDResNet
combines the advantages of both so as to achieve
the most stable and accurate diagnosis results.
Fig. 6 gives the confusion matrix, and Table 2 gives the prediction
results of the RGB-TDResNet on classifying different
fault types in the test samples. Fig. 6a, Fig. 6b, Fig. 6c and Fig.
6d correspond to the noise interference of 10 dB, 5 dB, 2 dB and
0 dB, respectively. We can see that the RGB-ResNet model provides
a 98.67% prediction accuracy for the test samples with
noise of 10 dB, 91.56% of 5 dB, 71.55% of 2 dB and 59.55% of 0
Fig. 5. Diagnosis results under noise interference.
IEEE Instrumentation & Measurement Magazine
45

Instrumentation & Measurement Magazine 26-2

Table of Contents for the Digital Edition of Instrumentation & Measurement Magazine 26-2

Instrumentation & Measurement Magazine 26-2 - Cover1
Instrumentation & Measurement Magazine 26-2 - Cover2
Instrumentation & Measurement Magazine 26-2 - 1
Instrumentation & Measurement Magazine 26-2 - 2
Instrumentation & Measurement Magazine 26-2 - 3
Instrumentation & Measurement Magazine 26-2 - 4
Instrumentation & Measurement Magazine 26-2 - 5
Instrumentation & Measurement Magazine 26-2 - 6
Instrumentation & Measurement Magazine 26-2 - 7
Instrumentation & Measurement Magazine 26-2 - 8
Instrumentation & Measurement Magazine 26-2 - 9
Instrumentation & Measurement Magazine 26-2 - 10
Instrumentation & Measurement Magazine 26-2 - 11
Instrumentation & Measurement Magazine 26-2 - 12
Instrumentation & Measurement Magazine 26-2 - 13
Instrumentation & Measurement Magazine 26-2 - 14
Instrumentation & Measurement Magazine 26-2 - 15
Instrumentation & Measurement Magazine 26-2 - 16
Instrumentation & Measurement Magazine 26-2 - 17
Instrumentation & Measurement Magazine 26-2 - 18
Instrumentation & Measurement Magazine 26-2 - 19
Instrumentation & Measurement Magazine 26-2 - 20
Instrumentation & Measurement Magazine 26-2 - 21
Instrumentation & Measurement Magazine 26-2 - 22
Instrumentation & Measurement Magazine 26-2 - 23
Instrumentation & Measurement Magazine 26-2 - 24
Instrumentation & Measurement Magazine 26-2 - 25
Instrumentation & Measurement Magazine 26-2 - 26
Instrumentation & Measurement Magazine 26-2 - 27
Instrumentation & Measurement Magazine 26-2 - 28
Instrumentation & Measurement Magazine 26-2 - 29
Instrumentation & Measurement Magazine 26-2 - 30
Instrumentation & Measurement Magazine 26-2 - 31
Instrumentation & Measurement Magazine 26-2 - 32
Instrumentation & Measurement Magazine 26-2 - 33
Instrumentation & Measurement Magazine 26-2 - 34
Instrumentation & Measurement Magazine 26-2 - 35
Instrumentation & Measurement Magazine 26-2 - 36
Instrumentation & Measurement Magazine 26-2 - 37
Instrumentation & Measurement Magazine 26-2 - 38
Instrumentation & Measurement Magazine 26-2 - 39
Instrumentation & Measurement Magazine 26-2 - 40
Instrumentation & Measurement Magazine 26-2 - 41
Instrumentation & Measurement Magazine 26-2 - 42
Instrumentation & Measurement Magazine 26-2 - 43
Instrumentation & Measurement Magazine 26-2 - 44
Instrumentation & Measurement Magazine 26-2 - 45
Instrumentation & Measurement Magazine 26-2 - 46
Instrumentation & Measurement Magazine 26-2 - 47
Instrumentation & Measurement Magazine 26-2 - 48
Instrumentation & Measurement Magazine 26-2 - 49
Instrumentation & Measurement Magazine 26-2 - 50
Instrumentation & Measurement Magazine 26-2 - 51
Instrumentation & Measurement Magazine 26-2 - 52
Instrumentation & Measurement Magazine 26-2 - 53
Instrumentation & Measurement Magazine 26-2 - 54
Instrumentation & Measurement Magazine 26-2 - 55
Instrumentation & Measurement Magazine 26-2 - 56
Instrumentation & Measurement Magazine 26-2 - 57
Instrumentation & Measurement Magazine 26-2 - 58
Instrumentation & Measurement Magazine 26-2 - 59
Instrumentation & Measurement Magazine 26-2 - 60
Instrumentation & Measurement Magazine 26-2 - 61
Instrumentation & Measurement Magazine 26-2 - 62
Instrumentation & Measurement Magazine 26-2 - 63
Instrumentation & Measurement Magazine 26-2 - Cover3
Instrumentation & Measurement Magazine 26-2 - Cover4
https://www.nxtbook.com/allen/iamm/26-6
https://www.nxtbook.com/allen/iamm/26-5
https://www.nxtbook.com/allen/iamm/26-4
https://www.nxtbook.com/allen/iamm/26-3
https://www.nxtbook.com/allen/iamm/26-2
https://www.nxtbook.com/allen/iamm/26-1
https://www.nxtbook.com/allen/iamm/25-9
https://www.nxtbook.com/allen/iamm/25-8
https://www.nxtbook.com/allen/iamm/25-7
https://www.nxtbook.com/allen/iamm/25-6
https://www.nxtbook.com/allen/iamm/25-5
https://www.nxtbook.com/allen/iamm/25-4
https://www.nxtbook.com/allen/iamm/25-3
https://www.nxtbook.com/allen/iamm/instrumentation-measurement-magazine-25-2
https://www.nxtbook.com/allen/iamm/25-1
https://www.nxtbook.com/allen/iamm/24-9
https://www.nxtbook.com/allen/iamm/24-7
https://www.nxtbook.com/allen/iamm/24-8
https://www.nxtbook.com/allen/iamm/24-6
https://www.nxtbook.com/allen/iamm/24-5
https://www.nxtbook.com/allen/iamm/24-4
https://www.nxtbook.com/allen/iamm/24-3
https://www.nxtbook.com/allen/iamm/24-2
https://www.nxtbook.com/allen/iamm/24-1
https://www.nxtbook.com/allen/iamm/23-9
https://www.nxtbook.com/allen/iamm/23-8
https://www.nxtbook.com/allen/iamm/23-6
https://www.nxtbook.com/allen/iamm/23-5
https://www.nxtbook.com/allen/iamm/23-2
https://www.nxtbook.com/allen/iamm/23-3
https://www.nxtbook.com/allen/iamm/23-4
https://www.nxtbookmedia.com