Introduction reading analysis

作者：张文博 学号： M202310564

Journal
期刊： IEEE TRANSACTIONS ON INDUSTRIAL
INFORMATICS

JCR 分区： Q1

影响因子： 12.3

Title:
Deep Residual Shrinkage Networks for Fault
Diagnosis
神经网络算法在机械智能故障诊断中的应用
 Original text

ROTATING machines are integral to manufacturing, power supply, transportation, and aerospace industries.
However,because these rotating machines operate under harsh working environments, failures unavoidably
occur in their mechanical transmission systems and can result in accidents and economic losses. Accurate
fault diagnosis of mechanical transmission systems can be used to schedule maintenance and extend service
time as well as ensure human safety [1]–[3].The existing fault diagnosis algorithms for mechanical
transmission systems can be classified into two categories, i.e.,signal analysis-based methods and machine-
learning-powered methods [4]. In general, signal-analysis-based fault diagnosis methods identify faults by
detecting the fault-related vibration components or characteristic frequencies. However, for large rotating
machines, the vibration signals are often composed of many different vibration components, including
meshing of gears and rotation of shafts and bearings. Furthermore, when faults are in their early stage, the
fault-related components are often weak and can easily be overwhelmed by other vibration components
and harmonics. As a result, it is often difficult for the traditional signal analysis-based fault diagnosis
methods to identify the fault-related vibration components and characteristics frequencies.
 Original text

Machine-learning-powered fault diagnosis methods, on the other hand, are able to diagnose faults without
identifying the fault-related components and characteristics frequencies. A number of statistical parameters
(e.g., kurtosis, root mean square, energy, and entropy) can be extracted to represent the health states, and
then a classifier (e.g., multiclass support vector machines, one-hidden-layer neural networks, and naïve
Bayes classifier) can be trained to diagnose the faults. However,the extracted statistical parameters are often
not discriminative enough to distinguish the faults, which can lead to low diagnostic accuracy. As a
consequence, finding a discriminative feature set has become a long-standing challenge for machine
learning-powered fault diagnosis [5].
 Original text

In recent years, deep learning methods [6], which refer to the machine learning methods with multiple
levels of nonlinear transformations, have become a useful tool in vibration-based fault diagnosis. To replace
the traditional statistical parameters,deep learning methods automatically learn features from raw vibration
signals, which can yield higher diagnostic accuracy.A variety of deep learning methods have been used in
machine fault diagnosis [7]–[14]. For example, Ince et al. [7] employed a 1-D convolutional neural network
(ConvNet) to learn features from current signals for real-time motor fault diagnosis. Shao et al. [9] applied a
convolutional deep belief network for fault diagnosis of electric locomotive bearings. However, parameter
optimization is often a difficult task for traditional deep learning methods. The gradients of the error
function, which have to be back-propagated layer by layer, gradually become inaccurate after flowing
through a number of layers. As a result, the trainable parameters in the beginning layers (i.e., the layers
close to the input layer) cannot be optimized effectively.
 Original text

Deep residual networks (ResNets) are an attractive variant of ConvNets, which use identity shortcuts to ease
the difficulty of parameter optimization [15]. In ResNets, the gradients not only are back-propagated layer by
layer, but also directly flow back to the beginning layers through identity shortcuts [16]. Due to the improved
parameter optimization, ResNets have been applied for fault diagnosis in a few recent papers [17]–[20].For
example, Ma et al. [17] used a ResNet with demodulated time–frequency features to diagnose planetary
gearboxes under nonstationary running conditions. Zhao et al. [18] used ResNets to fuse multiple sets of
wavelet packet coefficients for fault diagnosis. The advantages of ResNets over classical ConvNets
have been validated in these papers.
Vibration signals collected from large rotating machines, such as wind turbines, manufacturing machines,
and heavy trucks,often contain large amounts of noise. The feature learning ability of ResNets often
decreases when dealing with highly noised vibration signals. The convolutional kernels used in ResNets,
which serve as local feature extractors, may fail to detect the fault-related features due to the interference
of noise. In such a case, the learned high-level features at the output layer are often not discriminative
enough to correctly classify the faults.Therefore, it is necessary to develop new deep learning methods
for vibration-based fault diagnosis of rotating machines under strong background noise.
 Original text

This article develops two deep residual shrinkage networks (DRSNs), i.e., a DRSN with channel-shared
thresholds(DRSN-CS) and a DRSN with channel-wise thresholds (DRSNCW), to improve the feature learning
ability of ResNets from highly noised vibration signals, with the ultimate goal of yielding high diagnostic
accuracy. The major contributions are summarized as follows.
1) Soft thresholding (i.e., a popular shrinkage function) is inserted into the deep architecture as nonlinear
transformation layers, in order to effectively eliminate the noise-related features.
2) The thresholds are adaptively determined using specially designed subnetworks so that each piece of
vibration signal can have its own set of thresholds.
3) Two kinds of thresholds, namely, channel-shared thresholds and channel-wise thresholds, are considered
in soft thresholding, which is the reason for the terms DRSN-CS and DRSN-CW.
The remainder of this article is organized as follows. Section II provides a brief overview of classical ResNets
and a detailed elaboration on the developed DRSN-CS and DRSN-CW. Experimental comparisons are given in
Section III, and Section IV concludes this article.

Length of the abstract:1461 words

 General background of the study

 ROTATING machines are integral to manufacturing, power supply, transportation,

and aerospace industries. However,because these rotating machines operate under
harsh working environments, failures unavoidably occur in their mechanical
transmission systems and can result in accidents and economic losses. Accurate fault
diagnosis of mechanical transmission systems can be used to schedule maintenance
and extend service time as well as ensure human safety [1]–[3].The existing fault
diagnosis algorithms for mechanical transmission systems can be classified into two
categories, i.e.,signal analysis-based methods and machine-learning-powered
methods [4].
 General background of the study

 In general, signal-analysis-based fault diagnosis methods identify faults by detecting

the fault-related vibration components or characteristic frequencies. However, for
large rotating machines, the vibration signals are often composed of many different
vibration components, including meshing of gears and rotation of shafts and
bearings. Furthermore, when faults are in their early stage, the fault-related
components are often weak and can easily be overwhelmed by other vibration
components and harmonics. As a result, it is often difficult for the traditional signal
analysis-based fault diagnosis methods to identify the fault-related vibration
components and characteristics frequencies.
 General background of the study

 Machine-learning-powered fault diagnosis methods, on the other hand, are able to

diagnose faults without identifying the fault-related components and characteristics
frequencies. A number of statistical parameters (e.g., kurtosis, root mean square,
energy, and entropy) can be extracted to represent the health states, and then a
classifier (e.g., multiclass support vector machines, one-hidden-layer neural
networks, and naïve Bayes classifier) can be trained to diagnose the faults.
However,the extracted statistical parameters are often not discriminative enough to
distinguish the faults, which can lead to low diagnostic accuracy. As a consequence,
finding a discriminative feature set has become a long-standing challenge for
machine learning-powered fault diagnosis [5].
 General background of the study-interpretation

In this paper, the importance of rotating machinery is described in the

background of rotating machinery. Then it is pointed out that the
rotating machinery will inevitably fail, which emphasizes the importance
of fault diagnosis, and focuses on the author's research object and
research method.
 Review of previous research

 In recent years, deep learning methods [6], which refer to the machine learning
methods with multiple levels of nonlinear transformations, have become a useful
tool in vibration-based fault diagnosis. To replace the traditional statistical
parameters,deep learning methods automatically learn features from raw vibration
signals, which can yield higher diagnostic accuracy.A variety of deep learning
methods have been used in machine fault diagnosis [7]–[14]. For example, Ince et al.
[7] employed a 1-D convolutional neural network (ConvNet) to learn features from
current signals for real-time motor fault diagnosis. Shao et al. [9] applied a
convolutional deep belief network for fault diagnosis of electric locomotive
bearings. However, parameter optimization is often a difficult task for traditional
deep learning methods. The gradients of the error function, which have to be back-
propagated layer by layer, gradually become inaccurate after flowing through a
number of layers. As a result, the trainable parameters in the beginning layers (i.e.,
the layers close to the input layer) cannot be optimized effectively.
 Review of previous research

 Deep residual networks (ResNets) are an attractive variant of ConvNets, which use identity
shortcuts to ease the difficulty of parameter optimization [15]. In ResNets, the gradients not
only are back-propagated layer by layer, but also directly flow back to the beginning layers
through identity shortcuts [16]. Due to the improved parameter optimization, ResNets have
been applied for fault diagnosis in a few recent papers [17]–[20].For example, Ma et al. [17]
used a ResNet with demodulated time–frequency features to diagnose planetary gearboxes
under nonstationary running conditions. Zhao et al. [18] used ResNets to fuse multiple sets of
wavelet packet coefficients for fault diagnosis. The advantages of ResNets over classical
ConvNets have been validated in these papers.
 Review of previous research-interpretation

It is pointed out that machine learning algorithms have been used in fault
diagnosis in recent years, and many scholars have begun to use deep learning to
solve some problems in public industry fault diagnosis, and deep residual
network is also a relatively powerful deep learning algorithm, but it will have
problems under the interference of a lot of noise, thus leading to the algorithm
proposed in this paper.
 Research gaps and Importance of the study

 Vibration signals collected from large rotating machines, such as wind turbines, manufacturing
machines, and heavy trucks,often contain large amounts of noise. The feature learning ability
of ResNets often decreases when dealing with highly noised vibration signals. The
convolutional kernels used in ResNets,which serve as local feature extractors, may fail to
detect the fault-related features due to the interference of noise. In such a case, the learned
high-level features at the output layer are often not discriminative enough to correctly classify
the faults.Therefore, it is necessary to develop new deep learning methods for vibration-
based fault diagnosis of rotating machines under strong background noise.
 Research gaps and importance of the study-interpretation

This paper points out research gaps where ResNet still faces some challenges.
And the need to develop new deep learning algorithms to solve rotating
machinery fault diagnosis problems.
 Research objectives and methodology

 This article develops two deep residual shrinkage networks (DRSNs), i.e., a DRSN with
channel-shared thresholds(DRSN-CS) and a DRSN with channel-wise thresholds
(DRSNCW), to improve the feature learning ability of ResNets from highly noised
vibration signals, with the ultimate goal of yielding high diagnostic accuracy. The major
contributions are summarized as follows.
 1) Soft thresholding (i.e., a popular shrinkage function) is inserted into the deep
architecture as nonlinear transformation layers, in order to effectively eliminate the
noise-related features.
 2) The thresholds are adaptively determined using specially designed subnetworks so
that each piece of vibration signal can have its own set of thresholds.
 3) Two kinds of thresholds, namely, channel-shared thresholds and channel-wise
thresholds, are considered in soft thresholding, which is the reason for the terms DRSN-
CS and DRSN-CW.
 The remainder of this article is organized as follows. Section II provides a brief overview
of classical ResNets and a detailed elaboration on the developed DRSN-CS and DRSN-CW.
Experimental comparisons are given in Section III, and Section IV concludes this article.
 Research objectives and methodology-interpretation

In this paper, two kinds of deep residual shrinkage networks are developed
for rotating machinery fault diagnosis to improve the ability of learning
features from high noise vibration signals.
 Summary

Through the study of this section, I believe that a good introduction

should include the general background, review of previous research,
research gaps, research importance, research objectives and methods, so
as to make the introduction of the article more organized and better
explain their valuable parts. After the study of these two classes, I learned
to read the introduction of an article and wanted to apply these methods
to my future paper writing.