A Deep Learning Model with Adaptive Learning Rate for Fault Diagnosis
Xiaodong Zhai
1
, Fei Qiao
1
1. School of Electronics and Information Engineering, Tongji University, Shanghai 201804
E-mail: 1710332@tongji.edu.cn, fqiao@tongji.edu.cn
Abstract: With the increasing amount of data in the field of equipment fault diagnosis, deep learning is playing an increasingly
important role in the process of fault diagnosis, during which the timeliness requirement is high and the fault diagnosis results
need to be obtained accurately and timely. However, with the increase of network layers, the training time of deep learning model
becomes longer. Learning rate in the deep learning model plays an important role in the process of model training, and a
well-designed learning rate adjustment strategy can effectively reduce the training time and satisfy the requirements of fault
diagnosis. At present, some deep learning models usually adopt a globally uniform learning rate strategy, which is unreasonable
for different parameters. This paper has designed an adaptive learning rate strategy for the parameters of weight and bias
respectively in deep learning model. Specifically, the strategy contains a learning rate strategy based on stochastic gradient
descent method for weight, and a power exponential learning rate strategy for bias. Experiments are carried out to validate the
effectiveness of proposed learning rate strategy. Results suggest that the strategy can reduce the training time and reconstruction
error rate of deep learning model, and improve the classification accuracy of fault diagnosis.
Key Words: Deep learning, Learning rate, Adaptive, Fault diagnosis
1 Introduction
With the development of modern industrial technology,
the safety, stability, reliability and operation efficiency of
equipment have become the core competitiveness of
manufacturing enterprises [1], and equipment management
has become an important field in enterprise management. In
the process of production, the performance of equipment
deteriorates with the increase of service time, and various
faults will occur in the process of equipment operation.
When the equipment fails, the production efficiency will be
reduced. More seriously, the equipment will be shut down,
and malignant accidents such as machine damage and
human death will occur. Therefore, it is particularly
important to find and identify the types and locations of
faults in time. With the development of computer
technology, many artificial intelligence algorithms have
been applied in the field of equipment fault diagnosis. It is
predicted that the growing Internet of Things will connect
30 billion devices by 2020 [2], and the huge amount of data
will also promote the innovation of the monitoring process
of the physical network system of industrial 4.0. With the
increasing amount of data, the advantages of deep learning
using in dealing with large-scale data are highlighted.
The motivation of deep learning is to build and simulate
the neural network of human brain for analysis and learning.
It imitates the mechanism of human brain to interpret data,
such as images, sounds and texts [3-5]. Deep learning is a
multi-layer neural network model essentially. By combining
low-level features, we can get a higher-level and more
abstract feature representation to discover the distributed
feature representation of data. At the same time, it weakens
the adverse effects of unrelated factors and improves the
accuracy of classification and prediction [6]. Meanwhile, the
excellent performance of deep learning is mainly based on a
*
This work is supported by National Natural Science Foundation,
China(No. 71690234, 61873191), National Science and Technology Major
Project (2017-V-0011-0063) and the National Key R&D Program, China
(No. 2017YFE0101400).
large number of training data and deep-level network
structure, as a result, the training time of deep learning
model is longer than other machine learning algorithms
generally [7]. Therefore, how to speed up the training time
of deep learning model is a problem which is worth of
intensive study, especially when it is applied in engineering
practice.
2 Related Work
Traditional fault diagnosis methods include model driven
methods, knowledge driven methods, and data driven
methods. However, the first two methods are often limited
by professional technology, expert experience and other
knowledge. In addition, with the continuous development of
equipment status monitoring technology, more and more
equipment status data can be utilized. As a result, data
driven methods based on machine learning and artificial
intelligence have attracted people's attention in recent years
[8-9]. Data driven methods can discover the intrinsic law of
equipment status trends and estimate the fault types of
equipment by advanced methods based on equipment status
data. With the increasing amount of equipment status data,
more and more attention has been paid to the deep learning
method in machine learning.
There are two significant parameters in deep learning
model, which are weight and bias. However, traditional
deep learning models often use a global uniform constant
parameter for these two parameters, and the setting of this
constant parameter requires previous experience.
Meanwhile, it should be noted that there are a large number
of weight and bias parameters in deep learning model, and
they are two different types of parameters. Different
parameters play different roles. With this in mind, it is
unreasonable to provide the same learning rate strategy for
different parameters. A global uniform learning rate is not
necessarily suitable for all parameters, and it will reduce the
iteration efficiency and increase the model training time of
deep learning model.
At present, there have been some studies on the
adjustment strategies of the learning rate in the deep
2020 IEEE 9th Data Driven Control and Learning Systems Conference
November 20-22, 2020, Liuzhou, China
978-1-7281-5922-5/20/$31.00 ©2020 IEEE
668
Authorized licensed use limited to: Carleton University. Downloaded on June 20,2021 at 01:24:34 UTC from IEEE Xplore. Restrictions apply.