Full length articleResearch on fault diagnosis method of MS-CNN rolling bearing based on local central moment discrepancy
Introduction
In rotating machinery and equipment, a rolling bearing is an important supporting component [1], [2], [3]. The easiest way to prevent irreparable damage to rotating machinery caused by abnormal rolling bearing operating states is to detect the problem early and distinguish between the fault kinds [4], [5], [6]. Equipment failure detection and diagnosis is a very challenging problem [7], [8]. Researchers have proposed a variety of methods to solve the problem of equipment fault diagnosis, including time–frequency analysis [9], [10], fuzzy-based control [11], sparse signal decomposition [12], etc. In current years, the intelligent fault detection approach based on a data-driven has received considerable attention [13], [14]. The data-driven intelligent fault detection method could gain knowledge from the collected equipment operation data. Then applied the learned knowledge to perform an intelligent fault diagnosis of the equipment. Effectively converting the fault diagnosis problem, which previously required professional skills and extensive experience, into a classification problem [15]. Until present, researchers have presented a variety of intelligent defect diagnostic models based on deep learning (DL), such as convolutional neural networks (CNN) [16], Continuous Deep Belief Network (CDBN) [17], Residual building unit, Soft thresholding and Global context (RSG) [18]. The DL model can better adapt to diverse learning tasks and extract feature information from fault data in an adaptable approach.
The traditional DL approach offers high accuracy in fault detection. Unfortunately, to train a failure diagnostic model based on DL, a substantial amount of labeled data from target equipment is required. In an actual industrial environment, bearing equipment is frequently subjected to different working conditions. This makes obtaining a large number of bearing data sets with comprehensive fault categories challenging [19], [20]. When the rotational speed and load of the equipment change, for example, the characteristic distribution of the gathered data will depart from the prior state's characteristic distribution. The diagnostic system's detection performance will suffer as a result of these alterations [21], [22]. The abovementioned limitations prevent intelligent fault techniques from being used in real-world situations. The huge quantity of labeled data that may be taught in practice is usually derived from data collected during operations in a variety of settings. As a result, researchers have developed intelligent fault diagnoses based on transfer learning (TL) [23], [24].
The approach of TL has been successfully confirmed in fault diagnosis at this time [25], [26], [27]. The goal of TL is to improve the generalization task between source domain (Ds) and target domain (Dt) [28]. The model trains on data from the Ds try to learn certain common data features, and then builds a model with better generalization abilities to test data from the Dt [29]. A definition of transfer learning was put out by Pan et al. in their review [28]. TL's goal is to provide models the ability to apply information acquired from one activity to another. TL involves two key ideas, namely domain and task. Typically, we use D for the domain and T for the task. With the help of the knowledge in the given Ds, the learning of Dt is realized. Under specific situations, the domain adaption (DA) strategy can help the model extract transferable information in the Ds, according to the researchers [30], [31], [32], [33]. In the field of fault diagnostics, the Maximum Mean Discrepancy (MMD) has been frequently employed [34]. DA approaches based on MMD commonly use a kernel function to map the Ds and Dt data into a high-dimensional space and then minimize the distance between the two domains. Schwendemann presented a Layered MMD and CNN based bearing defect diagnostic system [35]. To accomplish the cross functioning condition diagnostics of bearings, Che et al. suggested a DBN network based on multi-kernel MMD (MK-MMD) [36]. To overcome the problem of data offset between Ds and Dt, Wu et al. presented an adversarial domain adaptive CNN structure based on MMD [37]. Zheng et al. proposed a locality-preserving projection-based transfer method for embedding data into subspaces and using MMD to reduce distributional differences between different datasets [38].
Rolling bearings usually work under complex conditions, such as changes in speed and load. When the working conditions change, the condition monitoring and fault diagnosis of the equipment will face more difficult challenges. Although the above intelligent defect detection approach based on transfer learning has a high detection ability, finding a suitable mapping kernel function when utilizing MMD for measurement learning is critical. The relevant kernel function must be updated when the data methods of the Ds and Dt are adjusted. Furthermore, measuring in a high-dimensional space will lengthen the model's computation time. To address the above-mentioned issues, this research offers a TL fault diagnostic technique based on the Central Moment Discrepancy (CMD) [39]: multi-scale CNN (MS-CNN) migration fault detection of rolling bearings based on the Local CMD (LCMD). The approach locates the same subdomain, in which samples with the same label but from distinct domains are found. In this domain, the difference in central moment across domains is therefore reduced. The following are the primary contributions of this paper.
- (1)
The approach based on MS-CNN-LCMD extracts information that can be transferred between Ds and Dt more effectively.
- (2)
The suggested technique makes greater use of transfer learning across multiple domains by avoiding the need for numerous kernel parameter selections.
- (3)
LCMD performs better in cross-domain diagnosis than classic CMD metric learning.
The rest of this manuscript is laid out as follows. Some fault diagnostic research on CNN and DA are reviewed in Section 2. The proposed approach is described in detail in Section 3. Section 4 delves into the details of the experiments and their outcomes and provides a discussion of the results. Finally, Section 5 is the conclusion.
Section snippets
Description of domain alignment problem in class subspace
The specific transfer problem is defined as follows. Knowledge related to Dt and learning task Tt is acquired on Ds and source domain task Ts. Use this knowledge to complete the study of Tt, under the condition of Ds ≠ Dt. The dataset of the Ds data can be represented as throughout the migration fault diagnostic procedure, is the ith sample of the Ds, and is the label of the ith sample. The Dt data is expressed as, is the jth sample of the target domain,
Establishment of LCMD
A DSAN diagnostic approach based on LMMD suggested by Zhu [45] influenced the LCMD measuring method. The author presents the idea of a sub-domain in this technique. The circumstances between the existing Ds and Dt separate the samples from both domains into a common sub-domain. Then, in the subdomain, align the distribution of Ds and Dt. In the training phase, standard CMD-based measurement learning directly evaluates the global difference between Ds and Dt data and minimizes the loss between
Experiment with migration fault diagnostics and experimental analysis
This section discusses the data processing methods utilized in the experiment and details the data used in the experiment. Two typical cross-working condition migration experiments are used to validate the suggested technique.
Conclusion
The MS-CNN migration learning approach, which is based on the difference in local central moment, is utilized in this study to detect the cross working condition defect of rolling bearings at various speeds and loads. By sharing parameters, the model projects the two-domain data into the same subspace. To lessen the distribution variation between the two domains, the data of the two domains are aligned according to the category connection in the subspace. The LCMD technique for local category
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgments
This work was supported in part by the National Natural Science Foundation of China under Grant 52075470, in part by the Central Government Guides Local Science and Technology Development Foundation under Grant 206Z4301G, in part by the Introduction of Foreign Intellectual Project of Hebei Province, in part by the Cultivation Project for Basic Research and Innovation of Yanshan University under Grant 2021LGZD006.
References (47)
- et al.
Multi-mode data augmentation and fault diagnosis of rotating machinery using modified ACGAN designed with new framework
Adv. Eng. Inf.
(2022) - et al.
Intelligent fault diagnosis for rotating machinery using deep Q-network based health state classification: A deep reinforcement learning approach
Adv. Eng. Inf.
(2019) - et al.
Bearing fault diagnosis via generalized logarithm sparse regularization
Mech. Syst. Sig. Process.
(2022) - et al.
Event-driven NN adaptive fixed-time control for nonlinear systems with guaranteed performance
J. Franklin Inst.
(2022) - et al.
Sparse representation based on parametric impulsive dictionary design for bearing fault diagnosis
Mech. Syst. Sig. Process.
(2019) - et al.
Exponential stability of nonlinear state-dependent delayed impulsive systems with applications
Nonlinear analysis: Hybrid systemsl.
(2021) - et al.
Multiscale inverted residual convolutional neural network for intelligent diagnosis of bearings under variable load condition
Measurement.
(2022) - et al.
Intelligent diagnosis of mechanical faults of in-wheel motor based on improved artificial hydrocarbon networks
ISA Trans.
(2022) - et al.
Intelligent fault diagnosis of rotating machinery using support vector machine with ant colony algorithm for synchronous feature selection and parameter optimization
Neurocomputing.
(2015) - et al.
An improved convolutional neural network with an adaptable learning rate towards multi-signal fault diagnosis of hydraulic piston pump
Adv. Eng. Inf.
(2021)
“Rolling bearing fault detection using continuous deep belief network with locally linear embedding
Comput. Ind.
A novel RSG-based intelligent bearing fault diagnosis method for motors in high-noise industrial environment
Adv. Eng. Inf.
Partial transfer learning in machinery cross-domain fault diagnostics using class-weighted adversarial networks
Neural Networks.
A deep domain adaption model with multi-task networks for planetary gearbox fault diagnosis
Neurocomputing.
A multi-source ensemble domain adaptation method for rotary machine fault diagnosis
Measurement.
Fault diagnostics between different type of components: A transfer learning approach
Appl. Soft Comput.
A Novel Deep Multi-Source Domain Adaptation Framework for Bearing Fault Diagnosis Based on Feature-level and Task-specific Distribution Alignment
Measurement.
A new deep domain adaptation method with joint adversarial training for online detection of bearing early fault
ISA Trans.
A multi-representation-based domain adaptation network for fault diagnosis
Measurement.
A novel deep convolution multi-adversarial domain adaptation model for rolling bearing fault diagnosis
Measurement.
A novel gas turbine fault diagnosis method based on transfer learning with CNN
Measurement.
Bearing fault diagnosis with intermediate domain based Layered Maximum Mean Discrepancy: A new transfer learning approach
Eng. Appl. Artif. Intell.
Domain adaptive deep belief network for rolling bearing fault diagnosis
Comput. Indus. Eng.
Cited by (12)
A subtle defect recognition method for catenary fastener in high-speed railroad using destruction and reconstruction learning
2024, Advanced Engineering InformaticsSubdomain adaptation joint attention network enabled two-stage strategy towards few-shot fault diagnosis of LRE turbopump
2024, Advanced Engineering InformaticsAdaptive variational autoencoding generative adversarial networks for rolling bearing fault diagnosis
2023, Advanced Engineering InformaticsIntelligent fault diagnosis of ultrasonic motors based on graph-regularized CNN-BiLSTM
2024, Measurement Science and TechnologyCross-domain open-set rolling bearing fault diagnosis based on feature improvement adversarial network under noise condition
2024, Journal of Intelligent and Fuzzy Systems