Abstract:As one of the representative unsupervised data augmentation methods, generative adversarial networks (GANs) have the potential to solve the problem of insufficient samples in fault diagnosis of rotating machinery. However, the existing unsupervised GANs are usually incapable of simultaneously generating multi-mode fault samples and have some shortcomings such as mode collapse and gradient vanishing. To overcome these deficiencies, a supervised model called modified auxiliary classifier GAN (MACGAN) designed with new framework is proposed in this paper. Firstly, a new ACGAN framework is developed by adding an independent classifier to improve the compatibility between the classification and discrimination. Secondly, the Wasserstein distance is introduced in the new loss functions to overcome mode collapse and gradient vanishing. Finally, to achieve stable training, a spectral normalization is used to replace the weight clipping to constrain the weight parameters of discriminator. The proposed method is applied to fault diagnosis of bearing and gear. Compared with the existing GANs, the proposed method can more efficiently generate multi-mode fault samples with higher qualities, which can be used to assist the training of deep learning-based fault diagnosis models with high accuracy and good stability.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is the problem of insufficient samples in the fault diagnosis of rotating machinery. Specifically, the paper proposes solutions to the following challenges: 1. **Multi - mode data augmentation**: Existing unsupervised generative adversarial networks (GANs) are usually unable to generate fault samples of multiple modes simultaneously, and have some defects, such as mode collapse and gradient vanishing. These defects limit their performance in practical applications. 2. **Over - fitting problem caused by insufficient samples**: In engineering practice, it is difficult to obtain a sufficient number of labeled fault samples, which will lead to a serious over - fitting problem in supervised learning models. Therefore, how to use a limited number of labeled fault samples to train an accurate and reliable deep - learning fault - diagnosis model is an urgent problem to be solved. To solve these problems, the paper proposes an improved Modified Auxiliary Classifier GAN (MACGAN), and its main innovations include: - **Introducing a new framework of independent classifier**: By adding an independent classifier to improve the compatibility between classification and discrimination, so as to generate high - quality multi - mode fault samples more efficiently. - **Introducing Wasserstein distance**: Introduce Wasserstein distance in the new loss function to effectively solve the problems of mode collapse and gradient vanishing. - **Using spectral normalization (SN)**: Replace weight clipping with spectral normalization to constrain the weight parameters of the discriminator, so as to achieve a stable training process. Through these improvements, MACGAN can generate high - quality multi - mode fault samples more efficiently in the fault diagnosis of rotating machinery, thereby assisting the training of deep - learning fault - diagnosis models and improving the accuracy and stability of diagnosis. ### Formula summary The following are the main formulas involved in the paper: 1. **The basic loss function of ACGAN**: \[ L_{\text{Source}}=\mathbb{E}_{x\sim P_r(x)}[\log D(x)]+\mathbb{E}_{z\sim P_z(z)}[\log(1 - D(G(z, c_g)))] \] \[ L_{\text{Class}}=\mathbb{E}_{x\sim P_r(x)}[-\log P(c = c_r|x)]+\mathbb{E}_{z\sim P_z(z)}[-\log P(c = c_g|G(z, c_g))] \] 2. **Wasserstein distance**: \[ W(P_r(x), P_g(y))=\inf_{\gamma\sim\Pi(P_r(x), P_g(y))}\mathbb{E}_{(x,y)\sim\gamma}[\|x - y\|] \] 3. **The loss function of MACGAN**: \[ L_D=\mathbb{E}_{x\sim P_r(x)}[D(x)]-\mathbb{E}_{z\sim P_z(z)}[D(G(z, c_g))] \] \[ L_G=-\mathbb{E}_{z\sim P_z(z)}[D(G(z, c_g))]+0.5\times L_C^R+0.5\times L_C^G \] \[ L_C=\lambda_1 L_C^R+\lambda_2 L_C^G \] 4. **Spectral normalization**: \[ W_{\text{SN}}(W)=\frac{W}{\sigma(W)} \] where \(\sigma(W)\) is the maximum of the weight matrix \(W\)

Multi-mode data augmentation and fault diagnosis of rotating machinery using modified ACGAN designed with new framework

A Bearing Fault Diagnosis Method for High-Noise and Unbalanced Dataset

Categorical Feature GAN for Imbalanced Intelligent Fault Diagnosis of Rotating Machinery

Leveraging Transfer Learning for Data Augmentation in Fault Diagnosis of Imbalanced Time-Frequency Images

Data Augmentation for Intelligent Mechanical Fault Diagnosis Based on Local Shared Multiple-generator GAN

A Trackable Multi-Domain Collaborative Generative Adversarial Network for Rotating Machinery Fault Diagnosis

An Improved Auxiliary Classifier Generated Adversarial Network for Bearing Fault Diagnosis

Generative adversarial networks for data augmentation in machine fault diagnosis

Improvement of Generative Adversarial Network and Its Application in Bearing Fault Diagnosis: A Review

Ensemble Data Augmentation for Imbalanced Fault Diagnosis.

A Novel Method for Imbalanced Fault Diagnosis of Rotating Machinery Based on Generative Adversarial Networks

Enhanced generative adversarial networks for fault diagnosis of rotating machinery with imbalanced data

FMRGAN: Feature Mapping Reconstruction GAN for Rolling Bearings Fault Diagnosis under Limited Data Condition

Classification of Multi-Type Bearing Fault Features Based on Semi-Supervised Generative Adversarial Network (GAN)

An Efficient Method for Imbalanced Fault Diagnosis of Rotating Machinery

An interpretable data augmentation scheme for machine fault diagnosis based on a sparsity-constrained generative adversarial network

ICoT-GAN: Integrated Convolutional Transformer GAN for Rolling Bearings Fault Diagnosis under Limited Data Condition

Two-Stage Multi-Scale Fault Diagnosis Method for Rolling Bearings with Imbalanced Data

Enhanced generative adversarial network for extremely imbalanced fault diagnosis of rotating machine

An Efficient Method Based on Conditional Generative Adversarial Networks for Imbalanced Fault Diagnosis of Rolling Bearing

LOSGAN: Latent Optimized Stable GAN for Intelligent Fault Diagnosis with Limited Data in Rotating Machinery