Fault Diagnosis Based on Space Mapping and Deformable Convolution Networks

Abstract

The data-driven method based on deep learning is one of the popular issues in the field of fault diagnosis. The completeness and representativeness of the feature matrix from massive and highdimensional fault data have a great impact on fault diagnosis performance. In addition, the ability of deep networks to extract the spatial characteristics between fault data is especially important for the accuracy of fault diagnosis. Therefore, we propose a method based on space mapping and deformable convolution networks (DCN) to ensure diagnostic accuracy by improving the spatial resolution and spatial constraint characteristics, and both the size and shape of the convolution kernel, one of the key steps in DCN, are adjusted adaptively according to the input of different sizes. Original data are projected into a more discriminative space by the combination of CN and PCA (i.e., space mapping). Then, DCN extract spatial constraints between fault data by training. The Case Western Reserve University (CWRU) bearing dataset and Xi'an Jiaotong University and Changxing Sumyoung Technology Co., Ltd. (XJTU-SY) datasets are used as benchmarks to perform experiments. The results demonstrate that the fault diagnosis method proposed in this paper performs well and can achieve 100% accuracy in the first several epochs. Comparative experiments based on 3 deep learning methods that combine preprocessed and unprocessed data with a convolutional neural network (CNN), residual networks (ResNets) and DCN are carried out to further show the advantages of the fault diagnosis method based on space mapping and DCN.

Bearing fault diagnosis based on inverted Mel-scale frequency cepstral coefficients and deformable convolution networks

“…The CWRU data set selected in this paper is identical to that in [21]. That method uses CN+kernel principal component analysis (KPCA)+DCNs, and the results can reach 100% accuracy in several epochs.…”

“…First, fault diagnosis methods based on traditional signal processing methods, such as wavelet transform (WT) [9][10][11], empirical mode decomposition (EMD) [12,13], variational mode decomposition (VMD) [14], bandpass filter [15,16], and filter bank [17,18]. Second, datadriven methods based on deep learning [19], such as convolutional neural networks (CNNs) [20], deformable convolution networks (DCNs) [21], deep auto-encoders [22], and stacked denoising auto-encoders [23].…”

Recent deep learning models for diagnosis and health monitoring: A review of research works and future challenges

“…Feature-level fusion is performed on the high-level features extracted by the deep model. To overcome the impact of high-dimensional signals on the classification effect, PCA-based algorithms first reduce the dimensionality of the data from multiple channels, and then the CNN model further extracts features to construct a fault diagnosis framework (Guo et al 2021;Qiu et al 2019;Zhao et al 2020c). To integrate the feature-fusing into the network, some publications have designed CNN-based fault diagnosis models with different structures (Hoang et al 2021;Wang et al 2019bJiao et al 2019;Fu et al 2020).…”

Deep Learning based Approaches for Intelligent Industrial Machinery Health Management & Fault Diagnosis in Resource-Constrained Environments

“…Each category has a substantial number of publications in recent years which we have tried to include in the review. Signal processing techniques manipulate, analyze and transform signals, such as Fourier transforms and wavelet analysis [17], [18], [19], [20], [21], [22]. Feature extraction involves identifying and selecting relevant features or characteristics from raw data that are informative for a particular task.…”