Feature Space Transformation for Fault Diagnosis of Rotating Machinery under Different Working Conditions

Abstract

In recent years, various deep learning models have been developed for the fault diagnosis of rotating machines. However, in practical applications related to fault diagnosis, it is difficult to immediately implement a trained model because the distribution of source data and target domain data have different distributions. Additionally, collecting failure data for various operating conditions is time consuming and expensive. In this paper, we introduce a new transformation method for the latent space between domains using the source domain and normal data of the target domain that can be easily collected. Inspired by semantic transformations in an embedded space in the field of word embedding, discrepancies between the distribution of the source and target domains are minimized by transforming the latent representation space in which fault attributes are preserved. To match the feature area and distribution, spatial attention is applied to learn the latent feature spaces, and the 1D CNN LSTM architecture is implemented to maximize the intra-class classification. The proposed model was validated for two types of rotating machines such as a dataset of rolling bearings as CWRU and a gearbox dataset of heavy machinery. Experimental results show the proposed method has higher cross-domain diagnostic accuracy than others, therefore showing reliable generalization performance in rotating machines operating under various conditions.

Keywords: attention mechanism; domain adaptation; fault diagnosis; feature space transformation; gearbox; vibration measurement.

Figure 1

Sensor mounting location and measurement vibration signals.

Figure 2

The vibration signal for each status in various operation conditions.

Figure 3

Visualization of the distribution concept for each domain dataset. (a) Before domain transformation; (b) the goal of domain adaptation; (c) training dataset of the proposed model.

Figure 4

The architecture of the proposed model including a latent vector shifting algorithm.

Figure 5

A data distribution plot for the failure modes of a gearbox with two working conditions in a real machine dataset. The source domain data is denoted S_Normal, S_Fault_0, S_Fault_1; the target domain data is denoted T_Normal, T_Fault_0, and T_Fault_1.

Figure 6

Attention mechanism-based encoding module.

Figure 7

Decoding process for reconstructing the input.

Figure 8

The classifier based on 1D CNN LSTM.

Figure 9

2D representation of the semantic analogies. Word analogies for (a) vector operation, (b) latent vector operation, and (c) latent vector operation of unknown input data.

Figure 10

Measuring position of each vibration signal in the CWRU dataset.

Figure 11

The confusion matrix of the classification results of the proposed model or without attention model.

Figure 12

The t-SNE visualization of the feature at the latest hidden layer for each domain data, such as the source and target, and each class in CWRU.

Figure 13

The t-SNE visualization of feature at latest hidden layer for each domain data such as source and target and each class in real machine.

Figure 14

Verification on a real machine using the embedded software that included the proposed learned model.