Operation stage division and RUL prediction of bearings based on 1DCNN-ON-LSTM

Abstract

Remaining useful life (RUL) prediction of bearings is significantly important to ensure reliable operation of bearings. In practice, it is routinely impossible to obtain the full life cycle degradation data of bearings that needs to be used in prediction. The accuracy of the RUL prediction of bearings is often affected by incomplete degradation data. Regarding this situation, this paper proposes a multi-sensor three-stage RUL prediction framework based on the one-dimensional convolutional ordered neuron long short-term memory (1DCNN-ON-LSTM) neural network. Firstly, 1DCNN is used to extract spatial features adaptively from multi-sensor’s data and fuse them into one-dimensional feature. Next, the unsupervised hierarchy mechanism of time series information based ON-LSTM is developed to determine the ‘initial degradation stage point’ and ‘rapid degradation stage point’ of the bearing from the one-dimensional feature. Once the signal features collected by sensors input to the model reach the degradation stage point, select the corresponding sensitive features as input and construct the 1DCNN-ON-LSTM model that performs RUL prediction after the degradation stage point to improve the prediction accuracy of the model. Based on the proposed hierarchy mechanism, the bearings’ operation process is divided into three operation stages: normal stage, initial degradation stage and rapid degradation stage. Finally, the experiments verify that the proposed method can effectively divide the operation stages of bearings to predict the RUL and improve the generalization ability and prediction accuracy of the model.

Ultra-long-term spatiotemporal feature fusion network for bearing remaining life prediction under strong interference

“…Despite these improvements, this approach lacks adaptive attention capabilities for different types of features and struggles to overcome the exponential decay problem in historical data due to inherent flaws in the recurrent structure. Guo et al [19] proposed a 1DCNN-ON-LSTM network to adaptively extract spatial features from multi-sensor data through 1DCNN. Subsequently, the bearing degradation stages were divided based on the ON-LSTM network.…”

Prediction model of bearing fault remaining useful life based on weighted variable loss degradation characteristics

“…(a) Reconstruction loss function L R The reconstruction loss function L R , as shown in align (3), minimizes the difference between the original signal and the reconstruction signal, which is an important standard to ensure the noise reduction capability of DAE. According to the idea of manifold learning, L R helps DAE to learn vector fields pointing to low-dimensional manifolds.…”

“…In order to ensure the safe and reliable early fault monitoring, not only the accurate and efficient fault diagnosis technology is needed, but also the demands on the remaining useful life (RUL) prediction method of the fault parts is increasing. Data-driven method is one of the commonly used life prediction methods, which has been widely used in the absence of prior knowledge of fault prediction [3]. The data-driven prediction method mainly focuses on two key issues: 'How to establish a health indicator reflecting the degradation process' and 'choosing a suitable prediction algorithm to predict the development trend' [4].…”

RUL Prediction for Piezoelectric Vibration Sensors Based on Digital-Twin and LSTM Network

“…The LSTM network is a special kind of RNN, mainly used to solve the gradient vanishing and gradient explosion problems while training long sequences; thus, LSTM can work better in long sequences. LSTM has been widely used for RUL prediction because it is suitable for time series prediction [37][38][39]. The primary cell structure of the LSTM network is shown in Figure 12, which effectively controls the consequences caused by accumulation by introducing cell states, forgetting gates, input gates, and output gates.…”

“…The primary cell structure of the LSTM network is shown in Figure 12, which effectively controls the consequences caused by accumulation by introducing cell states, forgetting gates, input gates, and output gates. Based on the currently given input x and the hidden layer output h from the previous moment, the LSTM network update process is shown as follows [38,40]:…”