Abstract:Deep learning is gaining attention in data-driven remaining useful life (RUL) prediction of machinery because of its powerful representation learning ability. With the help of deep learning techniques, the machine degradation information can be mined more sufficiently and some promising RUL prediction results have been achieved in several studies recently. These deep learning-based prognostics approaches, however, have the following weaknesses: 1) Their prediction performance largely depends on the hand-crafted feature design. 2) The correlations of different sensor data are not explicitly considered in representation learning. To overcome the above weaknesses, a new deep prognostics network named deep separable convolutional network (DSCN) is proposed in this paper for RUL prediction of machinery. In the proposed DSCN, the monitoring data acquired by different sensors are directly used as the inputs of the prognostics network. Then, a separable convolutional building block with a residual connection is built based on separable convolutions and squeeze and excitation operations. Through stacking multiple separable convolutional building blocks, the high-level representations are automatically learned from the input data. Finally, the RUL is estimated by feeding the learned representations into the fully-connected output layer. The proposed DSCN is validated using the vibration data from accelerated degradation tests of rolling element bearings and the public degradation simulation data of turbine engines, respectively. The experimental results show that the proposed DSCN is able to provide accurate RUL prediction results based on the raw multi-sensor data and is superior to some existing data-driven prognostics approaches. (C) 2019 Elsevier Ltd. All rights reserved.

A Novel Deep Learning Approach For Machinery Prognostics Based On Time Windows

A Novel Deep Learning-Based Encoder-Decoder Model for Remaining Useful Life Prediction

A novel multi-scale CNN and attention mechanism method with multi-sensor signal for remaining useful life prediction

A Remaining Useful Life Prediction Framework Integrating Multiple Time Window Convolutional Neural Networks

A Multi-Head Neural Network with Unsymmetrical Constraints for Remaining Useful Life Prediction

A Temporal Pyramid Pooling-Based Convolutional Neural Network for Remaining Useful Life Prediction

A Novel Combination Neural Network Based on ConvLSTM-Transformer for Bearing Remaining Useful Life Prediction

Deep Separable Convolutional Network for Remaining Useful Life Prediction of Machinery

Machine remaining life prediction based on multi-layer self-attention and temporal convolution network

Remaining Useful Life Prediction of Mechanical Equipment based on Time-series Auto-Correlation Decomposition and CNN

A novel spatio-temporal hybrid neural network for remaining useful life prediction

Recurrent Convolutional Neural Network: A New Framework for Remaining Useful Life Prediction of Machinery.

A novel deep convolutional neural network-bootstrap integrated method for RUL prediction of rolling bearing

Memory-enhanced hybrid deep learning networks for remaining useful life prognostics of mechanical equipment

A novel remaining useful life prediction based on transfer hybrid deep neural network under variable working conditions

Few-shot RUL prediction for engines based on CNN-GRU model

A Novel Framework for Machine Remaining Useful Life Prediction Based on Time Series Analysis

Multi-dimensional recurrent neural network for remaining useful life prediction under variable operating conditions and multiple fault modes

Adaptive Spatio-Temporal Graph Convolutional Neural Network for Remaining Useful Life Estimation

A concise self-adapting deep learning network for machine remaining useful life prediction

Remaining Useful Life Prediction Using Temporal Deep Degradation Network for Complex Machinery with Attention-based Feature Extraction