Abstract:Remaining Useful Life (RUL) estimation plays a critical role in Prognostics and Health Management (PHM). Traditional machine health maintenance systems are often costly, requiring sufficient prior expertise, and are difficult to fit into highly complex and changing industrial scenarios. With the widespread deployment of sensors on industrial equipment, building the Industrial Internet of Things (IIoT) to interconnect these devices has become an inexorable trend in the development of the digital factory. Using the device's real-time operational data collected by IIoT to get the estimated RUL through the RUL prediction algorithm, the PHM system can develop proactive maintenance measures for the device, thus, reducing maintenance costs and decreasing failure times during operation. This paper carries out research into the remaining useful life prediction model for multi-sensor devices in the IIoT scenario. We investigated the mainstream RUL prediction models and summarized the basic steps of RUL prediction modeling in this scenario. On this basis, a data-driven approach for RUL estimation is proposed in this paper. It employs a Multi-Head Attention Mechanism to fuse the multi-dimensional time-series data output from multiple sensors, in which the attention on features is used to capture the interactions between features and attention on sequences is used to learn the weights of time steps. Then, the Long Short-Term Memory Network is applied to learn the features of time series. We evaluate the proposed model on two benchmark datasets (C-MAPSS and PHM08), and the results demonstrate that it outperforms the state-of-art models. Moreover, through the interpretability of the multi-head attention mechanism, the proposed model can provide a preliminary explanation of engine degradation. Therefore, this approach is promising for predictive maintenance in IIoT scenarios.

Remaining Useful Life Prediction of Parts and Components Based on BiLSTM Autoencoder with Attention Mechanism and Skip Connection

Aero-Engine Remaining Useful Life Prediction Via Physics-Informed Self-Attention Encoder

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

Remaining Useful Life Prediction Based on Multi-channel Attention Bidirectional Long Short-term Memory Network

Remaining Useful Life Prediction for Multi-Sensor Systems Using a Novel End-to-end Deep-Learning Method

Remaining Useful Life Prediction of Rolling Bearing Based on Multi-Head Attention Embedded Bi-LSTM Network

Machine Remaining Useful Life Prediction Via an Attention-Based Deep Learning Approach

Convolutional neural network based on attention mechanism and Bi-LSTM for bearing remaining life prediction

A dual attention LSTM lightweight model based on exponential smoothing for remaining useful life prediction

An Improved Method for Predicting the Remaining Useful Life Using a Spatial–Temporal Feature Extraction Network With Attention Mechanism

Long Short-Term Memory Neural Network with Transfer Learning and Ensemble Learning for Remaining Useful Life Prediction

Bi-LSTM-Based Two-Stream Network for Machine Remaining Useful Life Prediction

Predicting remaining useful life of a machine based on embedded attention parallel networks

Attention-based sequence to sequence model for machine remaining useful life prediction

Remaining Useful Life Prediction Based on Deep Residual Attention Network

Attention Sequence to Sequence Model for Machine Remaining Useful Life Prediction

Long-term Temporal Attention Neural Network with Adaptive Stage Division for Remaining Useful Life Prediction of Rolling Bearings

Remaining useful life prediction based on spatiotemporal autoencoder

Multi-Dimensional Self Attention based Approach for Remaining Useful Life Estimation

Structural-Guided Interaction and Attention-Enhancing Compensation Network for Machine Remaining Useful Life Prediction

Remaining Useful Life Prediction of Turbofan Engine Using Hybrid Model Based on Autoencoder and Bidirectional Long Short-Term Memory