Abstract:Remaining useful life (RUL) prediction is a key technology in prognostics and health management (PHM), which is crucial for achieving predictive maintenance. A number of spatiotemporal learning methods based on graphs and recurrent neural networks have been used to predict RUL of rotating machinery systems. However, the information interaction between spatial sensors has not been properly constructed and fully extracted. In terms of temporal features, there is critical information loss during the process of degradation feature extraction. To address these issues, a novel spatiotemporal learning network is proposed for the RUL prediction of turbofan engines, which is based on a structural-guided interaction (SGI) mechanism and an attention-enhanced compensation (AEC) mechanism. The SGI mechanism leverages the actual sensor structure to ensure the real existence of sensor information interaction, followed by applying high-order Chebyshev graph convolutional networks to extract multiorder neighbor interaction information that simulates cascade effects. The AEC mechanism captures critical degradation features on the temporal sequence by considering multihead attention in self-attention convolutional long short-term memory network. It introduces a global information compensation strategy to significantly reduce degradation information loss. The experimental results on the Commercial Modular Aero-Propulsion System Simulation (C-MAPSS) turbofan engine dataset demonstrate that the proposed method consistently outperforms the state-of-the-art methods in terms of RUL prediction performance. The interpretable analysis is provided to reveal the model learning process through comparison experiments and feature map visualization.

Prediction of remaining useful life based on bidirectional gated recurrent unit with temporal self-attention mechanism

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

A BiGRU Method for Remaining Useful Life Prediction of Machinery

A Bidirectional Recursive Gated Dual Attention Unit Based RUL Prediction Approach

Data-driven remaining useful life prediction via multiple sensor signals and deep long short-term memory neural network

A Parallel Hybrid Neural Network With Integration of Spatial and Temporal Features for Remaining Useful Life Prediction in Prognostics

An Autoencoder Gated Recurrent Unit for Remaining Useful Life Prediction

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

Deep Bidirectional Recurrent Neural Networks Ensemble for Remaining Useful Life Prediction of Aircraft Engine

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

Prediction Interval Estimation of Aeroengine Remaining Useful Life Based on Bidirectional Long Short-Term Memory Network

A Hybrid Prognostics Deep Learning Model for Remaining Useful Life Prediction

Adaptive Feature Utilization with Separate Gating Mechanism and Global Temporal Convolutional Network for Remaining Useful Life Prediction

A dual path hybrid neural network framework for remaining useful life prediction of aero‐engine

Time-Gated Remaining Useful Life Prediction with Non-Periodical Inspection Data

Multi-granularity Cross-Domain Temporal Regression Network for Remaining Useful Life Estimation of Aero Engines

Multi-Resolution LSTM-Based Prediction Model for Remaining Useful Life of Aero-Engine

Health status assessment and remaining useful life prediction of aero-engine based on BiGRU and MMoE

Remaining useful life prediction of bearings by a new reinforced memory GRU network

A Spatio-Temporal Attention Mechanism Based Approach for Remaining Useful Life Prediction of Turbofan Engine

Remaining Useful Life Prediction Using a Novel Feature-Attention-Based End-to-End Approach