Abstract:Abstract The pump serves as crucial power component in the hydraulic system. In the actual working process, it often runs in time-varying speed conditions such as startup, shutdown, acceleration, and deceleration, which makes the rotor-bearing system in the pump often work in an unstable environment and prone to failure. Most of the existing fault diagnosis researches are based on vibration signals. However, variations in rotational speed over time will also induce changes in vibration signal, which will cover the signal change caused by the fault of the system itself, resulting in confusion of signal characteristics and misjudgment of fault diagnosis. To perform precise fault diagnosis under time-varying speed conditions, a diagnosis method based on Frequency Temporal Series (FTS) graph is proposed. The time domain vibration signal is converted into the frequency domain signal by Fast Fourier Transform (FFT). The frequency values corresponding to the top k maximum amplitudes in the spectrum are selected as the node features, and the nodes are connected in chronological order through edges. The fault signal is constructed as the FTS graphs to map the time-varying fault information. Subsequently, a synchronous extraction model of spatiotemporal information is built, and the fault feature information embedded in graph nodes and the time dependence information among internal nodes of the graph are extracted at the same time to generate multi-scale fusion features. Finally, the fully connected and Softmax layers are employed as the classifier to categorize the multi-scale fusion features that have been extracted spatiotemporal information. Through experiments under time-varying speed conditions, it is verified that the proposed method can effectively identify the speed change during the operation of the system. The diagnostic accuracy reaches 99.17 %, outperforming other deep learning diagnostic methods.

Timed failure propagation graph construction with supremal language guided Tree-LSTM and its application to interpretable fault diagnosis

Spatially-temporally online fault detection using timed multivariate statistical logic.

Deep Learning With Tensor Factorization Layers For Sequential Fault Diagnosis And Industrial Process Monitoring

A Graph Representation Learning-Based Method for Fault Diagnosis of Rotating Machinery under Time-Varying Speed Conditions

A Fault Prognosis Strategy Based on Time-Delayed Digraph Model and Principal Component Analysis

STAGED: A Spatial-Temporal Aware Graph Encoder–Decoder for Fault Diagnosis in Industrial Processes

Interpretable temporal degradation state chain based fusion graph for intelligent bearing fault detection

Interpretable Fault Diagnosis of Rolling Element Bearings with Temporal Logic Neural Network

Fault Diagnosis of Pump Rotor-Bearing System Based on Frequency Temporal Series Graph under Time-Varying Speed Conditions

A Novel Label-Aware Global Graph Construction Method and Spiking-Coded Graph Neural Network for Intelligent Process Fault Diagnosis

Spatial-temporal graph feature learning driven by time–frequency similarity assessment for robust fault diagnosis of rotating machinery

TFN: An Interpretable Neural Network with Time-Frequency Transform Embedded for Intelligent Fault Diagnosis

Temporal Logic Inference for Fault Detection of Switched Systems With Gaussian Process Dynamics

A novel rolling bearing fault diagnosis method based on time-series fusion transformer with interpretability analysis

Fault Diagnosis of Timed Event Graph

Interactive spatiotemporal LSTM approach for enhanced industrial fault diagnosis

Explainable fault prediction using learning fuzzy cognitive maps

Graph structure embedded with physical constraints-based information fusion network for interpretable fault diagnosis of aero-engine

A Gated Recurrent Generative Transfer Learning Network for Fault Diagnostics Considering Imbalanced Data and Variable Working Conditions

A novel fault diagnosis framework empowered by LSTM and attention: A case study on the Tennessee Eastman process

Spatial-Temporal Bearing Fault Detection Using Graph Attention Networks and LSTM