Aero-engine prognosis strategy based on multi-scale feature fusion and multi-task parallel learning

Zhou, Liang; Wang, Huawei; Xu, Shanshan

doi:10.1016/j.ress.2023.109182

Cited by 17 publications

(5 citation statements)

References 33 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…MFSMTP is a multi-task learning network that can extract features from data using three different sizes of convolutional kernels to capture more feature information [ 32 ]. The parameters of MFSMTP were as follows: {Input (1, 75, 14), convolution1 (5, 1, 2), convolution2 (3, 1, 1), convolution3 (1, 1, 0), linear (192, Class)}.…”

Section: Resultsmentioning

confidence: 99%

“…The multi-feature fusion strategy is a solution for capturing the comprehensive features of samples across different scales [ 31 , 32 , 33 ]. Three types of multi-feature fusion algorithms are commonly used nowadays.…”

Section: Introductionmentioning

confidence: 99%

“…The third approach involves fusing features from different perspectives or scales. CNNs with convolutional kernels of different sizes were used in [ 32 , 33 ] to extract multi-scale features simultaneously. In [ 38 ], LSTM and CNN were used to extract both the temporal and spatial features of the samples used.…”

Section: Introductionmentioning

confidence: 99%

“…This mechanism can effectively improve the identification of fault features using a model in a signal with noise. Unfortunately, each of these methods [ 29 , 30 , 31 , 32 , 33 ] was only trained for a single task, which limits their recognition accuracy and efficiency. Multi-task learning is a paradigm in machine learning that can process multiple tasks simultaneously and improve the generalization performance and classification accuracy of a model [ 41 ].…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Gas Sensor Array Fault Diagnosis Based on Multi-Dimensional Fusion, an Attention Mechanism, and Multi-Task Learning

Huang,

Wang,

Chen

et al. 2023

Sensors

View full text Add to dashboard Cite

With the development of gas sensor arrays and computational technology, machine olfactory systems have been widely used in environmental monitoring, medical diagnosis, and other fields. The reliable and stable operation of gas sensing systems depends heavily on the accuracy of the sensors outputs. Therefore, the realization of accurate gas sensor array fault diagnosis is essential to monitor the working status of sensor arrays and ensure the normal operation of the whole system. The existing methods extract features from a single dimension and require the separate training of models for multiple diagnosis tasks, which limits diagnostic accuracy and efficiency. To address these limitations, for this study, a novel fault diagnosis network based on multi-dimensional feature fusion, an attention mechanism, and multi-task learning, MAM-Net, was developed and applied to gas sensor arrays. First, feature fusion models were applied to extract deep and comprehensive features from the original data in multiple dimensions. A residual network equipped with convolutional block attention modules and a Bi-LSTM network were designed for two-dimensional and one-dimensional signals to capture spatial and temporal features simultaneously. Subsequently, a concatenation layer was constructed using feature stitching to integrate the fault details of different dimensions and avoid ignoring useful information. Finally, a multi-task learning module was designed for the parallel learning of the sensor fault diagnosis to effectively improve the diagnosis capability. The experimental results derived from using the proposed framework on gas sensor datasets across different amounts of data, balanced and unbalanced datasets, and different experimental settings show that the proposed framework outperforms the other available methods and demonstrates good recognition accuracy and robustness.

show abstract

Section: Resultsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Gas Sensor Array Fault Diagnosis Based on Multi-Dimensional Fusion, an Attention Mechanism, and Multi-Task Learning

Huang,

Wang,

Chen

et al. 2023

Sensors

View full text Add to dashboard Cite

show abstract

“…Liu et al 11 employed the channel attention mechanism (CAM) to assign larger weights to important features. Zhou et al 12 proposed a comprehensive framework based on multiscale feature fusion and parallel learning to predict RUL. Zhang et al 13 designed a dual-aspect selfattention based on a transformer (DAST) model for aero-engine RUL estimation, which also obtains promising performance.…”

Section: Introductionmentioning

confidence: 99%

A novel aeroengine remaining useful life prediction method considering degradation starting point

SUO,

WANG,

LIN

et al. 2024

Journal of Advanced Manufacturing Science and Technology

View full text Add to dashboard Cite

The degradation of aeroengines can be divided into two stages: the healthy stage and the unhealthy stage. Remain useful life (RUL) prediction should be triggered from the start time of the unhealthy stage to ensure safe operation.Nevertheless, many existing RUL prediction methods simply assign a fixed DSP to any aeroengine, limiting further improvement as the DSP is uncertain and varies with individual differences of aeroengines. To address this issue, a novel two-stage deep residual long-short term memory (Dual-DRLSTM) is developed, which integrates DSP detection and RUL prediction into one framework, and associates them through degradation health index (HI). First, DRLSTM is employed as the backbone to extract representative degradation features from multi-dimensional time-series monitoring data. Second, the Dual-DRLSTM relaxes the strong assumption of the fixed degradation start point (DSP) and performs DSP detection for each aeroengine. Then, the Dual-DRLSTM predicts the RUL of the aeroengine beyond the DSP in the unhealthy stage. Finally, the outstanding performance of Dual-DRLSTM is validated through a series of experiments on a public C-MAPSS dataset.

show abstract

The Dual-Encoder Transformer for Prediction of Aero-Engine Remaining Useful Life with Uncertainty Quantification

Xiang,

Zhang,

Zhang

2024

Mechanisms and Machine Science

View full text Add to dashboard Cite

Aero-engine prognosis strategy based on multi-scale feature fusion and multi-task parallel learning

Cited by 17 publications

References 33 publications

Gas Sensor Array Fault Diagnosis Based on Multi-Dimensional Fusion, an Attention Mechanism, and Multi-Task Learning

Gas Sensor Array Fault Diagnosis Based on Multi-Dimensional Fusion, an Attention Mechanism, and Multi-Task Learning

A novel aeroengine remaining useful life prediction method considering degradation starting point

The Dual-Encoder Transformer for Prediction of Aero-Engine Remaining Useful Life with Uncertainty Quantification

Contact Info

Product

Resources

About