A Novel Bearing Fault Diagnosis Method Based on Improved Convolutional Neural Network and Multi-Sensor Fusion DOI Creative Commons
Zhongyao Wang,

Xiao Xu,

Dongli Song

и другие.

Machines, Год журнала: 2025, Номер 13(3), С. 216 - 216

Опубликована: Март 7, 2025

Bearings are key components of modern mechanical equipment. To address the issue that limited information contained in single-source signal bearing leads to accuracy fault diagnosis method, a multi-sensor fusion method is proposed improve reliability diagnosis. Firstly, feature extraction process convolutional neural network (CNN) improved based on theory variational Bayesian inference, which forms inference (VBICNN). VBICNN used obtain preliminary results single-channel signals. Secondly, considering redundancy multi-channel signals, voting strategy fuse model final results. Finally, evaluated by an experimental dataset axlebox high-speed train. The show average can reach more than 99% and has favorable stability.

Язык: Английский

Machine learning-driven power prediction in continuous extrusion of pure titanium for enhanced structural resilience under extreme loading DOI Creative Commons
Ahmed Ghazi Abdulameer,

Muhannad M. Mrah,

Maryam Bazerkan

и другие.

Discover Materials, Год журнала: 2025, Номер 5(1)

Опубликована: Янв. 11, 2025

Abstract The increasing demand for advanced materials capable of withstanding extreme loading conditions, such as those encountered during impact or blast events, underscores the need innovative approaches in material processing. This study focuses on leveraging machine learning (ML) to enhance predictive accuracy continuous extrusion CP-Titanium Grade 2, a vital structural resilience critical applications. Specifically, an Artificial Neural Network (ANN) model optimized using Stochastic Gradient Descent (SGD) was introduced forecast power requirements with high precision. analysis utilized published dataset that comprises theoretical, numerical, and experimental calculations robust foundation validation comparison. A visualization highlighted influence process parameters, feedstock temperature wheel velocity, performance align thematic focus resilient design. ANN-SGD achieved RMSE 0.9954 CVRMSE 11.53% which demonstrated significant improvements prediction compared traditional approaches. By achieving superior alignment results, validated its efficacy reliable efficient tool understanding optimizing complex manufacturing processes. research emphasizes potential ML revolutionize processing conditions contribute broader goals sustainable manufacturing.

Язык: Английский

Процитировано

0
A Novel Bearing Fault Diagnosis Method Based on Improved Convolutional Neural Network and Multi-Sensor Fusion DOI Creative Commons
Zhongyao Wang,

Xiao Xu,

Dongli Song

и другие.

Machines, Год журнала: 2025, Номер 13(3), С. 216 - 216

Опубликована: Март 7, 2025

Bearings are key components of modern mechanical equipment. To address the issue that limited information contained in single-source signal bearing leads to accuracy fault diagnosis method, a multi-sensor fusion method is proposed improve reliability diagnosis. Firstly, feature extraction process convolutional neural network (CNN) improved based on theory variational Bayesian inference, which forms inference (VBICNN). VBICNN used obtain preliminary results single-channel signals. Secondly, considering redundancy multi-channel signals, voting strategy fuse model final results. Finally, evaluated by an experimental dataset axlebox high-speed train. The show average can reach more than 99% and has favorable stability.

Язык: Английский

Процитировано

0