The key to intelligent fault diagnosis is find relevant characteristics with the capability of representing different types faults. However, engineering problem that a few simple empirical features (EFs) cannot obtain high classification accuracy, and complex feature requires strong professional knowledge, which leads limited applications on general scale. In addition, extraction methods without prior knowledge guarantee model learned used for classification, its robustness generalization...

The rolling bearing is the key component of rotating machinery, and it also a failure-prone component. intelligent fault diagnosis method has been widely used to accurately diagnose faults. However, in engineering practice, difficult obtain sufficient sample data train model. Therefore, this paper, fusion model CGAN-2D-CNN that combines conditional generative adversarial network (CGAN) two-dimensional convolutional neural (2D-CNN) proposed for with small samples. Considering problem...

Class imbalance issue has been a major problem in mechanical fault detection, which exists when the number of instances presenting class is significantly fewer than that another class. This article focuses on zero-shot detection rolling bearing, extreme case imbalance. Aiming at this problem, two-stage recognition method proposed. First, inspired by conditional generative adversarial network, novel feature generating network composed extractor, discriminator, and generator designed to...

To reduce the economic losses caused by bearing failures and prevent safety accidents, it is necessary to develop an effective method predict remaining useful life (RUL) of rolling bearing. However, degradation inside difficult monitor in real-time. Meanwhile, external uncertainties significantly impact degradation. Therefore, this paper proposes a new RUL prediction based on long-short term memory (LSTM) with uncertainty quantification. First, fusion metric related runtime (or degradation)...

The intelligent diagnosis model based on deep learning method can effectively and accurately diagnose the health state of bearings widely used in various mechanical equipment. However, engineering practice, it is difficult to obtain sufficient labeled fault data, which would reduce performance. In order solve above problem, this article proposes a self-attention metric multiscale feature fusion classify bearing with few shots. proposed few-shot mainly contains extraction module module....

Abstract Fault diagnosis plays a crucial role in maintaining mechanical equipment reliability. Deep Neural Networks (DNNs) have exhibited superior performance fault under closed sample space assumptions. However, the deployment of neural network models practical industrial environments faces significant challenges due to emergence out-of-distribution (OOD) data, particularly novel categories that deviate substantially from initial training To address these limitations, we propose robust...

Abstract Rail damages can pose tremendous hazards for high-speed trains, making damage diagnosis critical in the field of engineering. Currently, deep learning enables an end-to-end approach rail diagnosis. However, training and test data real applications are often out distribution, or even represent fault categories that previously unseen. To address this situation, unseen framework (UDDF) effectively embeds mechanism features from simulation signals all possible has been proposed. In...

The implementation of condition monitoring and fault diagnosis is special importance for ensuring wind turbine (WT) operation safely stably. In practice, however, the data WT are limited, which makes it hard to identify faults accurately using existing intelligent methods. To address this, signals augmented self-taught learning network (SASLN) proposed generator, one most important parts in WT. SASLN, signal samples generated by Wasserstein distance guided generative adversarial networks...

The ultrasonic guide wave (UGW) has good application prospects in steel rail damage diagnosis, but the features of implied UGW are complex. Deep learning enables an end-to-end approach to fault diagnosis. Nevertheless, a large amount diversity data is needed for training, whereas signals simulation and repeated experiments lack diversity. Therefore, this paper, diagnostic framework based on transfer developed tackle problems mentioned above. proposed deep with pretraining strategy build...

Bolted joints play an important role in aerospace, machinery manufacturing, weapons and other fields, the contact pressure distribution at connection interface seriously affects service performance of bolts. Contact analysis is essential basis reference for structural design, calibration, inspection, safety monitoring bolted structures. Considering preload force frictional between joint components, a block mapping hexahedron mesh generation finite element modeling method connections...

Developing a trustworthy framework for intelligent fault diagnosis (IFD) of machines has two major challenges: confidently recognizing known faults and precisely detecting novel faults. However, current IFD frameworks are typically based on the closed-world assumption tackle issues independently, making it hard to meet expectations reliable in complicated working situations. In this paper, <italic xmlns:mml="http://www.w3.org/1998/Math/MathML"...

High-speed train structures, operating under time-varying conditions, present a significant challenge in the field of Structural Health Monitoring (SHM) due to uncertainties introduced during extraction damage indexes from signals. In this paper, novel method for quantifying fatigue crack based on improved energy-power entropy dynamic temperature environments is introduced. Two new types information entropy, namely energy singular spectral (ES) and power (PS) are proposed. A baseline...

Abstract Abstract: The operation of high-speed trains in dynamic temperature environments presents significant challenges for crack detection critical aluminum alloy components. It has been demonstrated that fluctuations significantly impact the performance Lamb wave-based structural health monitoring (SHM) systems, thereby compromising reliability damage protocols. In response to these challenges, a novel methodology is introduced, leveraging GMM-Wasserstein distance metrics address...

The damage states of multibolts are not only diverse but also difficult to obtain measured multibolt looseness signals due time and economic costs. Therefore, the imbalance between healthy damaged samples poses a challenge for monitoring looseness. A synthetic minority oversampling technique (SMOTE) based on finite element model is proposed solve problems acquisition imbalanced samples. small amount measurement data large simulation constitute dataset, their statistical characteristics...