Abstract Distance-based domain adaptation methods have received extensive application in the transfer learning field. Different distances different characteristics due to various data processing principles. Therefore, choosing appropriate distance can accomplish tasks more efficiently. Domain adversarial neural networks extract invariant features through game confrontation, but it is not capable of extracting hidden gear under speed fluctuations, and only using mechanism for feature...

The emergence of large Vision Language Models (VLMs) has broadened the scope and capabilities single-modal Large (LLMs) by integrating visual modalities, thereby unlocking transformative cross-modal applications in a variety real-world scenarios. Despite their impressive performance, VLMs are prone to significant hallucinations, particularly form inconsistencies. Building on success Reinforcement Learning from Human Feedback (RLHF) aligning LLMs, recent advancements have focused applying...

Abstract The generalized nonlinear sparse spectrum (GNSS), as an improved fast kurtogram (FK) method, effectively suppresses the interference of abnormal signals through preprocessing and enhancement. However, GNSS method inherits shortcoming traditional FK using finite impulse response filters to process nonstationary signals, which limits accuracy fault extraction. Therefore, more precise should be developed further improve performance features. Inspired by this, this paper introduces...

Abstract Currently, most fault diagnosis methods can achieve desired results from a single signal source. However, sensor has limited features and adaptability to the working environment, which will greatly affect results. To overcome this weakness, multichannel deep adaptive adversarial network (MCDAAN) based on fusing acoustic vibration signals is proposed in paper. The training process of MCDAAN primarily includes following aspects. First, extracted by neural feature extraction are fused...

In recent years, transfer learning has been widely used in mechanical fault diagnosis with some achievements. However, most methods do not perform well when the speed and load change simultaneously. Inspired by adversarial mechanism, a method named attention mechanism-guided domain network (AMDAN) is proposed this paper. AMDAN regards convolutional neural networks (CNNs) as generator of to learn mutually invariant features classifier discriminator network. Attention mechanism introduced take...

Abstract The extraction of gearbox fault features under shock interference is an exceedingly difficult and valuable subject. effective usage the resonance frequency band one solutions for this However, existing fast kurtogram (FK) method prone to misdiagnosis due sensitivity aperiodic shocks. To overcome FK irrelevant shock, paper proposes a nonlinear (NFK) method. First, Z-score normalization performed on signal. Then, Sigmoid used improve representation interference. Third, signal divided...

Abstract Improving bearing fault diagnosis accuracy under speed fluctuation is a challenge in engineering applications. With the development of big data processing technology, new solution, multi-sensor complementary information, has emerged. However, single-scale dimension compression, which adopted most fusion methods, captures only small amount valuable information. To deal with this deficiency, multi-scale dynamic network (MSDFN) proposed. First, considering existence non-stationary...

Large discrepancy of sample distribution resulting from speed fluctuation is a great challenge to mechanical equipment health monitoring. Existing fault diagnosis methods are often limited by the acquisition mechanism single-modal measurement. Considering above problems, multidimensional features dynamically adjusted adaptive network (MFDAAN) fused vibro-acoustic modal signals proposed in this paper. The MFDAAN considers context information activation Funnel (FReLU) function activate signal...

Abstract Machine learning is gradually applied to the fault diagnosis system of rotating machinery. However, can only classify and identify types previously trained by model in system. If required more faults, all untrained new need be input into retrain. Under current background big data, upgrade time will relatively long. To solve this problem, a parallel network based on intrinsic component filtering (PICF) proposed, which each type sample separately, then training reduced dimension,...

Changes in operating conditions often cause the distribution of signal features to shift during bearing fault diagnosis process, which will result reduced diagnostic accuracy model. Therefore, this paper proposes a dual-channel parallel adversarial network (DPAN) based on vision transformer, extracts from acoustic and vibration signals through networks enhances feature robustness training fusion process. In addition, Wasserstein distance is used reduce domain differences fused features,...

Abstract Many of the current fault diagnosis methods rely on time-domain signals. While richest information are contained in these signals, their complexity poses challenges to network learning and limits ability fully characterize them. To address issues, a novel Multi-channel Fused Vision Transformer Network (MFVTN) is proposed this paper. Firstly, Overlapping Patch Embedding (OPE) module introduced overlap map with edge information, preserving global continuous features adding positional...

In this paper, the number and location of deployment indoor small base stations are used as optimization variables to ensure coverage reduce interference between conditions. Based on corrected model, a differential evolution hybrid particle swarm algorithm is solve objectives. process each iteration, solution given by covered prediction model. The signal degree in target area calculated fed back historical experience data for learning, final planning scheme output, which has certain...

Abstract Variational mode decomposition (VMD) is widely used in the fault diagnosis of rotating equipment. The kurtosis index typically utilized as objective function to determine optimal conventional VMD. However, easily interfered with by abnormal impact signals, seriously affecting accuracy VMD extraction process. Therefore, instability for anomalous disturbances should be addressed. In addition, number modal decompositions <?CDATA $K$?> <mml:math...

Many of the current fault diagnosis methods for intelligence rely on either a signal in time domain or frequency domain. However, frequency-domain signals often fail to fully capture operational condition bearings, despite their intuitive nature. On other hand, time-domain contain valuable equipment status information, but complexity poses challenges learning by networks. To overcome challenge, this paper proposes dual-domain fusion adversarial network (DSFAN) based Vision Transformer. The...