In the realm of marine aquaculture, the netting of cages frequently accumulates marine fouling, which impedes water circulation and poses safety hazards. Traditional manual cleaning methods are marked by inefficiency, high labor demands, substantial costs, and considerable environmental degradation. This paper initially presents the current utilization of net-cleaning robots in the cleaning, underwater inspection, and monitoring of aquaculture cages, highlighting their benefits in enhancing operational efficiency and minimizing costs. Subsequently, it reviews key technologies such as underwater image acquisition, visual recognition, adhesion-based movement, efficient fouling removal, motion control, and positioning navigation. Ultimately, it anticipates the future trajectory of net-cleaning robots, emphasizing their potential for intelligence and sustainability, which could drive the marine aquaculture industry towards a more efficient and eco-friendly era.

As a classic nonlinear filtering method, Spectral Amplitude Modulation (SAM) is widely used in the field of bearing fault characteristic frequency identification. However, when the vibration signal contains high-intensity noise interference, the accuracy of SAM in identifying fault characteristic frequencies is greatly reduced. To solve the above problems, a Data Enhancement Spectral Amplitude Modulation (DA-SAM) method is proposed. This method further processes the modified signal through improved wavelet transform (IWT), calculates its logarithmic maximum square envelope spectrum to replace the original square envelope spectrum, and finally completes SAM. By highlighting signal characteristics and strengthening feature information, interference information can be minimized, thereby improving the robustness of the SAM method. In this paper, this method is verified through fault data sets. The research results show that this method can effectively reduce the interference of noise on fault diagnosis, and the fault characteristic information obtained is clearer. The superiority of this method compared with the SAM method, Autogram method, and fast spectral kurtosis diagram method is proved.

To accurately predict the remaining useful life (RUL) of rolling bearings under strong noise interference, this paper proposes an RUL prediction method based on adaptive Variational Mode Decomposition (VMD) and a hybrid Temporal Convolutional Network-Gated Recurrent Unit-Self-Attention (TCN-GRU-SA) framework. First, a parameter-optimized VMD algorithm is developed by integrating the Grey Wolf Optimizer (GWO) with VMD to extract effective intrinsic mode components (IMFs) and reconstruct denoised signals, thereby mitigating the impact of strong background noise. Subsequently, time-domain degradation features are extracted from the reconstructed signals to generate more representative feature datasets. These degradation features are then fed into a parallel TCN-GRU-SA prediction model. To enhance the model’s generalization capability and RUL prediction performance, the proposed hybrid architecture combines a Temporal Convolutional Network (TCN), which captures local temporal patterns, a Gated Recurrent Unit (GRU) for modeling long-term dependencies, and a Self-Attention (SA) mechanism to prioritize critical degradation-related features. Experimental validation on the PHM2012 rolling bearing accelerated lifetime dataset demonstrates that the proposed method achieves superior noise robustness and prediction accuracy compared to existing approaches. Specifically, it reduces the root mean square error (RMSE) by 18.7 % and improves the coefficient of determination (R2) by 12.3 % under high-noise conditions, confirming its effectiveness in industrial predictive maintenance applications.

Early bearing faults are often difficult to identify because their characteristic components are weak and easily masked by strong interference. To improve weak-fault feature extraction, this paper proposes a particle-swarm-optimization-based FitzHugh–Nagumo stochastic resonance (FHN-SR) method for bearing vibration signals. The raw signal is first preprocessed by de-meaning, Hilbert envelope demodulation, and standardization to construct a stable stochastic resonance (SR) input. Then, the key model parameters are adaptively optimized by maximizing the output signal-to-noise ratio around the target fault characteristic frequency. To evaluate the proposed method comprehensively, comparisons are carried out with classical SR, underdamped bistable stochastic resonance (UBSR), and a Fast-Kurtogram-based envelope-analysis scheme. Experimental validation is performed on three fault cases, including the rolling element fault case from the Case Western Reserve University (CWRU) dataset and the inner-race and outer-race fault cases from the Machinery Comprehensive Diagnostics Simulator (MCDS) platform. The results show that FHN-SR produces a clearer concentration of fault-related energy and achieves a higher output signal-to-noise ratio (SNR) than the compared methods in most cases. In particular, under degraded noise conditions, FHN-SR maintains more stable enhancement performance, indicating stronger robustness to interference. These results demonstrate that the proposed method provides an effective approach for extracting weak bearing fault features under complex noise backgrounds.