Bearing signal denoising is a pivotal task in predictive maintenance and condition monitoring of industrial machinery. However, conventional denoising methods often face difficulties in simultaneously suppressing noise and preserving essential physical features. To address this challenge, we propose a novel denoising framework that incorporates physical feature priors derived from low-dimensional manifold-based simulated data. Specifically, a regression branch—constructed using convolutional and residual neural networks—is integrated into the main denoising model to exploit the intrinsic structure of bearing signals. By embedding manifold-informed priors, the regression branch enhances denoising performance and ensures the retention of critical physical features. Experimental results demonstrate that the proposed approach surpasses traditional methods in both signal denoising and bearing fault diagnosis. Notably, the incorporation of manifold-derived priors improves the model’s capability to capture the underlying physical characteristics of the signals, indicating that the network has learned their inherent features rather than simply minimizing the loss function. Overall, this study introduces a robust denoising paradigm for complex industrial environments where bearing signals exhibit significant variability.

It is of great practical significance to study intelligent robots for crack detection of urban underground pipes instead of manual detection, which not only can save time and cost, but also can improve the accuracy rate of identification. The system consists of two parts: mobile platform and system detection platform. The mobile platform is obstacle-avoiding by ultrasonic module, segmenting the pipeline color by camera and issuing platform steering command. The system detection platform adopts Raspberry Pi as the core control system, and collects the pipeline leakage audio through the pickups, and the collected audio is passed through Gaussian mixture model-Hidden Markov model (GMM -HMM) and Deep neural networks- Hidden Markov model (DNN -HMM) respectively in Raspberry Pi model (DNN -HMM) for the purpose of detection. The accuracy and running speed of the two methods were tested separately through simulation experiments. GMM -HMM achieves 74.1% accuracy and 14 (frame/s) running speed, while DNN -HMM achieves 80.6% accuracy and 9 (frame/s) running speed, DNN -HMM has higher accuracy and running speed to meet the real-time requirements for pipeline leakage audio detection.

To address the issues of image quality degradation and single feature expression in tunnel face images, an intelligent rock mass evaluation method integrating multi-scale progressive enhancement and deep semantic modeling was proposed. A multi-scale image enhancement framework combining spatial domain filtering and transform domain denoising was employed, by which the signal-to-noise ratio was increased by 12.7 dB with significantly enhanced visibility of key geological structures including fractures and joints. A four-dimensional comprehensive evaluation index system was constructed based on the gray-level co-occurrence matrix, improved local binary pattern, and RGB color moments for quantitative rock mass characterization.The ResNet architecture was improved through integration of multi-scale feature extraction and dual attention mechanisms (channel attention + spatial attention), while a weighted cross-entropy-label smoothing composite loss function was adopted to address class imbalance. Based on a constructed database of 5,000 tunnel face images, the model accuracy of 94.27% was achieved on the test set (4.93% higher than the baseline model) with a computational complexity of 4.8 G FLOPs. Practical case verification indicated that the proposed method could provide real-time, objective geological decision support for tunnel construction, significantly improving the intelligence level of rock mass classification.