Data imbalance and large data probability distribution discrepancies are major factors that reduce the accuracy of remaining useful life (RUL) prediction of high-reliability rotating machinery. In feature extraction, most deep transfer learning models consider the overall features but rarely attend to the local target features that are useful for RUL prediction; insufficient attention paid to local features reduces the accuracy and reliability of prediction. By considering the contribution of input data to the modeling output, a deep learning model that incorporates the attention mechanism in feature selection and extraction is proposed in our work; an unsupervised clustering method for classification of rotating machinery performance state evolution is put forward, and a similarity function is used to calculate the expected attention of input data to build an input data extraction attention module; the module is then fused with a gated recurrent unit (GRU), a variant of a recurrent neural network, to construct an attention-GRU model that combines prediction calculation and weight calculation for RUL prediction. Tests on public datasets show that the attention-GRU model outperforms traditional GRU and LSTM in RUL prediction, achieves less prediction error, and improves the performance and stability of the model.

The rotating parts of large and complex equipment are key components that ensure the normal operation of the equipment. Accurate fault diagnosis is crucial for the safe operation of these systems. To simultaneously extract both local and global valuable fault feature information from key components of complex equipment, this study proposes a fault diagnosis network model, named MultiDilatedFormer, which is based on the fusion of transformer and multi-head dilated convolution. The newly designed multi-head dilated convolution module is sequentially integrated into the transformer-encoder architecture, constructing a feature extraction module where the complementary advantages of both components enhance overall performance. Firstly, the sample is expanded into a two-dimensional feature map and then input into the newly designed feature extraction module. Finally, the diagnostic output is performed by the designed patch feature fusion module and classifier module. Additionally, interpretability research is conducted on the proposed model, aiming to understand the decision-making mechanism of the model through visual analysis of the entire decision process. The experimental results on three different datasets indicate that the proposed model achieved high accuracy in fault diagnosis with relatively short data windows. The highest accuracy reached 97.95%, which was up to 10.97% higher than other models. Furthermore, the feasibility of the model is also verified in the actual dataset of the rotating parts of the injection molding machine. The excellent performance of the model on different datasets demonstrates its effectiveness in extracting comprehensive fault feature information and also proves its great potential in practical industrial applications.

This paper addresses the challenging task of determining the position and posture of small-scale thin metal parts with multi-objective overlapping. To tackle this problem, we propose a method that utilizes instance segmentation and a three-dimensional (3D) point cloud for recognizing the posture of thin special-shaped metal parts. We investigate the process of obtaining a single target point cloud by aligning the target mask with the depth map. Additionally, we explore a pose estimation method that involves registering the target point cloud with the model point cloud, designing a registration algorithm that combines the sample consensus initial alignment algorithm (SAC-IA) for coarse registration and the iterative closest point (ICP) algorithm for fine registration. The experimental results demonstrate the effectiveness of our approach. The average accuracy of the instance segmentation models, utilizing ResNet50 + FPN and ResNet101 + FPN as backbone networks, exceeds 97%. The time consumption of the ResNet50 + FPN model is reduced by 50%. Furthermore, the registration algorithm, which combines the SAC-IA and ICP, achieves a lower average consumption time while satisfying the requirements for the manufacturing of new energy batteries.

Aiming at the sound source localization of mechanical faults in a strong reverberation scenario with multiple sound sources, this paper investigates a mechanical fault source localization method using the U‐net deep convolutional neural network. The method utilizes the SRP‐PHAT algorithm to calculate the response power spectra of the collected multichannel fault signals. Through the utilization of the U‐net neural network, the response power spectra containing spurious peaks are transformed into “clean” estimated source distribution maps. By employing interpolation search, the estimated source distribution maps are processed to obtain location estimations for multiple fault sources. To validate the effectiveness of the proposed method, this paper constructs an experimental dataset using mechanical fault data from electromechanical equipment relays and conducts sound source localization experiments. The experimental results show that the U‐net network under 0.2 s/0.5 s/0.7 s reverberation time can effectively eliminate spurious peak interference in the response power spectrum. As the signal‐to‐noise ratio decreases, it can still distinguish the sound sources with a distance of 0.2 m. In the context of multifault source localization, the method is capable of simultaneously locating the positions of four fault sources, with an average localization error of less than 0.02 m. The method in this paper effectively eliminates spurious peaks in the response power spectra under conditions of multisource strong reverberation. It accurately locates multiple mechanical fault sources, thereby significantly enhancing the efficiency of mechanical fault detection.

In new energy battery manufacturing, machine vision is widely used in automated assembly scenarios for key parts. To improve the accuracy and real‐time multi‐target positioning recognition of flexible thin parts, this paper proposes a multi‐level feature fusion template matching algorithm based on the Gaussian pyramid. Firstly, the algorithm constructs a Gaussian pyramid by multi‐scale image construction. Secondly, considering the image features of each layer of the pyramid, this paper uses the grey‐based Fast Normalized Matching algorithm to obtain coarse positioning coordinates on the upper layer, and the improved Linemod‐2D algorithm is applied to the bottom layer image to get accurate positioning coordinates. Finally, the positioning coordinates returned from each layer are fused to obtain the final positioning coordinate. The experimental results show that the proposed algorithm achieves excellent performance in nickel plate positioning and recognition. It exhibits satisfactory performance in nickel sheet localization and recognition. In terms of angular error, repeat accuracy, and matching speed, it competes favourably with Halcon, VisionMaster, and SCISmart. Its positioning error closely approximates that of Halcon, effectively meeting the practical production demands for high‐speed feeding and high‐precision positioning.

Chromatic confocal sensing provides a noncontact approach for thickness measurement of transparent media with multiple interfaces. In practice, however, unavoidable sample tilt causes simultaneous degradation of return-signal coupling and spectral response, which leads to significant thickness errors. In this work, the error formation pathway from tilt to spectral distortion and finally to thickness deviation is systematically investigated, and a compensation method is proposed. First, based on the weighted superposition of entrance-pupil energy contributions, the effects of tilt-induced off-axis return and off-axis dispersion-mapping variation on spectral peak position and profile stability are analyzed. A Zemax non-sequential ray tracing model is then established to quantify the evolution of the two-dimensional irradiance distribution, return power, and dual-interface spectral features under different tilt angles. The results show that stable dual-interface responses can still be maintained at small tilts, whereas beyond a critical tilt range the system exhibits rapid coupling loss and reduced peak separability. Experimental measurements using a CL3-MG70 probe and transparent quartz plates further confirm the simulation trends and show that the tilt-induced thickness error increases approximately in proportion to the sample thickness. On this basis, a unified magnification model is established for the investigated probe and quartz samples and applied to tilt-error compensation. Using the thickness deviation relative to the 0° condition as the evaluation metric, the proposed method reduces both the maximum drift and RMSE by more than 90%, demonstrating a substantial improvement in measurement stability. These results provide a practical reference for tilt error analysis and correction in chromatic confocal thickness measurement of rigid transparent plates.