Search

Star sensors are widely employed in spacecraft, yet their observation accuracy is seriously affected by the thermal deformation of the sensor bracket. A high-precision self-calibration method for measuring a star sensor bracket thermal deformation by the quasi-common-path polarized-light difference based on laser autocollimation was first proposed, and the measurement equipment was developed. Zemax and Comsol software were used to simulate the influence of window glass and glue layer on the measurements, respectively, then laboratory experiments and field tests were performed. The simulation and experimental results show that the proposed method can greatly reduce the influence of mechanical drift, vibration, and other factors on the measurements, and greatly improve the accuracy and stability of long-term measurements. A comprehensive measurement accuracy of better than $0.45''$ was obtained in the vacuum test environment over $- 2~^{\circ }$ C to $- 6~^{\circ }$ C. Long-term vacuum thermal cycling of the star sensor bracket of “Aoke-1” satellite showed that the thermal stability of the yaw and pitch were $0.09''$ /°C and $- 0.45''$ /°C, respectively, which met the on-orbit operation stability requirements of $1.0''$ /°C. This developed measurement method and equipment ensures the future observation accuracy of the “Aoke-1” satellite and provides a new method for high-precision measurement of the relatively small angle change between two surfaces in vacuum and high-low temperature environments.

The development of electromagnetic ultrasonic guided wave testing technology has provided effective solutions for highly accurate positioning and quantification of rail defects and online real-time monitoring of rail health status. However, the principal limitation of the guided wave testing method is the difficulty in accurately obtaining a defect’s depth and normal shift at the rail web. Therefore, we investigate the normal energy distribution characteristics of hybrid high-order SH-guided waves and propose a highly sensitive baseline-free depth quantification and normal location method for buried defects based on the guided wave relative energy coefficient $G_{\text {rec}}$ and guided wave relative order coefficient $G_{\text {roc}}$ . These methods can rapidly and accurately locate and quantify buried defects by obtaining reflected and transmitted wave energies of different guided wave orders. Furthermore, the simulation and experimental results and the proposed quantification theory are mutually verified. The results indicated that the proposed damage index $G_{\text {rec}}$ without time-domain reconstruction can improve the accuracy of depth quantification by at least 15%, and the performance of $G_{\text {roc}}$ -based defect normal shift quantification surpassed commercial scanning ultrasonic body wave detectors and eddy current detectors. In general, this work has the potential to promote high-sensitivity holographic real-time monitoring of rail defects in the future.

Using ultrasonic guided wave technology to quantify defects has always been challenging, especially for nonarray electromagnetic acoustic transducers (EMATs). This is because EMATs are always implemented with complicated defect feature extraction and signal processing algorithms and have a low signal-to-noise ratio. Therefore, we propose a high-intensity unilateral oblique focusing EMAT (UOF-EMAT) to address these problems. Based on this new EMAT, the principle of guided wave defect quantification was studied, and a theory-informed progressive residual deep convolutional network (ProResNet) was developed. The results demonstrate that the proposed UOF-EMAT can achieve unilateral focusing of the guided wave signal and increase the signal intensity at the preset focal position by 11 dB compared with the traditional bilateral focusing EMAT. Moreover, compared with the quantification results of general shallow and deep neural networks (NNs), the defect quantification error of ProResNet is reduced by approximately 9–10 and 1.6–5.2 times, respectively. Furthermore, to explain the internal functions of ProResNet visually and verify the training logic of the network, the Grad-CAM++ method was used. Compared with the unexplainable direct visual method, the Grad-CAM++ method can effectively interpret the function of ProResNet visually, and the effectiveness of the theory-informed deep ProResNet in defect quantification is verified. The source code (with description) and database used in this article are available at (https://github.com/ShyTHU/ProResNet.git).

Wheelset fault detection with high accuracy is challenging due to poor image quality. Specifically, the wheelset images are collected dynamically outdoors and suffer from diffuse reflection and environmental interference. Thus, the images contain light stripe adhesions (light flairs) and local fractures to be inpainted. The existing inpainting models are inapplicable to restore grayscale wheelset images. They are also too heavy to be deployed in an embedded wheelset monitoring equipment. In this paper, we propose a lightweight high-precision inpainting model that consists of a recurrent similarity network with the ghost convolution (RSG-Net) to remove light flairs and repair local fractures. RSG-Net replaces standard Pconv (partial convolutional) layers with soft-coding ones that can improve the feature representational ability. To reduce the influence of the background region features on image restoration, an asymmetrical similarity measure is designed to calculate not only the angle difference between the target and the source feature vectors but also the activation of the source ones. The multi-scale structural similarity (MS-SSIM) loss term is introduced to precisely guide the structural information restoration, such as the stripe edges. Moreover, the ghost convolution is introduced in RSG-Net to realize the model compression that can retain the core features of wheelset images and remove the redundant features. We conduct three groups of experiments that demonstrate the accuracy superiority of the proposed RSG-Net over the baseline methods, and the number of parameters is reduced by about 50%.

To detect the wheelset deformation and wear, an intuitive method is to first collect wheelset multi-line laser stripe images by monitors, and then extract the centerlines to construct 3D contours that are transferred to the cloud data center by 6G communication. The images, however, contain flairs and fractures due to the influence of environmental interference and reflected light on the smooth surfaces. The image defects affect the accurate extraction of stripe centerlines. To segment the defects and inpaint them, we propose a new lightweight U-shaped segmentation model URS. A target-shaped receptive field is designed to efficiently extract the details of the local, the ring-shaped, and the cross-shaped context around the local, which facilitates segmenting various defects. A scale-select sub-module is designed to adjust the weights of features from different receptive fields. To train the model, a multi-line laser image defect segmentation dataset MLIDSD is constructed. Experiments demonstrate that the performance of our model surpasses twelve SOTA models explicitly and can achieve a balance between the accuracy and the lightweight requirement.