Sijue Li

The monocular vision-based pose estimation of non-cooperative target spacecraft is vital for tasks like on-orbit servicing and debris removal. While deep learning has improved monocular spacecraft pose estimation, existing methods suffer from limitations. First, the prevailing two-stage methods separate object detection and pose estimation processes, lacking end-to-end training and involving redundant feature extraction. Second, an over-reliance on convolutional neural networks (CNNs) can result in excessive dependence on texture and inadequate long-range dependency modeling. To address these drawbacks, we propose a Deformable Transformer-based Single-stage End-to-end SpaceNet (DTSE-SpaceNet). This network dynamically fuses features from multiple scales to predict keypoints, from which pose parameters are derived using the Perspective-n-Points (PnP) method. Furthermore, a novel shape loss function improves keypoint geometric accuracy and reduces outliers and enhancing performance. Extensive experiments on multiple public benchmark datasets demonstrate competitive performance and strong generalization capability, with computation and parameter advantages over two-stage methods.

Transient impact loads can easily cause structural damage and destruction, especially for composite structures that are susceptible to collisions. An accurate impact source localization method is crucial for evaluating the potential structutre damage location. The impact response signals, on the other hand, contain varing degrees of high and low frequency noise, and the structure is not limited to a single dimension. In this paper, the robust localization method with empirical model decomposition combined wavelet denoising (EMD-WD) and multi-output support vector regression (MSVR) is proposed. Following the achievement of intrinsic mode function (IMF) of distinct frequency bands, the last order IMF is abandoned to filter the low frequency disturbance. Following that, the discrete wavelet transform is employed to treat high frequency noise. Then, multi-dimensional features of impact responses, including time domain (TT), frequency domain (FF), time-frequency (TF) and distance domain (DD), are extracted. Finally, MSVR trained with multi-dimension feature vectors is used to predict various impact location coordinate values. The results on plate structure and cylinder composite monitoring data prove that proposed method can realize acceptable localization performance on both 2D/3D structure compared with other methods, with certain abilities to resist noise interference.

Structural health monitoring (SHM) technology is of great importance to ensure the long-term and reliable operation of engineering structures. The monitoring systems for large structures are characterized by large numbers of measurement points and complex signal. In recent years, deep learning has been widely used in many fields, such as intelligent fault diagnosis and health monitoring due to its excellent feature extraction and nonlinear fitting ability. However, the existing deep networks for condition monitoring have some drawbacks. First, the multisensor information fusion is rarely considered, which is quite common in SHM systems. Second, most of the proposed models by current research are with large parameters and high complexity. To solve the problems mentioned above, we propose a lightweight SHM framework suitable for the multisensor monitoring tasks. In other words, the adaptive multisensor fusion network (ASF-Net) is proposed to dynamically mix with multichannel input samples and extract the efficient feature on different scales. Moreover, a joint loss function is established for network compression by transferring the knowledge from bulky teacher network to lightweight student network. Finally, to reduce the performance degradation during model compression, a multigeneration knowledge distillation (MGKD) strategy is proposed. The experiments on two SHM datasets show that the proposed method can achieve excellent classification accuracy, and model performance degradation of less than 2% can be obtained with at most 40.24 times less parameters and 128.82 times less computation, which has a great prospect on resource-constrained platforms.

To prevent unmanned aerial vehicles (UAVs) from threatening public security, anti-UAV object tracking has become a critical issue in industrial and military applications. However, tracking UAV objects stably is still a challenging issue because the scenarios are complicated and the targets are generally small. In this article, a novel long-term tracking architecture composed of a Siamese network and re-detection (SiamAD) is proposed to efficiently locate UAV targets in diverse surroundings. Specifically, a new hybrid attention mechanism module is exploited to conduct more discriminative feature representation and is incorporated into a Siamese network. At the same time, the attention-based Siamese network fuses multilevel features for accurately tracking the target. We further introduce a hierarchical discriminator for checking the reliability of targeting, and a discriminator-based redetection network is utilized for correcting tracking failures. To effectively catch up with the appearance changes of UAVs, a template updating strategy is developed in long-term tracking tasks. Our model surpasses many state-of-the-art models on the anti-UAV benchmark. In particular, the proposed method can achieve 13.7% and 16.5% improvements in success rate and precision rate, respectively, compared with the strong baseline SiamRPN++.

With the growth of computing power, deep learning methods have recently been widely used in machine fault diagnosis. In order to realize highly efficient diagnosis accuracy, people need to know the detailed health condition of collected signals from equipment. However, in the actual situation, it is costly and time-consuming to close down machines and inspect components. This seriously impedes the practical application of data-driven diagnosis. In comparison, the full-labeled machine signals from test rigs or online datasets can be achieved easily, which is helpful for the diagnosis of real equipment. Thus, we introduced an improved Wasserstein distance-based transfer learning method (WDA), which learns transferable features between labeled and unlabeled signals from different forms of equipment. In WDA, Wasserstein distance with cosine similarity is applied to narrow the gap between signals collected from different machines. Meanwhile, we use the Kuhn–Munkres algorithm to calculate the Wasserstein distance. In order to further verify the proposed method, we developed a set of case studies, including two different mechanical parts, five transfer scenarios, and eight transfer learning fault diagnosis experiments. WDA reached an average accuracy of 93.72% in bearing fault diagnosis and 84.84% in ball screw fault diagnosis, which greatly surpasses state-of-the-art transfer learning fault diagnosis methods. In addition, comprehensive analysis and feature visualization are also presented.

When performing fault diagnosis tasks on bearings, the change of any bearing’s rotation speed will cause the frequency spectrum of bearing fault characteristics to be blurred. This makes it difficult to extract stable fault features based on manual or intelligent methods, resulting in a decrease in diagnostic accuracy. In this paper, a two-stage, intelligent fault diagnosis method (order-tracking one-dimensional convolutional neural network, OT-1DCNN) is proposed to deal with the problem of fault diagnosis under variable speed conditions. Firstly, the order tracking algorithm is used to resample the monitoring data obtained under different rotation speeds. Then, the one-dimensional convolutional neural network is adopted to extract features of the fault data. Finally, the fault type of collected data can be obtained by fully connected networks based on the features extracted. In the time domain, while the proposed algorithm only relies on the fault data collected under one speed as the training dataset, it is capable of doing fault diagnosis under different speed conditions. In the condition with the largest difference in speed with each dataset, the accuracy of the proposed method is higher than the baseline methods by 0.54% and 11.00%—on CWRU dataset and our own dataset respectively. The results show that the proposed method performs well in dealing with the fault diagnosis under the condition of variable speeds.