In the real scenario of engineering, the failure time of rotating machinery is generally much less than when it is in a healthy condition. Considering the cost, it is unrealistic to conduct the large-sample and long-time failure tests. This results in the problem of data imbalance in fault diagnosis, i.e. the number of normal samples far exceeds that of the fault ones, which seriously affects the accuracy and stability of fault diagnosis. For the settlement of the above problem, an auxiliary classier Wasserstein generative adversarial network with gradient penalty (ACWGAN-GP) is proposed in this paper, which is capable of generating high-quality samples for the minority classes stably utilizing an imbalanced training set. In the experiment of fault diagnosis, the generated samples firstly go through the availability verification, and then are employed to augment the imbalanced dataset gradually. The final results show that the proposed method is competent for the generation of data, which is highly similar to real samples, and the accuracy of fault diagnosis has effective improvement as the imbalanced dataset is gradually expanded to equilibrium. Additionally, the ACWGAN-GP model presents better performance in sample generation than other widely-used methods.

Rolling bearing is one of the core parts of modern industrial equipment, with high safety and reliability. Therefore, predicting its remaining useful life (RUL) is essential and far-reaching. The critical difficulty of RUL estimation technology is how to extract the features reflecting the bearing degradation trend and how to select the RUL estimation algorithm. Although the RUL estimation algorithms based on deep learning (such as CNN and RNN) have achieved remarkable results recently, these methods are not good at analyzing and predicting long-time series. We propose a transformer prediction model with a multi-scale gated convolution network (MSGCNN-TR) to solve this question to predict the RUL of rolling bearings. This model mainly includes sample feature extraction and bearing RUL estimation. The first part is a feature extractor composed of a depth convolution path with residual and a GRU path. It can effectively extract the spatial and temporal features that reflect the bearing degradation trend in the rolling bearing life data. The second part is an RUL estimation model composed of a four-layer transformer encoder. The multi-head self-attention mechanism in the Transformer encoder can effectively capture the long-term or short-term dependency relationship in the rolling bearing life data. Finally, we performed comparative and ablation experiments on the IEEE PHM2012 dataset. The results prove that the performance of the proposed model is better than other methods in the comparative experiment, and the ablation experiment proves the necessity and effectiveness of each module.

As important equipment, rotating machinery has been widely used in many industrial fields. Because rotating machinery is prone to failure, its fault diagnosis will be of great significance. In the industrial scene, rotating machinery is usually in normal operation, so it is difficult to accumulate fault samples. Therefore, it will face the problem of fault diagnosis with small samples, which seriously affects the accuracy and stability of fault diagnosis. To solve the above problems, the author proposes a fault diagnosis method based on ACWGAN-GP and homogeneous stacking ensemble learning. Firstly, the method utilizes the argmax multi-class classification idea to construct multiple different training subsets. Secondly, these constructed training subsets are used to train multiple base learners based on ACWGAN-GP, Finally, the meta learner based on Softmax Regression is used to fuse these trained basic learners. So far, the complete fault diagnosis is realized. In this paper, the proposed method is applied to the gearbox data set and the bearing data set. Through a series of experiments, it is proved that this method can not only effectively solve the problem of fault diagnosis with small samples, but also effectively improve the classification accuracy and stability.

Virtual-to-real registration is a crucial aspect of 3D registration, which presents a more challenging multimodal 3D registration problem due to the different data structures between virtual and real models. In this paper, we utilize point cloud registration algorithm to align virtual and real models, transforming the multimodal 3D registration problem into a cross-source point cloud registration problem. We propose a method for extracting macro and micro structures to represent the shared features of virtual and real models, combined with a multi-constraint registration algorithm, to achieve high-accuracy virtual-to-real registration tasks. This method can register unseen 3D objects using virtual prior knowledge, and allow partial point cloud registration without the need for a 360-degree scan of the model. Our approach can effectively resist interference from typical cross-source point cloud registration problems such as varying densities, missing data, and distribution changes. Furthermore, by processing only 0.2% of the original number of point clouds through downsampling, we can effectively diminish the effects of noise and outlier, as well as significantly decrease time consuming. Experimental results show that our algorithm outperforms other advanced point cloud registration algorithms in cross-source point cloud registration for virtual-to-real registration.

The D-S evidence theory is extensively applied to manage uncertain information. However, the theory encounters challenges related to conflicts during the fusion process, impeding the precise identification of multi-subset focal elements. This paper introduces a novel method for conflicting evidence fusion that incorporates the Bray–Curtis dissimilarity, cosine distance of the included angle, and belief entropy. The method comprehensively evaluates three aspects—evidence similarity, evidence distance, and the amount of information—while considering factors like the credibility and uncertainty of evidence. Initially, the evidence undergoes conversion into single-subset focal element evidence through the improved Pignistic probability function. Subsequently, the credibility between pieces of evidence is established using the Bray–Curtis dissimilarity and angle cosine distance, while the uncertainty of the evidence is computed using belief entropy. The weighted correction coefficient of the evidence is determined by integrating the credibility and uncertainty of the evidence. Subsequently, the corrected evidence is fused using the D-S evidence theory to derive the final judgment. An analysis of two sets of arithmetic examples, considering both single-subset and multi-subset focal elements, demonstrates the faster convergence and enhanced accuracy and reliability of the proposed method in comparison to existing approaches.

Addressing the challenges posed by the complexity of the structure and the multitude of sensor types installed in space application fluid loop systems, this paper proposes a fault diagnosis method based on an improved D-S evidence theory. The method first employs the Gaussian affiliation function to convert the information acquired by sensors into BPA functions. Subsequently, it utilizes a pignistic probability transformation to convert the multiple subset focal elements into single subset focal elements. Finally, it comprehensively evaluates the credibility and uncertainty factors between evidences, introducing Bray–Curtis dissimilarity and belief entropy to achieve the fusion of conflicting evidence. The proposed method is initially validated on the classic Iris dataset, demonstrating its reliability. Furthermore, when applied to fault diagnosis in space application fluid circuit loop pumps, the results indicate that the method can effectively fuse multiple sensors and accurately identify faults.