In response to the growing demand for complex thin-walled lightweight alloy components in the automotive and aerospace industries, this study investigates the limitations of traditional gas pressure forming technologies. Using 2A12 aluminum alloy thin sheets as the research material, hot high-speed gas bulging experiments were conducted to study the effects of rapid inflation and rapid deflation processes on the forming accuracy, wall thickness, and strain distribution of bulged components. This aims to provide guidance for theoretical research and validate the superiority of the rapid deflation process. The results show that: (1) When forming cup-shaped components at 400 °C, the die-fitting degree of the component formed by the rapid deflation process reaches 89.5% and the minimum corner radius is 2.5 mm. Overall, the forming accuracy of this process is significantly superior to that of the rapid inflation process. (2) Within the temperature range of 400–450 °C, the rapid deflation process successfully formed a spherical-bottom component with a depth of 30 mm, overcoming the cracking defects induced by localized cooling and non-uniform temperature fields in the rapid inflation process, thereby improving the forming limit. (3) Under consistent conditions, the wall thickness uniformity of the sheet formed by the rapid deflation process is significantly higher than that of the sheet formed by rapid inflation, and the wall thickness uniformity improves with increasing temperature. Future work is expected to further enhance the repeatability and stability of forming accuracy and the forming limits of extreme geometries by further optimizing process parameters and expanding the material applicability range. This will provide practical technical support for the manufacturing of lightweight, high-performance aerospace equipment and automotive components.

Multi-view learning can effectively exploit the consistency and complementarity among multiple data sources and has become a major research direction in semi-supervised classification. However, the existing methods commonly suffer from several limitations, including the loss of view-specific information caused by premature feature fusion, interference from redundant inter-view noise, and the limited discriminative capability of consensus representation. These issues severely restrict classification performance under low-label settings. To address these limitations, this paper proposes Dual-branch Multi-view Learning with Dual-contrastive Information Bottleneck. The proposed framework constructs a decoupled dual-branch graph convolutional architecture to explicitly separate view-specific representations from cross-view consensus representation, thereby alleviating feature homogenization at the structural level. Furthermore, we design a dual-contrastive information bottleneck optimization mechanism, where the CLUB constraint minimizes redundant mutual information across views to suppress noise, while the InfoNCE constraint maximizes the mutual information between consensus representation and labels to enhance discriminative capability. Additionally, we employ an adaptive attention fusion module to dynamically integrate the dual-branch representations, further refining task-relevant features. The experiments conducted on nine public datasets demonstrate that the proposed method achieves favorable performance improvements over most of the selected comparison methods in semi-supervised classification tasks.