To optimize the crawling phenomenon of slides under circumstances of low speed and a heavy load, a composite lubrication structure is adopted to alleviate the crawling phenomenon. The response surface optimal-design method establishes a quadratic mathematical model for multistage composite lubrication structure parameters, including crawling time and average friction coefficient. The optimal combination parameters of multistage composite lubrication structures have been determined. The optimal ratio of lubricating oil to molybdenum disulfide (MoS₂) has been identified, and a composite lubrication structure has been proposed to enhance the crawling phenomenon and friction performance of sliding guide rails under medium-speed and medium-load conditions. These research outcomes indicate that when low speed and a heavy load are present, the crawling time and friction coefficient initially diminish and subsequently augment as the width, spacing, and cycle length of the sinusoidal texture and the diameter of the hexagonal pit expand. The optimum configuration of multistage composite lubrication structures is as follows: The width of the sine-wave texture b amounts to 0.15 mm, the cycle length e is 2 mm, the spacing c is 1.5 mm, and the diameter of the hexagonal pit d is 0.2 mm. When the mass ratio of guide oil to MoS₂ is 2:1, it exhibits supreme crawling resistance and antifriction attributes. In circumstances involving a medium load and speed, multistage composite lubrication structures manifest pre-eminent friction performance. These data can steer the design of multistage composite lubrication structures on the surface of slide rails.

PCB defect images suffer from tiny defects, subtle morphological differences and complex background wiring, making traditional single-feature classification unstable. This paper proposes a dual-branch image classification method combining a Transformer and CNN, which jointly models local anomalies and global semantic relationships. The model uses a convolutional branch and a Transformer branch to extract local defect features and global wiring dependencies, respectively. A cross-layer semantic interaction mechanism is adopted for multi-level information fusion, and a discriminative feature enhancement module is applied to highlight key defect regions and suppress background interference. Experiments show that the model improves overall accuracy by over 2%, with an F1-score of 0.930 and defect identification coverage of 0.927. It performs stably across different defect types and background complexities without obvious bias, providing new insights for hybrid deep model design in industrial defect image classification.