Accurate and efficient detection of small and complex surface defects on hot-rolled steel plates remains a significant challenge in industrial quality assurance. Current deep learning detectors often exhibit limitations in detection accuracy and training convergence speed, particularly for small objects, which limits their practical deployment in real-time industrial inspection systems. To overcome these deficiencies, this paper proposes a multi-scale deformable transformer iterative query refinement network (MDT-Net). MDT integrates three key innovations: a Swin Transformer backbone for robust multi-scale feature representation, a deformable attention mechanism to significantly reduce computational complexity and accelerate convergence, and an iterative bounding box refinement strategy for precise localization. Extensive experiments on the NEU-DET dataset demonstrate MDT’s superior performance, achieving 82.7% mAP₅₀. Crucially, MDT significantly outperforms other mainstream detectors in small object detection, reaching an mAP_50:95 of 0.55, and exhibits remarkably faster training convergence. These findings confirm MDT’s effectiveness and robustness for accurate and efficient steel surface defect detection, thereby providing a crucial tool for enhancing sensor-based quality control and offering a promising solution for industrial quality management.

Laser assistance offers a promising pathway for high-efficiency and low-damage ultraprecision grinding for difficult-to-machine hard-brittle semiconductors. This study employs atomistic simulation to investigate the surface removal and subsurface damage mechanisms of C-, M-, and A-plane AlN workpieces during single-grit laser-assisted nanogrinding (LAG). The results indicate that LAG reduces material pileup, thereby decreasing the grit–workpiece contact area and grinding resistance. By leveraging laser-induced thermal effects to enhance atomic plastic flow, LAG evidently achieves a higher material removal rate than conventional grinding (CG). Grinding the C-plane along a orientation yields the lowest surface roughness, although this improvement is not useful for the M- and A-planes. Tangential force increases linearly with grinding depth in both methods, but LAG exhibits a lower rate of increase. LAG consistently produces lower grinding forces and friction coefficients and results in lower dislocation densities in C- and A-plane AlN workpieces at nearly all grinding depths. The C-plane exhibits the thinnest damage layer, followed by the M-plane, with the A-plane the thickest. Increasing the laser power density lowers the grinding force and enhances the removal efficiency. Optimal power density minimizes subsurface damage and improves surface quality; however, excessive power density exacerbates damage. This work provides valuable insights for developing high-efficiency, low-damage LAG techniques for hard-brittle semiconductors.