Surface defect automatic detection has great significance for copper strip production. The traditional machine vision for surface defect automatic detection of copper strip needs artificial feature design, which has a long cycle, and poor ability of versatility and robustness. However, deep learning can effectively solve these problems. Therefore, based on the deep convolution neural network and the transfer learning strategy, an intelligent recognition model of surface defects of copper strip is established in this paper. Firstly, the defects were classified in accordance with the mechanism and morphology, and the surface defect dataset of copper strip was established by comprehensively adopting image acquisition and image augmentation. Then, a two-class discrimination model was established to achieve the accurate discrimination of perfect and defect images. On this basis, four CNN models were adopted for the recognition of defect images. Among these models, the EfficientNet model through transfer learning strategy had the best comprehensive performance with a recognition accuracy rate of 93.05%. Finally, the interpretability and deficiency of the model were analysed by the class activation map and confusion matrix, which point toward the direction of further optimization for future research.

Seamless flatness rollers are the core equipment for cold-rolled strip flatness detection. The change in wear resistance of the roller surface-strengthened layer has important guiding significance for the prediction of the roller surface life. The experimental results of the five rounds of friction and wear testing show that the volume–wear rate of the original roller surface is only 1.85 × 10⁻⁶ mm³ (N·M)⁻¹. After the first grinding, the wear resistance of the roller surface decreased and the wear rate was approximately 2.76 × 10⁻⁶ mm³ (N·M)⁻¹. After the second grinding, the wear morphology, wear rate and wear resistance of the samples in each round showed little difference. The average wear rate in the last three rounds of the test is 3.56 × 10⁻⁶ mm³ (N·M)⁻¹ and the wear resistance is significantly lower than that of the previous two rounds of the test. The calculation shows that the original surface wear resistance is 32.92% higher than that of the surface after the first grinding, and the average wear resistance of the surface after the first grinding is 22.59% higher than that of the surface after the second, third and fourth grinding. With long-term use and repeated grinding repair, the roller surface will show a trend of wear resistance decline. When the wear resistance decreases to a certain degree, it will reach a stage of stability. These phenomena provide regular and quantitative references for the study of the life prediction of flatness meter rollers.

The surface of a whole-roll flatness roll is in long-term contact with the steel strip, leading to slipping and wear and placing higher demands on the performance of the roll surface. This study establishes a finite element model for moving induction quenching and a phase transformation hardness numerical model by generating multi-field simulations and hardness predictions for the flatness roll during induction quenching. First, the thermal–physical properties of the roll material, MC3, are calculated using JMatPro V13.0. The dynamic domain and moving mesh techniques are applied in COMSOL Multiphysics to simulate time-varying boundary conditions, and the JMAK and K-M phase transformation models are used for electromagnetic–thermal–microstructure field simulations. Subsequently, the Taguchi method is used to optimize the induction quenching process of the flatness roll. After optimization, the martensitic hardened layer depth along the axial direction of the roll becomes uniformly distributed near the target value of 3 mm. Finally, through the modified Maynier hardness model, the corrected formula for the Vickers hardness of MC3 is obtained. The calculated hardness value of the roll surface in the simulation model reaches 950 HV, which agrees well with the experimental hardness results, validating the ability of the numerical model to guide specific processes.

This paper investigates collaborative trajectory planning and tracking control for semi-trailer-mounted lifting vehicles used in wind turbine blade transport on complex mountainous roads. The proposed approach addresses the challenging problem of coordinating the articulated chassis with a three-degree-of-freedom upper operating device during ultra-long blade transport. First, a unified kinematic model of the articulated chassis and the three-degree-of-freedom upper operating device is established for collaborative trajectory planning and tracking control under low-speed transport conditions. In addition, a dynamic formulation is provided to describe the inertial coupling characteristics of the vehicle-blade system. Second, a continuous trajectory planning method driven by task gradients is proposed to generate collision-free trajectories satisfying multiple constraints, including kinematic feasibility, obstacle avoidance, and actuator limits. A damping injection strategy is introduced to improve numerical conditioning during the iterative optimization process. Third, a linear time-varying model predictive controller is designed based on local linearization along the reference trajectory, which handles system constraints explicitly and achieves high-precision trajectory tracking. Simulation and hardware-in-the-loop experiments show that the proposed method generates smooth and feasible trajectories, while the tracking controller achieves engineering-level position accuracy and joint-angle tracking errors below 3 degrees, satisfying practical requirements for wind turbine blade transport on mountainous roads.