The edge of a hot rolling strip corresponds to the area where surface defects often occur. The morphologies of several common edge defects are similar to one another, thereby leading to easy error detection. To improve the detection accuracy of edge defects, the authors of this paper first classified the common edge defects and then made a dataset of edge defect images on this basis. Subsequently, edge defect recognition models were established on the basis of LeNet-5, AlexNet, and VggNet-16 by using a convolutional neural network as the core. Through multiple groups of training and recognition experiments, the model’s accuracy and recognition time of a single defect image were analyzed and compared with recognition models with different learning rates and sample batches. The experimental results showed that the recognition model based on the AlexNet had a maximum accuracy of 93.5%, and the average recognition time of a single defect image was 0.0035 s, which could meet the industry requirement. The research results in this paper provide a new method and thought for the fine detection of edge defects in hot rolling strips and have practical significance for improving the surface quality of hot rolling strips.

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.

Cold-rolled non-oriented silicon strip is widely used, and users have strict requirements for its transverse thickness difference. It is of great significance to study the quantitative relationship between the transverse thickness difference and incoming section profile of cold-rolled silicon strip and to formulate appropriate control indexes of the hot-rolled profile. To achieve the above purpose, this paper first proposes a method to describe the section profile of hot-rolled strip. A mechanism model for predicting the transverse thickness difference of cold-rolled silicon strip is established. Based on the characteristics of neural network transfer learning, the calculated results of the mechanism model are combined with actual production data, and the PSO-LM-BP neural network is trained by using the strategy of pre-training + retraining to obtain the mechanism-intelligence model for the prediction of the transverse thickness difference of cold-rolled silicon strip. The innovation of this paper is the combination of physical model and neural network. The prediction accuracy of the model is improved by two orders of magnitude on average, and the operation time is reduced. The relationship between the hot-rolled strip section crown, wedge and cold-rolled strip transverse thickness difference is quantitatively analysed, and the control strategy diagram of the key parameters of the hot-rolled section is finally obtained. The production of cold-rolled silicon strip with 1420 mm UCM shows that this strategy has a beneficial effect on the transverse thickness difference control of a cold-rolled strip.

To determine the exact distribution of tensile stress in a cold rolled strip, tensile stress evolution caused by lateral uneven velocity distribution outside the deformation zone of the strip is examined. The finite difference method is employed to solve the problem. Different factors, including the strip width and the form and amplitude of the initial stress distribution on the stress evolution, are analyzed. On the one hand, to improve the calculation speed, a “Gaussian” curve is proposed to fit the results calculated by the finite difference method. Simulation results show that the stress evolution calculated by the fitting equation is in good agreement with that obtained by the finite difference method. On the other hand, to verify the exactness of the model, an experiment is conducted, and the comparison between the calculated results and experimental values is discussed.