摘要
轧辊表面形貌直接影响板带钢产品的表面质量甚至其织构组成。为了定量研究轧辊磨损过程中其表面形貌特征的变化规律,采集了1个磨辊周期内不同轧制阶段的轧辊表面图像信息,基于MATLAB平台,利用图像处理方法对不同阶段的下机轧辊表面图像进行了预处理,提取样本图像的几何形状及纹理特征等17维图像特征参数和分形维数,经过对这些参数的属性约简建立了用于轧辊磨损形貌状态识别的BP神经网络模型。结果表明:等效面积圆半径、圆形度、纹理熵及二阶矩等轧辊形貌图像特征参数和分形维数可以作为描述轧辊表面形貌的定量指标,并可用BP网络模型对轧辊磨损形貌进行识别预测。为建立轧辊磨损形貌的定量评价体系提供了新途径。
The surface morphology of roll is directly related to the strip products＇ surface quality and even the texture component. Quantitative monitoring of the morphological changes in the wear process of roll was researched in this study. Based on the MATLAB tool, the method of image processing was used to make pretreatrnent of roll surface morphology in different stages, and the study got 17 dimensional features of geometry, texture characteristics and fractal dimension. BP neural network model using to recognize the roll wear states was also established after parameters simplification. The results show that the image features, such as equal- area- cirde radius, circularity, texture entropy and second-order moment, and fractal dimension can be used as quantitative parameters to describe the surface morphology of roll. It＇ s also useful to use BP network model to recognize and judge surface morphology of roll wear. This study will supply a new approach to establish the quantitative evaluation system for the wear morphology of roll.
出处
《钢铁》
CAS
CSCD
北大核心
2015年第4期95-100,共6页
Iron and Steel
关键词
轧辊磨损
表面形貌
图像特征
分形
roll wear
surface morphology
image feature
fractal
