基于改进HRNet和椭圆拟合的PDC钻头复合片磨损定级方法

打开文本图片集

中图分类号：TE921 文献标识码：ADOI：10.12473/CPM.202407043

Abstract：The current PDC cuter wear grading has problems of low efficiency，poor standardization and insuffcient accuracy.This paper presents a wear grading method based on improved HRNet and elipse fitting. First，the contour of the PDC cuter was extracted.The depthwise separable convolution was used to replace the general convolution of the HRNet network，andthe four diffrent resolution outputs of the network feature extractionlayer were fused bytheCGAF module.Meanwhile，the EMA attention module was introduced into the fusion feature map. Second，the RANSAC ellpse fiting algorithm was used for data preprocessing，fiting ellpse screening and interior point supplement，thereby improving the fiting accuracy of the segmented PDC cutter.Finally， the wear grading was conducted for the PDC cuter. The study results show that the improved HRNet network achieves the mean intersection over union and the mean pixel accuracy of category of 97. 60% and 98.79% ，respectively，with the number of parameters reduced by 72% ，the model size reduced by 70% ， and the grading accuracy reaching 96. 5% at the mean time of only O.261 s required for grading a single PDC cutter. The research findings provide theoretical reference for drilling cost reduction in oilfields.

Keywords： PDC cutter; wear grading; HRNet network model; elipse fitig; CGAF module; semantic seg. mentation

0 引言

随着设计制造技术的日臻成熟，聚晶金刚石复合片（polycrystalline diamond compact，PDC）钻头现已成为全球油气钻探不可或缺的破岩工具[1]。（剩余13612字）