石油地质与工程2026，Vol.40Issue(5)：50-55,6.DOI:10.26976/j.cnki.sydz.202605008
基于多模态特征融合的岩屑识别系统设计与实现
Design and implementation of cuttings recognition system based on multimodal feature fusion:a case study of Changqing Oilfield in Ordos Basin
摘要
Abstract
Aiming at the severe cuttings fragmentation induced by PDC drilling technology and the low identification accuracy of conventional manual lithology discrimination,an intelligent lithology identification system based on multimodal data fusion is proposed in this paper.A high-resolution acquisition platform is constructed for collecting cuttings images and X-ray elemental data,which integrates image features including color,texture and fluorescence of cuttings as well as geochemical elemental characteristics.An improved MobileNetV2 multi-task learning framework incorporated with a channel attention mechanism is employed to extract bimodal features separately,and intelligent classification of cuttings is further accomplished via multimodal feature concatenation and deep fusion.Experimental validation demonstrates that the proposed multimodal fusion-based lithology identification system improves the identification accuracy from 84.5%achieved by single-modal approaches to 91.5%.The field coincidence rates of lithology identification reach 90.6%for oil wells and 82.0%for gas wells respectively.The developed system overcomes the core technical bottleneck of field cuttings identification,provides an efficient solution for intelligent mud logging,and improves the decision-making reliability of oil and gas exploration.关键词
岩屑智能识别/多模态数据融合/岩屑图像数据/元素数据Key words
intelligent identification of cuttings/multimodal data fusion/cuttings image data/elemental data分类
信息技术与安全科学引用本文复制引用
刘杰,黄子舰,吴汶芮,杨占伟,赵志明,刘波..基于多模态特征融合的岩屑识别系统设计与实现[J].石油地质与工程,2026,40(5):50-55,6.基金项目
辽宁省教育厅2023年度科学研究项目"数字化岩屑智能识别技术研究"(JYTMS20231420). （JYTMS20231420）
