融合注意力与残差网络的石油管材失效宏观影像智能识别方法

Intelligent Analysis of Macro Failure Image of Oil Country Tubular Goods

基于深度学习的石油管材失效影像智能识别方法是目前的研究热点。针对传统的识别网络/模型存在精度不佳、泛化能力差等问题,基于残差块堆叠设计思路,建立了融合注意力与残差网络的石油管材失效宏观影像智能识别模型/算法。该算法赋予网络对石油管材失效典型宏观影像(断裂、腐蚀、磨损、形变)特征权重分配功能,有效提升神经网络捕获宏观影像关键特征信息的能力。与传统BottleNeck结构网络87.52%的平均分类精度相比,融合注意力与残差网络的AXBlock网络模型的平均分类精度达到94.93%。此研究工作将为石油管材失效智能诊断、失效分析预测与预防提供技术支撑。

The intelligent recognition method of oil country tubular goods(OCTG)failure images based on deep learning is currently a research hotspot.In response to the problems of poor accuracy and poor generalization ability of traditional recognition networks/models,based on the residual block stacking design and an intelligent recognition model/algorithm,an integrating attention and residual network for OCTG failure images classification was established.This algorithm endows the network with the function of assigning weight to typical macroscopic images of OCTG,such as fracture,corrosion,wear,deformation,effectively improving the ability of the neural network to capture key feature information in macroscopic images.Compared with traditional BottleNeck structured networks,the average classification accuracy of the AXBlock has been improved from 87.52%to 94.93%.This research work will provide technical support for intelligent diagnosis,failure analysis,prediction,and prevention of OCTG failures.