摘要
将多传感器信息融合技术应用于抽油管缺陷在线检测系统。油管缺陷定量检测的多传感器信息融合模型的建立分别在数据层、特征层和决策层三个融合层次上进行;选取4路传感器信号进行信号直接融合,通过硬件直接形成油管的偏磨信号;建立了基于插值法的偏磨缺陷的定量分析方法,并给出了实测结果。对于坑状缺陷,通过对28路传感器所观测的目标进行统一的特征融合,提取特征向量,利用神经网络的决策模型完成了坑状缺陷的量化分析。
The data fusion models that we used quantitative recognition to detects ol oil-well tubing were based on data fusion layer, feature fusion layer, decision fusion layer. 28 hall sensors were contributed on oil-well tubing to collect signals. Signals from 4 channels were fusioned together directly through hardware and output signals of abrasion on oil-well tubing. Lagrange interpolating formula was applied estimate to the size of abrasion defects by experiments. Signals from 28 channels were used to pickup feature information to pits on oil-well tubing. The decision model based on neural-network was applied to estimate the size of pit defects.
出处
《中国机械工程》
EI
CAS
CSCD
北大核心
2005年第17期1512-1515,共4页
China Mechanical Engineering
关键词
抽油管
无损检测
多传感器融合
缺陷
oil-well tubing
nondestructive testing
multi-sensor fusion
defect