基于红外热成像与CNN的压裂装备故障精准识别及预警

Accurate identification and early-warning of faults of fracturing equipments based on infrared thermal imaging and convolutional neural network

关键词:  红外热成像  卷积神经网络  压裂泵  状态监测  故障识别

Key words:  infrared thermal imaging  convolutional neural network  fracturing pump  condition monitoring  fault identification

摘要:

页岩气大规模压裂作业过程中,以压裂泵为代表的压裂装备的安全性、可靠性直接关系到整体压裂作业的顺利进行。考虑到复杂工况及作业环境对振动分析的影响,且设备内部不便安装振动传感器,可引入红外热成像技术进行运行状态的监测。由于页岩气压裂设备外部壳体较厚,加之内部液体的降温作用,使得泵头体等常见故障区域温度表征不明显。针对此问题,引入卷积神经网络实现压裂装备故障精准识别和早期预警的智能化、无人化。通过模拟现场压裂工况,开展室内试验。结果表明,提出的压裂装备故障识别方法能够达到94.8％准确率,同时将预警时间提前了10 s,对于降低事故后果严重度有借鉴作用。

Abstract:

During the large-scale shale gas fracturing operation, the safety and reliability of the fracturing equipment represented by the fracturing pump is directly related to the smooth progress of the overall fracturing operation. Considering the impact of complex working conditions and operating environment on vibration analysis and the inconvenience of installing vibration sensors inside the equipment, infrared thermal imaging technology is introduced to monitor the operating status. Due to the thick outer shell of the shale gas cracking equipment and the cooling effect of the internal liquid, the temperature characteristics of common fault areas such as the pump head are not obvious. In view of this problem, convolutional neural network (CNN) was introduced to realize the intelligent and unmanned precision identification and early warning of fracturing equipment faults. By simulating on-site fracturing conditions and conducting laboratory tests, the analysis results show that the fracturing equipment fault identification method proposed in this paper can achieve an accuracy rate of 94.8%, and advance the warning time by 10 s, which is of great significance to reduce the severity of the accident consequences.