航空发动机滑油磨粒浓度在线检测技术研究  

Research on On-line Detection Technology of Aero-engineLubricating Oil Abrasive Particle Concentration

作　　者：侯媛媛 李江红[2] HOU Yuan-yuan;LI Jiang-hong(School of Computer Science,Xi'an Aeronautical University,Xi'an 710077 China;School of Power and Energy,Northwestern Polytechnical University,Xi'an 710072 China)

机构地区：[1]西安航空学院计算机学院,陕西西安710077 [2]西北工业大学动力与能源学院,陕西西安710072

出　　处：《自动化技术与应用》2025年第2期123-126,154,共5页Techniques of Automation and Applications

基　　金：陕西省重点研发计划项目(2023-YBGY-131);陕西省重点研发计划项目(2023-YBGY-014)。

摘　　要：针对滑油中磨粒形状复杂且尺寸大小不一,传统滑油磨粒检测方法存在时效性差、检测尺度小、精度低、非铁磁性磨粒不能检测等缺点。研究航空发动机滑油磨粒浓度在线检测技术。设计连续流微流控芯片图像采样方案,构建滑油图像采样系统;基于YOLOv3模型,设计SER算法,优化模型的推理置信度阈值;通过滑油磨粒浓度测试分析滑油图像采样偏差和模型误差。结果表明,该系统能够检测较宽尺度范围的磨粒浓度,且单次检测耗时较快,仅13.42 s。在置信度0.95下,总体磨粒浓度误差能够控制4.16%内。In view of the complex shape and different sizes of abrasive particles in lubricating oil,the traditional methods for detecting abra-sive particles in lubricating oil have shortcomings such as poor timeliness,small detection scale,low precision,and inability to de-tect non-ferromagnetic abrasive particles.It studies the online detection technology of obrasive particle concentvation in aircaft engine lubricating oil.In this paper,a continuous flow microfluidic chip image sampling scheme is designed,and a lubricating oil image sampling system is constructed,based on the YOLOv3 model,a SER algorithm is designed to optimize the inference confi-dence threshold of the model,the lubricating oil image sampling deviation and model are analyzed through the lubricating oil abrasive particle concentration test.The results show that the system can detect the concentration of abrasive particles in a wide range of scales,and the single detection takes only 13.42 s.At a confidence level of 0.95,the overall abrasive concentration error can be controlled within 4.16%.

关 键 词：航空发动机滑油 磨粒浓度 在线检测 SER算法 深度学习 YOLOv3模型 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程] TP391[自动化与计算机技术—控制科学与工程]