【作者】 李绍成；

【导师】 左洪福；

【作者基本信息】 南京航空航天大学 ， 机械电子工程， 2009， 博士

【摘要】 油液磨粒监测是机械设备磨损状态监测的重要手段之一,可以为机械设备故障诊断和视情维修提供分析依据。本文在对现有油液磨粒监测技术深入分析的基础上,针对现有监测技术的不足,结合油液磨粒监测的实际要求,研究了基于静电感应和显微图像的油液磨粒监测方法。论文的主要工作及创新点如下:1.设计了一种油液磨粒在线监测静电传感器。给出了静电传感器的结构模型,采用有限元方法对传感器静电场进行了数值求解,得到了静电传感器的灵敏度空间分布,并对影响静电传感器灵敏度的主要因素进行了仿真研究;对静电传感器的频率响应特性进行了分析和实验研究;提出了静电传感器结构尺寸优化设计方法。2.研究了基于显微图像分析的油液磨粒在线监测方法。设计了基于显微图像的油液磨粒在线监测系统;分析了油液磨粒图像的特点,给出了磨粒目标提取的主要流程;分析了磨粒图像模糊退化机理,研究了磨粒图像模糊退化模型,提出了基于微分图像自相关的磨粒图像模糊尺度计算方法,在此基础上,研究了磨粒图像模糊恢复方法;提出了基于差值图像粗分割和Otsu算法相结合的磨粒图像分割方法;利用粗糙集理论对磨粒特征描述体系进行了优化处理;利用最小二乘支持向量机设计了磨粒综合分类器,并采用粒子群算法对支持向量机模型中的基本参数进行了优化选取。3.提出了一种数字摄像机和数码相机相结合的大磨粒显微图像检测方法。设计了大磨粒图像检测系统,给出了大磨粒图像采集与处理的方法和步骤;研究了基于彩色空间转换和小波分析的磨粒多聚焦图像融合方法;研究了基于相位相关配准和小波融合的磨粒图像拼接方法;利用最小二乘支持向量机设计了铁磁性磨粒分类器。4.构建了油液磨粒静电监测实验平台,进行了初步的磨损静电监测实验研究;构建和开发了基于显微图像的油液磨粒在线监测系统,并对系统性能进行了验证;构建和开发了大磨粒图像检测系统;给出了基于静电感应和显微图像的油液磨粒集成监测方法。更多还原

【Abstract】 Oil wear particle monitoring is one of the important means for condition monitoring of mechanical equipments, which can provide analysis basis for fault diagnosis and condition based maintenance of mechanical equipments. Based on the study of the current technologies for oil wear particle monitoring, for the shortage of the current oil wear particle monitoring technologies and the demand of oil wear particle monitoring, a new oil wear particle monitoring method based on electrostatic sensor and microscopical image analysis is proposed. The main contents are listed as follows.1. An electrostatic sensor with ring-shaped electrode is designed for wear particle on-line monitoring. The structural model of the electrostatic sensor is given. The analysis method for electrostatic field based on finite element method is studied. Numerical solution of the sensor electrostatic field is carried out by using the finite element analysis software ANSYS. The spatial sensitivity of the sensor is derived. The effect factors of the sensor sensitivity are researched in detail. The frequency response characteristic of the electrostatic sensor and its effect factors are investigated theoretically and experimentally. The optimal method on structural dimension design of the electrostatic sensor is proposed.2. An on-line oil wear particle monitoring system based on microscopical image analysis is designed. The main procedure of wear particle object extraction is given according to the image characteristic. The degrading mechanism of the blurred wear particle image is analyzed. The degradation model of the blurred wear particle image is derived. The method on the blur parameters calculation of the blurred wear particle image based on difference and autocorrelation is proposed. On this basis, the restoration method of the blurred wear particle image is studied. The wear particle image segmentation method based on oil background image and Otsu is proposed. The characteristic parameter system of wear particle is optimized based on rough sets. The classifier for two kinds of wear particle is designed based on the least squares support vector machines, and the parameters of this model are optimized by particle swarm optimization algorithm. Based on this classifier, the integrative wear particle classifier is designed according to the wear particle recognition system.3. The detection method for large wear particle based on the digital video recorder and common camera is proposed. A large wear particle detection system based on microscopical image is designed .The image fusion method based on color space conversion and wavelet analysis is studied. Image mosaic method based on phase correlation and wavelet fusion is researched. The classifier for ferromagnetic wear particle is designed using the least squares support vector machine.4. The experiment platform for wear electrostatic monitoring is established, and the preliminary experiment research is carried out. An on-line oil wear particle monitoring system based on microfluidic chip and microscopical image is developed, and the system’s performance is tested by the particle counter and the ferrography technology. A large wear particle detection system based on microscopical image fusion and mosaic is developed. The integrated monitoring method based on electrostatic sensor and microscopical image analysis is studied.更多还原