油液在线监测系统中磨粒识别技术研究

针对磨损状态监测要求，构建了基于显微图像分析的油液在线监测系统。根据系统光路特点，对磨粒图像进行了基于彩色特征的转换，并通过与背景图像的差值处理来快速提取磨粒目标。基于最小二乘支持向量机设计了磨粒两类分类器，并利用粒子群优化算法对最小二乘支持向量机模型中的参数进行了优化选取；根据磨粒识别体系，设计了基于最小二乘支持向量机的磨粒综合分类器。最后，利用铁谱分析技术对系统性能和识别效果进行了检验，结果表明本系统具有较高的检测精度和识别效果。

Research on Technologies of Wear Debris Recognition for the On-line Oil Monitoring System

For the demand of on-line wear monitoring, an on-line oil monitoring system is constructed based on micro-image analysis. The image of wear debris is transformed as gray image according to its colorful feature based on the system light route characteristic, and the wear debris object is gotten by subtracting the background image from the wear debris image. The classifier for two kinds of wear debris is designed based on the least squares support vector machines, and the parameters of this model are optimized by particle swarm optimization algorithm. The integrative wear debris classifier is designed by the least squares support vector machines according to the wear debris recognition system. The performance of this on-line oil monitoring system is tested by the ferrography technology. The result shows that the detection precision of this system is high.