磨粒的计算机识别方法的研究与评述

铁谱技术是以磨损颗粒分析为基础的诊断技术，目前是监测机械磨损工况与诊断磨损故障最为有效的方法。而磨粒识别又是铁谱诊断技术的关键环节。本文综合分析计算机技术在磨粒识别中的应用，将磨粒识别方法分为人－机交互磨粒识别法、半自动磨粒识别法和自动磨粒识别法３种，并对此进行详细分析与比较。     

发动机铁谱磨粒分析与磨粒识别研究

从发动机摩擦副及其摩擦磨损特点着手，提出了用于发动机磨损状态监测与诊断的主要10类磨粒，依据其识别特征提出了用于磨粒识别的14个磨粒形状、表面纹理与颜色特征；最后应用BP网络分层识别策略进行了磨粒识别。     

基于磨粒表面信息的磨损表面特征评估

建立了基于磨粒表面信息的磨损表面评估方法。首先选择合理的磨粒和磨损表面特征参数，通过识别磨粒类型，获得磨损过程中具有典型性和代表性的磨粒类型，然后选取这些具有代表性的磨粒类型，得到磨粒的表面特征向量，进而来研究磨损表面和磨粒表面的映射关系，实现基于磨粒表面信息的磨损表面特征评估。实例表明，根据磨粒表面特征评估磨损表面特征是可行的。     

磨损微粒识别的模糊决策树方法

磨损微粒识别是在磨损工况监控中确定机器磨损类型的关键步骤。本文基于非单调模型推理方法，依据磨粒识别过程中所依赖特征的层次关系，建立了磨粒识别的模糊推理决策树。这一决策树开发的软件试验结果表明，本文所提出的方法在实现磨粒的自动识别方面是有效的。     

基于模糊神经网络的磨粒识别专家系统(续)

二、磨粒的分类及识别特征 铁谱分析技术将油液中的固体颗粒分为正常磨损、疲劳磨损、球状、层状、严重滑动、切削、腐蚀磨损、红色氧化物、暗金属氧化物、有色金属、非金属和晶体、污染物等12类。 领域专家根据磨粒的形态（总体形状、边缘细节和表面纹理）、颜色、粒度和厚度信息来识别铁谱磨粒,它们大都是模糊信息,一般用语义形式描述,因此需要采用模糊神经网络识别,识别结果为磨粒类型,系非模糊集合。磨粒特征见表1。     

基于链码的金属磨粒分形参数计算

介绍了基于链码的磨粒分形参数及其计算方法，采用编写的磨粒识别程序，对不同磨损阶段测得的磨粒分布数进行分析，可以得到磨粒二重分形参数和拐点数值，发现磨粒分布分形维数的变化与磨损状态改变相对应。该方法用于磨粒分形特征与磨损状态相关性规律的研究识别，简便快捷。     

基于显微图像的在线润滑油中磨粒识别

针对现有润滑油磨粒监测技术的不足,提出一种将微流控技术与铁谱技术相结合的磨粒在线检测方法。该系统用微流控芯片取代传统铁谱片并显微成像;在图像的模糊恢复上,采用基于运动模糊图像微分与自相关的方法进行模糊尺度计算,应用维纳滤波方法对模糊图像进行模糊恢复,得到纹理及边界更加清晰的图像;在磨粒特征的提取上,运用颜色特征用来识别磨粒的成分,包括铁质磨粒、有色金属磨粒和氧化物磨粒,运用尺度和形状等特征识别磨粒产生的磨损机制;在颗粒的识别上,应用灰色关联分析方法实现磨粒的自动识别。该磨粒在线监测系统不仅能对磨粒进行计数,也可以对磨粒类型进行判断,其准确率达到90%。     

基于神经网络的工程机械磨损故障诊断研究

根据工程机械结构特点、摩擦磨损规律和磨粒特征建立了标准磨粒谱 ;提出了基于神经网络的磨粒识别技术 ,设计了智能磨粒识别系统 ,诊断实例表明 ,用神经网络方法可以准确地识别工程机械磨损故障类型、程度和部位     

磨损磨粒的BP神经网络识别

介绍了铁谱分析技术对设备状态监测与故障诊断的方法;通过机械润滑油或液压油中微观磨损颗粒的分析来判断机器当前的工作状态.铁谱的计算机图像分析技术是近年来研究的热点.基于BP神经网络对磨损磨粒进行识别,提出了磨粒的分步识别策略,并以磨粒样本对网络进行训练,取得了较好的识别效果.     

图像颜色特征提取在铁谱图像分类及磨粒识别中的应用研究

铁谱图像所包含的彩色信息对磨粒识别、磨损形式分析非常重要。对图像分类技术进行研究，主要对铁谱图像的颜色特征进行研究，提出了聚类树分析、模糊聚类技术与统计分析相结合的定量研究方法，对铁谱图像进行背景、磨粒区域分割，以获得可进行定量分析的磨粒。计算的颜色特征为铁谱图像的进一步处理和识别以及磨粒的机器自动识别、磨损形式分析奠定了基础。     

磨粒轮廓特征变换研究

本文提出了一个磨粒轮廓特征变换的方法，该方法解决了磨粒图像的平移、旋转和大小不变性，同时，该变换所得的结果能较为全面地反映了磨粒的轮廓特征，这使磨粒识别具有较高的准确率成为可能。文中给出了一个利用此变换方法所得的轮廓特征作为神经网络输入的磨粒识别实例。     

基于自动磨粒分析技术的直升机在用发动机油界限值的研究

运用自动磨粒分析仪(LNF)测定直升机在用发动机油的切削磨损、滑动磨损、疲劳磨损和非金属颗粒,通过对数据的频数分布直方图的计算分析,研究得到切削磨损、滑动磨损、疲劳磨损和非金属颗粒的界限值,可以作为直升机在用发动机油监测的又一补充,弥补了原子发射光谱不能监测较大磨损金属颗粒的缺陷,为直升机的合理换油提供技术依据。     

Gradation of oxidational wear of metals

This paper considers mild-oxidational wear of metals by studying their behavior under friction with different loads. Low carbon, steel and copper are chosen as the model materials. We show that tribo-oxidation and the structure of surface layers of materials, both formed in the process of plastic deformation during friction, provide the boundary conditions of mild and severe wear. Oxidational wear is predominant when structural changes are minimal. As the load increases, oxidational wear is at first accompanied by metallic wear and afterwards the oxidational wear accompanies the metallic wear. The structure of the metal surface layers changes gradually during these processes, so that the strengthening of the metal is high enough to withstand friction forces. When the magnitude of frictional forces becomes higher than the maximal strength of the plastically deformed metal, the transition to severe wear occurs.The composition of different types of oxides and the fineness of wear particles varies with the friction conditions. Under light load friction conditions, fine wear particles are formed. These particles contain oxides of high oxygen content. As the friction conditions become tougher, in particular when the load increases, large-sized wear particles are formed. These particles contain oxides of a higher metal content. Phase composition and fineness of wear particles are used for gradation of mild wear.Analyses of phase composition of oxides and estimation of the fineness of wear particles are suggested as a method of wear character diagnostics. The electron diffraction method of the study of wear particles is used for this analysis in order to evaluate and choose appropriate friction and wear conditions.     

油润条件下磨粒分形行为与摩擦状态相关性的研究

以45#钢-铜配副在浸油状态下进行磨损试验,采用自行构建的磨粒分析系统考察不同磨损阶段产生的磨粒形态,测量多种磨粒表征参数,并就各表征参数与摩擦力的关联性进行了对比分析.分析结果表明:磨粒群体中轮廓分形维数的分布呈正态分布;磨粒分形维数相对于其它表征参数而言,和摩擦力相关性更大.     

铁谱磨粒多重分形特征研究

为有效描述铁谱磨粒特征,提出用多重分形谱参数表达磨粒形态特征的新方法。选择盒计数法计算磨粒图像的多重分形谱,研究磨粒多重分形谱的有效性,分析磨粒多重分形谱参数的不变性和鲁棒性;确定磨粒图像预处理方法,并对4类典型磨粒的多重分形谱参数进行统计分析。结果表明:将多重分形谱参数应用于磨粒识别,总识别率为82.5%。磨粒具有明显的多重分形特性,可用多重分形谱参数来描述磨粒的形态特征;多重分形谱参数具有平移不变性,但对灰度变化和噪声干扰的鲁棒性较差,在提取多重分形谱参数时,需要对磨粒图像做严格的预处理。     

基于解析方法的轮胎磨损颗粒物运动轨迹分析

轮胎磨损颗粒物的运动轨迹分析对磨损颗粒物的捕集具有非常重要的理论意义。建立了轮胎磨损颗粒物的运动微分方程,分析了车速和磨损颗粒物粒径对其运动轨迹的影响,设计试验验证了模型的正确性。结果发现：较小粒径（10μm）的磨损颗粒物与较大粒径（500μm）的磨损颗粒物的飞扬区间变化较小,粒径100μm左右的磨损颗粒物的飞扬区间变化较大。覆盖件150～250mm的高度区域是轮胎磨损微小颗粒物（小于100μm）的散射聚集区域。     

联合光度立体视觉和图像校正的磨粒三维形貌重建

分析铁谱技术在机械装备磨损状态检测中发挥了重要作用,但其仅能提供磨粒二维图像,导致磨粒形貌信息不足。为实现磨粒三维形貌的精确重建,联合光度立体视觉和图像校正,为精确重建磨粒的三维形貌,联合光度立体视觉和图像。该方法首先采用大津阈值法由全光源图像识别磨粒与背景区域;然后为消除LED发光强度差异对磨粒形貌重建的影响,结合平面形状和朗伯反射模型确定背景区域的理想成像亮度,并校正各光度图像序列的亮度;最后根据光度立体视觉方法,由校正后光度图像序列重建磨粒的三维形貌。以不同类型的磨粒为测试样本,将所提出的方法的重建结果与激光共聚焦显微镜的测量结果进行对比。结果表明：重建磨粒的形貌参数误差小于15%,表明提出的方法能够精确重建磨粒的三维形貌。     

Three-body wear of UHMWPE acetabular cups by PMMA particles against CoCr, alumina and zirconia heads in a hip joint simulator

Three-body abrasive wear of ultra-high molecular weight polyethylene (UHMWPE) acetabular cups by loose polymethyl methacrylate (PMMA) bone cement particles is an important mechanism responsible for elevated wear debris generation in total hip arthroplasty. The resistance of the femoral head material to third-body damage has been considered critical for the wear performance of the polyethylene component. This study examines the effect of loose bone cement particles on the wear rate of UHMWPE acetabular cups against both metal and ceramic counterfaces in a hip joint simulator. Against the CoCr head, the UHMWPE cup showed a strong dependence of wear rate on the concentration of the PMMA particles in the lubricant. At a concentration less than 5 g/l, the presence of the PMMA particles had no detrimental effect on the wear rate; higher concentrations of the PMMA particles greater than 5 g/l led to an accelerated wear of the acetabular cups. Mild scratching damage was observed on the CoCr heads after testing with all PMMA-containing lubricants. However, no increased UHMWPE wear rate was found against these damaged femoral heads in a fresh lubricant without PMMA particles, indicating that femoral head scratching was not a major cause for the elevated wear observed under the three-body abrasive conditions. Against both alumina and zirconia ceramic heads, the wear rate of the UHMWPE was independent of the concentration of the PMMA particles. It was observed that a significant portion of the CoCr heads was covered with loose patches of PMMA particles. The higher the concentration of the PMMA particles, the greater the area of the head covered with PMMA particles. The attachment of PMMA particles to the ceramic heads was much reduced compared to the CoCr heads. It is therefore concluded that ceramic femoral heads are effective against potential run-away wear of the UHMWPE acetabular cups when an excessive amount of loose PMMA particles are present in the lubricant.