基于油液全息在线监测的风电变速器摩擦学系统全寿健康状态分析方法

With the background of wind turbine gearbox, the project proposes a new life-cycle health analysis theory and method based on the lubricant hologram monitoring. And researches will be carried out centering on four scientific issues. First, morphological characterization of on-line fatigue wear debris is studied via the crack developing slip lines under rolling-sliding contact, and the theory and method of wear condition characterization are established based on the on-line wear debris features through fundamental and bench tests. Second, an oil hologram model coupled with physicochemical, contamination and wear information will be built by referring to lubricant degradation model, wear and lubrication theory. The on-line oil hologram monitoring will be realized by extracting the measurable components. Third, multi-intelligent-body method for multi-information fusion is studied to realize the characterization of oil state. Fourth, a new health pattern including normal, sub-health, breakdown and failure, is adopted for the wind gearbox life-cycle health description. With oil state fuzzy mapping method, an identification model capable of life-cycle self-learning would be constructed. By training in real wind turbine, a life-cycle health monitoring and predication system will be established for wind turbine gearboxes, which will provide basic theory and method for the safe, stable, and reliable performance for large wind turbines.

本项目以风电变速器系统为背景，提出了基于油液全息监测的全寿健康分析理论及方法，围绕三个科学问题:油液全息耦合模型与解耦、典型磨损机理与在线磨粒特征的映射关系、系统全寿命健康模式表征开展研究。具体内容包括4个方面：采用滚滑接触的疲劳裂纹滑移线研究在线疲劳磨粒的形态学特征，并通过基础和台架试验，建立在线磨粒特征表征磨损状态的理论及方法；结合润滑剂衰变模型、磨损和润滑理论，建立理化、污染、磨损的油液全息耦合模型，并提取可测特征分量实现油液全息在线监测；研究基于多智能体技术的信息融合方法，建立油液状态表征模型；采用正常、亚健康、故障、失效四个健康模式描述风电变速器系统的全寿健康模式，并结合强化试验分析，采用油液状态模糊映射方法建立风电变速器全寿健康模式识别的自学习模型。通过风机实机试验训练，构建风电变速器全寿命健康状态监测及预测系统，为大型风机的安全、健康、稳定运行奠定理论和技术基础。

课题以风电变速器为工程对象，研究了基于油液全息在线监测的摩擦学系统全寿健康状态分析方法，针对轴承及齿轮两种摩擦副，从在线磨粒特征与典型磨损机理的映射关系、润滑油全息特征监测传感技术开发、多信息融合表征润滑油理化和衰变规律、设备全寿命健康状态动态建模四个方面展开研究，主要成果包括：①研究了滾滑摩擦副疲劳裂纹拓展与微切削过程表征模型，建立了典型疲劳和切削磨粒的形态分析理论模型；②结合multi-SVDD、K-means和SVM智能算法建立了氧化、切削、球形、疲劳和滑动磨粒的在线识别模型，通过动态磨粒的数量、颜色和形态特征等建立了磨损状态在线综合描述模型；③针对油液全息特征的可测参量，研究了水分、颗粒、在线磨粒图像和动态磨粒视频采集传感技术，并实现了多传感技术的集成；④通过润滑油衰变机理及可测性分析，研究了油液理化特性的耦合关系及其全寿命磨损周期的演变规律，并采用Mean-Shift算法建立润滑油状态表征模型；⑤根据润滑油的衰变规律，融合油液理化信息表征参量，采用隐马尔可夫方法建立了润滑油衰变表征模型；⑥设计了风电变速器润滑油在线监测传感系统，完成了传感器、采集控制与传输三个模块的集成，加工并装配了监测系统样机并成功安装在3个风场的4台风机上，借助四球摩擦磨损全寿命监测数据，采用Mean-Shift算法建立了磨损状态转变辨识与全寿命演变描述模型，验证了在线磨粒特征在机器全寿命磨损状态监测中的有效性。课题成果的应用将为风电变速器磨损状态在线监测与健康分析技术提供必要的研究基础，对实现重大装备的视情维护具有重要的工程意义。