齿轮微点蚀损伤的磨粒浓度动态量化表征及识别算法研究

Micropitting is a kind of early fatigue wear, and it is difficult to be recognized and monitored in real-time. Dynamic quantitative characterization and recognition of gear micropitting failures based on debris concentration of in-used lube oil by on-line monitoring are proposed. Firstly, a numerical method of modified Archard’s wear model with Palmgren-Miner theory for predicting wear mass of micropitting with mild wear will be investigated. Then, attenuation rules of debris with different dimensional sizes are studied. A multi-dimensional analysis method with time domain and debris dimensional size domain is adopted to mapping debris concentrations of characteristic dimensional sizes with wear rates of specified wear mechanisms, in which removal of wear debris and dissipation of lube oil are taken into account. A failure evolution of gear micropitting is monitored by the wear rate with the debris concentration of the characteristic dimensional size in time sequence. Hence, dynamic quantitative characterization of gear micropitting failures is achieved. Moreover, an optimal estimation of unscented Kalman filter algorithm with sliding time window will be applied to the monitored wear rates to filter high-frequency noises, and a curve of micropitting evolution under changing working conditions can be obtained. Lastly, a warning of gear micropitting failures is set according to time-series pattern matching based on calculating dynamic time warping distance. The proposal breaks barriers in gear micropitting failure monitoring via vibration methods. It is one of key technologies of ensuring high efficiency, safety in services and long-life operations of gear transmission systems, as well as very important for gear lubrication and life-cycle design.

齿面微点蚀是一种齿轮早期疲劳损伤，难以实时监测与识别。本项目提出基于齿轮在用润滑油磨粒浓度在线监测，实现齿轮微点蚀损伤的动态量化表征与识别。首先，结合Archard模型和Palmgren-Miner疲劳理论，研究齿轮微点蚀伴随轻微磨损的磨损量数值计算方法。其次，研究磨粒衰减和润滑油消耗时，不同粒径磨粒浓度在润滑系统中的动态特性，建立多维度磨粒浓度与磨损率之间的映射关系。第三，提出采用“时域-粒度域”多维度分析方法，获取微点蚀演变过程中特征粒径磨损率，实现齿轮微点蚀动态量化表征。第四，应用滑动时间窗和无迹卡尔曼滤波，对变工况下的齿轮微点蚀磨损率进行最优估计，滤除监测数据的高频噪声，获得齿轮微点蚀的演化曲线。第五，基于动态时间弯曲距离的时序模式匹配方法，实现齿轮微点蚀失效预警。本项目对于齿轮润滑和寿命设计，保障重大装备齿轮传动系统高效、安全和长寿命运行具有重要意义。

齿面微点蚀的实时监测与识别是一个研究热点，也是一项急需发展的重大装备监测关键技术，尚未形成完整的理论与方法。本项目主要研究了齿轮疲劳及黏着磨粒生成的模拟计算，润滑系统中动态时变磨粒浓度建模，磨粒浓度动态降噪、磨粒多维度分析与磨损失效匹配、以及润滑油磨粒采样概率分析技术，并提出了相应的预测技术。具体进展包括：.1）提出了疲劳磨粒生成机理及数值仿真方法。.2）基于齿轮胶合磨损的机理及Rabinowicz黏着磨损理论，计算了软齿面齿轮发生黏着磨损的磨粒最小粒径，通过获取在线图像可视铁谱磨粒百分覆盖面积指数和特征粒径大于黏着磨损最小粒径的大磨粒数量，实现了齿轮胶合磨损的在线监测。.3）提出了对于不同磨损率激励的润滑系统，磨粒浓度可以表示为摩擦副磨损率与磨粒衰减函数的卷积除以润滑油的总量。.4）基于磨粒在取样油管中的运动分析与仿真计算，提出了采样油液流速的设置规则，保证目标磨粒能够到达传感区域。采用VOF和DPM模型对监测对象油液中的磨粒分布情况进行数值仿真，实现了根据设备类型和工况选择合适的采样位置，减小了取样油液中磨粒信息的损失。基于数据的统计学分析方法，对OLVF在监测过程中特征磨粒的采样概率进行建模，建立了特征磨粒采样概率与磨粒数量、采样间隔以及采样数量之间的数学关系。.5）基于磨粒浓度“时域-粒度域”多维度分析，实现了微点蚀动态识别。.6）基于磨粒沉积率的磨粒浓度量化表征指数及其抗饱和抗沉积非线性提取算法，实现了OLVF监测磨粒浓度指标的动态表征降噪。.7）采用动态时间弯曲距离定量分析当前磨损模式与正常磨损模式的匹配程度，实现了齿轮微点蚀在线失效预警。.相关理论研究进行了大量试验研究，并且获得了企业的认可，部分研究内容已经进入产品，预计未来5年能够产生一定的经济效益和社会效益。.上述工作是实现齿轮磨损监测的重要理论和技术基础。