针对风电齿轮箱齿面微点蚀损伤在线难监测，难预警的困境，本项目拟开展齿面微点蚀损伤在线油液磨粒-SCADA数据协同感知与预警方法研究。首先，提出基于SCADA数据的风电齿轮箱实况载荷蒙特卡罗模拟方法，结合部分弹流和细观多轴疲劳理论，建立随机瞬时极端载荷诱变下齿面微点蚀损伤模型，分析损伤过程中微点蚀磨粒的产生规律。其次，研究了磨粒在润滑系统中动态特性。建立多过滤循环强润滑风电齿轮箱磨粒衰减模型，分析微点蚀损伤过程润滑系统中磨粒浓度及其粒径分布的演变规律。再次，提出了基于磨粒沉积率的磨粒浓度指数及抗饱和非线性提取算法，开发了基于运动目标识别和深度学习的磨粒粒径及形态特征的抗气泡分割与识别算法。最后，提出了风机齿轮箱磨损健康指数的构建方法，建立了在线油液磨粒监测磨损状态评估框架，开发了齿轮微点损伤状态在线自适应预警算法。该研究为风电齿轮箱齿面微点蚀损伤的在线监控提供了技术途径和理论依据。

Aiming at the difficulty of on-line monitoring and early warning for the micro-pitting failure of tooth flanks in a wind power gearbox, this proposal is going to carry out the research on the collaborative perception and early warning method of micro-pitting failure based on both on-line oil debris and SCADA data. Firstly, an SCADA data based Monte Carlo simulation method for the practical load of gearbox is proposed. Combined with partial film EHL theory and multi-axial fatigue theory at the microscopic scale, the micro-pitting failure model of tooth flanks under stochastic transient extreme load is established to analyze the development rule of micro-pitting debris over time. Secondly, the dynamic characteristic of wear debris in lubrication systems is investigated. A wear debris attenuation model for a strong lubrication with multi-filtration cycle system in wind turbine gearbox is determined, to analyze the evolution of debris concentration and size distribution in the lubrication system over the micro-pitting damage process. Thirdly, a deposition rate based index of debris concentration with anti-saturation nonlinear is proposed. A matching algorithm is also developed. Moreover, an anti-bubble segmentation and recognition algorithm based on moving target recognition and deep learning is proposed. Finally, a wear health index construction method is proposed. In view of this, an evaluation framework of wear state based on on-line oil debris monitoring is established. On-line early warning algorithm for the micro-pitting failure of tooth flanks in a wind power gearbox is developed. The result will provide a technical and theoretical basis for on-line monitoring of tooth flanks micro-pitting in wind turbine gearboxes.