中文摘要：

齿轮箱故障是造成风电机组停机时间最长的一种故障，对其故障进行早期预警，对保证整机的可靠运行和减少维修费用具有重要意义。文章提出了一种基于确定性随机子空间方法的齿轮箱故障预测算法，首先，该算法利用齿轮箱正常状态的实时监测振动和转速数据，建立齿轮箱的状态空间模型，并得到一组参考特征值：然后利用这组参考特征值与实际监测数据所求特征值进行比较，利用均方根误差（RMSE）作为齿轮箱故障预警指标，并结合统计过程控制原理定义该指标的门槛值，来实现对齿轮箱运行状态的监控。通过对实际监测数据的仿真验证．表明了所提方法的正确性和有效性。

英文摘要：

The longest downtime of wind turbine is due to failures of the gearbox. So early fault predic-tion of gearbox is meaningful for ensuring reliable running and reducing maintenance costs. A fault pre-diction algorithm for gearbox is proposed based on an identification method of deterministic combiningwith stochastic subspace. Firstly, a state space model of gearbox is built up by using vibration data andspeed data under normal operational condition. And model＇s eigenvalues, which are obtained by com-puting characteristic polynomial are defined as a reference eigenvalues. Then new eigenvalues are cal-culated by the real-time vibration data and these new eigenvalues are compared with the referenceeigenvalues. Their root-mean-square error is defined as an index of fault prediction of gearbox.Whether the index exceeds a threshold determined by SPC principle indicates the running state ofgearbox. The correctness and effectiveness are verified by the simulation.