中文摘要：

为了减少风机齿轮箱严重故障的发生,提出了一种基于随机子空间识别方法的齿轮箱故障预测算法。该算法首先建立齿轮箱的随机状态空间模型,并利用正常运行时的振动监测数据计算模型的参数矩阵的特征值,并将其作为参考特征值;然后将由实际振动数据所求得的特征值与参考特征值进行比较,如果两者误差很小,则说明齿轮箱正常,反之则异常。为了减少计算量,引入均方根误差（RMSE）作为齿轮箱故障判别指标,并利用统计过程控制（SPC）原理定义该指标的阈值。最后,对一台实际风机的振动监测数据进行仿真,结果表明了所提出算法的有效性。

英文摘要：

A gearbox fault prediction algorithm based on the stochastic subspace identification is proposed to reduce the serious faults of wind turbine. A stochastic state space model of gearbox is built and the vibration monitoring data in normal operation are applied to calculate the eigenvalues of its parameter matrix,which are taken as the references. The eigenvalues calculated based on the actual vibration monitoring data are compared with the references and their differences are used to detect the gearbox fault. The RMSE（Root-Mean-Square Error） is introduced as a fault detection criterion to reduce the calculation load and its threshold is defined by SPC（Statistical Process Control） principle. Simulation is carried out with the vibration monitoring data of an actual wind turbine and the simulative results verify the effectiveness of the proposed algorithm.