风力机叶片蒙皮初始裂纹的盲信号提取

针对风力机叶片初始裂纹特征难以提取的问题,提出了一种逐步提取并消减噪声源信号从而获得微弱裂纹故障特征的盲提取方法.首先基于卷积混合模型极小化改进代价函数推导自适应学习迭代算式,在仿真实验中确定非线性激励函数和滤波器的传输函数,根据输出信号的性能参数证明了改进算法对尖峰噪声的异常点更加敏感稳健.在风力机叶片疲劳实验台上模拟叶片蒙皮的初始横向裂纹,通过声发射信号采集系统获得观测信号,分析噪声源的特性并提取了初始裂纹的声发射信号特征,为风力机叶片状态监测和预警提供了依据.

Blind signal extraction for initial crack on blade skin of wind turbine

A blind extraction method is presented through progressive extraction and reducing noise signal to obtain the weak crack fault feature,which can solve the difficult feature extraction problem of initial crack on wind turbine blade.Firstly,an adaptive iterative learning algorithm is derived from minimizing the improved cost function based on convolutional mixing model.In simulation experiment,the non-linear activation function and filter transferring function are determined;and the performance parameters of the output signal approve that the algorithm is more sensitive and robust for the abnormal points of the peak noise.Finally,the initial transverse crack was simulated on the fatigue experiment test rig of wind turbine blade.The mixing signals were collected using the acoustic emission acquisition system,and the weak crack feature is analyzed and extracted based on the improved algorithm.The proposed method provides the foundation for the condition monitoring and early warning of wind turbine blade.