论文部分内容阅读
针对常规Hurst指数估计法计算赫斯特(Hurst)指数时受拟合长度与直线拟合方式影响较大的问题,提出一种改进标度长度选取方法及进行分段拟合的优化措施。其中标度长度采用2的指数方式选取,而将重标极差序列曲线分为4段,选取其中合适的1段曲线进行直线拟合,所得的直线斜率即是Hurst指数估计值。通过数值仿真验证了该方法比常规Hurst指数估计法具有更好的估计精度。分析了声发射信号的Hurst指数与近似熵作为表征碰摩发生及强度的特征参数的有效性。在转子试验台上采集不同碰摩状态下的声发射信号,将提取的Hurst指数与近似熵以单独和组合的方式分别作为BP神经网络的输入来识别不同的碰摩状态。实验结果表明:Hurst指数与近似熵相结合具有较高的识别准确率,为旋转机械碰摩故障识别提供了一种新的方法。
Aiming at the problem that the Hurst exponent is influenced by the fitting length and the line fitting method when the conventional Hurst exponent estimation method is used, a method of improving the selection of scale length and the optimization of piecewise fitting are proposed. Wherein, the scale length is selected by the exponential method of 2, and the curve of the reprogramming difference sequence is divided into 4 sections. One of the suitable section curves is selected for straight line fitting, and the obtained slope of the straight line is the Hurst index estimation value. The numerical simulation shows that this method has better estimation accuracy than the conventional Hurst index estimation method. The Hurst exponent and approximate entropy of acoustic emission signal are analyzed as validity parameters to characterize the occurrence and intensity of rubbing. The acoustic emission signals of different rubbing states were collected on the rotor test rig. The extracted Hurst exponents and approximate entropies were separately and combined as the input of BP neural network to identify different rubbing states. The experimental results show that the combination of Hurst exponent and approximate entropy has high recognition accuracy, which provides a new method for the rubbing fault identification of rotating machinery.