基于模糊熵的转子碰摩声发射信号的识别

Recognition of Rub-impact Acoustic Emission Signal Based on Fuzzy Entropy

利用模糊熵理论来度量转子碰摩声发射信号的特征参数相对于不同碰摩状态识别模式的不确定度。根据碰摩声发射信号的特点,选用平均信号电平、幅度、幅度动态范围以及小波包分解信号前四个节点重构信号的能量值作为声发射信号识别的特征参数,由训练样本确定各特征参数针对不同碰摩类别的基于高斯形式的隶属度函数,并由隶属函数得到特征参数与类别之间的模糊关系矩阵。由于各特征参数对于声发射信号识别的有效性不同,因此在计算模糊关系矩阵时引入有效度系数,提出一种利用模糊熵定义有效度系数的方法。结合该系数得到修正的模糊关系矩阵并计算综合评价模糊集合,选择隶属度最大的类别作为识别结果。在转子试验台上采集的不同碰摩状态的声发射信号进行验证,试验结果表明,模糊综合评价方法是一种有效的声发射识别手段,并可以利用参数有效性的差异来提高识别效率。

Fuzzy entropy is used to determine the uncertainty of characteristic parameters of rub-impact acoustic emission （AE） signal in the different rub-impact modes. Average signal level, amplitude, dynamic amplitude range and energies of four nodes reconstruction signals decomposition by wavelet packets are chosen as the characteristic parameters of recognition of AE signal. The membership function of each characteristic parameter belonged to different rub-impact mode based on Gaussian model is obtained from the training samples respectively. Subsequently the fuzzy relation matrix of characteristic parameters and modes is calculated with membership functions. According to different effectiveness in recognizing AE signal by the characteristic parameters, an algorithm based on fuzzy entropy is presented to calculate effectiveness coefficient. The integrated evaluate fuzzy sets is calculated with the new fuzzy relation matrix modified by effectiveness coefficient, and the mode which has the most degree of membership is chosen as the recognition results. The experiments indicate that the integrated fuzzy evaluation is effective in analysis and recognition of AE, and the differences of parameter effectiveness can be used to improve recognition efficiency.