Study on Recognition Characteristics of Acoustic Emission Based on Fractal Dimension

It is difficult to recognise acustic emission (AE) signal because of serious pollution by noise. Fractal dimension is a new method to describe the characteristics of AE singal. According to the complex computation and low precision of box counting dimension, correlation dimension and Katz  dimension, an algorithm of logarithmic fractal dimension based on waveform length was proposed in this paper the deduction process was introduced. The experiment datas was rub impact AE signals sampled from rotating test stand. Gaussian white noise and non-stationary noise were added to simulate the field AE signal which polluted seriously by noise. Then, three algorithms based on box counting dimension, Katz dimension and logarithmic fractal dimension were compared in AE signal recognition. The results show that the algorithm of logarithmic dimension distinguishes rub impact AE signals from strong noise is more effectively, and has lower computation and higher precision than others. It provides a new approach to identify characteristics of  AE signal and detect rub impact fault of rotating machinery.