An Analysis Method For Rotating Machinery Fault Diagnosis Based On Mathematical Morphology Fractal Dimension

With economic globalization and the rapid development of science and technology, the industry is put forward higher requirements for the stablity and efficient operation of mechanical equipment, so there are more and more attentions paid to equipment fault diagnosis technology. Rolling bearing one of the most commonly used parts in the mechanical equipment, but it is especially easy to be damaged, which is very adverse for the life of the whole equipment system and normal production. So it is very necessary for the fault diagnosis and state detection of the mechanical equipment bearing.Mechanical fault diagnosis is actually signal processing. Modern equipment is more and more large and complicated, and often show some nonlinear behavior, there are some limitations on the analysis of the nonlinear characteristics for the traditional signal analysis methods. Nonlinear processing algorithm can solve this problem. Mathematical morphology is a relatively new discipline, It mainly use a probe named "structural element" to explore the signal details and morphological characteristics, which is a very efficient fault diagnosis method by nonlinear thought to portrary and depict signal. What is different from the time-frequency analysis method is mathematical morphology do not need analyse in in time domain and frequency domain of signal, but only need to calculate the fractal dimension of signal, simple and intuitive.Firstly, the structural element have effects on the accuracy of the morphological is investigated. There are a comparison among cosine type, semi-circular and triangle structure element in this paper, the comparison shows selection characteristics of structural elements and its influence on the algorithm. To realize the algorithm by select the appropriate structural element, and distinguish the morphological spectrum entropy of signal by probabilistic neural network(PNN), which achieve a good results.And then, because of sensitivity to mechanical noise, it is necessary to choose local characteristic decomposition(LCD) scale to remove noise before using morphological fractal. Empirical mode decomposition(LCD) can appear the mode mixing phenomenon serious end effect, local mean decomposition(LMD) has a large amount of calculation, but LCD has its own advantages. This paper apply the method combining LCD and mathematical morphology to diagnose machinery faults, and can obtain more accurate results than the box dimension.Finally, the fractal interval is applied to distinguish the mechanical fault state based on the characteristics of single fractal state. This part, through studying on volatility of fractal dimension, and considering the overlapping phenomenon in the line state, abandoning the value of larger and smaller values, introducing the harm of the state overlap phenomenon in detail, apply interval estimation to divide the state of signal. The method ’s validity is verified according to the measured sample, which can be very fast to judge the fault state.