Fault Stage Feature Extraction Method Based on Antinoise Gradient Operator Morphological Filters

Abstract

Abstract The cyclical impact signal aroused by mechanical fault contains feature information, such as fault type and fault stage. However, a transient impact component is usually modulated on the fundamental frequency and mixed with a large amount of noise. Thus, it is difficult to be extracted effectively. A feature extraction method of faults stage is presented based on morphological filter theory. An antinoise morphological gradient operator is applied to signal filtering. This operator can weaken the noise and clearly highlight the impact characteristics of different fault stages. The kurtosis of the spectrum band, including the passing frequencies of mechanical components, is used to differentiate fault stages. The applications on simulation signals and bearing test data indicate that this method can realize accurate extraction of fault characteristic frequency submerged in noise and effective distinction of different fault stages.