As the most commonly used mechanical transmission components,faults are very common, including gear faults, bearing outer ring, inner ring and rolling element faults. These faults are caused by periodic impact. However, when the fault occurs, due to the complex environment, the vibration signal collected by the sensor often contains a lot of background noise, and these pulse signals are often submerged by noise. Therefore, reasonable and effective noise reduction methods play an important role in the accuracy of fault diagnosis.
Because Med can detect the pulse component in the fault signal, it is widely used in the fault diagnosis of rotating machinery. Considering that Med can only extract a single pulse, a new method of maximum correlation kurtosis deconvolution (MCKD) is proposed. MCKD has good deconvolution performance and can extract multiple pulse signals, but the noise reduction accuracy of MCKD is limited by multi parameters and resampling process. In this paper, an improved maximum correlation kurtosis deconvolution (imckd) method is proposed, and its effectiveness is verified by simulation and experimental analysis. In the early stage of rotating machinery fault, the kurtosis index increases obviously, but with the emergence of periodic pulse, the kurtosis value will gradually decrease. Therefore, the kurtosis value of a single pulse is larger, when a series of pulse signals appear, the kurtosis value will become smaller. In conclusion, if we want to get better results, we need to find a parameter that can reflect more periodic pulses to replace the objective function of MED, so as to improve the Med.
ZHY gear proposed a new kurtosis spectral entropy (KSE) index. By calculating the kurtosis spectral entropy of the intrinsic mode function of each layer, taking its maximum value as the objective function, the improved PSO is used to optimize, and the noise reduction effect of MED is further improved, which is called mksed. The experimental results show that mksed based on PSO has better noise reduction effect than Med and MCKD, and it is applied to gearbox fault diagnosis.
Because minimum entropy deconvolution can only highlight a single effect, it has great limitations in fault diagnosis. Therefore, a maximum kurtosis spectral entropy deconvolution method based on improved local particle swarm optimization is proposed
(1) kurtosis spectral entropy is constructed as an index to measure the influence of continuous pulse. Kurtosis spectral entropy is the ratio of kurtosis and envelope spectral entropy. This definition not only keeps the original characteristics of MED, but also improves the uniformity of periodic pulse.
(2) kurtosis spectral entropy is constructed as a performance index to improve the limitation of MED method, which can reflect more periodic pulses and provide better information extraction ability for fault diagnosis.
(3) by calculating the kurtosis spectral entropy of the intrinsic mode function of each layer, taking its maximum value as the objective function, the improved particle swarm optimization algorithm is used for optimization, which effectively overcomes the influence of filter length on mksed and further improves the noise reduction effect of MED. The experimental results show that the proposed method is effective and superior in fault diagnosis.