Research On Theories And Applications Of Ensemble Empirical Mode Decomposition

Hilbert-Huang transform(HHT)is a self-adaptive time-frequency analysis for non-linear,unstable signals. It contains empirical mode decomposition(EMD)and its correspondingHilbert spectrum. The key of HHT is the empirical mode decomposition, which breaks amulti-frequency signal down into a series of single frequency component signals essentially.There are few problems, in which the key one is mode mixing. For solving the mode mixingproblem, ZHAOHUA WU proposed a new approach called ensemble empirical modedecomposition(EEMD), which was very suitable for applying to the signal analysiscontaining impact components and fault diagnosis. EEMD was proposed without enoughtime, so there were few problems in theory and calculation method. EEMD is mainlyapplied in aerospace and other high-tech industries, and few applied in marine. Based onthis background, the EEMD theory and its application in the fault diagnosis of ship powerequipment are discussed in this paper.Based on HHT and EEMD, the two key parameters and self-adaptive necessaryconditions of EEMD were studied in this paper. Then, starting with the adding-modes wayof white noise, the paper proposes an improved EEMD method, which can improve theaccuracy of signal decomposition. By introducing the iterative EMD interval-thresholding(EMD-IIT) denoising method, its principles were discussed detailedly. The simulationresults show that denoising effect of EMD-IIT for steady state signals is good, but it is badfor signals containing impact components. Based on EMD-IIT and improved EEMD, thispaper puts forward a new adaptive EEMD method. Its decomposition effect is good forsteady signals and the method reduces the residual white noise. Considering conditions ofadded white noise and the probability statistical characteristics of white noise, the paperproposes an improved EEMD method, which can restrain mode mixing of signal andimprove the accuracy of signal decomposition, and is more suitable for application in shippower equipment fault diagnosis.By performing an experiment for simulating gear pitting fault and bearing inner ringfault, the experimental datum were measured and acquired. By using self-adaptive EEMDbased on probability and statistic on gear and bearing fault signal analysis, faultcharacteristic frequencies are not obvious. Proceed from the view of energy, doing furtheranalysis of IMF, this paper proposes a new fault diagnosis method based on EEMD andinstantaneous energy density spectrum. That the gear and bearing fault signals of thesimulation analysis show failure frequencies of the gear and bearing can be got by this method. Compared with other fault diagnosis methods, it takes an advantage oncomputational cost. Then, the proposed method is applied in fault diagnosis for misfire andabnormal clearance in a diesel engine, getting clear fault characteristics. The results showedthat it is very sensitive to the misfire fault and able to distinguish whether the valve trainclearance is too loose or too tight. The tight clearance has some difficulties to bedistinguished from the normal condition.