水轮机空化声发射信号的优化VMD特征提取

Feature Extraction for Cavitation Acoustic Emission Signals of Hydraulic Turbines Based on Optimized VMD

针对变分模态分解(VMD)算法中分解层数和惩罚因子2个参数对分解结果有着显著影响且不易确定的问题,提出了灰狼和布谷鸟混合优化VMD算法(简称优化VMD算法)。该方法以包络熵差异互相关系数作为适应度函数,以全局最小适应度值为优化目标,筛选出最佳的VMD参数组合。将优化VMD算法用于水轮机空化声发射信号的特征分析,通过分解得到本征模态函数(IMF),建立了IMF能量随空化系数的变化关系,反映了水轮机空化的发展状态。结果表明:随着空化系数的减小,各主要IMF能量增大,反映了水轮机空化从无到有,从弱到强的变化过程,验证了优化VMD算法用于水轮机空化分析的正确性。

To solve the problem that the number of decomposition modes and the penalty factor in VMD algorithm have significant influence on the decomposition result, and the above two parameters are difficult to be determined, a hybrid search optimization VMD method by grey wolf optimization(GWO) and cuckoo search(CS) algorithm was proposed, which takes the envelope entropy difference cross-correlation number as the fitness function, the minimum global fitness value as the optimization goal, to select the best combination of VMD parameters. Using the optimized VMD in characteristic analysis of acoustic emission signals under hydraulic turbine cavitation conditions, the relationship between the energy of intrinsic mode function(IMF) and the cavitation coefficient was obtained to reflect the development of turbine cavitation. Results show that: with the decrease of cavitation coefficient, the energy of major IMF increases, exhibiting the development process of cavitation in hydraulic turbines growing out of nothing and from weak to strong, proving the optimized VMD method to be effective in cavitation analysis of hydraulic turbines.