基于EMD分解在电火花数据分析的应用

Application of EMD-based decomposition in electric spark data analysis

经验模态分解(empirical mode dccomposition,EMD)是处理非平稳、非线性信号一种有效的新方法。运用EMD分解法将脉冲典型负荷—电火花信号中本征模态函数(intrinsic mode function,IMF)分量逐级分离出来,得到原信号的多尺度振荡特性;分析各个分量与原序列的显著性检验和相关系数以及各个分量自身的周期性。用IMF分量和趋势项合成原序列并作误差分析,得到信号的总误差率通过理论计算为4.3%。利用bior4.4和db2两种小波基在3层分解系数基础上对电火花数据展开小波包变换。最后借助MATLAB平台对EMD和小波包分解方法进行仿真并做理论对比分析。结果表明,EMD分解法在提取机床运行放电特性及不同工况下信号更具优越性和有效性,提取结果更能满足实际工程应用需求。

Empirical mode decomposition is an effective new method in the treatment of non-stationary and nonlinear signal. EMD method is used to separate typical pulse energy step by step,which is the inherent Intrinsic Mode Function in EMD signal and to obtain multiple scales oscillation characteristics of original signals. The author analyzes on the significance tests of each component and the original sequence and the correlation coefficient and respective periodicity of each component. With IMF and the synthesis of original sequence with trend term component,error analysis is made. By theoretical calculation,the total error rate between the signal and the original signal is 4. 3%. Based on the three decomposition coefficients,two kinds of Wavelet basis i. e. bior4. 4and bd2 are used to make Wavelet Packet Transform on electric spark data and make actual simulative comparative analysis with the help of Matlab platform. The experimental results show that EMD possesses more superiority and effectiveness in the extraction of machine tool operation＇s discharge characteristics and extraction of different signals under various working conditions. Furthermore,the extraction results can better meet the needs of practical application.