Design and Implementation of a Time-frequency Analysis System for Non-stationary Vibration Signals Using Mixed Programming

Abstract

At present, in the feature extraction of non-stationary vibration signals, many timefrequency analytic methods have emerged to meet the further need of non-stationary signal analysis such as Wavelet Transform (WT), Short Time Fourier Transform (STFT), WignerVille Distribution (WVD), Empirical Mode Decomposition (EMD), Ensemble Empirical Mode Decomposition (EEMD) and so on. However, these time-frequency analysis methods can only be carried out on Matlab platform and the processes are of low efficiency, bringing a lot of inconvenience in signal analysis because they can not run on Windows system independently. According to the defects, a novel method of mixed programming which combines Matlab with Delphi based on COM (component Object Model) module technology is proposed to design a time-frequency analysis system for non-stationary vibration signals assembling fourteen methods. The advantages of two languages are shared with each other. In design of the system, a combined method which combines two or three analytic methods is put forward for relational analytic modules to make the analytic results more perfect and clearer, like WAVELET-FFT, EMD-WVD etc.. This paper uses the EMD-WVD combined method as an example to present the programming thoughts and processes. The effectiveness of the system is validated using a simulation signal and an experimental signal. The system can be executed separated from MATLAB environment.