Research On The Signal Processing Method For Balancing Measurement System Based On EMD

With the development of society, the mechanical rotor has been throughout allareas of life and production. The rotor balancing problems have been the focus of theresearchers. The precise detection of unbalancing problems not only can eliminate thesecurity risks but also bring huge economic benefits. And these methods are subject tothe constraints of the insufficient of Fourier transform itself.This paper mainly focuses on the signal processing method for automobileflywheel-based balancing measurement system. The vibration signal which collectedby the sensors usually contains the pulse and harmonic interference noise signal.Traditional time-frequency analysis method does not have a good performance ineliminating the pulse interference. If the frequency distribution of the vibration signalto be processed is very irregular, an aliasing phenomenon may be occurred withFourier transform. Using the Wigner-Ville distribution method will producecross-term, and there will be significant unwanted signals by using waveletdecomposition. In addition, due to the fixed basis function which has to be chosen bythe time-frequency analysis method based on Fourier transform theory, it results in itsrelatively poor adaptability and a heavy computational burden.To solve the problem, this paper applied mathematical morphology filter toextracting the vibration signal as preprocessing and then use the method of empiricalmode decomposition (EMD) to extract the frequency, amplitude and phase of thesignal. Essentially, the EMD method is to make a non-stationary signal smooth. Itspurpose is to decompose the fluctuations or trends present in the signal at different scales under progressively so that the Intrinsic Mode Function (IMF) can be obtained.The instantaneous amplitude and instantaneous frequency are available by the Hilberttransform of each IMF. Each IMF is stationary through the EMD. Therefore the IMFsare able to show the real physical process of the signal. However, when the signalwith a jumping time scale is processed by EMD, the mode mixing problem will occur.To solve this problem, the paper employed a noise-auxiliary analysis method,ensemble Empirical Mode Decomposition (EEMD). The method makes use of theuniform distribution of the white noise. On the one hand, it can eliminate theabnormal disturbance which is generated by hopping transition; on the other hand, theinsertion of a uniformly distributed random scale noise can suppress mode mixingeffectively.For the pulse interference present in the signal often making the empirical modedecomposition occurs errors and distortions, the paper applied the method ofmathematical morphology.Mathematical morphology is a nonlinear signal processingand analysis method, which is often used for image information processing.Mathematical morphology describes a signal with set, by moving a structural elementto detect the target signal’s characteristics. In this way, the target signal is matched soas to achieve the purpose of extracting the signal, maintaining details and noisesuppression. For the pulse interference in the vibration signal, the morphologicalopening and closing combined filter can get rid of it. The simulation results provedthe effectiveness of the method.Based on the two points above, this paper applied mathematical morphologyfilter to extracting the vibration signal as preprocessing. Then for the mode mixingproblem which exists in EMD, EEMD is utilized to extract the frequency and phase ofthe unbalance signal.