CN102855408A - ICA (independent component analysis)-based EMD (empirical mode decomposition) improvement process IMF (intrinsic mode function) judgment method - Google Patents

Abstract

The invention relates to a novel ICA (independent component analysis)-based EMD (empirical mode decomposition) improvement process IMF (intrinsic mode function) judgment method. Aiming at the problem of excessive IMF components generated in the frequency band filtering EMD improvement process, the method introduces ICA into the EMD improvement process to automatically separate real IMF components. The method includes: improving EMD to decompose a structural response signal to obtain IMFs of each frequency band; and respectively taking the IMFs as an input matrix, and using a FastICA algorithm in the ICA for separation to automatically separate the real IMF out. The method can well process multi-degree-of-freedom, non-linear and unsteady-state response signals, can be combined with other methods (such as Hilbert transform) for modal parameter recognition, can be used for signal processing and modal parameter recognition in the fields of civil engineering, aerospace, automatic control, mechanical engineering and the like, and has the advantages of increase of signal to noise ratio of data, anti-jamming capability and the like.

Description

Based on IMF decision method in the improvement EMD process of ICA

Technical field

The present invention relates to the Time-Frequency Analysis technical field of structural vibration response, particularly a kind of new for IMF(intrinsic mode function in the improvement EMD process of ICA) decision method.

Background technology

The EMD method be widely applied in each field, but there has been the mode aliasing in the EMD method since formal proposition in 1998, and this has limited its application in practice greatly.The appearance of mode aliasing is relevant with the algorithm of EMD itself on the one hand, also is subjected on the other hand the impact of original signal frequecy characteristic.The method that Huang had once proposed to interrupt detecting solves the mode aliasing, namely directly the result is observed, and then again decomposes if there is the mode aliasing, and this method needs artificial posteriority to judge.2009, the research group of Huang is by decomposing the as a result large quantity research of statistical property of white noise to EMD, population mean empirical mode decomposition (Ensemble Empirical Mode Decomposition by the plus noise assistant analysis is proposed, EEMD) method, the method utilizes white Gaussian noise to have the equally distributed statistical property of frequency, make the signal that adds behind the white Gaussian noise have continuity at different scale, thereby effectively solve the mode mixing problem.Other scholars then propose the method for various improvement EMD, that wherein gives prominence to the most improves one's methods as at first utilizing the approximate range of FFT guestimate frequency, then allow the bandpass filter of signal by assigned frequency band, carry out again classical EMD process, to solve the mode aliasing, but the defective of the method is to produce too much, false IMFs, there is the scholar to propose to utilize the related coefficient of IMFs and original signal to judge real IMF, but this kind method needs the in advance threshold values of artificial assigned I MF judgement, can not realize the automatic identification of IMF.

In recent years, blind source was separated in every field and was widely used, and its basic thought is independently to pass a plurality of observation signals in principle optimized algorithm according to statistics to be decomposed into several independent components, thereby realizes enhancing and the analysis of signal.ICA(Independent Component Analysis) is relatively ripe a kind of method of blind signal processing, carries out the separation of Independent sources signal mainly for aliasing signal.Consider that any observation signal is decomposed into some IMF(Intrinsic Mode Function through EMD) and discrepance after, in theory, each IMF is the single mode of oscillation of any time, and each constantly has single instantaneous frequency, and is separate between each IMF component.Therefore, FastICA algorithm among the definite and ICA of improvement EMD decomposition and IMF component can be combined, the differentiation of real IMF just converts solution procedure to the isolated component of natural mode of vibration mixed signal among the EMD like this, and the component of ICA gained is exactly real IMFs.

The FastICA algorithm is blind signal processing a kind of algorithm commonly used, and it can eliminate mutual information and information redundancy between each input quantity, isolates the separate composition in inside of hiding between the information, automatically identifies real IMF component thereby eliminate false mode.

Summary of the invention

The object of the present invention is to provide a kind of based on IMF decision method in the improvement EMD process of ICA, the method is conducive to solve band filter and improves the decision problem that produces too much false mode and real IMF in the EMD process, thereby realizes the automatic separation of real IMF.

The objective of the invention is to be achieved through the following technical solutions: a kind of based on IMF decision method in the improvement EMD process of ICA, utilize the FastICA algorithm among the ICA to eliminate the too much false mode that produces in the band filter improvement EMD process, automatically isolate real IMF component, the method may further comprise the steps:

Step 1: at first response signal is improved the EMD process, namely utilize the frequency range of FFT guestimate signal, and make signal by the bandpass filtering of different frequency range, broadband signal is decomposed into some narrow band signals; Then utilize EMD respectively to each narrow band signal x( t) decompose, obtain the IMFs input matrix of different frequency range c( t);

Step 2: utilize FastICA algorithm among the ICA respectively to the IMFs input matrix of each frequency range c( t) separate, obtain output matrix S, i.e. real IMF component.

The invention has the beneficial effects as follows that utilizing ICA to solve band filter improves the decision problem that produces too much false mode and real IMF in the EMD process, thereby realize the automatic separation of real IMF.The method can be processed multiple degrees of freedom, non-free vibration, non-linear and astable response signal well, and can combine with additive method (such as Hilbert transform) and carry out the Modal Parameter Identification of structure, have the signal to noise ratio (S/N ratio) and the antijamming capability that improve data, strengthen the characteristics of Modal Parameter Identification accuracy.

Description of drawings

Fig. 1 is workflow diagram of the present invention.

Fig. 2 is the acceleration responsive timeamplitude map of the embodiment of the invention one actual measurement.

Fig. 3 is that the acceleration responsive of the embodiment of the invention one actual measurement decomposes 12 IMFs and 1 the remainder figure that obtains through classical EMD.

Fig. 4 is that the embodiment of the invention one centre frequency is that the subband signal of 122.9Hz decomposes the part IMFs figure obtain through EMD.

Fig. 5 is the real IMFs figure that the embodiment of the invention one obtains.

Embodiment

The present invention is based on IMF decision method in the improvement EMD process of ICA, produce the decision problem of too much false mode and real IMF when improving classical EMD mode capacity of decomposition deficiency for band filter, utilize the FastICA algorithm among the ICA to eliminate the too much false mode that produces in the band filter improvement EMD process, automatically isolate real IMF component.As shown in Figure 1, the method may further comprise the steps:

Step 1: at first the structural response signal that records is improved the EMD process, namely utilize the frequency range of FFT guestimate signal, and make signal by the bandpass filtering of different frequency range, broadband signal is decomposed into some narrow band signals; Then utilize EMD respectively to each narrow band signal x( t) decompose, obtain the IMFs input matrix of different frequency range c( t);

Step 2: utilize FastICA algorithm among the ICA respectively to the IMFs input matrix of each frequency range c( t) separate, obtain output matrix S, i.e. real IMF component.

The present invention is further illustrated below in conjunction with specific embodiment.

Concrete, stride for a certain 7 layers, 2

The 1 steel frame scaled model of striding, the acceleration-time curve that actual measurement obtains under the excitation of power hammer, as shown in Figure 2.Before adopting this method to carry out signal analysis, utilize first classical EMD process that this acceleration signal is decomposed, can obtain as shown in Figure 3 12 IMFs and a remainder, from Fig. 3, can find out significantly the aliasing of mode.Then use method provided by the present invention and carry out signal analysis, its implementation is as follows:

At first, in order to solve the problem of above-mentioned mode aliasing, EMD to this response signal application enhancements analyzes, namely use the frequency range of FFT estimated signal, can obtain 7 crest frequencies by the FFT spectrogram, respectively with the centre frequency of these 7 frequencies as each frequency band, allow the bandpass filtering of signal by assigned frequency band.Then, each narrow band signal being carried out EMD decomposes.Subband signal take mid-band frequency as 122.9Hz is example, after decomposing through EMD, this subband signal obtains 10 IMFs and a remainder (part as shown in Figure 4), this explanation exists too many false mode, so use method proposed by the invention, real IMF component is extracted, be about to the larger IMFs of front 7 energy distribution as the input matrix of FastICA, utilize the FastICA algorithm to eliminate mutual information and information redundancy between each input quantity, isolate the separate composition in inside of hiding between the information, thereby remove false mode and automatically identify real IMF component, can obtain IMF1 component as shown in Figure 5 by analysis, by that analogy, in like manner can obtain the IMF component of other each sub-bands as shown in Figure 5.

More than be preferred embodiment of the present invention, all changes of doing according to technical solution of the present invention when the function that produces does not exceed the scope of technical solution of the present invention, all belong to protection scope of the present invention.

Claims (1)

1. one kind based on IMF decision method in the improvement EMD process of ICA, it is characterized in that: utilize the FastICA algorithm among the ICA to eliminate the too much false mode that produces in the band filter improvement EMD process, automatically isolate real IMF component, the method may further comprise the steps:

Step 1: at first response signal is improved the EMD process, namely utilize the frequency range of FFT estimated signal, and make signal by the bandpass filtering of different frequency range, broadband signal is decomposed into some narrow band signals; Then utilize EMD respectively to each narrow band signal x( t) decompose, obtain the IMFs input matrix of different frequency range c( t);

Step 2: utilize FastICA algorithm among the ICA respectively to the IMFs input matrix of each frequency range c( t) separate, obtain output matrix S, i.e. real IMF component.

Images