In wood damage in early detection and the actual environment wood defect, weak damage acoustic emission signals are often submerged in the background noise, to purified damage acoustic emission signal from the background noise characteristics, for carrying wood early damage detection and quality evaluation is of great significance. Material of acoustic emission signal processing has been research hot spots, including wood/non_wood materials research, acoustic emission signal. Study found that the material in the process of the acoustic emission signal sampling, the measured acoustic emission signal is often a variety of signal superposition, and contains various constituents, belong to the non_stationary signal, so the data analysis, the first thing to consider noise removal problem. The traditional filtering method based on Fourier transform in improving signal_to_noise ratio and spatial resolution contradiction on the two indicators. Low_pass filter can restrain noise through a smooth, but due to the distortion process of signal distortion; High_pass filter can make the edges more steep, but the background noise also be strengthened at the same time, some high frequency components of which is likely to signal and noise filter instrument, lost signal of high frequency information.
In this paper, the de_noising algorithm based on wavelet packet transform to purified wood damage acoustic emission signal, can not only remove noise effectively, and can retain the signal mutation parts, reducing the loss of signal of high frequency information. After compared with the original sampling signal, the denoising signal damage characteristics of enhanced obviously, the results show that the method has good ability of de-noising.
A wavelet packet signal de-noising processing
1.1 principle of wavelet packet de-noising
Threshold denoising principle is: the signal wavelet decomposition, if the noise energy significantly less than the signal energy, and the corresponding wavelet coefficients and noise will be significantly less than the corresponding wavelet coefficients and signal, choose a suitable threshold processing of wavelet coefficient, set below the threshold of wavelet coefficients to zero, higher than the threshold of wavelet coefficients shall be maintained or contract. The denoising steps as follows:
1, signal wavelet packet decomposition, choose a wavelet and determine a wavelet decomposition level N, and then the signal s N layer wavelet packet decomposition.
2, calculate the best tree (i.e., the optimum wavelet base). For a given standard of entropy, calculate the best tree.
3, threshold of wavelet packet decomposition coefficient quantification. For each wavelet packet decomposition coefficient (especially low frequency decomposition coefficient), choose a proper threshold value and the coefficient of threshold quantization.
4, wavelet packet reconstruction. According to the N layer of coefficient of wavelet packet decomposition coefficient and through quantitative processing, wavelet packet reconstruction.
1.2 wavelet packet threshold denoising algorithm is proposed
To estimate the wavelet coefficient must be to select threshold functions and threshold, the commonly used threshold function with hard threshold function, soft thresholding function and some improved threshold function such as compromise between hard and soft threshold function, semi-soft threshold function, etc.
Hard threshold is as follows:
Soft threshold is as follows:
Because of the hard threshold of discontinuity at some point, there was a certain oscillation in signal reconstruction. Soft threshold good continuity, but poor, and there is a constant value, make the reconstructed signal of low signal noise ratio (SNR), mean square error is bigger. How to choose the threshold value in the threshold function THR, directly related to the pros and cons of denoising. Is too small, threshold value will keep some after denoising signal noise information; If the threshold is selected very big, will be lost the useful signal, therefore in the process of denoising threshold selection is the key.
1.3 in the wavelet packet de-noising threshold selection
1, forced de_noising processing. This method all the high frequency coefficients of wavelet decomposition structure becomes zero, namely the high frequency part all filter to remove, and then to reconstruct the signal processing. This method is simpler, reconstruction after denoising signal is more smooth, but easy to lose the useful signal components.
2, the default threshold de-noising processing. The method USES ddencmp function signal the default threshold, then using wdencmp function de-noising processing.
3, given the soft or hard threshold de-noising processing. The noise in the actual process, the threshold is often can be gained through empirical formula, and the threshold is more credible than the default threshold. Wthresh functions available in the threshold quantization processing according to wood acoustic emission signal characteristic, this research USES the default threshold de-noising processing, namely the default threshold by ddencmp function signal generation, and then the de-noising processing.
2 wood acoustic emission signal de-noising processing
2.1 acoustic emission signal test
This experiment chooses from Inner Mongolia daxing AnLin larch, 32 years old, about 32 cm diameter at breast height, made size is 240 mm x 20 mm x 20 mm specimen 30 of them, the length of the direction for the texture direction, the moisture content of 12%.
In RG - 3 type electronic universal testing machine to carry on the three point bending test, loading until timber specimen, the loading rate of 2 mm/min. Experiment, the sand on the specimen polished surface is 80 mm from center place, until the light smooth, then use absorbent cotton to wipe clean. To ensure sufficient coupling, evenly on the acoustic emission sensor butter, burnish of sensor coupling in the specimen, and magnetic compression, and then use Beijing PengXiang technology company's digital acoustic emission instrument to collect the whole process of acoustic emission signal, sampling frequency for each channel 500 kHz, sampling time is 0.02 s.
2.2 in the Matlab wavelet packet de-noising
This study based on the principle of wavelet packet de-noising, using Matlab software compiled the acoustic emission signal de-noising, the experiment collected wood acoustic emission signal de-noising processing.
Figure 1 is made by wood acoustic emission testing, acoustic emission of the original signal were collected from figure 1 b shows that AE signal in the process of wood bending frequency range is roughly around 40-200 kHz, and contain random noise.
igure 2 is obtained by wavelet de-noising processing wood acoustic emission signal, the results show that the wavelet analysis can well retain useful feature information in signal.
Figure 3 is based on wavelet packet analysis method, the use of Matlab functions in the filtering algorithm for signal wavelet packet de_noising effect after processing, compared with the wavelet analysis can clearly, acoustic emission signal of wood treated retain all original signal features, well preserved the useful signal of mutations and rush, but irrelevant information and noise has been greatly reduced, and the center frequency in 150_200 kHz. Shows that using wavelet packet analysis method for the wood of the acoustic emission signal can achieve very good denoising effect.
3 conclusion
For or in the actual environment in the early wood damage detection of wood defect, a faint wood acoustic emission signal contains lots of background noise, the noise reduction purified original acoustic emission signal is the precondition of wood damage defect detection. This paper applied the wavelet function in MATLAB toolbox of larch wood produced by the three point bending wavelet packet de_noising treatment was conducted on the acoustic emission signal, noise reduction and comparison of the wavelet packet analysis and wavelet analysis of the effect of noise reduction is different.
Through the analysis of experimental data, using wavelet packet in the wood the default threshold de_noising treatment method for acoustic emission signal de_noising processing, noise elimination is complete, and is a good way to save the useful signal of spike and mutation, to improve the signal-to-noise ratio. Shows that using wavelet packet analysis can further improve the acoustic emission signal processing ability, but also for wood damage defects diagnosis has very important practical significance.
Is adopted in this paper the author's innovative: wood acoustic emission detection based on wavelet packet analysis method, for the early wood defect and wood structure damage real_time detection provides effective and reliable method. The project economic benefits (2 million yuan).