A Study On The Wood Hole-defect Recognition Model Based On Fuzzy Pattern Recognition

Nowadays, the demand for wood is increasing. However, the wood resources is limited, so more attention is paid on wood nondestructive testing that can improve the utilization rate of wood. Moreover, it is also necessary to improve the precision of wood nondestructive testing in order to protect wooden ancient building, high value trees, and wooden bridges.In recent years, there are decades of testing methods used in wood testing area, including ultrasonic wave, stress wave, nuclear magnetic resonance, X-ray, impulse resistance method, microwave etc. Acoustic radiation method namely hammering method is widely used in nondestructive testing field like watermelon internal defect including deep seam, decay and collision, as well as egg shell crack detection, but rarely used in the field of wood nondestructive testing. Therefore, acoustic radiation method is applied in nondestructive testing of wood hole defects in this paper, which provides a new method and train of thought for wood nondestructive testing field.The fuzzy pattern recognition technique already has been widely used in nondestructive testing such as pipe defect detection, plate and shell structure damage detection and space truss structure damage identification and so on. It is an inexorable trend to introduce the fuzzy pattern recognition technique into wood detecting field.Statistical analysis software of SPSS and MATLAB were applied in this paper to make time domain analysis on test data. The result shows that:the four waveform characteristics of average full-wave width, the large amplitude, coefficient of kurtosis and average peak value conform to the distinguishability of good characteristics. Average waveform full-wave width and coefficient of kurtosis meet the demand of independence for excellent characteristics while the large amplitude and average peak value don’t. Average waveform full-wave width and kurtosis coefficient meet the demand of reliability for excellent characteristics while the large amplitude and average peak value don’t. Therefore, the optimal characteristic indicators that are extracted in this paper are resonant frequency, average waveform full-wave width and kurtosis coefficient. In addition, the large amplitude and the average peak value are also employed because of meeting the demand of differentiation between groups. It is also proved that3-characteristic model has better recognition function than5-characteristic model by experiments.In order to identify the internal hole defects in the wood, a new method based on fuzzy pattern recognition and acoustic non-destructive testing of wood was proposed in this paper. The experiment was carried out on timber samples with holes defects at different locations and in different diameters. Acoustic signals were collected with hammering method, and time- frequency feature vectors were extracted as the sample data. Cluster analysis was made on the training samples using fuzzy similar matrix based on the transitive closure, after which different classes of fuzzy patterns were created. The test samples were then identified by "maximum membership degree" principle. Thereby the fuzzy pattern recognition model for timber hole defects is established. It can be concluded by experiment analysis that:(1) The fuzzy pattern recognition model for Acer mono wood hole defect built in this paper is simple and smart, which is suitable for the recognition of Acer mono wood defect location and size, and the accuracy is also ideal. The resonance frequency of samples without holes was the maximum, followed by that of samples with the end holes, and that of samples with middle holes was the minimum. The resonance frequency in the same order of samples with hole defects decreases with the increasing bore diameter except that of samples with a hole dimensioning18mm in diameter, and the specific reason needs further research and investigation.(2) Whether the drilling direction and wood texture direction is tangent,45°or vertical has no effect on the resonance frequency of timber sample and fuzzy pattern recognition model for the timber hole defects. Although different tap strength can inspire different order resonant frequencies, the tap strength, under the condition of inspiring the same order resonance frequency, has greater effects on5-characteristic model, and almost no effects on3-characteristic model.