The Mechanism Of Detecting Defects In Timber Using Ultrasonic Test Based On Wavelet Neural Networks

Wood inner-defect detecting technology is a newly arisen and many subjects intersectional technology which has been successfully applied in wood production and many other fields. This paper is around wood inner-defect detecting technology, ultrasonic non-destructive testing for wood defects is studied using the energy spectrum variety of the ultrasonic signal by wavelet transform, coefficient of wavelet node and the Artificial Neural Network (ANN).First, wood specimens of standard and different types are made. Wood specimens are nondestructive tested by advanced ultrasonic test system, and copy relevant originality ultrasonic signal; db1, db5, db10, coif5 wavelet radix are analyzed in ultrasonic test. Select db5 wavelet radix as principal wavelet radix in disposing ultrasonic signal; wavelet signal different test which is prevail apply in mechanical failure; can't apply in wood ultrasonic test.Secondly, originality signal of different pieces are dispelled by wavelet bundle, and receive crunode's energy of No.5 layer signal wavelet bundle, so get energy varieties of defect pieces and fine pieces; energy changers of defect wood pieces mostly are decided by degree of wood defect; in analyzing crunode's energy varieties of No.5 layer signal wavelet bundle, find (5,0) of 32 crunodes is the biggest. This shows defect character information is the most, so taking (5, 0) crunodes as the inputs of the artificial neural network.Finally, taking 32 crunode's energy varieties of No.5 layer and (5,0) crunode's wavelet radix as the character inputs of the artificial neural network. Tow network analyze the ability of identifying wood defect, through train network by calculated stylebook. The results show that neural network check result, which taking (5,0) crunode's wavelet radix as the character inputs, is more efficiency. The precision has attained 99% above.