Font Size: a A A

The Study, Based On Ultrasound Detection Of Adhesive Properties

Posted on:2008-12-19Degree:MasterType:Thesis
Country:ChinaCandidate:F XueFull Text:PDF
GTID:2191360212478681Subject:Weapons systems, and application engineering
Abstract/Summary:PDF Full Text Request
Adhesive structure and nondestructive testing technologies are widely used in aerospace, nuclear, chemistry, petroleum industry and other national defence and civil fields. The ultrasonic testing for the defects of complete disbond in a multi-layered structure with lower acoustic impedance beneath a high acoustic impedance overburden is one of difficult problems in the ultrasonic nondestructive testing field.The ultrasonic testing for debonded defects in a multi-layered steel-rubber adhesive composite structure is discussed in this paper. As the acoustic impedance of the steel layer differs far from the laminated rubber, the signals reflected by debonded rubber layers are very weak, and they are masked in the strong echoes from the first interface of steel-rubber. The echoes caused by the delaminated rubber can't be identified by the traditional pulse-echo method. It's necessary to solve the problem by advanced signal processing methods. The testing objects and models are introduced at first. The advantages and shortcomings of some existing signal processing methods used to solve such problems are described. Two kinds of methods are discussed in detail in this paper.One method is based on extracting echoes by adaptive signal processing methods. Adaptive filtering method is discussed in the paper, the first, second and third interface echoes from the debonded defects in steel-rubber are extracted successfully by the method.The other method is to solve the problem by pattern recognition. Eigenvectors are extracted by wavelet in certain time windows, in this paper, BP and RBF neural network are trained with eigenvectors, tested in practical project and results of the experiment are satisfied, also the error rate of testing results about two neural networks, rate of convergence , combination property of two networks are compared.The methods discussed in this paper are based on the practical application. They have instructional values for testing the defects of complete disbond in a multi-layered composite structure.
Keywords/Search Tags:adhesion, ultrasonic, adaptive filtering, wavelet neural network
PDF Full Text Request
Related items