| The stability and safety of the aerospace equipment is threatened by the existing interfacial bonding defect of thin composite materials in existence. It's a burning question that discovering and recognizing the harmfulness of bonding defect. In this paper, aiming at the problems of bonding defect recognition of thin composite materials, the pattern recognition system is established with the fuzzy neural network, then recognized the echo signal derived from ultrasonic detection for bonding defect of thin composite materials.First, multidimensional eigenvectors of quantitatively describing the bonding defect were established by analysising the characteristic information of echo signal, to realize the accuracy and quantification recognition of the bonding defect. Then aiming at the recognition of bonding defect, the fuzzy neural network pattern recognition system was established via combination of intelligible expression "if-then" of the fuzzy system and learning capability and adaptability of the neural network to fuse the fuzzy system and neural network. In the realization of system, the membership grade about eigenvector of bonding defect to debonding degree will be obtained by BP neural network, and then the applicability of fuzzy rules is obtained, hence the debonding degree of testing sample will be calculated. According to the accuracy of recognition for every eigenvector, select two eigenvector to build the combined applicability of fuzzy rules, and then recognition result will be obtained. The results of the experiment showed that the echo signal derived from ultrasonic detection for bonding defect of thin composite materials can be recognized in good result by the pattern recognition system of fuzzy neural network. It makes a better foundation for realization the quantification recognition and quantification detection, and provides advantaged for online detecting about ultrasonic detection for bonding defect of thin composite materials.
|
PDF Full Text Request