Research On Damage Detection For Composite Material Based On Fuzzy Neural Network

Because of the excellent performance, composite material was used in various domains, especially in the aeronautics and astronautics. Development of the damage detection became a key technology for the extent of the application of composite material. Determining the degree and location of damage in the composite materials was one of most important aims of structure health monitoring technology. The paper proposed a method that using the lamb wave as the carrier of damage information in the composite materials, extracting the characteristic parameters with time-frequency domain analysis technology, and applying fuzzy neural network to identify the damage of composite materials.Firstly, the design and realization of a PC-based integrated software system for structure health monitoring was studied. Virtual instrument technology and Labview were employed to design the structure health monitoring software system. With the use of multifunction data acquiring panels PCI-1714 and PCI-1721, any actuating signal outputting, high-speed data acquisition and plenty of signal processing were achieved.Secondly, the multiple modes and dispersion nature of ultrasonic Lamb wave were theoretically investigated with the Rayleigh-Lamb equations and its numerical solution to decide the size of the composite laminate plate and the distribution of PZT sensors. The time-frequency transform methods were employed to process the signals acquired from the PZT sensors and the characteristic parameters were extracted as the inputs of fuzzy neural network.Thirdly, the advantage and flaw of fuzzy control and neural network control were analyzed. And T-S fuzzy neural network controllers, which combined the advantages of fuzzy theory and neural network, were developed to determine the location and the degree of damage in the composite material. The algorithms of fuzzy neural network system were systematically studied and realized by Matlab.Finally, the damage detection tests and analysis were studied with the T-S fuzzy neural network controller. After trained by the locale data and validated by simulation, the FNN was employable.