关键词: 超声检测, 裂纹深度, 智能识别, 定量方法

Abstract:

In order to quantitatively recognize crack dep th and effectively imp rove the accuracy of ultrasonic tes-ting, the wavelet analysis and artificial neural network (ANN) are introduced to the intelligent recognition of crack dep th. The p roposed recognition method is then discussed in detail in terms of the basic p rincip le of ultrasonic tes-ting, the determination of defect dep th characteristics, the abstraction of dep th characteristics bywavelet, the struc-tural parameters of ANN as well as the training and testing networks. Moreover, the feasibility of the intelligent quantitative recognition of crack dep th by ANN is discussed, and an intelligent experimental system is finally set up, by which a flaw samp le is quantitatively recognized and analyzed. The results indicate that the introduction of the wavelet analysis and ANN p rovides a feasible app roach to the quantitative recognition of cracks in ultrasonic tes-ting feasible.

Key words: ultrasonic testing, crack depth, intelligent recognition, quantitative method