Research On Time-Frequency Analysis And Feature Extraction For Composite Plate Bonding Flaw Ultrasonic Echo

Metal/non-metal bonding composite plate is applied in field of aeronautics, astronautics and national defense because of its special physical and mechanic properties. Bonding quality of composite plate fatally affects quality and durability of products. Detection for bonding flaw of this material is finished by Industrial ultrasonic flaw detector, and results of detection only provide qualitative judgments: bonding or de-bonding, without quantitative evaluations for non-good bonding between bonding and de-bonding, which is one of puzzled problems interest more and more researchers.The research firstly uses ultrasonic testing system to sample stable ultrasonic echo, deeply analyzes two properties of ultrasonic echo:Low-frequency increasing and High-frequency decreasing, and reveals potential relationships between these two properties and de-bonding degree. Secondly, the paper applies wavelet (pack) analysis and Hilbert-Huang transformation to process ultrasonic echo, discusses time-varying and time-frequency distribution, and extracts features which indicate de-bonding degree as well as construct classified (10-classfication) feature vectors. Finally, fuzzy pattern recognition is proposed to verify quantitative recognition effect of extracted feature vectors (3-deminsion), and quantitative recognition for de-bonding of composite plate is implemented at length.The paper is focused on systematic experiments and basic research on quantitative recognition for de-bonding of metal/non-metal composite plate with certain of meaningful results in theory and method; it lays foundation of intellectualization and automation for quantitative recognition for de-bonding of composite plate, and has important theoretical meaning and practical value.