Ultrasonic Flaw Classification in Seafloor Petroleum Transferring Pipeline through Chaotic Optimization and Support Vector Machine

摘要

Aiming at seafloor pipeline flaw identification from ultrasonic signals, a new flaw identification method based on chaotic optimization and support vector machine (SVM) is studied in this paper. The ultrasonic inspecting signals of two kinds of artificial flaws are acquired by experiments. Their features were extracted using wavelet packet decomposition (WPD) and incorporated into the feature set. The complete feature set is encoded in a chromosome and then optimized by chaotic optimization with respect to identification accuracy and number of discarded features. The experimental results show that the feature selection based on chaotic optimization improves the performance of the SVM classifier greatly because it can select the features having high separating power and discard redundant or irrelevant features.