Abstract:Aimed at the problem of routine operations in ship pipeline system of valve and pump adjustment of false alarms of leak detection system were very easy led, by made use of Support Vector Machine (SVM) suitable to process with wavelet data collection, leakage recognition was carried out for the pipeline system in various working conditions. The SVM parameters were optimized by using the Artificial Bee Colony (ABC) with searching power of local and complete area optimal solution, so as to avoid local area optimal solution and improve accuracy rate in recognition. By wavelet analysis to extract eigenvectors in working condition, the classifier of ABCSVM was designed and built up to implement leakage recognition. As compared with the commonly used algorithm of BP neural network, the ABCSVM algorithm has a better accuracy and adaptability in leakage recognition.