Research on Fault Diagnosis of PCCP Based on Support Vector Machine

Recent years, many utilities have experienced catastrophic rupture of critical Prestressing Concrete Cylinder Pipe (PCCP) lines throughout the world. Much attention has been focused on reliably assessing the condition of PCCP mains. However, assessment of embedded prestressing wire is difficult. Continuous acoustic monitoring can provide a means of locating problematic pipe sections. In this paper the application of support vector machine (SVM) in acoustic signal detection is proposed. And the effect of this method is investigated. Some key parameters of SVM and kernel functions are surveyed. SVM methods are more effective, especially for the case of lack of training samples. The experiment shows that the SVM method has good classification ability and robust performances. The techniques can provide the opportunity to identify problematic pipe sections and repair the pipe prior to failure. Therefore it can help to prolong the life of a suspect pipeline while minimizing the potential for catastrophic failure.