Pipeline abnormal classification based on support vector machine using FBG hoop strain sensor

Pipelines function as blood vessels serving to bring life-necessities, their safe usage is one of the foremost concerns. In our previous work, fiber Bragg grating (FBG) hoop strain sensor with enhanced sensitivity was developed to measure the hoop strain variation on a pressurized pipeline. In this paper, combined with multi-class support vector machine (SVM) learning method, the hoop strain information is used to classify pipeline abnormal working conditions, including cases of external impact, normal leakage and small rate leakage. The parameters of different kernel functions are optimized through 5-fold cross validation to obtain the highest prediction accuracy. The result shows that when taking radial basis kernel function (RBF) with optimized C and y values, the classification accuracy for abnormal condition reaches up to 95%. The error appears only in separating the small leakage cases from normal working conditions. This pipeline abnormal classification approach using FBG hoop strain sensor combined with multi-class SVM shows potential prospective in pipeline accident monitoring and safety evaluation.