Condition Monitoring of Pipelines

Abstract

Structural components are often subjected to life limiting service conditions, such as vibrational fatigue, thermal fatigue, and corrosion. Monitoring the condition of such equipment is critical for its continued safe operation. Monitoring permits operators not only to assess the accumulated damage caused by system operation, but also to identify and avoid or minimise those operating conditions which result in significant damage or reductions in remaining life. The condition of any critical component can be monitored. Thus, people with heart conditions often have electro-cardiograms to monitor a variety of heart abnormalities. An industrially critical component is more likely to be a gearbox, a motor, a transformer or a pipeline. This report deals with condition monitoring of pipelines for a utility services company. Data from ultrasound scans is analysed in order to detect the presence of pipeline defects. The approach uses data visualisation to aid feature extraction. The main tools employed are PCA, FFTs, AR models and neural network classifiers. The intent of the project is to solve the problem of detection. Indeed, most of the time when the alarm system is raised the region is not actually defective. This report explains a part of the work which has been done before, then the approaches we tried this year.