Pipeline networks are economical means of transporting oil and gas. Most pipelines have been designed with a typical life span of 25 to 30 years. Existing pipelines are aging and quite susceptible to failure, due to poor construction of joints, corrosion, fatigue and material cracks. Accidents, terrorist activities, sabotage or theft can also cause leak disasters. Prevention of catastrophic failures of pipelines is critical for public safety and the environment. In order to maintain the healthy state of pipelines, continuous and accurate monitoring of pipelines is crucial, especially for corrosion and leakage. In order to reduce the impact of petroleum product spills, quick and effective leakage detection is needed to mitigate the problem.

The focus of this study is the development of a novel pipeline monitoring scheme to detect faults such as corrosion and leakage. Dynamic properties, such as frequency response functions (FRFs) and transient responses, of the pipeline system can change when faults occur. In the case of a corroded pipeline or leaky pipeline system, the mass and stiffness of the pipeline changes, thereby changing the FRFs of the pipeline system. The variation in the dynamic parameters is used to predict the presence of the corrosion or leakage in the pipeline system.

Experimental modal analysis was performed using a small-scale pipeline setup in various operating and physical conditions. In parallel, finite element analysis was undertaken for the same physical and operating conditions of the pipeline. A novel adaptive neuro fuzzy inference system (ANFIS) based scheme was developed for pipeline monitoring using fuzzy rules and neural networks. The trained ANFIS architecture was tested with a number of sets of dynamic parameters that were obtained by simulating leakage and corrosion in the pipeline and by varying the pressure and flow rate.

The results show that the ANFIS-based pipeline monitoring system can accurately predict the corrosion or leakage in the pipeline with an acceptable band of errors. The predicted results validates that the ANFIS-based pipeline monitoring system is less time-consuming and more flexible through the use of fuzzy rules incorporated with real-world systems.