In this paper, we propose a data-driven approach to find the optimal inspection interval for a petroleum pipeline that is subject to long-term corrosion failure. The approach accounts for the determination of both the failure frequency and consequences of failure. Three forms of corrosion are studied; these are uniform corrosion, pitting corrosion and stress corrosion cracking. Failure frequency is estimated by fitting historical failure data of pipeline into either a homogenous Poisson process or power law. The consequences of corrosion attack is calculated in terms of economic loss, environmental damage and human safety and determined for small leaks, large leaks and rupture of pipeline. Both failure frequency and consequences are utilized to estimate total loss due to pipeline operation. A risk based integrity maintenance optimization of the pipeline is obtained by minimizing the economic loss of pipeline, taking human risk and maintenance budget as constraints. The approach is very robust and well validated.
Petroleum pipelines are generally subject to different degrees of failure and degradation during operation and in their entire life cycle. Common failure mechanisms include corrosion, mechanical damage, third-party damage, and design imperfections. Corrosion reduces the integrity of an operating pipeline, and consequently lowers its service life. Corrosion failure could eventually lead to leaks or rupture, of a pipeline carrying huge human, financial, and environmental loss. In reality, maintenance resources are usually limited and maintenance managers are constantly seeking cost effective means of allocating maintenance budgets without compromising on public safety. This explains why there has been an increased awareness both in the industry and in academia on integrity maintenance optimization. However, it appears data-driven approach using real data from existing pipelines to model integrity maintenance has been less investigated. The approach is generally more robust than purely analytical methods due to use of historical data.