The cost of corrosion to society is enormous, variously placed at 2% to 5% of the gross domestic product (GDP). As our systems age or we explore increasingly aggressive environmental conditions seeking new energy sources, corrosion looms as an important consideration to both safety and economic viability of projects. Sustainability is a buzz word used in many organizations and is defined as meeting the needs of the current generation without endangering those of the future generations. This means that we seek to reduce our environmental footprint while maintaining economic viability and generational equity. Corrosion management plays an important role in sustainability in many ways by reducing the need to manufacture replacement materials resulting in reduced emissions, increasing energy efficiency through the use of light-weight materials and other measures, and ensuring the safety of systems for current and future generations. However, our decisions about how to manage corrosion are made under uncertain conditions, arising from both errors in data and models used to predict corrosion. Risk is the currency used by decision makers; therefore, corrosion engineers and scientists must be able to integrate their knowledge into risk management.
To assess the risk of corrosion, we need to understand the fundamental mechanisms of corrosion and link such understanding to probabilistic assessments. Purely empirical knowledge is insufficient in predicting events for which we have scant prior experience. The corrosion field is replete with examples of failures that were never anticipated, and, after the failures occurred, considerable effort was spent understanding their mechanisms. Examples include the intergranular stress corrosion cracking (SCC) of line pipe steel (1970s), transgranular SCC of line pipe steel (1990s),1 and SCC of Ni-based alloys in high-purity water (early 1960s).2 Corrosion engineers, who worked on oil and gas pipelines and nuclear power plants before these failure happened, likely did not anticipate that these failure modes posed major financial and safety risks. We envision that by coupling our increasing understanding of corrosion processes with quantitative risk management tools we can increase our ability to prognosticate failure risks beyond our past experiences.
NACE International sponsored a two-day conference in June 2013 in Washington, DC, to explore risk assessment methods used in different industries and to bring different perspectives to the practice of risk management of corrodible systems. The first of its kind, this conference brought together risk and corrosion specialists from diverse industries. The conference was kicked off by a keynote talk by John Garrick on quantitative risk assessment approaches and their uses for predicting catastrophic risk. Readers of this journal can refer to his classic paper on risk assessment as well as the comprehensive book on this subject.3–4
This special issue of Corrosion presents selected papers from that conference. Although Corrosion has published several papers regarding risk of corrodible systems in the past, it is hoped that a special issue and the references therein stimulate further research in this important area. The paper by Ayello, et al., describes a Bayesian network approach to integrate diverse quantitative corrosion models and expert inputs in one frame work, with examples of pipeline threat assessment for validation of the approach. Although the Bayesian network model is useful for prioritizing the locations of highest threat probability, derivation of the input and conditional probabilities requires considerable effort in modeling failure processes. The paper by Case describes a stochastic approach to modeling carbon dioxide (CO2) corrosion incorporating mechanistic understanding of the corrosion process. The effect of inhibitor injection on damage probability is also modeled using probability estimates. McCallum, et al., describe a Markov chain process for computing the pit depth distribution as a function of time and compare this to a kinetic Monte Carlo method. They apply this approach to both aluminum and steel corrosion, exemplifying aircraft and pipeline systems, respectively. While these methods are useful for describing the probability of different pit depths in a given system, the necessary parameters have to be derived from previously observed results, restricting their predictive aspects. Sánchez and Sagüés describe a model for reinforced concrete corrosion based on the diffusion of chloride from the external surface to the rebar surface. They compare the effect of a potential-dependent critical chloride versus potential-independent critical chloride model on damage function. While their model of damage function does not include a full probabilistic calculation, the mechanistic modeling allows projections of damage as a function of a number of observable parameters. Valor, et al., present a method to estimate the reliability of oil and gas pipelines that cannot be inspected using in-line tools (also called non-piggable pipelines). Their approach relies on extensive field investigations of pipelines to catalog external corrosion defects, which are then used to derive defect probability density functions. A reliability index is then developed based on these and historical data. Although, such observation-based probability distributions of corrosion defects are essential for extrapolating from small size scales to larger scale, extrapolating in time would require mechanistic understanding of the complex external corrosion processes. Macdonald, et al., outline a neuro-fuzzy logic-based approach to estimating the risk of microbiologically influenced corrosion (MIC) in oil and gas pipelines. The novel approach relies on both subject matter expert opinion and quantitative data to estimate the risk of MIC, which is an inherently complex process.
The papers presented in this special issue span a range of techniques and are limited to a few applications. But, they highlight the need for corrosion researchers to combine their knowledge with risk assessment techniques tuned for specific systems. One risk assessment method is not necessarily sufficient just as one corrosion model or experimental data set may not adequately represent the behavior of a complex system. Furthermore, validation of the risk assessments through historical data is essential and difficult since the needed data are often missing. It is hoped that further improvements in both the models and validation will continue to be made and presented in future issues.
