Pipeline-Life Assessment Based on a Leak-Growth Model

Abstract

Summary. An empirical model is used to forecast the frequency of pipeline leaks and to assess pipeline failure rate. The model, which is based on a simple exponential-growth relationship, predicts the cumulative number of leaks, n, as a function of time: n=A exp(mt), where A and m are empirical constants. The model is applicable to essentially all carbon-steel pipelines exposed to serious corrosion attack, either internal or external. The model is useful in evaluating the technical and capital alternatives in pipeline design, operation, and maintenance. Introduction The literature on pitting is quite extensive. A review of the pertinent literature on pitting up to 1963 was carried out by Kolotyrkin. Szklarska-Smialowska reviewed the literature between 1960 and 1970, and Rowe compiled a list of the relevant pitting literature from 1960 to 1974. A number of investigators demonstrated that the cumulative number of pitting leaks, the maximum pit depth, the corrosion rate, and the time to first leak are all randomly distributed quantities. A number of statistical correlations using the extreme-value statistical approach developed by Gumbel have been proposed to quantify the maximum pit depth. Weibull statistical analysis was used by Finley and Toncre to model the time to first leak on 2,500 miles [4023 km] of pipeline submerged in Lake Maracaibo. Their results indicated that the time to first leak could be represented by the Weibull distribution function. None of these models considered the effects of pipe dimensions, corrosivity, or corrosion allowance on pipeline-leak frequency. This paper describes methods for using a conceptually simple pipeline-leak-frequency model to choose among technical/capital alternative in pipeline design, maintenance, and operation. The model is applicable to essentially all carbon-steel pipelines, provided that they are exposed to serious corrosion attack, either internal or external. An analysis of historical leak-frequency data from water-injection pipeline systems with the proposed empirical model is presented. Statistical Nature of Pitting Most metals and alloys are polycrystals consisting of grains of various shapes and orientations, on the boundaries of which various chemical elements may be concentrated. Imperfections like voids and cracks occur inside the metal. Thus, the properties of the metal are subject to random variation from one portion of the metal to another. Systems in which the elements or elementary volumes possess random properties are often studied with statistical analysis. In the consideration of the corrosion propagation process, the metal may be visualized as consisting of elementary volumes with randomvariation properties. In particular, their resistivities to corrosion penetration vary randomly. Corrosion begins at the surface and proceeds from one region to the next. When corrosion reaches a given region in the metal, it acquires a certain propagation rate, depending on the properties of that region. Because these properties are random, the corrosion rate is also random even though the external conditions (the corroding medium) may remain unchanged. Two mechanisms of corrosion must be considered in an assessment of the effects of corrosion: general and pitting corrosion. General Corrosion. General corrosion results in uniform wall thinning. The major effect of general corrosion on a pipeline is to increase the pressure stress. When the pressure stress, up, equals or exceeds the yield strength, ay, of the pipe wall, failure can occur. This limiting criterion can be expressed as follows: (1) where b = original nominal wall thickness, C = corrosion allowance, R = general corrosion rate, t = time during which corrosion takes place, FD = pipeline-pressure design factor (F =0.72 for cross-country pipelines), and The time during which an acceptable degree of general corrosion takes place is normally greater than or equal to the design life. From Eq. 1, with the limiting case where sigma /sigma = 1, the time during which such corrosion takes place is given b (2) Serious damage resulting from general corrosion of pipelines, either internal or external, has rarely been observed. Atmospheric corrosion rates are usually so low that they have no impact oil the structural integrity of pipelines; exceptions may be found in industrial areas subject to periodic, severe atmospheric pollution. Pitting Corrosion. Pitting corrosion causes nonuniform metal loss. In the case of internal corrosion, it is usually found to be most damaging in the lower portions of a pipeline. Pitting corrosion rates are statistical in nature. The time to first leak depends on the extreme pitting rate rather than on the average pitting rate. Extreme pitting rates are usually two to five times larger than the average rate. Because of this statistical nature, an assumed extreme rate that is three times the average will have an error of no more than 30% in calculation of the total number of leaks in a pipeline system over a given service life. SPEPE P. 656^