Full Description

API 581 is the industry standard for Risk-Based Inspection (RBI) in the refining and petrochemical industry. This bulletin will start by reviewing and explaining the methods used in the 1st and 2nd Editions of API 581 to calculate the damage factor (normalized probability of failure) of a tube or pipe subject to plastic collapse under internal pressure with a wall that is thinning due to corrosion at some uncertain corrosion rate. This will be followed by a detailed discussion of some improvements that are planned for the upcoming 3rd Edition.

In the 1st Edition, a single table of damage factors is provided, indexed by the ratio of expected metal loss to the last known thickness, known as the Art parameter, taking into account the number and effectiveness of any inspections during which the uncertain corrosion rate was measured. The damage factors in this table were calculated using a simple limit-state based reliability method known as the Mean Value First Order Second Moment (MVFOSM) method. A limit state is a function defined in terms of all the input parameters such that its negative values correspond to presumed failure states while all of its positive values correspond to presumed safe states. The specific limit state chosen for the 1st Edition is simply the difference between the flow stress and the hoop stress. By treating the uncertain input parameters such as metal loss, pressure, and flow stress as random variables described by normal probability distributions, the MVFOSM method results in a normally distributed limit state with a mean and variance that is determined from the mean and variance of each of the input parameters. The probability of failure is then equal to the probability that the limit state is negative, which can be obtained quite easily from its normal distribution.

The uncertainty in the corrosion rate is accounted for by assuming three possible corrosion states, each with a mean corrosion rate that is equal to, double, or quadruple the measured rate C . The probability of being in each corrosion state is described by a discrete probability distribution function, but a continuous corrosion rate distribution is assumed within each state. The total probability of failure is the weighted average of the probability of failure for each of these corrosion states, determined using MVFOSM with the corrosion rate distribution assumed for each state, weighted by the probability of being in each state. Some prior probabilities are assumed, and then Bayes Theorem is used to update these state probabilities after each inspection, depending on how effective the inspection was at measuring the true corrosion rate. The final damage factors are then manually smoothed in order to even out artificial fluctuations caused by choosing the corrosion rate uncertainty within each state to be too small when compared to the uncertainty arising from the corrosion state distribution itself.

Although the 1st Edition method had been used successfully for many years as the basis for inspection planning within the RBI framework, it did suffer from some serious limitations. First and foremost, the damage factor table was generated using only one specific set of input parameter values corresponding to a representative base case. Inaccurate results are obtained for cases deviating significantly from this base case. Specifically, it is not possible to properly account for any amount of excess metal thickness, in addition to the minimum required wall thickness tmin plus the corrosion allowance, beyond what was considered in the base case. Another problem is that the manual method of smoothing the damage factor curves was never fully documented, and so the final tabulated values could never be exactly reproduced.

The 2nd Edition used a modified form of the Art parameter to index into the same damage factor table in order to account for the effect of having excess metal thickness, which is a common occurrence for seamless pipes ordered by schedule. This simple change greatly increased the applicability of this approach, although it still had its limitations since it was based on the same underlying methodology and damage factor table as in the 1st Edition.

In the 3rd Edition, this and other limitations will be fully resolved by replacing the damage factor table with an improved version of the underlying procedure used to generate it. The equations are written more compactly in terms of a newly identified dimensionless parameter, Y, which is the ratio of the initial hoop stress to flow stress. The damage factor is completely determined in terms of Y, the original Art parameter, and the coefficient of variation of all the uncertain input parameters. The limit state is also expanded slightly to account for additional wall shapes.

Two new reproducible methods for smoothing the damage factor curves are also presented and discussed. Manual smoothing no longer suffices when using a procedure, which necessitates an automated approach. Only the simplest of these smoothing methods will be incorporated into the 3rd Edition, but both are discussed here in some detail.

A more efficient and straightforward way of handling multiple inspections is also developed, by evaluating the newly-defined inspection factors once rather than recursively applying Bayes Theorem for each new inspection.

It is also shown that the original damage factor table can be extended by creating multiple new damage factor tables, each corresponding to a different value of the Y parameter (the original damage factor table corresponded to the single value Y= 0.215 ). Comparisons are made between these new tables and the original table from the 1st and 2nd Editions.

The procedure proposed here is shown to have many desirable properties that were once the motivation for using the modified Art parameter from the 2nd Edition. In particular, it is shown that the damage factor for a pipe with a smaller tmin is always less than that for a pipe with a larger tmin, all other things being equal. Additionally, it is shown that the damage factor remains small whenever there is excess metal thickness, except in unusual cases where the pipe is many times thicker than tmin , because the uncertainty in the metal loss increases with time. For this reason, the modified form of the Art parameter from the 2nd Edition is no longer required.

Product Details

Published: 2014 Number of Pages: 170 File Size: 1 file , 5.8 MB