Proposed API 581 Inspection Plan Optimization

API 581 is a good recommended practice. It's methodology allows the Owner / User to calculate risk by damage mechanism and component and use that information to prioritize inspection tasks. As good as it is, does it really do the best job optimizing the individual tasks and by extension Maintenance / Shutdown planning? My personal opinion is no. There are further ways to optimize the process and make it better.

In order to understand my position, we first need to go through how risk-based inspection prioritization is currently done. Risk per API 581 Version 3, Part 1, Section 4.3.1, 1st paragraph is calculated thus:

4.3.1 Determination of Risk

In general, the calculation of risk is determined in accordance with Equation (1.6), as a function of time. The equation combines the POF and the COF described in Sections 4.1 and 4.2, respectively.

R(t) = P_f(t) ⋅ C_f

(1.6)

The risk calculated from the equation is compared to a "Risk Target" set by the Owner / User. The overall inspection planning concept is that if the risk target is reached before the planned date, then damage mechanism tasks are generated with the due date being the date which the target was reached.

So far so good, but the details matter. As we all know, consequence of failure (COF), once calculated for the four hole sizes is unlikely to vary, while probability of failure (POF), based upon the inputs for determining degradation rate for each damage mechanism, can vary with each inspection. Understanding this, allows us to concentrate on the POF as the most frequently changing value for risk. So, what constituents make up P_f(t)?

Let us look at how P_f(t), the total POF, is calculated and used for risk. From Part 2, Section 3.1:

3.1 Overview

The POF is computed from Equation (2.1).

P_f(t) = gff_total ⋅ F_MS

(2.1)

In this equation, the POF, P_f(t), is determined as the product of a total generic failure frequency (GFF), gff_total, a damage factor (DF), D_f(t), and a management systems factor, F_MS.

We will not get into F_MS here as it is a relatively static value. Likewise, gff_total are static values and also tabularized in API 581. Therefore, we will concentrate on D_f(t) which can vary with each inspection. According to Part 2, Section 3.4.1, paragraph 2, D_f(t) consists of the following:

DF estimates are currently provided for the following damage mechanisms.

1. Thinning - D^thin_f-gov
2. Stress Corrosion Cracking (SCC) - D^scc_f-gov
3. External Damage - D^extd_f-gov
4. High Temperature Hydrogen Attack (HTHA) - D^htha_f
5. Mechanical Fatigue (Piping Only) - D^mfat_f
6. Brittle Fracture - D^brit_f-gov

The above are combined to calculate the total Df(t) according to Part 2, Section 3.4.2, Subsection a), this way:

3.4.2 Damage Factor Combination for Multiple Damage Mechanisms

1. Total DF, D_f-total - If more than one damage mechanism is present, the following rules are used to combine the DFs. The total DF is given by Equation (2.2) when the external and/or thinning damage are classified as local and therefore, unlikely to occur at the same location.

   D_f-total = max[ D^thin_f-gov, D^extd_f-gov ] + D^scc_f-gov + D^htha_f + D^brit_f-gov + D^mfat_f

   (2.2)

   If the external and thinning damage are general, then damage is likely to occur at the same location and the total DF is given by Equation (2.3).

   D_f-total = D^thin_f-gov + D^extd_f-gov + D^scc_f-gov + D^htha_f + D^brit_f-gov + D^mfat_f

   (2.3)

   Note that the summation of DFs can be less than or equal to 1.0. This means that the component can have a POF less than the generic failure frequency.

Whether or not you believe that external and/or internal thinning damage occurs at the same location, the rest of either equation (2.2) or (2.3) remains as shown below:

D^scc_f-gov + D^htha_f + D^brit_f-gov + D^mfat_f

Both equations imply that all of the other damage mechanisms do occur in the same location. The first issue is using the summation of damage factors to drive inspection task due dates. My plant experience tells me that different damage mechanisms rarely occur in the same location.

The second issue is that summing all of the damage factors provides a composite damage factor and by extension a composite POF value that is, by default, applied to each damage mechanism. It is then obvious that the due dates recommended by the algorithm for each damage mechanism will all be the same.

Extending this concept further, each damage mechanism's damage factor, when individually combined with the COF, will likely reach the Risk Target at different due dates. This concept is the antidote for the two issues defined above. Therefore, Equation (2.1) above would change to be:

where gff_total and F_MS would remain as currently defined but D_f(Ind) would be substituted with each individual damage mechanism's damage factor:

DF estimates are currently provided for the following damage mechanisms.

1. Thinning - D^thin_f-gov
2. Stress Corrosion Cracking (SCC) - D^scc_f-gov
3. External Damage - D^extd_f-gov
4. High Temperature Hydrogen Attack (HTHA) - D^htha_f
5. Mechanical Fatigue (Piping Only) - D^mfat_f
6. Brittle Fracture - D^brit_f-gov

This would also cause each damage mechanism's risk calculation to become:

In turn, R(Ind) would be compared to the Risk Target for each individual damage mechanism, thereby establishing an independent due date and extent for each. Logically, this appears to be a better solution for establishing due dates and extents for inspection tasks.

My personal belief is that this method would better optimize inspection resources while maintaining the same level of risk based upon the Risk Target. I believe this should be an optional workflow with API 581 in addition to the current one. I am interested in your feedback on this proposal. Please comment below or contact me directly to start the conversation.

Comment on this article