The **Selective Corrosion Rate Analysis**^{ Pat. Pending} technique utilizes a unique combination of data validation checks, corrosion
models, statistical techniques and advanced trend identification tools to highlight
specific corrosion mechanisms and provide accurate remaining life projections from UT
thickness data. The analysis is completely data-driven and requires only a single analysis parameter - the system's required reliability.

**Selective Corrosion Rate Analysis**
reports are designed to enable rapid review by Inspectors and Engineers. Re-inspection
schedules are optimized to provide a practical minimum level of inspection to achieve the
required reliability.

Since **Selective Corrosion Rate Analysis** utilizes state-of-the art
statistical principles and does not require the use of multiple, subjective analysis
parameters, the results are completely defensible from a regulatory standpoint.

At the heart of the *Selective
Corrosion Rate Analysis Report* is the *Corrosion Rate Reliability Plot*. This
plot shows the corrosion rates for all ‘active’ TML’s in the system
(horizontal axis) vs. the cumulative occurrence of the corrosion rate (vertical axis). The
vertical axis on a *Corrosion Rate Reliability Plot* is labeled
"Reliability", since the corrosion rate distribution is projected to the
system’s required reliability for determining remaining life.

Similar plots have been used for many years by Rotating Equipment Engineers for predicting equipment failures. Only recently have these plots been used to analyze the reliability of fixed equipment. For *Corrosion Rate Reliability Plots, *the upper (right) end of the plot provides the value used for determining remaining life.

**Selective Corrosion Rate Analysis** finds the best TML/corrosion rate relationship and statistical distribution in order make accurate remaining life projections and to identify trends. This is achieved by obtaining the best statistical "fit"** **for the corrosion rate data. A visual indication of the *fit* may be made by observing how well the corrosion rate data clusters around the best straight line on the *Corrosion Rate Reliability Plot*. Statistical indicators are used to accurately measure this "goodness-of-fit".

Several factors may cause an unacceptable *fit* on *a Corrosion Rate Reliability Plot*. There may be more than one corrosion mechanism (i.e. localized corrosion) in the system or an incorrect statistical distribution may have been selected. If corrosion rate data follows a Weibull distribution, for example, a plot based on normally distributed data may yield an unacceptable fit. There may also be significant thickness variations within components, or perhaps components were replaced with those having a different nominal thickness (e.g. a different pipe schedule for piping systems) between inspections. **Selective Corrosion Rate Analysis** correctly identifies and corrects for these real-life situations.

In the following *Corrosion Rate Reliability Plot*, the measured long-term corrosion rates range from about 3 mils/year to just over 20 mils/year. The data clusters nicely around the best straight (solid) line for the (log normal) statistical distribution. In this example, **Selective Corrosion Rate Analysis**identified the best distribution and found a single corrosion mechanism, as evidenced by the single solid line. Projecting the line to 95% reliability (typical for API 570 Class 2 systems) yields a corrosion rate near the maximum measured rate of 20 mils/year. The dotted line represents the short-term corrosion rate. The change in slope for the short-term data indicates a possible recent change in the system, since the corrosion has become less uniform (occurs over a wider range).

By running a series of* Corrosion Engineering Models*,** ****Selective Corrosion Rate Analysis** finds the best *fit* and displays the results on a *Corrosion Rate Reliability Plot*. If, for example, a system has higher corrosion rates in elbow components than the rest of the system, the data will be grouped per this *Corrosion Engineering Model* and fitted to the best statistical distribution. The *Corrosion Rate Reliability Plot* would then have two solid lines (one for pipe, one for elbows) and the *fit* will improve significantly over the *All TMLs* case.

Selective Corrosion Rate Analysis Reports always show the *Corrosion* *Rate Reliability Plot* for *All TMLs*. When a *Corrosion Engineering Model* identifies improvements over the *All TMLs* case, a separate plot showing these improvements is also shown on the report.