In API 510, it offers two ways for corrosion rate determination, one is using" point-to-point" method, in which long-term and short-term corrosion rates can be compared so that the corrosion rate which best reflects the current process can be chosen.1
The other way is "statistical analysis method", in which owner can establish a representative and suitable model to monitor and analyze pressure vessel corrosiveness behavior in order to maintain its mechanical integrity. This paper, based on the experience of thousands of pressure vessels, will present the model using linear regression which has been used successfully to monitor corrosion rate. And to maintain the quality of thickness measurements, statistical approaches like control chart have been applied effectively to reduce inspection errors.
Metal loss due to corrosion is a universal phenomenon in refineries which could in turn cause leakage or explosion if not well monitored. There are several units in a refinery such as crude distillation unit, hydro-processing unit, acid alkylation unit, etc. In each unit, there are hundreds of pressure vessels which have different potential damage mechanisms.2 Hence, it's critical to establish an effective and efficient way to monitor thickness changing behavior.3
Considering two facts that:
1. Most pressure vessels’ surface area in refinery is over 10 ft2, some are even over 100 ft2.
2. Traditional thickness monitoring ultrasonic testing (UT) probe's diameter is about 1 inch or smaller.
The concept of these 2 facts results in relative dimension between whole pressure vessel's surface area and thickness monitoring location (TML) is immense as shown in (Figure 1).
Therefore, to use ultrasonic thickness measurement (UTM) as a tool to determine the mechanical condition of pressure vessel is similar to sampling.
Although there are some thickness monitoring techniques such as thickness scanning or radiography profile, it's not economically feasible to apply these techniques to every pressure vessel. Instead of fixing "exactly" same locations to monitor thickness of specified point, "approximately" same location to monitor thickness changing behavior would be a better option in a broad sense.
