How Playing Battleship Helps with CML Optimization

  • Home
  • How Playing Battleship Helps with CML Optimization

How Playing Battleship Helps with CML Optimization

Author: Brandon Stucky, CorrSolutions Team Leader, Senior Engineer

Imagine knowing the exact rate of corrosion for every pipe and every corrosion monitoring location (CML) in a plant. This knowledge would increase piping reliability and integrity beyond any currently known level. CorrSolutions has taken steps forward to make this a reality with their patented BOAR (Bayesian Optimized Asset Reliability) assessment. The BOAR assessment utilizes a hierarchal Bayesian model combined with data sampling to determine CML corrosion rates, while also utilizing expert opinion and measurement error to provide both more accurate corrosion rates and their confidence intervals.

Let’s explore a bit more about why we complete inspections and why we have CMLs in the first place. At the very core of inspection, there are two reasons why we take and use thickness readings; to determine a corrosion rate and to identify areas of local corrosion. Imagine a piping circuit with 100% general corrosion where every piping component uniformly corrodes at the same rate. If this fictitious circuit were to truly corrode uniformly, the entire circuit would only need a single inspection point, with enough inspections to determine an accurate corrosion rate. With this information, maintenance personnel would utilize that rate, compare it to the thickness of each piping component, and plan replacements prior to a failure. Unfortunately, 100% uniform corrosion is a rare occurrence, and we’ve learned this the hard way. In many instances, plants had CMLs on their piping circuits when failures occurred, but the failure did not occur at the location of the CML. So, more CMLs were added to try to identify localized corrosion. By adding more CMLs, plants tried to be smart about where they placed the extra CMLs, typically adding them to areas that may experience different corrosion rates than the initial CMLs. However, over multiple iterations of this method, plants now have significantly more data to review, calculate, and analyze.

Origins of Thickness Management

When API 570 was first published in 1993, the API committee had multiple battles to fight. First, they were trying to standardize thickness management, which was traditionally handled by each company independently, to comply with the new process safety management (PSM) guidelines. Thus, anything API passed had to be generally agreed upon and accepted by the whole industry. This can be a major challenge when companies have been building their own processes for decades and now an industry standard must be agreed upon. The second issue was the existing technology. Computers were not a mainstay on inspectors’ desks; rather, they had pencil and paper. As a result, all thickness management methods needed to be simple enough that the calculations could be completed by hand. Today, because of the original technological limitations, the API 570 methodology only looks at each individual CML without considering neighboring CMLs. Looking back at why extra CMLs were set, we remember it was to determine a corrosion rate, and to identify localized corrosion (i.e., identify places where different corrosion rates exist). Unfortunately, by only looking at CMLs independently of one another, inspectors and plants lose sight of the extent of localized corrosion throughout a circuit.

New CML Solution

In early 2022, CorrSolutions received a patent for its BOAR data analysis method, which combines inspection data and expert opinion with the concept of Bayesian updating. API 581 methodology also uses a Bayesian approach in its calculations. If you’ve ever played the game Battleship, then you have utilized Bayesian updating. In its simplest form, Bayesian updating means using new information to affect your next action. In Battleship, before choosing where to launch your fictitious weapon, you scan the board and inherently have some predetermined probabilities of the ships’ locations.  Once the weapon is fired, you hear either “HIT” or “MISS”; your brain takes that information and recalculates where the opponent’s vessel may be hidden. Using Bayesian updating, your brain uses its prior knowledge of the game board and the attack on your opponent to learn if you connected with a plastic piece or landed in the vast grid of water. Once you know the outcome, you reset your bearings and launch another attack. Similar to Battleship, the CorrSolutions BOAR assessment looks at each CML, considers expert opinion, and analyzes the data between the CMLs to identify trends for localized corrosion.

The Power of Similar CMLs

For example, CML-1 has two data points with low corrosion rates and the adjacent CML-2 has five data points that show significantly higher corrosion rates. Applying a BOAR assessment will allow the two CMLs to “talk” to each other. From a practical standpoint, the two CMLs are in similar service and physical configurations and should be experiencing similar levels of corrosion, unless something unexpected is occurring such as local corrosion. The BOAR assessment reviews both CMLs and sees that CML-1 only has two data points whereas CML-2 has five data points, which provides more proof to support the higher corrosion rate. The BOAR assessment puts higher confidence in the CML-2 data and applies it to CML-1 by slightly increasing the corrosion rate of CML-1, until there is proof that CML-1 is not experiencing the same level of corrosion as CML-2. By increasing the corrosion rate for CML-1, the inspection interval will decrease. By changing the priority for CML-1, the plant will be required to gather the data and confirm the corrosion rate.

Apply Expert Opinion

Conducting data analysis in a vacuum can be extremely dangerous, especially if not enough care is taken to understand the context of the data. In instances when only limited data is available, the BOAR model will rely on the expert opinion rather than the limited data. Once the model has enough data to either confirm or contradict the expert opinion, it will shift its analysis to prioritize the data. In addition, when the measured data has limited data points and does not agree with the expert rate, then the BOAR model will infer that to be uncertainty and recommend conducting an inspection to collect additional data and reduce the uncertainty in the corrosion rate. The BOAR assessment interprets both expert opinion and CML data to fully understand context and apply it across all piping circuits, regardless of the quality of data collection.

BOAR Rate

Once a BOAR rate is determined, then the real use cases present themselves. A BOAR rate is not only a single corrosion rate but a distribution of thicknesses at any point in time that, when combined, create corrosion rates. The BOAR assessment utilizes the estimated corrosion rates and the Inspection Data Management System (IDMS) thickness data to determine probable thicknesses in time. With these thickness distributions created, the corrosion rates can be determined with confidence. Use of confidence intervals in the thicknesses and corrosion rates allows the next step in the use case; CMLs with a large spread in the corrosion rate confidence intervals require more data and are prioritized for inspection. Conversely, CMLs with tight, consistent data become the top candidates for either longer intervals or optimization.

When looking at the entire circuit, remember all the CMLs in a circuit are interconnected, and we can see how the individual CML distributions compare to one another and with the overall circuit. In places where the CMLs are consistent across the circuit, you can consider the circuit to have general corrosion and then select the appropriate inspection methods. However, when there are large peaks and valleys in the CML comparison, it is clear the area is experiencing localized corrosion, so using radiography or scanning becomes more important. Remember, extra CMLs help to determine where local corrosion is occurring; therefore, when areas are identified, that information needs to be used. In addition, those areas of general corrosion help to describe how tight the variances are, which determines the level of confidence in the uniform rate. When confidence is high, the number of inspections can be greatly reduced.

Conclusion

The CorrSolutions BOAR assessment combines your expert knowledge with the thickness data you have spent so much money on obtaining. That data shouldn’t go to waste, nor should you discredit the your experts. The BOAR methodology is the best of both worlds. By implementing a Bayesian updating methodology with real data and expert opinion, CorrSolutions takes circuit level data input and provides CML level analysis. This analysis provides recommended review of circuitization, optimization of CML placement or inspection intervals, and even an individual circuit analysis that can be input into an RBI program.

To learn more about the CorrSolutions BOAR assessment, please contact R. Branden Stucky by filling in the form below.

Newsletter Archive

Access all of our previously published Industry Insights Newsletter articles:

Recently Published

Manage the Damage

Author: Paul J. Kowalski, Staff Engineer I

In this World Fertilizer article, Paul Kowalski explains that damage mechanism reviews (DMRs) are one of the first steps in developing many key process safety management (PSM), mechanical integrity (MI), and equipment reliability programmes in fertilizer processing units. Completing a DMR can help ensure long-term reliability of fixed equipment and piping, allow for quicker run, repair, and replace decisions to be made, and is a critical component for proper material selection.

Read More »

Equity Engineering Practices: Best Practices for Capital Projects, Repurposing Projects, and Lifecycle Management

Authors: Joel Andreani, Senior Vice President of Consulting Engineering, Principal Engineer II; Susie Szymanski, Practices Director
The EEPs help owner/users manage equipment and facilities from “cradle to grave,” over the entire lifecycle of their equipment and infrastructure. In this article, Joel Andreani and Susie Szymanski explore how the history of the EEPs and E2G’s dedication to expanding the collections has helped owner/users improve operational reliability and reduce costs.

Read More »

Capturing Corporate Memory with Internal Best Practices – Retaining Knowledge After the Great Retirement/Resignation

Authors: Joel Andreani, Senior Vice President of Consulting Engineering, Principal Engineer II; Susie Szymanski, Practices Director
We knew the great retirement/resignation was coming; however, the last two years has expedited a rapid loss of technical knowledge. Think about your organization, if your experienced employees’ knowledge and expertise are not recorded, their 40-plus years of practical experience and knowledge will be lost. Read how the Equity Engineering Practices (EEPs) integrate legacy documents and employee knowledge to support effective asset lifecycle management.

Read More »
Is Advanced Statistical Analysis the Next Step in Your Piping Inspection Journey?

Is Advanced Statistical Analysis the Next Step in Your Piping Inspection Journey?

Advanced statistical analysis can provide measurable benefits and new insights into corrosion management, especially when the knowledge is shared across an organization. In this Inspectioneering Journal article, Chris Massengale and Robert Clarke discuss how the implementation of this advanced analysis with other best practices will provide a better understanding of corrosion risks, optimize maintenance costs, and create safer facilities.

Read More »
Generic selectors
Exact matches only
Search in title
Search in content
Post Type Selectors
Pages
Industry Insights Newsletter Articles
Events
Library Items