Likelihood Assessment

TL;DR

Likelihood assessment is about figuring out how probable a future event, like a component failure, is to happen. It's not just a guess; it uses data and logic to estimate frequency or probability. Getting this right helps you make smart decisions about inspections and maintenance.

1. The Mental Model

Think of likelihood assessment as trying to predict the future based on past information and current conditions. You're trying to put a number or a category on how often something will occur, not just if it could happen. It helps you focus your resources where they're most needed.

2. The Core Material

When assessing likelihood for RBI (Risk-Based Inspection), you're primarily estimating the Probability of Failure (PoF). This isn't about a single component failing once, but rather the frequency or probability of a specific damage mechanism leading to failure within a given timeframe.

Qualitative Likelihood

Photo by Stanislav Kondratiev on Pexels

This approach uses descriptive terms and expert judgment to categorize likelihood. It's often used when detailed data is scarce or the situation is unique.

• Process: Experts review available information (design, operating history, inspection results, industry experience) and assign a likelihood category (e.g., High, Medium, Low, or a more granular scale like "Improbable" to "Almost Certain").
• Benefits: Quick, flexible, and good for initial screening.
• Drawbacks: Subjective, can vary between assessors, and lacks numerical precision.

Quantitative Likelihood

Photo by RDNE Stock project on Pexels

This method uses numerical data and statistical models to calculate a specific probability or frequency of failure.

• Process: Involves collecting and analyzing data such as historical failure rates, operating hours, material properties, inspection effectiveness, and damage growth rates. You might use statistical distributions (like Weibull or Exponential) or reliability block diagrams.
• Benefits: Objective, precise, and allows for direct comparison and trending.
• Drawbacks: Requires significant data, can be complex, and assumptions need careful validation.

Semi-Quantitative Likelihood

Photo by Negative Space on Pexels

This approach blends elements of both qualitative and quantitative methods. It often uses a scoring system, where different factors contributing to likelihood are assigned numerical scores, which are then combined to produce a likelihood index. This index is then mapped to a qualitative category.

• Factors considered:
  • Damage Mechanism Susceptibility: How prone is the material to a specific damage type?
  • Effectiveness of Inspection: How well do current inspections find damage before failure?
  • Operating Conditions: Are conditions (temperature, pressure, corrosivity) severe?
  • Maintenance History: Has the component failed before or had issues?
  • Time in Service: How long has the component been operating?

graph TD
    A["Identify Damage Mechanism"] --> B["Assess Susceptibility (e.g., Material, Environment)"]
    B --> C["Evaluate Operating Conditions (e.g., Temp, Pressure, Corrosivity)"]
    C --> D["Review Past Inspection History (Findings, Effectiveness)"]
    D --> E["Consider Time in Service"]
    E --> F["Combine Factors (e.g., using a scoring matrix or algorithm)"]
    F --> G["Assign Likelihood Category (e.g., High, Medium, Low)"]
    G --> H["Calculate PoF (if quantitative data available)"]

Probability of Failure (PoF)

Photo by DS stories on Pexels

PoF is usually expressed as a probability (e.g., 1x10^-3 failures per year) or a frequency (e.g., 0.001 failures per year). It's crucial to specify the timeframe. A higher PoF means a greater chance of failure.

Data Sources for Likelihood

• Internal Plant Data: Your facility's own maintenance records, inspection reports, and failure histories. This is often the most relevant.
• Industry Databases: Organizations like API, ASME, or specific industry groups collect and share failure data for various equipment types (e.g., OREDA for offshore reliability).
• Expert Judgment: Experienced engineers, metallurgists, and operations personnel provide valuable insights, especially for unusual situations or lack of data.
• Manufacturer Data: Equipment reliability data provided by the manufacturer.

3. Worked Example

Let's consider a carbon steel pipe segment in a crude oil service. We're assessing the likelihood of internal corrosion leading to a leak.

1. Identify Damage Mechanism: Internal corrosion.

2. Qualitative Assessment:

   • Susceptibility: Crude oil can be corrosive, especially with water and sulfur content. Initial thought: Medium.
   • Operating Conditions: Temperature is 80°C, pressure is 10 bar. Process analysis shows potential for water accumulation and H2S. Increases likelihood.
   • Inspection History: Last UT scan 3 years ago showed general wall loss, but within acceptable limits. No recent internal inspections. Suggests increasing likelihood with time.
   • Time in Service: Pipe is 15 years old.
   • Combined Qualitative View: Based on potential corrosivity, lack of recent detailed internal inspection, and age, the likelihood is assessed as Medium-High.

3. Semi-Quantitative Scoring (Simplified):

   • Corrosion Potential (0-5, 5=high): Let's say we score 4 (due to known corrosives).
   • Inspection Effectiveness (0-5, 5=poor effectiveness): Last UT was external, 3 years ago. Let's score 3 (detects some damage but not great for internal).
   • Age/Time in Service (0-5, 5=old): 15 years. Score 3.
   • Operational Control (0-5, 5=poor control): Assume process controls are generally good but don't specifically mitigate this corrosion well. Score 2.
   • Total Score: 4+3+3+2 = 12.
   • Mapping to Likelihood Category: If a score of 1-5 is Low, 6-10 is Medium, 11-15 is High, then a score of 12 maps to High Likelihood.

This semi-quantitative approach helps validate or refine the purely qualitative judgment and provides a slightly more defensible basis for the assessment.

4. Key Takeaways

• Likelihood assessment predicts the probability or frequency of an event, like failure, within a specific timeframe.
• It can be qualitative (descriptive terms), quantitative (numerical values), or semi-quantitative (scoring systems).
• Data sources include your plant's history, industry databases, and expert knowledge.
• A higher Probability of Failure (PoF) means a greater chance of the event occurring.
• Always specify the timeframe when discussing PoF, e.g., "failures per year."
• Effective inspections can significantly reduce the likelihood of undetected damage leading to failure.
• Likelihood is one half of the risk equation; the other is consequence.

Common Mistakes to Avoid:
- Confusing likelihood with consequence: These are distinct components of risk.
- Ignoring data limitations: Don't treat qualitative assessments as precise quantitative numbers.
- Over-reliance on "gut feeling": Always try to back up judgments with available data or reasoning.
- Not defining the timeframe: "Likely" without a timeframe is meaningless.
- Failing to update assessments: Likelihood changes as conditions, inspections, and time in service evolve.

5. Now Try It

Take a common piece of equipment in your plant (e.g., a pump, a specific vessel, or a section of piping) and identify one potential damage mechanism for it (e.g., cavitation in a pump, stress corrosion cracking in a vessel, external corrosion on piping). Spend 15 minutes trying to qualitatively and semi-quantitatively assess the likelihood of that damage mechanism leading to a failure in the next 5 years. What information would you need to make a more accurate assessment? What score would you assign to different factors? What would a "high" or "low" likelihood mean for that specific scenario?