Many of the data gathered through inspection, acquired in non-destructive testing (NDT), and recorded in destructive tests in a materials testing laboratory are sample data, so that the synthesis of these data is inherently uncertain.  Nevertheless, plant engineers and operators need to act on the basis of these inspection data, making decisions to adjust inspection frequency or procedures, alter maintenance levels of service or replace the equipment.  Statistical analysis of the sample data is an effective means of reducing the uncertainty and thereby increasing the confidence in critical asset management decisions.

Statistical analysis of inspection and NDT data is important because:

• It is not practical to inspect or assess all surfaces and components of all high risk assets, so sampling and analysis often is the only feasible approach.
• Statistical analysis can determine where more detailed inspection or testing is warranted, and where it is not.
• Even where inspection has been performed, there is a measurable probability that some conditions of interest have not been detected – statistical analysis can determine the probability of detection and the significance of non-detection.
• Over time, the growing body of inspection and NDT data permits increasingly precise and insightful characterizations of plant condition; i.e., the database itself becomes an important asset management resource.

Knowledge-based services

• Our team of senior statisticians, metallurgists, structural integrity engineers and materials engineering specialists hold advanced degrees in their fields and have, on average, more than 20 years of experience in interpreting inspection and NDT data.
• The team is skilled in:
  • General multi-variate statistics
  • Multi-variate statistical transformations, trend analysis
  • Value-of-information analysis
  • Advanced probabilistic statistics
  • Extreme value analysis (click for detail and examples)
  • Bayesian statistics (conditional probabilistic models)

Software

The Quest Reliability team uses commercial statistical analysis programs such as Matlab and Mathcad, and specialized software tools such as ABAQUS and @Risk, as well as our proprietary FEA suite of products (Signal Fitness-For-Service™, FEACrack™, FEAFlaw™, LifeQuest™).

Examples

• Two years of inspection data are examined statistically to identify subtle trends in multiple damage mechanisms that indicate degradation of material strength. 
• Condition assessment and operating data from several locations in a large asset (e.g., fired heater) are analyzed statistically to guide future inspections.
• A Bayesian model is used to predict the cost-risk of asset failure over time.

For further information about Statistical Analysis Inpection Data please contact us.