NDE reliability and probability of detection (POD) evolution and paradigm shift

• Published in: AIP conference proceedings Vol. 1581; no. 1
• Main Author: Singh, Surendra
• Format: Conference Proceeding Journal Article
• Language: English
• Published: Melville American Institute of Physics 18.02.2014
• Subjects: ACCURACY; Bayesian analysis; CALIBRATION; CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSICS; Component reliability; Computer simulation; Damage tolerance; Data acquisition; DATA ANALYSIS; DEFECTS; DETECTION; EVALUATION; Flaw detection; FORECASTING; FRACTURE MECHANICS; HUMAN FACTORS; Life prediction; Materials engineering; Morphology; Nondestructive testing; PROBABILITY; RELIABILITY; Reliability aspects; Reproducibility; SIMULATION; Statistical analysis; Statistical methods; Variance analysis
• ISSN: 0094-243X; 1551-7616
• DOI: 10.1063/1.4865079

The subject of NDE Reliability and POD has gone through multiple phases since its humble beginning in the late 1960s. This was followed by several programs including the important one nicknamed “Have Cracks – Will Travel” or in short “Have Cracks” by Lockheed Georgia Company for US Air Force during 1974–1978. This and other studies ultimately led to a series of developments in the field of reliability and POD starting from the introduction of fracture mechanics and Damaged Tolerant Design (DTD) to statistical framework by Bernes and Hovey in 1981 for POD estimation to MIL-STD HDBK 1823 (1999) and 1823A (2009). During the last decade, various groups and researchers have further studied the reliability and POD using Model Assisted POD (MAPOD), Simulation Assisted POD (SAPOD), and applying Bayesian Statistics. All and each of these developments had one objective, i.e., improving accuracy of life prediction in components that to a large extent depends on the reliability and capability of NDE methods. Therefore, it is essential to have a reliable detection and sizing of large flaws in components. Currently, POD is used for studying reliability and capability of NDE methods, though POD data offers no absolute truth regarding NDE reliability, i.e., system capability, effects of flaw morphology, and quantifying the human factors. Furthermore, reliability and POD have been reported alike in meaning but POD is not NDE reliability. POD is a subset of the reliability that consists of six phases: 1) samples selection using DOE, 2) NDE equipment setup and calibration, 3) System Measurement Evaluation (SME) including Gage Repeatability &Reproducibility (Gage R&R) and Analysis Of Variance (ANOVA), 4) NDE system capability and electronic and physical saturation, 5) acquiring and fitting data to a model, and data analysis, and 6) POD estimation. This paper provides an overview of all major POD milestones for the last several decades and discuss rationale for using Integrated Computational Materials Engineering (ICME), MAPOD, SAPOD, and Bayesian statistics for studying controllable and non-controllable variables including human factors for estimating POD. Another objective is to list gaps between “hoped for” versus validated or fielded failed hardware.