Technical report | Guidelines for Interpretation of Published Data on Probability of Detection for Nondestructive Testing
This report provides general guidelines for the interpretation of published data on probability of detection (POD) for nondestructive testing. An overview is provided of the different types of probability of detection data, methods for statistical analysis and the assumptions that may be embedded in these analyses. Four key issues have been identified which need to be addressed when assessing the applicability of published probability of detection trial data to a new nondestructive testing application. Specific consideration should be given to the system boundary, which defines those elements of the inspection process and other factors potentially affecting inspection reliability that are considered to be under examination in the POD trial and those that are considered to be outside the scope of the trial.
Appropriate application of nondestructive testing (NDT) methods is dependent on knowledge of the minimum sizes of defects that the techniques are capable of reliably detecting, relative to the defect sizes that could be structurally significant. For some applications, the failure of NDT to detect a single defect could cause catastrophic failure including loss of life.
The reliability of NDT is commonly characterised as the probability of detection (POD) of a specific type of defect as a function of defect size. This report provides general guidelines for the interpretation of published data on POD. When probability of detection is estimated using a traditional POD trial in which field NDT technicians perform inspections on specimens with known defects, then the POD information obtained from the trial is strictly applicable only to the exact conditions and defect types for which the POD trial inspections were performed. Any broader application of the estimated POD to other inspection conditions is reliant on an engineering assessment that the change in inspection conditions will not reduce the POD. Four key questions have been identified which are designed to assist engineering staff to assess the applicability of published POD trial data for a new NDT application:
- How closely do the NDT technique and defect and material types used in the POD trial experiment match the new application, and how important are the differences?
- Where were the system boundaries for the POD trial?
- Who conducted the POD trial and for what purpose?
- What has not been said in the reporting of the POD trial results?
The purpose of a POD trial is to obtain an estimate of the POD by acquiring suitable experimental data and conducting an appropriate statistical analysis. Confidence limits are applied to the estimated POD to account for sampling variability inherent in any empirical statistical trial. It is not necessary to have a comprehensive understanding of the methods for statistical analysis of POD data in order to make use of published data. However, an understanding of different types of POD data and assumptions that may be embedded in the analysis methods may be helpful in interpreting the literature.