Advancements in evaluating reliability of nondestructive technologies for the detection of subsurface fracture damage in R.C. bridge decks
Metadata[+] Show full item record
During the last few decades, many efforts have been made to assess the reliability of nondestructive evaluation (NDE) technologies used for the detection of subsurface damage in concrete bridge decks. During these efforts, reliability of NDE technologies has either been described anecdotally, or been solely relegated to the probability of detection (POD) or accuracy estimation. Although these indices are important, most of the previous work did not take into account the probability of false alarm (POFA) of NDE technologies, nor did they investigate the reliability considering multiple threshold settings throughout test results. In addition, the existing body of research has used a limited physical sampling such as coring to validate NDE results. Consequently, the assessments were rather controversial, and there was no general agreement about the reliability of such technologies. Because most diagnosis systems are characterized by noisy data and less than perfect detection characteristics, reliability is to be carefully assessed considering all possible diagnosis output with multiple threshold settings within practical range of applications. In other words, when NDE data do not fall into either of the two obviously defined categories: true positive (TP), meaning the NDE data indicates a defect and there is a defect, or true negative (TN), meaning the NDT data indicates no defect and there is no defect, reliability analysis should also include the two types of incorrect indications: failure to give a positive indication in the presence of a defect (false negative, FN) and giving a positive indication when there is no defect (a false alarm or false positive, FP). The \three decades of NDI reliability assessments" report developed by Karta Technologies, Inc. in 2000 under supervision of the Air Force NDI Office stated that POD alone cannot describe the reliability of NDE technologies unless the probability of false alarm (POFA) is also considered in the analysis. POFA may be induced by noise with several possible sources: human, nature of phenomenon to be measured, and environmental conditions. The report covered nearly 150 reports and manuscripts from over 100 authors. However, a review of research literature reveals that little theoretical work on the reliability assessment in terms of both POD and POFA has been undertaken since then. In this research, the reliability of impact echo (IE), infrared thermography (IRT), and ground penetration radar (GPR) technologies for the detecting of subsurface damage in concrete plate-like members is assessed by using a statistical analysis method called receiver operating characteristic (ROC). The proposed analysis method has the capability to integrate POD and POFA indices over a wide range of decision threshold settings in a single curve, which is useful in assessing trade-off in choosing a threshold and for quantitatively comparing the performance of NDE technologies. This methodology for assessing NDE reliability is intended to provide a more effective means of comparing different technologies used in civil engineering applications, to make the evaluation process of a quantitative scheme, to reduce subjectivity and variability in interpreting NDE data, and to improve sensitivity to extract more information from NDE data. Area under ROC curve (AUC), which is interpreted as the probability of correctly classifying an arbitrarily pair of negative and positive test points, can provide for the desired quantitative reliability index, which can be used to compare the performance of one NDE technology to another. Results of this research obtained from ROC analysis indicate a great ability of IE and IR in detecting subsurface fracture damage such as delamination and debonding. In both technologies, there exist some threshold settings that can provide for a relatively high POD with very low POFA, and consequently, the areas under their ROC curves were very high. Data obtained from GPR testing, on the other hand, indicates that GPR technology has a very limited ability to detect physical damage such as subsurface delamination. This conclusion contrasts with that been argued by a large body of the previous work. However, GPR showed a good sensitivity to the presence of corrosive environments such as moisture and chloride when the concentrations of these factors are above some threshold values that may facilitate the initiation of steel reinforcement corrosion.
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivs 3.0 License.