Developing a Reliability Measure for Nondestructive Testing of Reinforced Concrete
Most Nondestructive Evaluation (NDE) Technologies are characterized by noisy data and less than perfect detection characteristics. As a result, NDE data do not necessarily fall into either of the two obviously defined categories of true positive (TP) or true negative (TN). TP is defined as a result which indicates a defect and there is a defect and TN is defined as a result that indicates no defects and there is no defect. NDT data can also include two types of incorrect indications; these are false negative (FN) or false positive (FP), also known as false alarm. FN is the failure to give a positive indication in the presence of a defect and a FP is a positive indication when there is no defect. Reliability with performance assessments have solely been relegated to probability of detection (POD) assessments through probability of detection (POD) analysis. Although POD (also known as true positive rate, TPR) is an important component, it is merely the first step of a reliability assessment. POD analysis has been used historically to describe the reliability of NDT technologies for the detection of one dimensional (1-D) targets (i.e., crack), but it has some drawbacks. First, and most importantly, it does not consider the probability of false alarm (POFA) accompanying NDE testing. False alarm index is induced by noise with several possible sources which might include, human, nature of phenomenon to be measured, environmental conditions and so on. Second, assessment of POD through POD analysis is estimated based on a single pre-set decision threshold criterion. The concept of decision threshold may be challenging to quantify here, but in most situations, it is important to assess the performance of NDE technologies based on multiple criteria. Third, for a reasonable estimation, POD analysis requires inspecting a large number of “positives” (i.e., cracks), which can be quite difficult with three-dimensional (3-D) targets (i.e., delaminations, honeycombing, etc.). Last, statistical analysis of the POD curve is quite complicated and a challenge to implement. In this research, the reliability of impact echo (IE), infrared thermography (IR), ground penetration radar (GPR), and sounding methods for detecting a corrosion induced delamination in a concrete plate-like members are assessed by using an alternative method of analysis called receiver operating characteristic analysis (ROC); this analysis method has the potential to overcome many barriers associated with POD analysis. ROC methodology applied for the assessment of the reliability of NDE technologies is intended to provide a more effective means of comparing different technologies applied to civil engineering applications.