On the Performance Quantification of Active Sensing SHM Systems Using Model-Assisted POD Methods

Abstract

System quantification is a critical and necessary step to provide an assessment of performance of any inspection technology, either classified to NDE or Structural Health Monitoring (SHM) techniques. For NDE techniques, the well-established method of probability of detection (POD) is widely accepted and adopted to quantitatively characterize and validate the reliability to detect a specified event (flaw) using the an NDE system. The resulting curves of system reliability correlate POD values with flaw sizes while statistical uncertainties are reflected by expressing the POD value together with a confidence level (CL). For most critical applications (like aerospace applications) 90/95 (POD/CL) is chosen as standard. Accordingly, a well defined and general applicable procedure need to be developed which allows appropriate quantification of SHM-based evaluation techniques, in particular with a view to active sensing systems. Along these lines, this contribution shows that the widely acknowledged procedure for NDE systems cannot be adopted directly for SHM systems without substantial changes. A broad variety of significant differences are inherent to the nature of SHM systems which occur in terms of main uncertainties (operator vs. environment), available test specimens (many vs. one unique), sensor configuration (moving vs. fixed), as well as test environment (fixed vs. varied). Based on these considerations, a model-assisted POD methodology for active SHM systems will be presented which allows to characterize the detection performance within a statistical framework mainly governed by the variability of environmental conditions that SHM systems will experience. In such a way, a similar POD/CL requirement can be established for SHM systems following a well defined protocol. To achieve this goal, missing key developments will be discussed as calibration techniques for model-assisted POD calculation, and appropriate damage simulators to mimic various damage scenarios in experimental tests.