Convergence of Multiple Statistical Methods for Calculating the Probability of Detection from SHM Sensor Networks

Abstract

The use of in-situ sensors for real-time health monitoring of a wide array of civil structures can be a viable option to overcome inspection impediments stemming from accessibility limitations, complex geometries, and the location and depth of hidden damage. The maturity of Structural Health Monitoring (SHM) sensors has evolved to the point where many networks have demonstrated sensitivities that meet or exceed current damage detection requirements. As a result, there is a growing need for well-defined methods to statistically quantify the performance of sensors and sensor networks. Statistical methods can be applied to laboratory and flight test data to derive Probability of Detection (POD) values for SHM sensors in a fashion that agrees with current nondestructive inspection (NDI) validation requirements. However, while there are many agreed-upon procedures for quantifying the performance of NDI techniques, there are no guidelines for assessing SHM systems. While the intended function of the SHM and NDI systems may be very similar, there are distinct differences in the parameters that affect their performance and differences in their implementation that require special consideration. Factors that affect SHM sensitivity include flaw size, shape, orientation and location relative to the sensors, operational and environmental variables and issues related to the presence of multiple flaws within a sensor network. The FAA Airworthiness Assurance NDI Validation Center (AANC) at Sandia Labs, in conjunction with the FAA WJH Technical Center, has conducted a series of SHM validation and certification programs aimed at establishing the overall viability of SHM systems and producing appropriate precedents and guidelines for the safe adoption of SHM solutions for aircraft maintenance. This paper will present the use of several different statistical methods, some of them adapted from NDI performance assessments and some proposed to address the unique nature of damage detection via SHM systems, and discuss how they can converge to produce a confident quantification of SHM performance. Comparisons of hit-miss, a versus ȃ, and One Sided Tolerance Intervals will provide valuable insights into how the characteristics of the collected SHM data affect the formulation of that system’s POD curve. Similarities between NDI and SHM assessments will be highlighted in order to provide a foundation in traditional flaw detection performance measures. In addition, considerations of the controlling factors to be considered when collecting SHM response data will be discussed