On the integrity of deformation monitoring

Abstract

In safety-critical applications, deformation monitoring systems are required to issue timely alerts when a deformation beyond a critical threshold occurs. Only a very small probability of failing to issue an alert when in fact one should have been given, is acceptable. This probability is referred to as integrity risk. In this contribution, we show how to evaluate this risk, thereby taking the intimate link between testing and estimation into account. Using a simple example, the basic integrity components of deformation monitoring are introduced and illustrated. The integrity risk is then formulated for the generalized case where multiple-hypothesis testing is involved. As monitoring systems, in addition to issuing timely alerts, are also required to provide deformation estimates, it is also crucial to assess their confidence levels. In doing so, the statistical testing, that preceded the estimation of the deformation parameters, needs to be accounted for. As this is not the customary procedure followed in practice, we show how the combined estimation and testing can be probabilistically accounted for, and thereby demonstrate that the customary practice can give a too optimistic outcome of the stated confidence levels. The presented methodology is worked out and numerically illustrated by means of two deformation examples.

An enhanced RANSAC-RTK algorithm in GNSS-challenged environments

“…Baarda (1968) proposed reliability and roughness detection theory in the 1960s. Teunissen also conducted a series of studies on this topic and established an outlier identification procedure in the geodetic literature (Teunissen 1997;Teunissen 1998;Teunissen 2018;Teunissen et al 2020). However, the presence of correlations between observations can lead to incorrect identifications, especially for multiple or continuous outliers.…”