Introducing data-centric engineering to instrumented infrastructure

Abstract

A variety of methods exist in the structural health monitoring literature that aim to combine the observed data and predicted outputs from physics-based models (e.g. model updating and calibration). Typically implemented on a case-by-case basis, there is currently no unifying procedure or method for formally synthesising this combination of information. Data-centric engineering (DCE) is an emerging class of analytical approaches that is aimed at studying engineered systems and assets through the synthesis of various data-driven and physicsbased models. DCE may also involve procedures for fusing sensor data from both experimental and operational systems. This paper introduces a DCE-based approach using data collected from experimental and operational railway structures (bridges and sleepers), which have been instrumented with advanced fibre optic sensors (FOS). This study provides a high-level definition of DCE approaches to studying instrumented infrastructure and discusses some of the challenges in implementing these methods in practice. An example of an implementation of a Gaussianprocess based DCE method is provided, with the goal of predicting the response of operational instrumented rail infrastructure (i.e. concrete sleepers) over time. Some practical issues of this implementation relate to the measurement systems, data acquisition rate, efficiently batch processing the data and accounting for uncertainty in the response predictions. In leveraging both the information gained from real-time measurement data and from traditional analytics or physics-based methods, a DCE-based modelling approach can provide unique a set information and insights into the operational performance of infrastructure.