New feature based on influence lines for anomaly detection in bridge monitoring

Bridge monitoring using displacement, strain, acceleration or rotation measure-ments during normal operation can be a good supplement to conventional manual bridge inspections. In order to detect damage from the measured signal, features must be extracted from the signal that are primarily damage-sensitive. In recent years, features based on the influ-ence lines have been increasingly investigated. The influence line represents the simplified signal that can be measured with a sensor at a fixed position when a single vehicle crosses the bridge. In this paper, we present the new feature normalised integrated influence lines (niil). We show that niil is sensitive to damage, independent of vehicle geometry, velocity, sampling rate, devi-ation from ideal lane and robust against noise using the analytical bending beam and strain data from a real bridge that has been successively damaged. niil can be calculated based on the curvature, rotation or displacement influence lines.