Application of AI tools in creating datasets from a real data component for structural health monitoring

Abstract

Structural health monitoring (SHM) based on dynamic methods is becoming a widely applied method, mainly due to its high accuracy combined with the fact that it is not necessary to limit service activities during monitoring. In order to accurately identify the vibration characteristics of a complex structure, such as frequency, mode shape, and damping ratio, it is necessary to arrange a dense network of acceleration sensors, which might be a challenge due to onsite conditions. Most sensors are fixed and can only be used for a single building during their lifetime. This may result in a waste of resources if there are not enough sensors for a specific structure or even if their layout is inefficient, thus resulting in insufficient or missing data. To overcome this, a new approach based on the application of artificial intelligence (AI) may be considered. Specifically, an artificial neural network (ANN) may be used to generate missing data to determined areas from the position of one or more fixed measurement points. For that case, an ANN model should be trained and tested on a number of projects to ensure accuracy during operation. The results show that the data generated is accurate and the data storage capacity is optimized. A major benefit is that a large number of sensors can be removed from the building to serve other purposes, optimizing the costs for SHM. This chapter presents the approach and a case study that was carried out on a cable-stayed bridge in Vietnam. The obtained results show the potential of applying AI in creating virtual data, serving larger goals in SHM.