SHM approach with indirect load measurements and Kalman update of state and model parameters

The present contribution is dedicated to the development of advanced approaches for structural health monitoring (SHM) of civil engineering structures. The approach consists of two principal parts, a continuous load monitoring system (LMS) and a continuous state and condition monitoring system based on Kalman Filter update. The overall approach can serve as a kind of Digital Twin for SHM. The measurement data assimilation with a physically based structural model by use of Kalman Filters can provide reliable state predictions if the actual load history is known. To overcome this problem, the present work proposes a two-step update procedure. The first step contains a cluster of Convolutional Neural Networks for the traffic load identification, including its magnitude, mixture, position and velocity. The obtained numerical results show a very good correlation between predicted and reference data. In the second step the identified loads are used as input for the Kalman Filter updates. The results of a second numerical example show that the model parameters like stiffness, damping and mass can be continuously predicted even in presence of sudden changes and permanent uncertainties.