Online Data-Driven Prediction of Spatio-Temporal System Behavior Using High-Fidelity Simulations and Sparse Sensor Measurements

Predicting the behavior or response for complicated dynamical systems during their operation may require high-fidelity and computationally costly simulations. Because of the high computational cost, such simulations are generally done offline. The offline simulation data can then be combined with sensors measurement data for online, operational prediction of the system's behavior. In this paper, a generic online data-driven approach is proposed for the prediction of spatio-temporal behavior of dynamical systems using their simulation data combined with sparse, noisy sensors measurement data. The approach relies on an offline–online approach and is based on an integration of dimension reduction, surrogate modeling, and data assimilation techniques. A step-by-step application of the proposed approach is demonstrated by a simple numerical example. The performance of the approach is also evaluated by a case study which involves predicting aeroelastic response of a joined-wing aircraft in which sensors are sparsely placed on its wing. Through this case study, it is shown that the results obtained from the proposed spatio-temporal prediction technique have comparable accuracy to those from the high-fidelity simulation, while at the same time significant reduction in computational expense is achieved. It is also shown that, for the case study, the proposed approach has a prediction accuracy that is relatively robust to the sensors’ locations.