New paper published in Reliability Engineering & System Safety
S. Schär, B. Sudret and S. Marelli published a new paper about functional autoregressive models
Our latest paper, "Surrogate modeling with functional nonlinear autoregressive models (F-NARX)", published in Reliability Engineering and System Safety, introduces a novel approach to the emulation of dynamical systems. F-NARX moves beyond traditional discrete-time NARX models by focusing on functional temporal features of inputs and outputs instead of individual time step values. This functional view addresses key challenges of classical NARX models: it improves prediction stability, reduces over-reliance on past autoregressive predictions, efficiently handles systems with long memory and makes the model robust to varying time discretization.
Demonstrated on an eight-story building under wind load and a steel frame under seismic excitation, F-NARX accurately predicts complex system responses over extended time periods at low computational cost. This work therefore provides a powerful and practical tool for surrogate modeling, particularly in engineering applications.
Please visit external page this link to get to the publication.