Data‐driven nonlinear autoregressive with external input model‐based compensation for real‐time testing

Abstract

Actuator control plays an important role for real-time testing based experimental techniques such as real-time hybrid simulation (RTHS). Accurate modeling of actuator dynamics can be challenging due to the complexity and nonlinearity of the servo-hydraulic system. In this study, nonlinear autoregressive with external input (NARX) modeling is introduced to emulate the servo-hydraulic dynamics. The command and measured displacements of the actuator are used as input and output of the NARX model. The coefficients of the NARX model of different orders are determined online through ordinary least square regression in a real-time manner. Data weight is further proposed to account for most recent variation in command and measured displacements. Real-time tests with predefined random and chirp signals are conducted to evaluate the performance of proposed NARX model-based compensation of different orders, window length, and data weight. The efficacy and robustness of proposed NARX model-based compensation are further verified through computational simulation of a RTHS benchmark model. Both numerical simulation and laboratory experiments demonstrate that the proposed method enables effective negation of servo-hydraulic dynamics for real-time testing.