Disturbance observer-based time-domain dynamic load identification

Dynamic load identification, as one of the inverse problems in structural dynamics, often suffers from ill-posedness when directly inverting transfer function matrices in traditional dynamic load identification approaches. Although numerous solutions have been proposed, most existing dynamic load identification methods still rely on matrix inversion. This leads to computational complexity, low accuracy, and failure to achieve real-time identification. We propose a novel time-domain identification method for arbitrary dynamic loads in time-domain inspired by the disturbance observer (DOB) of control theory. We innovatively treat dynamic loads as external disturbances of the structural dynamic system, and develop the DOB based on the state-space equation parameters of the system. In this way, the dynamic load identification is transformed into the estimation of external disturbances by using the DOB transformed from the structural dynamical equations. The convergence of the proposed method has been rigorously proven and its feasibility is validated through numerical simulations and experiments on a cantilever frame structure. With known structural parameters and vibration responses, the method achieves accurate identification of both single-point and multi-point arbitrary dynamic loads with good noise tolerance. In the simulation example, the normalized root mean square error (NRMSE) of the single-point dynamic load time history identification result is 4.946%, and the average NRMSE of the multi-point dynamic load identification result is 2.676%. The experimental results show that the NRMSE results of single-point and multi-point arbitrary dynamic loads on the cantilever frame structure are less than 10%.