A modified particle filter for parameter identification with unknown inputs

Abstract

Particle filter (PF) is usually used for identifying structural parameters in nonlinear systems. However, the traditional PF method requires that all the external excitations are available or can be measured, which may not be the case for many structures. In this paper, a modified PF method with unknown inputs, referred to as PF-UI, is proposed to identify the augmented states, including structural states and parameters, as well as the unknown excitation inputs. The augmented states are identified by the traditional PF algorithm, and the unknown excitations are simultaneous identified with the least-square algorithm. Such an analytical solution for PF-UI is not available in the previous literature. Numerical studies of two typical nonlinear systems are conducted to demonstrate that the proposed approach is capable of identifying the states and parameters with unknown excitation inputs.