Summary
System identification (SI) techniques can be used to identify the dynamic parameters of mechanical systems and civil infrastructures. The aim is to rapidly and consistently model the object of interest, in a quantitative and principled manner. This is also useful in establishing the capacity of a structure to serve its purpose, thus as a tool for structural health monitoring (SHM). In this context, input–output SI techniques allow precise and robust identification regardless of the actual input. However, one of the most popular and widely used approaches, the Rational Fraction Polynomial (RFP) method, has several drawbacks. The fitting problem is nonlinear and generally non‐convex, with many local minima; even if linearised via weighting, it can become severely ill‐conditioned. Here, a novel proposal for the broadband macro‐modelling of structures in the frequency domain with several output and/or input channels is presented. A variant of the vector fitting approach, the Fast Relaxed Vector Fitting (FRVF), applied so far in the literature only for the identification of electrical circuits, is translated and adapted to serve as a technique for structural SI and compared with other traditional techniques. A study about the robustness of the FRVF method with respect to noise is carried out on a numerical system. Finally, the method is applied to two experimental case studies: a scaled model of a high‐aspect‐ratio (HAR) wing and the well known benchmark problem of the three‐storey frame of Los Alamos laboratories. Promising results were achieved in terms of accuracy and computational performance.