An improved NExT-DMD for efficient automated operational modal analysis

The demand for identifying modal parameters under random excitation is rapidly increasing, but improving the efficiency of automated operational modal analysis (AOMA) remains a significant challenge, particularly during the time-consuming system identification stage. We propose an efficient AOMA method based on the integration of the natural excitation technique (NExT) with dynamic mode decomposition (DMD), termed as NExT-DMD, coupled with the density-based spatial clustering of applications with noise (DBSCAN). This improved NExT-DMD overcomes the reliance of original DMD on free-decay responses and allows operational modal analysis, i.e. modal parameter identification from random excitations. A systematic hyperparameter optimization strategy for DBSCAN is developed based on rank stability, facilitating automated identification. A numerical composite wing model with closely spaced modes was used to validate the proposed method, demonstrating a maximum frequency difference of 3.43 % against the finite element method. With the responses captured by the non-contact three-dimensional optical technique, a physical wing model was tested to show the capability of the proposed method to deal with real-world complex structures. The results demonstrate that our method can improve the identification efficiency by 41.10 % compared with the NExT-eigensystem realization algorithm (ERA) and by 53.83 % compared with covariance-driven stochastic subspace identification (Cov-SSI). The efficient NExT-DMD AOMA method is promising for operational modal analysis with large datasets.