Description
This contribution illustrates a new paradigm for the automatic output-only modal identification of linear structures under ambient vibrations, namely the intelligent automatic operational modal analysis (i-AOMA). The proposed approach relies on the covariance-based stochastic subspace identification (SSI-cov) method whereas a machine learning technique is implemented to automatically tune its control parameters. Two distinctive phases can be recognized in the proposed approach. First, quasi-random samples of the control parameters for the SSI-cov method are generated. Once the SSI-cov method is performed for each sample, the corresponding stabilization diagrams are processed to prepare a database that will be employed to train the intelligent core of the i-AOMA method. This is a random forest algorithm that predicts which combination of the control parameters for the SSI-cov method can provide good modal estimates. New quasi-random samples of the control parameters for the SSI-cov method are generated next, until a statistical convergence criterion is fulfilled. If the generic sample is classified as feasible by the intelligent core of the i-AOMA method, then the SSI-cov method is performed. Hence, stable modal results are carried out from the stabilization diagrams and relevant statistics are also calculated to evaluate the uncertainty due to the variability of the control parameters. Some applications are finally reported, and the links to freely download the corresponding Python code are provided.
