Intelligent Automatic Operational Modal Analysis: Application to a Tall Building

The stochastic subspace identification (SSI) technique is widely adopted for operational modal analysis of structural systems. Regardless of the specific implementation, its workflow basically consists of four main steps: definition of the control parameters governing the SSI algorithm; estimation of the system poles and construction of the stabilization diagram (SD); interpretation of the SD; quantification of the confidence level associated to the modal results. The definition of the control parameters ruling the modal identification via SSI algorithms play a crucial role. Manual selection procedures and rules-of-thumb have been largely employed for this task, but they are unsuitable for reliable automatic applications. Therefore, a new paradigm for automatic operational modal analysis is presented in this work. The proposed approach is named intelligent automatic operational modal analysis (i-AOMA) method and relies on the effective integration between the SSI algorithm and a machine learning technique. Initially, quasi-random samples of the control parameters for the SSI algorithm are generated. Once the SSI algorithm is performed for each sample, the corresponding SDs are processed to prepare a database for training the intelligent core of the i-AOMA method. This is a machine learning technique that drives intelligently the selection of the control parameters for the next applications of the SSI algorithm, which is performed until a convergence criterion is fulfilled. At the end, the uncertainty level about the modal estimates due to the variability of the control parameters is assessed. The potential of the proposed approach is demonstrated through the identification of a large structure