Prediction of the Dynamic Behavior of Beams in Similitude Using Machine Learning Methods

The dynamic behavior of structures can be investigated using concepts of complete (exact) and incomplete (distorted) similitudes.The incompleteness is much more of interest since the complete similitudes are difficult to be achieved and the experiments are often executed using distorted models as test articles. In this work, beams in similitude have been investigated using machine learning to establish degrees of correlation between similar systems, without invoking governing equations and/or solution schemes. Machine learning is based on algorithms that derive models from sample inputs providing data-driven predictions. The absence of an explicit algorithm, being the process totally data-driven, confers to the approach a high versatility which allows its application even in the vibroacoustic research fields and problems. In view to validate the machine learning predictions, numerical investigation of beams in similitude has been performed. The good predictions obtained with machine learning highlight the potentialities of these algorithms and open the way to analyses with more complex structures.