Casaburo, Alessandro (2021) An investigation on the vibroacoustic behavior of systems in similitude. [Tesi di dottorato]


Download (9MB) | Preview
[error in script] [error in script]
Item Type: Tesi di dottorato
Lingua: English
Title: An investigation on the vibroacoustic behavior of systems in similitude
Date: 8 July 2021
Number of Pages: 240
Institution: Università degli Studi di Napoli Federico II
Department: Ingegneria Industriale
Dottorato: Ingegneria industriale
Ciclo di dottorato: 33
Coordinatore del Corso di dottorato:
Franco, FrancescoUNSPECIFIED
Petrone, GiuseppeUNSPECIFIED
Date: 8 July 2021
Number of Pages: 240
Uncontrolled Keywords: Similitude, Machine learning, Structural dynamics
Settori scientifico-disciplinari del MIUR: Area 09 - Ingegneria industriale e dell'informazione > ING-IND/04 - Costruzioni e strutture aerospaziali
Date Deposited: 20 Jul 2021 13:41
Last Modified: 07 Jun 2023 10:40


Similitude theory allows engineers to establish the necessary conditions to design a scaled - up or down - model of a full-scale prototype structure. In recent years, the research on similitude methods, which allow to design the models and establish similitude conditions and scaling laws, has grown so that many obstacles associated with full-scale testing, such as cost and setup, may be overcome. This thesis aims at, on the one hand, expanding the possibilities of similitude methods by means of their application to new structural configurations; on the other hand, at the investigation of new approaches. Therefore, similitude conditions and scaling laws of thin aluminium plates with clamped-free-clamped-free boundary conditions, first, and aluminium foam sandwich plates with simply supported and free-free boundary conditions, then, are derived. Particularly, two sets of conditions are derived for the sandwich plates: the first by expliciting all the geometrical and material properties, the second by combining some parameters into just one with physical meaning, that is, the bending stiffness. These conditions and laws are successively validated by means of dynamic experimental tests, in which reconstructions of the natural frequencies and the velocity response of the prototype are attempted. Also the prediction of the radiated acoustic power is performed for the sandwich plates. All the tests highlight that these laws do not work fine when the models are distorted, i.e., when the similitude conditions are not satisfied. Therefore, the potentialities of machine learning are investigated and used to establish degrees of correlation between similar systems, without invoking governing equations and/or solution schemes. In particular, artificial neural networks are used in order to predict the dynamic characteristics, first, and the scaling parameters, then, of beams, as test (since they do not exhibit distorted models), and plates. In the latter case, the predictions of the artificial neural networks are validated by the results provided by the experimental tests. The networks prove to be robust to noise, very helpful in predicting the response characteristics, and identifying the model type. Finally, the similitude methods are used as a tool for supporting, and eventually validating, noisy experimental measurements, not for predicting the prototype behavior. In this way, they can help to understand if a set of measurements is reliable or not. Therefore, the sandwich plates are analysed with digital image correlation cameras. Then, with the help of an algorithm for blind source separation, the force spectra and velocity responses are reconstructed. It is demonstrated that the similitude results are coherent with the quality of the experimental measurements, since the curves overlap when the spatial patterns are recognizable. Instead, when the displacement field is too polluted by noise, the reconstruction exhibits discrepancies. This proves that the application of similitude methods should not be underestimated, especially in the light of the expanding range of approaches which can extract important information from noisy observations.


Downloads per month over past year

Actions (login required)

View Item View Item