Development of a prediction method of Rayleigh damping coefficients for free layer damping coatings through machine learning algorithms


Yilmaz I., Arslan E., Kiziltas E. C. , ÇAVDAR K.

INTERNATIONAL JOURNAL OF MECHANICAL SCIENCES, vol.166, 2020 (SCI-Expanded) identifier identifier

  • Publication Type: Article / Article
  • Volume: 166
  • Publication Date: 2020
  • Doi Number: 10.1016/j.ijmecsci.2019.105237
  • Journal Name: INTERNATIONAL JOURNAL OF MECHANICAL SCIENCES
  • Journal Indexes: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Academic Search Premier, Aerospace Database, Communication Abstracts, Compendex, INSPEC, Metadex, zbMATH, Civil Engineering Abstracts
  • Keywords: Rayleigh damping, Viscous layers, Machine learning, Free layer damping, SUPPORT VECTOR MACHINE, MECHANICAL-PROPERTIES, VISCOELASTIC MATERIALS, TOPOLOGY OPTIMIZATION, FINITE-ELEMENTS, VIBRATION, MODEL, COMPOSITE, PROPERTY, ENSEMBLE
  • Bursa Uludag University Affiliated: Yes

Abstract

Application of damping coatings on metal sheets is a commonly used method to suppress the undesirable vibration and noise levels in various industries. As numerical simulations have a vital role while designing a high-quality product with fewer costs, an accurate and practical way of modelling such type of structures is necessary. It was aimed to develop a methodology that helps to define damping parameters of such viscoelastic coating layers through Rayleigh damping coefficients. Machine learning tools were considered to find a prediction formula which yields Rayleigh coefficients based on thicknesses. For this purpose, several tests were conducted with different coating thicknesses on steel plates. In parallel, a great number of simulations were performed not only to make comparisons with the reference values from tests but also to feed the learning algorithms with the data sets. The results were compared including the ones from the Oberst equation. The results from the machine learning showed significantly better matching performance with the tests, as there seems to be a limitation problem for Oberst accuracy.