Inverse design of composite xylophone beams using finite element-based machine learning

Abstract

Wood, as one of the important materials to make music instruments, has many drawbacks, including its sensitivity to changes in temperature and humidity, the lack of available wood resources, the significant variability of wood, and the demanding level of expertise needed. New materials such as composites are needed to be developed to be substitutes for woods. In this study, a finite element (FE)-based machine learning method is proposed to design a xylophone using sandwich beam as a substitute to wood. A finite element (FE) procedure based on a higher order layer-wise beam theory is developed. In addition, a machine learning model is developed to predict the natural frequencies of sandwich beams. The model is designed and trained to consider the results obtained from developed FE procedure as input and predict accurate natural frequencies. The results recorded from this model are compared with the experimental values. Then, inverse analysis is performed to design sandwich beam of different geometric sizes for constrained natural frequencies using machine learning.