Adaptive momentum-based optimization to train deep neural network for computer simulation of hygro-thermo-mechanical vibration performance of laminated plates

Abstract

This article explores the vibration response of the hybrid composite structure with the aid of the adaptively tuned deep neural network (DNN). In order to find the features of the design-points, the semi-numerical technique is applied to the governing differential equations of the system acquired on the foundation of the kinematic theory with refined higher order terms. Considering higher order terms made the accuracy of this analysis suitable for moderately thick plates as well as thin ones. DNN is trained for vibrational characteristics of the design-points by employing adaptive learning rate method as the high-speed optimizer. Accuracy of the semi-numerical method (used for determining the design-points) is evaluated through the comparison with the results reported in the published studies. While, the validity of the DNN-based model in predicting the response of the system at design-points is tested and confirmed by analyzing the trend of mean squared error.

Keywords:

• Multiscale hybrid angle-ply laminated composite
• rectangular plate
• Halpin-Tsai homogenization
• hygrothermal environment
• deep neural network

Acknowledgments

This paper is supported by Research Foundation of the Nanchang Normal University for Doctors (NSBSJJ2018014).