Stochastic dynamic analysis of twisted functionally graded plates. (15th August 2018)

• Record Type: Journal Article
• Authors: Karsh, P.K.; Mukhopadhyay, T.; Dey, S.
• Is Part Of: Composites. Number 147(2018)
• Page Start: 259
• Page End: 278
• Publication Date: 2018-08-15
• Subjects: Stochastic dynamics -- Twisted functionally graded plate -- Artificial neural network -- Polynomial neural network -- Finite element method -- Sensitivity analysis; Composite materials -- Periodicals; Materials science -- Periodicals; Composite materials; Periodicals; Electronic journals; 620.118
• DOI: 10.1016/j.compositesb.2018.03.043
• Languages: English
• ISSNs: 1359-8368

Abstract

Abstract: This paper presents a stochastic dynamic analysis of functionally graded plates by following an efficient neural network based approach coupled with the finite element method. An isoparametric quadratic element having eight nodes is considered for the finite element analysis of pre-twisted functionally graded cantilever plates subjected to variation in geometric parameters, material properties and temperature. Both individual and compound effects of stochasticity in the uncertain input parameters are accounted to quantify their influence on the first three natural frequencies, mode shapes, and frequency response functions of functionally graded plates. A sensitivity analysis is conducted to ascertain the relative effects of various prospective sources of uncertainty. Latin hypercube sampling is utilised to train the efficient surrogate models, which are employed as a medium of uncertainty propagation. The comparative performance of artificial neural network and polynomial neural network is assessed in the stochastic dynamic analysis of the pre-twisted functionally graded plates from the viewpoint of accuracy and computational efficiency. The results are validated with respect to direct Monte Carlo simulation based on the finite element model of the functionally graded plates. It is observed that the artificial neural network based algorithm can achieve a significant level of computational efficiency without compromising the accuracy of results. The results presented in this article reveal that the source uncertainties of functionally graded plates have a significant effect on the dynamic responses of the structure.