Interval prediction of ultimate strength for laminated composite structures using back-propagation neural network

This paper proposes an interval prediction approach of ultimate strength for laminated structures. For the response prediction of complex structures, the presented method can ensure the necessary accuracy and greatly reduce the cost of computation. The method is different from the other traditional methods of prediction, which overcomes the limitation of perturbation methods in solving nonlinear problems, such as the Taylor series expansion, and avoids the enormous computation of the Monte Carlo simulation (MCS). In this paper, the output can be calculated by the finite element technology and mechanics of composite materials, and then the neural network (NN) is introduced to approximate the relation between the input and output, and finally the surrogate model is constructed. Subsequently, the issue of uncertainty propagation can be translated to optimal problem for extremum value. In addition, although the explicit expression of the established mapping relationship is unknown, the best fitness and worst fitness can be searched in the given bound of uncertain variables based on genetic algorithm (GA), and then achieve the upper bound and lower bound of the structural response depending on the capability of global search. After proposed technologies are given in detail, the engineering example of composite structure is presented and the results are discussed with common methods of uncertain propagation based on Taylor series expansion method and MCS, which demonstrates the validity and reasonability of the developed methodology.