In this paper, experimental and numerical investigations on cord–elastomer composites are presented. A finite-element model is introduced, which was developed within the framework of an industrial project. The model is able to simulate an elastomer matrix with inserted cords as load bearing elements and to predict the strains and stresses in cord and elastomer sections. The inelastic material behavior of the elastomer matrix and the yarns is described by corresponding material models suitable for large deformation processes. With the help of a specially developed demonstrator bellows, which is similar to an air spring, the simulation results are compared with experiments. For this purpose, the digital image correlation method is used to determine the deformations on the outer surface of the demonstrator bellows and to calculate the strains on and between the cords. The comparison of the results shows that the employed simulation method is very well suited to predict the strains in these cord–elastomer composites.

A decisive disadvantage of high dimensional optimization problems is the large number of parameters which make the identification process difficult. Many of this are reflected in the mathematical properties of the objective function used for parameter identification. Ideally, the objective function has a simple, paraboloid-like form with a single minimum whose location corresponds to the desired parameter set. Instead, unfavorable "landscape forms" occur, for example "bumpy" areas with several local minima. In literature, the dominant methods are attempt to simplify high-dimensional functions, or to research new, mostly stochastic optimization methods. This article examines a way to obtain information about the landscape of the objective function underlying an optimization problem. A path search algorithm based on the Dijkstra algorithm is presented. The purpose of the algorithm is to find the approximately deepest path between two local minima, which allows to obtain information about the characteristics of the objective function. It is tested on the Himmelblau and Rastrigin function and allows to examine the landscape between two local minima with respect to their topographic characteristics.

In this paper we demonstrate the application of the deepest‐path algorithm for better understanding of objective function landscapes resulting from the optimization process. For this purpose, the sensitivities of the Yeoh and Ogden material model parameters are compared for different load cases. This analysis shows a much higher variation of the material parameters for the Ogden model than for the Yeoh model at approximately constant objective function values. The reasons for this may be local minima or shallow gradients in the objective function landscape. Afterwards, the deepest‐path algorithm is performed between selected designs from the sensitivity analysis. It can be seen that the deepest‐path algorithm provides further information about local minima in the objective function landscape, which are not clear identifiable from a sensitivity analysis.

No abstract

No abstract