Experimental verification of nonlinear static neuro-finite element method using a planar truss model

In this paper a small truss model is tested for the purpose of verification of the “neuro-finite element method” which is a matrix and finite element method recently proposed by the second author for nonlinear static analysis of structures using neural network capabilities. The method has been tested numerically for structures assembled by one-dimensional elements such as trusses or frames under static loading successfully. The truss tested in this paper has 5 nodes and 8 members and is 1 degree indeterminate. The truss is loaded until one of its members buckles in compression so that the structure becomes unstable and collapses. The method is numerically applied too, and the results from the simulation are compared with tests. The results show the method and tests are in perfect comparison until the load drops in a result of the buckling. The results are discussed in detail.