The Application Of Intelligent Algorithm With Artificial Neural Network In Structure Optimization

Structural optimal design is a kind of discrete variable optimization problem. Because the displacement, stress and other constraints are usually considered, sometimes it is difficult to find the global optimal solution when using the traditional optimization algorithm. Through the API provided by the FEM software and the intelligent algorithm, this article carried out studies on the method of the optimal design for truss or frame structures, the major work and innovations are as follows:(1) Using the artificial neural network technology and the intelligent optimization algorithms, a more efficient routine was presented to find the solution to the structural optimal design problem. Firstly, some special training samples of structure are analyzed by the FEM software, and then, the results of nodal displacement are used to train the BP network. After that, in the evolution of a GA, re-analyzing of the structure is no longer required, and the structural displacements are predicted by the trained BP network. This procedure requires much less computing time than the original GA. Through the univariate and multivariate optimization design with a two-dimensional plane frame structure with 24 floors, the validation of presented optimization method was verified the result that it is suitable for the multivariable optimization problem. Some analysis results are also presented, including a 2D truss structure with 17 members and two 3D truss structure with 42 members and 72 members, respectively. About 80% of computing time is saved by applying the proposed method.(2) In order to improve the optimization convergence speed of truss structure, Matlab distributed cluster technique is employed in this paper. This technique supports analyzing the SAP2000 FEM models on two computers simultaneously, and consequently, the computing time for optimization iteration is saved. Compared with optimization results of the three truss cases which considering displacement,displacement and stress constraints, the results show that Matlab distributed cluster decreases more than 50% computing time compared with GA running on a single computer. Moreover, adding constraint conditions does not influence the convergence rate.Therefore, the method has a good applicability. Furthermore, the mathematical model of structural optimization with discrete variable is established for the example, the design variable is selectable in the stationary discrete and continuous interval.The results show that the iterative optimization process keeps stable convergence and maintains a high operational efficiency, so these method can also be applied to optimization problems with discrete variables.