Options
2015
Conference Paper
Titel
Use of swarm intelligence for topology optimization of truss structures with stochastic loading conditions
Alternative
Einsatz von Schwarmintelligenz zur Topologieoptimierung im Fachwerk unter stochatischen Lastannahmen
Abstract
The continuously growing variety of cars as well as increasingly shorter development periods necessitate new intelligent methods for supporting the planning and design process of production equipment, which is still based on experience and very time consuming at the moment. Moreover, there exist lots of change loops, because the calculation of the stiffness takes place after the design and construction process. Therefore, it is the objective to develop a planning-support software tool for an automatic design of an optimal truss structure. Here we understand a ""lightweight"" and ""stiff"" structure as optimal. So we have a multiobjective optimization problem with lots of pareto-optimal solutions. In addition to the optimization, fast stiffness calculations are performed using the beam theory. Usually, the finite element method used deterministic loading conditions. However in reality, it is often not possible to predict the ""real"" loading conditions. Therefore we used normally distributed random numbers for the forces, whereby the absolute value and angle of the force can vary. We develop a two-step approach for building an optimal truss structure using swarm intelligent techniques. Starting with a bionic preprocessing, ant colony optimization is used to detect suitable areas for truss structure in a given design space. The ACO is searching for a short path through all clamping and connection points. Also the stiffness of the truss structure is relevant. That means, after a path has been found, we calculate the stiffness and the value has a significant influence on the evaluation of the path. Thus, longer stiffer path are rated better than shorter unstable ones. Starting with the basic truss structure from preprocessing (ACO), the particle swarm optimization is used to determine the optimal node position for the remaining truss elements. The user can define some bounding boxes for the nodes of the basic truss structure. The nodes can move freely inside these bounds. For this problem, the PSO is an appropriate procedure. To deal with the several objectives, a non-dominated sorting method for the PSO is used to achieve a uniform distribution of the Pareto front. The optimal solution can be loaded into CATIA-V5 as parametric CAD model. Through the integrated calculation of the stiffness, the quality of the process is increased. The use of the methods for other applications (crane and bridge construction) is investigated.