Optimization of tuned mass dampers parameters using Artificial Neural Networks

作者：D. K. Monteiro, D. B. M. Matos, Patrick O. B. Costa, L. F. F. Miguel · 发表于：Ibero-Latin American Congress on Computational Methods in Engineering (CILAMCE) · 年份：2024 · DOI：10.55592/cilamce.v6i06.10268

Control systems are usually employed to minimize the dynamic effects on structures. These controllers must be designed according to the physical and geometric characteristics of the system and the environmental loads to which the structure will be subjected. Passive control is the most suitable for extreme events because it does not depend on an external power source to function properly. The use of a linear attenuator, such as the tuned mass damper (TMD), allows for dynamic analysis in the frequency domain, which provides a reduction in the computational cost of the analysis. Among the several methods used to determine the optimal parameters of the damper, analytical methods in their closed form can be applied to simpler structures. In structures with several degrees of freedom or complex geometry, the optimal TMD parameters, as well as its optimal position, can be acquired by methods based on stochastic optimization. This kind of problem is commonly solved in the state-of-the-art by Metaheuristic algorithms. However, in general, these procedures require high computational costs, which makes Artificial Neural Networks (ANN) a viable alternative to minimize the processing time. Thus, this paper proposed a methodology based on ANN in comparison with the Circle-Inspired Optimization Algorithm (CIOA) to determine the optimal parameter of tuned mass dampers, i.e., the damper period. Training of the neural networks was performed with a dataset of optimal TMD values, which were obt...