Artificial Intelligence-Based Prediction Models for Optimal Design of Tuned Mass Dampers in Damped Structures Subjected to Different Excitations

Abstract

Tuned mass damper (TMD) is a type of energy absorbers that can mitigate the vibrations of the main system if its frequency and damping ratios are well adjusted. By adopting simple assumptions on the structure and loadings, many analytical and empirical relationships have been presented for the estimation of the parameters for TMDs. In this research, methods based on the artificial intelligence (AI) techniques are proposed for optimal tuning of the TMD parameters of the main damped-structure for three kinds of loadings: white-noise base acceleration, external white-noise force, and harmonic base acceleration. For this purpose, a dataset using the cuckoo search (CS) optimization algorithm is created. The performance of the proposed methods based on the radial basis function (RBF) neural network, feed-forward neural network (FFNN), adaptive neuro-fuzzy inference system (ANFIS), and random forest (RF) techniques are evaluated by some statistical indicators. The results show the proper performance of these methods for the optimal estimation of the TMD parameters. Overall, the ANFIS method results in best matching with the observed dataset. Moreover, the simulation results indicate that the TMD’s optimal frequency ratio is reduced, while its optimal damping ratio is increased, against the increase in the TMD mass ratio of the main structure subjected to harmonic base acceleration. This trend with a less slope is observed for the optimal frequency ratio of the TMD in the main structure subjected to external white-noise force; however, the optimal damping ratio of the TMD is independent of its mass ratio in this case. Similar results are obtained for the main structure subjected to white-noise base acceleration.