Machine learning-based model for prediction of optimum TMD parameters in time-domain history

In this study intended for optimum design of tuned mass dampers (TMDs), which is one of the passive control systems, used with the aim of protection, and even retrofitting structures seismically, a hybrid approach, where metaheuristic methods were combined with machine learning technology, was presented to carry out the mentioned aim. With this respect, to obtain the mentioned TMD designs for a single degree of freedom systems, optimization analyses based on the dynamic design process were carried out with a metaheuristic method. The second step is also to develop a machine learning-based prediction model, and it was provided that the ensured optimum parameters were processed via artificial neural networks (ANNs), and the model was trained in this scope. Moreover, the performance, reliability and convergence success of the prediction model were measured with some error metrics, too. By this means, it also became possible that the optimum parameters were determined concerning different structure designs in a shorter time, rapidly in an effective way. Additionally, by using optimal results predicted via ANNs-based model, some formulations were developed that can calculate the optimum TMD damping and frequency ratios directly, and their validity was controlled on both single and multiple degrees of freedom structures.