Wavelet-Neural Network Based Robust Optimization of Self-Centering Viscous Damper for Steel MR Frame

ABSTRACT

This study implements robust design optimization (RDO) of a self-centering viscous damper in the presence of seismic uncertainty. The RDO minimizes the effect of random input parameters on the output response, thus providing a robust solution. The optimal phenomenological parameters of the self-centering viscous damper are evaluated to ensure its robustness. The objective function of the RDO is cast as a sum of the mean and standard deviation of maximum floor acceleration and a constraint function of the inter-storey drift ratio of the floors. The mean component of the objective function minimizes the structural responses due to external loads, whereas the other component reduces the propagation of uncertainty from the input parameters. To reduce the computational cost of the RDO, a wavelet-based neural network is used as a surrogate model. The proposed optimization method is implemented in a 16-storey steel moment resisting frame with consideration of two ensembles of ground motions, i.e. far-field and near-field records, which are taken from FEMA P695 (ATC-63). To highlight the effectiveness of the proposed self-centering damper in mitigating maximum and residual drift ratios, two additional optimizations are undertaken using viscous and self-centering dampers. The numerical study shows that the neural network with a Gaussian wavelet function produces better accuracy compared to other wavelet functions. Also, the numerical analysis clearly elucidates the advantages of the proposed damper in reducing the residual drift of the floors over a viscous damper and a self-centering damper, which may lead to minimizing the post-earthquake economic losses.

KEYWORDS:

• Self-centering viscous damper
• residual drift
• far-field and near-field earthquake
• robust design optimization
• wavelet-based neural network

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This study was funded by Natural Sciences and Engineering Research Council of Canada under the Discovery Grant programs (RGPIN-2019-05584).