Neuronal-Attenuation Laws for the Mexican Subduction: An Updating Effort

Abstract eng:
The most recent shaking experiences have demonstrated that the predictions of the seismic models are not always in agree with the registered responses. A deep examination of the current subduction attenuation laws for PGA (peak ground acceleration) has pointed out that most of them uses older information than fifteen years and the functions are not taking into consideration the latest source-station configurations and the aging of materials. In this paper a neural network NN that permits to estimate PGA (vertical, east-west and north-south components) via the magnitude M, the focal depth FD and the epicentral distance ED (as classes and numerical parameters), is presented. For constructing this renewed attenuation law, 1270 records collected from 1960 to 2015 at rock-like sites are considered. The obtained results show that calculated PGAs using the neuronal model are remarkably close to those recorded. The proposed attenuation curves are compared with Ground Motion Prediction Equations (GMPEs) using events from México, Japan, Chile and USA. This evaluation raises the question of regional dependence of ground-motion which is a highly debated issue. The results also show that the NN performs considerably better than the traditional equations so it could be considered as a good alternative in seismic hazard assessment.

• Export as: BibTeX | MARC | MARCXML | DC | EndNote | NLM | RefWorks
• Add to your basket