A PREDICTION OF GROUND-MOTION ACCELERATION BY ADAPTIVE NEURO-FUZZY INFERENCE SYSTEMS: AN EXAMPLE BASED ON THE NGA-WEST 2 DATA

Abstract eng:
Adaptive Neuro Fuzzy Inference System (ANFIS) is used here to obtain the robust Ground-Motion Prediction Model (GMPM). Avoiding a priori functional form., ANFIS provides fully data-driven predictive models. A large subset of the NGA-West 2 database is used, including 2335 records from 580 sites and 140 earthquakes. Only shallow earthquakes and recordings corresponding to stations with measured average shear- wave velocity between 0 and 30-meters depth (Vs30) properties are selected. Three basics input parameters are choice: the moment magnitude (Mw), the Joyner-Boore distance (RJB), and Vs30. The ANFIS model output is the Peak Ground Acceleration (PGA), Peak Ground Velocity (PGV) and 5%-damped Pseudo-Spectral Acceleration (PSA) at periods from 0.01 to 4 s. A procedure similar to the random-effects approach is developed to provide between and within event standard deviations. The total standard deviation (σ) varies between 0.3055 and 0.3586 (log10 unit) depending on the period. The ground motion predictions resulting from such simple, three explanatory variables, ANFIS models, are shown to be comparable to most recent NGA results (e.g. Boore et al. 2014 and Derras et al. 2016). The main advantage of ANFIS compared to ANN is its simple and one-off topology: five layers. Our results exhibit a number of physically sound features: magnitude scaling of the distance dependency, near-fault saturation distance increasing with magnitude and amplification on soft soils. The ability to implement ANFIS model using an analytic equation is demonstration.

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