Seismic attenuation relationship with homogeneous and heterogeneous prediction-error variance models

摘要: Peak ground acceleration (PGA) estimation is an important task in earthquake engineering practice. One of the most well-known models Boore-Joyner-Fumal formula, which estimates PGA using moment magnitude, site-to-fault distance and site foundation properties. In present study, complexity for this formula homogeneity assumption prediction-error variance are investigated efficiency-robustness balanced proposed. For purpose, a reduced-order Monte Carlo simulation algorithm Bayesian model class selection presented to obtain suitable predictive seismic attenuation relationship. approach, each (a with model) evaluated according its plausibility given data. The one highest robust since it possesses optimal balance between data fitting capability sensitivity noise. A database strong motion records Tangshan region China obtained from Earthquake Data Center analysis. proposed based on database. It shown that heterogeneous much simpler than suggested by Boore, Joyner Fumal (1993).