Validation of seismic response analyses using seismometer array recordings
A framework for the validation of computational models used to predict seismic response based on observations from seismometer arrays is presented. The framework explicitly accounts for the epistemic uncertainty related to the unknown characteristics of the ‘site’ (i.e. the problem under consideration) and constitutive model parameters. A mathematical framework which makes use of multiple prediction-observation pairs is used to improve the statistical significance of inferences regarding the accuracy and precision of the computational methodology and constitutive model. The benefits of such a formal validation framework include: (i) development of consistent methods for determination of constitutive model parameters; (ii) rigorous, objective and unbiased assessment of the validity of various constitutive models and computational methodologies for various problem types and ground motion intensities; and (iii) an improved understanding of the uncertainties in computational model assumptions, constitutive models and their parameters, relative to other seismic response uncertainties such as ground motion variability.