关键词: 渗流, 参数反演, 贝叶斯理论, 熵-盲数, 差分进化自适应Metropolis算法

Abstract: The existing parameter inversion methods based on Bayesian theory consider only the randomness in inversion process, neglecting the grey and unascertained uncertainties in seepage parameters. This paper presents an inversion method for estimating seepage parameters based on Bayesian theory and entropy-blind numbers. This new method introduces the entropy-blind number theory into the inversion model, taking the parameters to be inverted as blind numbers and fully considering the uncertainties in inversion process. We use a differential evolution adaptive metropolis (DREAM) algorithm to derive the posterior distribution of seepage parameters, and adopt a response surface model to replace the seepage simulation forward model, thereby avoiding a large number of forward model simulations in the traditional Bayesian seepage parameter inversion method. Our new method and its accuracy are validated through example analysis and a case study of permeability coefficient inversion for the foundation of a concrete gravity dam.

Key words: seepage, parameter inversion, Bayesian theory, entropy-blind number, differential evolution adaptive metropolis algorithm