Scholarlink

<p>To effectively characterize the spatial variability of soil properties using random field theory, it is crucial to determine the autocorrelation function (ACF) model and model parameters, including scale of fluctuation (<i>δ</i>). However, it is a challenging task when relying solely on monitoring data. This paper proposes a novel Bayesian model selection method to identify the optimal ACF model among a pool of candidate ACF models (i.e., single exponential, squared exponential, second-order Markov and binary noise ACFs) and to learn <i>δ</i> and spatially varying soil parameters using time-series monitoring data of pore water pressure (PWP). The prior knowledge of <i>δ</i> for saturated hydraulic conductivity (<i>k</i><sub>s</sub>) is determined through an extensive literature review of similar soils. The Bayesian Updating with Subset simulation (BUS) method is adopted to facilitate ACF model selection and parameter identification based on a large number of monitoring data of PWP. A real slope in Hong Kong, China, is used to illustrate the effectiveness of the proposed method. The results show that the proposed method can properly identify the optimal ACF model based on the estimated model evidences or occurrence probabilities. It can simultaneously learn the spatially variable soil parameters and ACF model parameters using the monitoring data. The proposed method provides a new perspective for random field modeling of the spatial variability of soil properties with monitoring data. Additionally, the second-order Markov ACF model is identified as the optimal ACF choice. The integration of time-series PWP monitoring data also significantly reduces the epistemic uncertainties of <i>δ</i> and spatially varying <i>k</i><sub>s</sub>. This type of uncertainty, representing a lack of knowledge about soil parameters, can be reduced. The PWP values predicted using the optimal ACF model and posterior samples of parameters match well with the monitoring data of PWP during the validation periods. Accounting for the uncertainty of <i>δ</i> and learning its distribution are extremely important for the Bayesian model selection and parameter identification using the monitoring data.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Bayesian autocorrelation model selection and parameter identification using time-series monitoring data of pore water pressure of the slope under rainfall

  • 期刊论文

  • 发表日期：

    2026年4月21日

查看全文
  • Shui-Hua Jiang,
  • Hong-Hu Jie,
  • Jiawei Xie,
  • Jian-Hong Wan,
  • Peng Lan,
  • Jinsong Huang

摘要

To effectively characterize the spatial variability of soil properties using random field theory, it is crucial to determine the autocorrelation function (ACF) model and model parameters, including scale of fluctuation (δ). However, it is a challenging task when relying solely on monitoring data. This paper proposes a novel Bayesian model selection method to identify the optimal ACF model among a pool of candidate ACF models (i.e., single exponential, squared exponential, second-order Markov and binary noise ACFs) and to learn δ and spatially varying soil parameters using time-series monitoring data of pore water pressure (PWP). The prior knowledge of δ for saturated hydraulic conductivity (ks) is determined through an extensive literature review of similar soils. The Bayesian Updating with Subset simulation (BUS) method is adopted to facilitate ACF model selection and parameter identification based on a large number of monitoring data of PWP. A real slope in Hong Kong, China, is used to illustrate the effectiveness of the proposed method. The results show that the proposed method can properly identify the optimal ACF model based on the estimated model evidences or occurrence probabilities. It can simultaneously learn the spatially variable soil parameters and ACF model parameters using the monitoring data. The proposed method provides a new perspective for random field modeling of the spatial variability of soil properties with monitoring data. Additionally, the second-order Markov ACF model is identified as the optimal ACF choice. The integration of time-series PWP monitoring data also significantly reduces the epistemic uncertainties of δ and spatially varying ks. This type of uncertainty, representing a lack of knowledge about soil parameters, can be reduced. The PWP values predicted using the optimal ACF model and posterior samples of parameters match well with the monitoring data of PWP during the validation periods. Accounting for the uncertainty of δ and learning its distribution are extremely important for the Bayesian model selection and parameter identification using the monitoring data.