Scholarlink

<p>In this paper we introduce two methods to using Gaussian Process Models to solve history matching/inverse problems and design optimisation problems. The first uses the full capabilities of Gaussian Process Models to iteratively create a map across the full parameter space that indicates where a solution to the problem is unlikely to be found. At each iteration this process contracts the space over which a search will be conducted. This leads to high confidence that the global optimum has been located. The second method uses the ability of a Gaussian Process Model to approximate a function across the whole of a parameter space to create an approximation to the unknown inverse function. This leads to a fast solution of the inverse problem. The methods were tested on the IC Fault Model and two variants of the PUNQ S3 model.</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

Gaussian process models for history matching

  • 期刊论文

  • 发表日期：

    2026年7月9日

查看全文
  • Jonathan Carter,
  • Ian Vernon

摘要

In this paper we introduce two methods to using Gaussian Process Models to solve history matching/inverse problems and design optimisation problems. The first uses the full capabilities of Gaussian Process Models to iteratively create a map across the full parameter space that indicates where a solution to the problem is unlikely to be found. At each iteration this process contracts the space over which a search will be conducted. This leads to high confidence that the global optimum has been located. The second method uses the ability of a Gaussian Process Model to approximate a function across the whole of a parameter space to create an approximation to the unknown inverse function. This leads to a fast solution of the inverse problem. The methods were tested on the IC Fault Model and two variants of the PUNQ S3 model.