Reducing forecast uncertainty by using observations in geotechnical engineering

详细信息    查看全文

• 作者：Maximilian Huber
• 关键词：Data assimilation ; Inverse analysis ; Geotechnical engineering ; Observational method
• 刊名：Probabilistic Engineering Mechanics
• 出版年：2016
• 出版时间：July 2016
• 年：2016
• 卷：45
• 期：Complete
• 页码：212-219
• 全文大小：1748 K

文摘

Especially in geotechnical engineering, a high level of uncertainty in the design of structures is present. Standards and guidelines recommend the observational method for projects with a high level of uncertainty and when the geotechnical behaviour is difficult to predict. The behaviour of a complex geotechnical problem is measured in each constructions step and these measurements are compared to simulation results. As a next step one compares the used model parameters and assumptions. In case of big differences, one adapts the system in order to improve it for the next construction step. However, this design approach is based on engineering judgement in combination of deterministic approaches, which are adapted sequentially whenever new observations are available. This formulates the need for a sound mathematical and statistical framework, which allows to combine measurement to quantify forecast uncertainty.

This paper explains two mathematical concepts for data assimilation. The sequential and variational data assimilation consider a stochastic system, which is updated by uncertainty observations. These concepts reduce the simulation uncertainty by using observations. Two case studies show the applications of both concepts in geotechnical engineering problems. The first case study is discussing the possibilities and limitations of the sequential data assimilation concept in a theoretical example, whereas the second case study is demonstrating the combination of settlement measurements and a stochastic subsoil model by means of variational data assimilation. Additionally, the concept of forecast uncertainty quantification is demonstrated in the second case study. At the end a brief review of the data assumption concepts and the given forecast uncertainty quantification approach is given together with conclusions for further research.