1. Evaluation and Implementation of Model Uncertainties in Geotechnical Design
Presenter: Prof. Dr.-Ing. Kerstin Lesny
HafenCity University  Hamburg  Germany
Geo-Risk 2017, Denver
C20a/ Reliability- and Risk-Based Code Developments, Part I

2. Members of the TC205/TC304 discussion group
Sami Oguzhan Akbas
Witold Bogusz
Sébastien Burlon
Kerstin Lesny (discussion leader)
Giovanna Vessia
Kok Kwang Phoon
Chong Tang
Limin Zhang
Discussion report available on the ISSMGE website:
Comments are welcome!
(Quelle:
Geo-Risk  Denver/Colorado

3. Definition of Model Uncertainty
Model: calculation model, not the whole geotechnical model including e.g. the ground model
Type of model: analytical, empirical, semi-empirical closed-form solutions numerical (e.g. FE) model
Deviation of:
ideal situation
real behavior
calculated behavior
calculated behavior
real behavior
Geo-Risk  Denver/Colorado

4. Model Uncertainty Assessment
Model factor approach:
applicable only where a unique design quantity X exists
Xmeas to be collected from a load test database which needs to cover the whole range of possible design situations and shall fulfill specific requirements to provide sufficient information.
Appropriate methods for interpretation of load-displacement curves for ULS and SLS problems!
MF = Xmeas/Xcal or /MF = Xcal/Xmeas
with: MF = model factor or bias
X = load, a resistance, a displacement, etc.
Geo-Risk  Denver/Colorado

5. Model Uncertainty Assessment
Representative quantity approach:
Examples: head displacements of a laterally loaded flexible pile, top displace-ments of retaining walls, lateral strain in neighboring buildings due to excavation
reduction of model uncertainty to a single quantity questionable in case of complex structures, e.g. retaining walls
Two-step approach:
Combination of numerical analyses and load test results to establish a generalized model factor for specific design model.
MF = Xmeas/Xcal or /MF = Xcal/Xmeas
same as before, but: X = representative quantity characterizing the behavior of the structure
Geo-Risk  Denver/Colorado

6. Model Uncertainty Assessment
Intrinsic uncertainties:
ground conditions: spatial variabilty
consideration of subsoil characteristics within design method (degree of simplification)
determination of soil parameters (correlations to field tests or lab tests)
load test execution: measurement errors, interpretation and evaluation of load test results, selection of load tests for database, lack of information, personal experience
Cannot be eliminated!
Implementation in design possible:
as a random model factor in RBD
as a deterministic model factor in LRFD (but not necessarily a constant value!)
Geo-Risk  Denver/Colorado

7. Our Wish List on the Path to an Ideal Code
to further close the gap between theoretically based, complex approaches and the daily design practice of 'normal' geotechnical engineers, i.e. to provide a bridge between different 'languages speaking' or different 'ways of thinking’
to promote “ease of use”
to promote gathering and to provide later access to data of successes and failures of geotechnical structures to learn and to verify our design models
Geo-Risk  Denver/Colorado

8. Conclusions
model uncertainties to be evaluated using extensive databases covering all possible design situations under natural, but also under laboratory conditions
intrinsic uncertainties (e.g. soil variability, measurement errors etc. ) can hardly be seperated
use of deterministic (not constant) model factors in LRFD formats is possible, at least for standard design situations
Proper transformation and acceptance in the engineering practice important!
Geo-Risk  Denver/Colorado