ABSTRACT
The calculated response from a numerical model will deviate from the measured one given the presence of modelling idealizations and real world construction effects. This deviation can be directly captured by a ratio between the measured and the calculated quantity. The ratio is also called a model factor in many design guides. The probabilistic distribution of the model factor is arguably the most common and simplest complete representation of model uncertainty. The characterisation of model uncertainty is identified as one of the critical elements in a geotechnical reliability-based design process in Annex D of ISO 2394:2015 “General Principles on Reliability of Structures”. This Spotlight paper reviews the databases for various geo-structures and determines their associated model statistics. Foundation load test databases are the most prevalent. A recent effort to compile a large generic database (PILE/2739) that contains 2739 field load tests conducted on various piles and installed in different soils and countries, is highlighted. This systematic compilation of load test data is part of a broader research agenda to digitalise foundation design for “precision construction”, which is targeted at characterising “site-specific” model factors and soil parameters based on both site-specific and generic data for further customisation of design to a particular site. The mean and COV of the model factor for a range of geo-structures, geomaterials, and limit states (both ultimate and serviceability) are summarized in a form suitable for adoption in design and codes of practice. Based on this summary, it is proposed that a model factor for a design model can be classified as: (1) moderately conservative (1 ≤ mean < 2), (2) highly conservative (2 ≤ mean < 3), or (3) very highly conservative (mean ≥ 3). The model uncertainty can be as: (1) low dispersion (COV < 0.3), (2) medium dispersion (0.3 ≤ COV < 0.6), (3) high dispersion (0.6 ≤ COV < 0.9), and (4) very high dispersion (COV ≥ 0.9). This summary represents the most extensive and significant update of Table 3.7.5.1 in the 2006 JCSS Probabilistic Model Code.
Acknowledgement
The authors are grateful to Zijun Cao; Yit-Jin Chen Peter Day, Mahongo Dithinde, Bengt H. Fellenius, Kerstin Lesny, Yoshihisa Miyata, Shadi Najjar, Chang-Yu Ou, Sukumar Pathmanandavel, Johan V. Retief, and Limin Zhang for their invaluable comments and assistance in the preparation of this paper.
Disclosure statement
No potential conflict of interest was reported by the authors.
ORCID
Kok-Kwang Phoon http://orcid.org/0000-0003-2577-8639