Determination of Input Parameters for a Fully Probabilistic Geotechnical Design Stanovení vstupních parametrů pro plně pravděpodobnostní geotechnický návrh Lumír MIČA 1 , Roman KOIŠ 2 , Jiří BUČEK 3 , Radoslav RUSÍNA 4 Abstract: A deterministic analysis is mainly used for a design of engineering structures nowadays. Fully probabilistic design is a new trend for analysing engineering structures. Determination of input parameters is a very important part for both types of analyses. Unlike to a deterministic analysis, the input parameters for a fully probabilistic design are defined as random values in stochastic approach. It means that statistical data set of an input parameter is needed from which probability distribution function (pdf) can be derived. This function can be then defined by various statistical parameters such as the mean value, the variance, etc. The paper deals with a determination of the pdf type and of its statistic parameters for an oedometric modulus, which is the input parameter for a modelling the interaction of a structure with a subsoil in Soilin software package. The statistical analysis is done for two cases. In the first case direct laboratory measurements of an oedometric modulus is used. This case shows an example of a data evaluation of loess soil at Brno region. The second case shows an evaluation done for an indirect measurement of the oedometric modulus – cone penetration test (locality – Klobouky u Brna). Abstrakt: Pro navrhování stavebních konstrukcí se v současné době nejvíce používá deterministická analýza. V posledních letech se však stále více rozvíjí plně pravděpodobnostní přístup. Obecně však pro oba přístupy je alfou a omegou stanovení vstupních parametrů. Na rozdíl od deterministického přístupu jsou u plně pravděpodobnostního přístupu vstupní parametry náhodné veličiny. To znamená, že je nutné mít k dispozici statistický soubor dat, z kterého může být určeno rozdělení pravděpodobnosti (pdf), které je definováno statistickými parametry jako je např. střední hodnota, rozptyl, atd. Z takto rozsáhlé problematiky se článek zabývá stanovením typu pdf a jeho statistickými parametry pro geomechanický parametr oedometrický modul. Oedometrický modul byl zvolen z toho důvodu, že je vstupním parametrem pro modelování interakce podloží se základovou konstrukcí pomocí podprogramu SOILIN. Statistická analýza je provedena pro dva případy. V prvním případě jsou použity výsledky z oedometrických laboratorních zkoušek (přímé měření). Tato analýza je ukázána na příkladu naměřených dat pro spraš. Druhý případ získání dat pro statickou analýzu vychází z nepřímého testování oedometrického modulu pomocí statické penetrační zkoušky. Jako příklad byla zvolena lokalita Klobouky u Brna. Keywords: Oedometric modulus, Oedometric laboratory test, Cone penetration test, Statistical analysis. Klíčová slova: Oedometrický modul, Oedometrická zkouška, Statická penetrační zkouška, Statistická analýza. 1 Ing. Lumír Miča, Ph.D., Brno University of Technology, Faculty of Civil Engineering, Veveří 95, 602 00 Brno, Czech Republic, +420541147234, [email protected]2 Ing. Roman Koiš, Statika Olomouc, s.r.o., Balbínova 374/11, 779 00 Olomouc, Czech Republic, +420585700702, [email protected]3 Ing. Jiří Buček, Ph.D., FEM consulting, s.r.o., Veveří 95, 602 00 Brno, Czech Republic, +420541147374, [email protected]4 Ing. Radoslav Rusina, Ph.D., FEM consulting, s.r.o., Veveří 95, 602 00 Brno, Czech Republic, +420541147374, [email protected]
9
Embed
Determination of Input Parameters for a Fully ... · PDF filelaboratory test which must be carried out by given procedure according to ČSN CEN ISO/TS 17892-5 [8]
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Determination of Input Parameters for a Fully Probabilistic Geotechnical Design
Stanovení vstupních parametrů pro plně pravděpodobnostní geotechnický návrh
Lumír MIČA1, Roman KOIŠ
2, Jiří BUČEK
3, Radoslav RUSÍNA
4
Abstract: A deterministic analysis is mainly used for a design of engineering structures nowadays. Fully
probabilistic design is a new trend for analysing engineering structures. Determination of input parameters is a very
important part for both types of analyses. Unlike to a deterministic analysis, the input parameters for a fully
probabilistic design are defined as random values in stochastic approach. It means that statistical data set of an input
parameter is needed from which probability distribution function (pdf) can be derived. This function can be then
defined by various statistical parameters such as the mean value, the variance, etc. The paper deals with a
determination of the pdf type and of its statistic parameters for an oedometric modulus, which is the input
parameter for a modelling the interaction of a structure with a subsoil in Soilin software package. The statistical
analysis is done for two cases. In the first case direct laboratory measurements of an oedometric modulus is used.
This case shows an example of a data evaluation of loess soil at Brno region. The second case shows an evaluation
done for an indirect measurement of the oedometric modulus – cone penetration test (locality – Klobouky u Brna).
Abstrakt: Pro navrhování stavebních konstrukcí se v současné době nejvíce používá deterministická analýza.
V posledních letech se však stále více rozvíjí plně pravděpodobnostní přístup. Obecně však pro oba přístupy je
alfou a omegou stanovení vstupních parametrů. Na rozdíl od deterministického přístupu jsou u plně
pravděpodobnostního přístupu vstupní parametry náhodné veličiny. To znamená, že je nutné mít k dispozici
statistický soubor dat, z kterého může být určeno rozdělení pravděpodobnosti (pdf), které je definováno
statistickými parametry jako je např. střední hodnota, rozptyl, atd. Z takto rozsáhlé problematiky se článek zabývá
stanovením typu pdf a jeho statistickými parametry pro geomechanický parametr oedometrický modul.
Oedometrický modul byl zvolen z toho důvodu, že je vstupním parametrem pro modelování interakce podloží se
základovou konstrukcí pomocí podprogramu SOILIN. Statistická analýza je provedena pro dva případy. V prvním
případě jsou použity výsledky z oedometrických laboratorních zkoušek (přímé měření). Tato analýza je ukázána na
příkladu naměřených dat pro spraš. Druhý případ získání dat pro statickou analýzu vychází z nepřímého testování
oedometrického modulu pomocí statické penetrační zkoušky. Jako příklad byla zvolena lokalita Klobouky u Brna.
An objective when defining an engineering structure is ensuring its reliability as well as
its economy. This task is most distinct when solving foundation structures. Those structures are
in interaction with both subsoil and superstructure and they generate so-called interaction
system. For ensuring reliability of engineering structures at first there were used an allowable
stress method or a factor of safety method. With new knowledge and application of theory of
probability there was created an ultimate limit state method or partial reliability coefficient.
Although this method is specified as probabilistic it is needed to keep in mind that it is about
deterministic calculation because all magnitudes are given with particular value. On the other
hand it is not corresponding with full reality because most of events yield to randomness. It is
because input parameters do have specific variability. In addition in geotechnics this is
amplified by fact that soil is not made by man (it was not dosed in exact recipe) but by nature,
hence there is also developing of space variability (but it is not a subject of the article). If we
want this randomness to be involved in the design then we have to apply some of the stochastic
methods (essential magnitude as a random magnitude, random function or by fuzzy sets theory).
We have to apply so-called fully probabilistic method. Although the application of this method
is in-process for longer time its signification has not been well appreciated. It was mainly
because of chances of getting input parameters as same as the efficiency of computers. In recent
years this tendency is changing and in near future the usage of this method will be common way
of designing engineering structures. A research project SISMO is also in a spirit of this
tendency. The project deals with stochastic analysis of interaction system “subsoil-foundation-
superstructure”. The part of this project is also statistic analysis of geomechanical input
parameters.
2 Statistic analysis of geomechanical data
Stochastic interaction problem is being solved by usage of computing core FEM of
system SCIA ENGINEER with a program unit SOILIN [1], [11] and probabilistic FReET [4].
For analysis of the interaction between subsoil and foundation structure the model called surface
model of subsoil is being applied. For this model is given that during its strain the same virtual
work as in 3D subsoil is being carried out whereas it is possible to establish all hierarchy of
parameters C1, C2, C3. This model of subsoil is described in detail in the book [2]. However the
parameters C are not characteristic of soil because they are dependant also on the geometry of
the foundation. It is needed to figure them out either experimentally (almost unrealistic) or by
calculations. For this reason the program SOILIN has been created. This program is able to find
out the process of settlement wherever and from which to find out the wanted parameters C.
This procedure is carried out on the basis of state of stress of elastic homogenous half-space and
standard model of soil. Physical model of subsoil in program unit SOILIN results from ČSN 73
1001 [9], where the settlement of surface of subsoil is being calculated from given state of stress
from the formulae:
, ,
1 ,
nz i i or i
i
i oed i
ms h
E
(1)
where z,i is vertical normal stress component in elastic isotropic homogenous infinite
half-space (or in stratum), or,i is analogous component of original geostatic state of stress, mi is
correcting coefficient of surcharge (the coefficient of structural rigidity) and n is the number of
layers with a thickness hi and oedometric modulus Eoed,i, in which so-called effective stress is
non-negative:
, , , , 0zú z i s i z i i or im (2)
Zones where the effective stresses are negative also the deformation is zero. It is mainly
in a greater depths where the subsoil is no more being deformed. The conditions of zero
effective stress determine so-called depth of deformation zone of subsoil.
Input parameters to the calculation is Eoed, which can be replaced by Edef, at present
definition of Poisson ratio . The usage of Eoed or Edef is dependent on its assignment.
Oedometric modulus is being defined by laboratory testing while deformation modulus is most
often gain from the results of in-situ tests. The next input parameter is a bulk density of soil , which is also being defined by laboratory testing (very difficult to determine) it is being taken
from the tables. The last parameter is a coefficient of structural rigidity which is not possible to
determine by measuring but only by use of tables.
In order to carry out of stochastic analysis of construction it is needed to carry out the
randomness of the problem. It means to think of the input geomechanical parameters as random
with specific theoretical distribution of probability „pdf“(normal, lognormal, Weibull etc.). The
type „pdf“ is commonly determining on the basis of statistic analysis of random selection Xi of
great range and on the basis of physical principle of a given effect. The most difficult task is to
have the data set of plotting parameter. Generally we do have two options available, which is
either the laboratory (direct) or field (indirect) testing of geomechanical data.
2.1 Statistic analysis form laboratory testing data
In the case of our solving problems it concerns particularly with determination of
oedometric modulus of deformability. Oedometric modulus characterises the state when soil
owing to vertical surcharge cannot be deformed to the sides. It is called one-axial deformation.
Oedometric modulus is being calculated for single intervals of loading according to the formulae
(3). Hence there must be stated a given range of stresses. This modulus is being determined by
laboratory test which must be carried out by given procedure according to ČSN CEN ISO/TS
17892-5 [8]. For its determination it is needed to remove undisturbed sample.
z
zoedE
(3)
In order to carrying out of statistic evaluation we must have data set available. In the
work experience it is very difficult because the range of site investigation is rather limited,
which is being reflected on the number of sufficient data. Hence the other option how to get
those data is archive. (e.g. Czech geological service or the archives of investigative
organisation). From those sources it is possible to get the data which can be comprised to
statistic analysis. The utilisation of those data is shown on the example of Brno town for the
loess which is most frequent soil you can find in this area.
Statistic characteristics of oedometric modulus of this kind of soil has been evaluated
from 40 results of oedometric test by usage of program IDENT. The results are summarised in
table 1.
Load
interval
MPa
Ēoed
MPa SEoed vEoed aEoed eEoed
Type of
pdf
0,05-0,10 6,56 2,69 0,41 0,40 -1,14 Bouned
normal
0,10-0,20 7,02 1,97 0,28 1,08 0,34 Weibull
0,20-0,30 8,00 1,85 0,23 0,24 -0,98 Weibull
0,30-0,40 19,21 5,38 0,28 0,25 -0,96 Weibull
0,40-0,50 22,60 5,04 0,22 -1,14 -0,68 Pearson III
Table 1. Statistical characteristics for Eoed [3]
In the table there is evaluated mean value Ēoed, standard deviation sEoed, variation
coefficient VEoed, coefficient of inclination aEoed and taperness eEoed and also appropriate type of
distribution of probability. All the time we have to keep in mind that each statistic
characterisation is valid for appropriate interval of loading.
In conclusion it is needed to mention, that the size of this modulus may be significantly
influenced during the extracting but also during the test of itself [7].
2.2 Statistical analysis with using penetration testing
The second possibility, how to get data set for statistical analysis, is using indirect
methods of geotechnical survey. The penetration testing is one of them. In these days four
testing methods are standardized. Those are:
Cone Penetration Test,
Dynamic Probing Test,
Standard Penetration Test,
Weight Sounding Method
The first and second methods are being used in the Czech Republic and from these tests
we can obtain information about subsoil:
Delineating soil stratigraphy
Type of soil
Mechanical properties of soil by using correlation formula
Compaction quality
and the others
Cone penetration test (CPT)
The principle of test consists of pushing cone tip into the ground continuously at 20
mm/s. The depth interval between readings is usually each 20 mm or 50 mm and the maximum
interval is 200 mm. The base CPT probe measures the resistance on the cone tip qc and the
sleeve friction fs but other probes with specific sensors can be used. The qc and fs is calculated
Table 2 Modified and empirical probability distribution functions
2.4 Conclusion
We believe that fully probabilistic calculations are going to be employed more frequently
in civil engineering practise. In order to use it, adequate data bank available for input data
selection will be needed.
SISMO project deals with stochastic interaction of foundation with subsoil. A part of this
project deals with the possibility of determination of input data for subsoil.
In the case of selected physical model of subsoil the concern is the probabilistic
modelling of oedometric modulus. This paper presents two options how to get the data set of
oedometric modulus for subsequent statistical analysis:
• The first option is the utilization of results from laboratory tests. The advantage of this
procedure is smaller error when determining it in comparison with the second option
(penetration testing – correlation relation). On the other hand we have to work with
limited number of data available in the range of survey. In case of the particular
construction it usually concerns about two up to three values.
• The second option described in this paper is the utilization of cone penetration testing
(CPT). The advantage of this testing is getting more representative sample of data set
due to methodology of testing.
For evaluation of data set the universal probabilistic software FreET has been used. This
software communicates with FEM analysis (SCIA ENGINEER) through very efficient interface
for solving fully probabilistic analysis of the interaction of foundation with subsoil [5]. The
interface is created within SISMO project.
Acknowledgements
This research was financially supported by the project of the Czech Science Foundation
(GA ČR) No. GA103/09/1262, No. GAČR č. 103/08/0752 and by the research project of The
Ministry of Education, Youth and Sports (MŠMT ČR) No. MSM0021630519. Authors
appreciate this support.
References
[1] Buček J., Kolář V., Obruča J. (1993): SOILIN Výpočet sedání a parametrů interakce
podle norem ČSN, DIN a zásad EC7 (SOILIN Calculation of settlement and interaction
parameters in accordance with ČSN, DIN and EC7 – in Czech). Brno, FEM consulting,
s.r.o, 56s.
[2] Kolář V., Němec I. (1989) Modelling of Soil-Structure Interaction. New York, Oxford,
London, Amsterdam, Tokyo, ELSEVIER, 334p.
[3] Koiš, R. (1995) Zpřesněné výpočetní modely při navrhování základových pasů, Vysoké
učení technické v Brně (Improvement computational models for designing of shallow
foundation – in Czech), Brno, 122s.
[4] Novák, D., Vořechovský, M., Rusina, R. (2007): FREET version 1.5 – User's Manual
and Theory Manual. Brno/Červenka Consulting, http://www.freet.cz
[5] Novák, D., Miča, L., Teplý, B., Vořechovský, M., Buček, J., Rusina, R. and Němec, I. (2009): Soil-structure interaction stochastic modelling. In: Furuta, Frangopol and
Shinozuka, M. (Eds.), Safety, Reliability and Risk of Structures, Infrastructures and