You are on page 1of 6

Developing a new bridge live load model for South Africa

P.F. van der Spuy & R. Lenner


Stellenbosch University, South Africa

ABSTRACT: South Africa's bridge design code, TMH7, was first introduced in 1981 and is based on the CEB-
FIP Model Code of 1978 as well as BS5400 and the National Building Code of Canada. All of these codes have
since been superseded. In deriving the code it was noted that a fully probabilistic approach was not possible
due to insufficient data at the time. The code was found to be unconservative for narrow and short span bridges
and hence some revisions were issued in 1988. In 1991 it was proposed to increase the axle load by 25 % to
180 kN, but the code was not amended. In 1996 the legal limits for Gross Vehicle Weights and Axle Loads in
South Africa were increased, but no study was conducted to determine the impact on the bridge loading code.
The current load model is widely regarded as cumbersome to apply and various different applications thereof
exist in industry. This study used one year of WIM data to calculate hogging and sagging moments as well as
shear forces for span lengths up to 50 m. The forces and axle weights were extrapolated normally to a return
period of 975.3 years. The extrapolation resulted in a characteristic axle weight of 168 kN which was applied
as a double axle spaced 1.2 m apart, similar to the Eurocode. The extrapolated load effects and the tandem axle
were used to calculate a distributed load of 15.5 kPa.

1 INTRODUCTION the time (Anderson, 2006). Since then a substantial


amount of WIM data has become available in South
South Africa's bridge design code, TMH7, was first Africa.
introduced in 1981 and is based on the CEB-FIP WIM data was used to derive an improved live
Model Code of 1978 as well as BS5400 and the Na- load model for bridge design in South Africa. This
tional Building Code of Canada. With time it was model is both up to date and simple to apply.
found that the code has various deficiencies, espe-
cially for narrow and short span bridges.
Revisions to the code were issued in 1988, but
Oosthuizen et al. (1991) showed that the code still un- 2 TMH7 LOAD MODELS
derestimated bending moments for spans between 4
m and 9 m as well as shear forces on span lengths be-
low 23 m. A proposal was made to increase the axle
2.1 NA loading
load from 144 kN to 180 kN, but this amendment was
never made. In 1996 the South African Department of NA loading is considered as normal traffic loading
Transport increased the legal gross vehicle weight and consists of a distributed component (Figure 1)
and axle loads on South African roads, but no study and an axle load. A dynamic allowance is included
was performed to determine the impact of this on the according to the Swiss formula (CSRA, 1981).
bridge loading code. The current load model is widely The distributed component of NA loading, Qdist, is
regarded as cumbersome to apply and various differ- 36 kN/m for the first 36 m loaded length and then re-
ent applications thereof exist in industry. duces according to Equation 1.
When deriving the code in 1978, Liebenberg noted
that a probabilistic study of extreme truck events was 180
Qdist = +6 (1)
not viable due to a lack of statistical information at √L
2.3 NC loading
where L = loaded length. Nominal NC loading is a loading representing multi-
wheeled trailer combinations with controlled hydrau-
lic suspension and steering intended to transport very
heavy indivisible payloads. The applied loading is 30
kPa with a configuration as shown in Figure 3.

3, 4 or 5 m

5, 10, 15 or 20 m 0, 5, 10, 15, 20 or 25 m 5, 10, 15 or 20 m

Figure 3. NC load configuration

NC vehicles shall move along the centre line of a


bridge only with an allowance of 1 m either side for
moving offline. No allowance is made for impact.

Figure 1. NA distributed load

3 REFERENCE TO EUROCODE
The axle load component of NA loading, Qaxle, is
given by Equation 2 in kN. South African building and bridge design codes have
historically been based on the British codes which
Qaxle = 144/√n (2) have since been superseded by the Eurocodes (Van
der Spuy, 2014). Many of the building design codes
where n = the number of the notional lane in which in South Africa incorporate parts of the Eurocode or
the axle load occurs. The axle load therefore de- have adopted the Eurocode entirely. In this study the
creases as the number of loaded notional lanes in- configuration of Load Model 1 of EN1991-2 (CEN,
creases. 2003) was adopted, but different values were calcu-
lated for the axle and distributed loads.
2.2 NB loading
Nominal NB loading is a unit loading representing a 4 WIM DATA
single abnormally heavy vehicle. Thirty six units of
NB loading are typically applied and referred to as
NB36. This corresponds to a wheel load of 90 kN, an 4.1 Data source
axle weight of 360 kN and a total weight of 1440 kN.
The NB load configuration is shown in Figure 2. National Route 3 between Johannesburg and Durban
is considered to be the heaviest freight route in South
Africa (Lenner et al., 2017) and for this reason a sta-
tion along this route was used for this study. The
1m

Roosboom WIM station has been measuring since


1m

November 2000 and the data from this station is con-


sidered to be of good quality (de Wet, 2010a). One
1m

year (2016) of data was used as recommended by


Sivakumar et al. (2008). This station measures traffic
2m 6, 11, 16, 21 or 26 m 2m in the slow lanes in both directions.
Figure 2. NB load configuration

Only one NB vehicle is allowed on a bridge at a time


without any other traffic loading acting in conjunc-
tion. No allowance is made for impact.
4.2 Traffic composition correct the magnitude of recorded axle loads retro-
spectively. The application of this method results in a
The number of vehicles in the year of data was 859
k-factor which is multiplied by all axle weights to
896. Figure 4 shows the distribution of vehicle types
suppress the systematic WIM error. The method has
indicating that six and seven axle vehicles dominate
been accepted by the South African National Roads
the GVW tail.
Agency Limited (SANRAL) and was applied in this
study as a way to calibrate the data.

5 METHODOLOGY

5.1 Span lengths investigated


Short to medium span lengths between 5 m and 50 m
were investigated based on the assumption that free
flowing traffic governs for these span lengths
(Caprani and OBrien, 2010). Span lengths within this
range cover the majority of bridges in South Africa.

Figure 4. Vehicle type distribution 5.2 Creation of vehicle convoy


Daily convoys of axles were created from the WIM
Lenner et al. (2017) showed that, when compared to file using axle distances as well as time stamps and
data from Auxerre measured in 1986, the South Afri- speeds to determine inter-vehicle gaps. The time
can axle loads are lower than those in Europe, but the stamp resolution from the Roosboom station is 0.01 s
GVW is higher. This serves as motivation not to which is preferable (Enright, 2010).
adopt the Eurocode directly, but to calculate country
specific axle and distributed loads for Load Model 1.
The data from Auxerre, France, is the data that was 5.3 Calculation of load effects
used to derive the Eurocode load models (Enright and The load effects that were investigated are sagging
OBrien, 2012; Sedlacek et al., 2008; Zhou, 2013; moments and shear forces in single span structures
Zhou et al., 2012). and hogging moments in two span structures
(Caprani, 2005; Enright and OBrien, 2012; Lenner,
2014; Lenner et al., 2014). The daily convoys were
4.3 Cleaning passed over varying span lengths for the different
load effects while recording the daily maximum val-
Slavik developed a technique called Golem (un- ues for each. The increment distance for the convoys
published) to specifically address sources of WIM er- was implemented as 0.444 m which corresponds to a
ror for South African data. Golem rejects any vehicle time step of 0.02 s at 80 km/h.
that meets the following criteria:

- Travelling at less than 5 km/h or more than 150 5.4 Return period
km/h A 5 % probability of exceedance in 50 years was used
- Truck length less than 4 m or longer than 26 m in this study, similar to the Eurocode (CEN, 2003).
- Fewer than two axles This relates to a return period of 975.3 years (Enright,
- GVW less than 3.5 t 2010).
- Individual axle weight exceeding 16 t
- Axle spacing shorter than 0.53 m or longer than
10 m

Number of vehicle records removed by this screening


is 33 855.

4.4 Calibration
De Wet (2010a, 2010b) developed the Truck Tractor
(TT) method which is a post-calibration procedure to
5.5 Extrapolation Return period (975.3 years)

With WIM data not being available for the full return
period extrapolation techniques were applied. Only
the upper tail of a parent distribution contributes sig-
nificantly to the extrapolated value at the return pe-
riod (Bailey, 1996; Zhou, 2013; Zhou et al., 2012) and
therefore only the upper 2√n of daily maximum val-
ues were of interest for this study. To extrapolate to
the return period a straight line was fitted to the upper
2√n of data points on Gaussian probability paper
(Kozikowski, 2009; Nowak, 1994, 1993; Nowak and
Hong, 1991). In South Africa heavy vehicles are al-
lowed to travel on weekends and holidays and there- 5 10 15 20 25 30 35 40 45 50 Spanlength (m)

fore for a return period of 975.3 years and 365 days


per year the corresponding value on the vertical axis
is 4.540. The Gaussian assumption was made in this Return period (975.3 years)
study as it is the method that was used in the deriva-
tion of the AASHTO code (Nowak, 1993) and the Eu-
rocode (Sedlacek et al., 2008).

6 RESULTS

6.1 Extrapolated load effects


Daily maximum load effects were plotted on Gauss-
ian probability paper and extrapolated to the return 5 10 15 20 25 30 35 40 45 50 Span length (m)

period. Hogging moments are shown in Error! Ref-


erence source not found., sagging moments in Er-
ror! Reference source not found. and shear forces
in Error! Reference source not found.. Table 1 Return period (975.3 years)

gives a summary of the load effect values at different


span lengths.

Table 1. Summary of extrapolated load effects.


______________________________________________
Span length
__________ Hogging
_______ Sagging
_______ Shear
_____
m kNm kNm kN
______________________________________________
5 264 378 340 Figure 5. Extrapolation of axle weights
10 820 1258 528
15 1664 2234 626
20 2318 3535 740
25 2474 4865 826
30 2973 6127 888 Span length (m)
10

15
20
25
30
35
40
45
50
5

35 3523 7451 937


40 4140 8877 975
45 5547 10331 1002
50 6864 11759 1018
_____________________________________________
6.2 Extrapolated axle weight
It is clear that for medium to longer span lengths the Daily maximum axle weights were extrapolated as
tails of the load effects do not strictly plot as straight per Sedlacek et al. (2008) who used a half-normal dis-
lines on Gaussian probability paper. This is an indi- tribution. Once again the upper 2√n of daily maxi-
cation that the Gaussian approximation to the distri- mum values were fitted with a straight line on Gauss-
bution of the load effect tails should be revisited in ian probability paper. This resulted in a characteristic
subsequent research. axle weight of 168 kN. The extrapolation of the axle
weights is shown in Figure 7. From Figure 7 it can be
seen that the tail of the axle weights plots as a straight
line on Gaussian probability paper, indicating that
axle weights closely resemble Gaussian behavior.
was used to calculate daily maximum hogging mo-
ments, sagging moments and shear forces for span
lengths up to 50 m. These forces were extrapolated on
Gaussian probability paper to a return period of 975.3
years.
The resulting load model is a double axle of 168
kN each, spaced at 1.2 m apart with a distributed load
of 15.5 kPa. When compared to the Eurocode the load
model has a lower axle load (300 kN for Eurocode),
but a larger distributed load (9 kPa in the Eurocode)
(CEN, 2003).

REFERENCES

Anderson, J.R.B., 2006. Review of South African


Live Load Models for Traffic Loading on Bridge
and Culvert Structures Using WIM Data.
6.3 Calculation of distributed loads University of Cape Town.
Bailey, S.F., 1996. Basic Principles and Load Models
The extrapolated axle weight of 168 kN was applied for the Structural Safety Evaluation of Existing
to the different span lengths as a double axle at 1.2 m Road Bridges. EPFL.
spacing, similar to the Eurocode (CEN, 2003). To re- Caprani, C.C., 2005. Probabilistic Analysis of
produce the extrapolated load effects from Table 1, a Highway Bridge Traffic Loading. University
College Dublin.
distributed load was calculated which acts in conjunc- Caprani, C.C., OBrien, E.J., 2010. Estimating
tion with the double axle. The distributed load is Extreme Highway Bridge Traffic Load Effects,
in: Safety, Reliability and Risk of Structures,
Infrastructures and Engineering Systems :
Proceedings of the Tenth International
Conference on Structural Safety and Reliability
(ICOSSAR2009), Osaka, Japan. pp. 3053–3060.
CEN, 2003. EN1991-2: Traffic loads on bridges.
Brussels.
Committee of State Road Authorities, 1981. TMH7
Parts 1 and 2: Code of Practice for the Design of
Highway Bridges and Culverts in South Africa.
doi:ISBN 0 7988 2158 2
de Wet, D.P.G., 2010a. Post-Calibration and Quality
Management of Weigh-in-Motion Traffic Data.
Stellenbosch University.
de Wet, D.P.G., 2010b. WIM Calibration and Data
Quality Management. J. South African Inst. Civ.
Eng. 52, 70–76.
Figure 6. Distributed loads Enright, B., 2010. Simulation of Traffic Loading on
Highway Bridges. University College Dublin.
shown in Figure 9 for different load effects and span Enright, B., OBrien, E.J., 2012. Monte Carlo
lengths. It can be seen that the hogging moment load Simulation of Extreme Traffic Loading on Short
effect for a 10 m span length produces the largest dis- and Medium Span Bridges. Struct. Infrastruct.
tributed load of 15.5 kPa. It is shown that the distrib- Eng. 9, 1267–1282.
doi:10.1080/15732479.2012.688753
uted load decreases with an increase in span length, Kozikowski, M., 2009. WIM Based Live Load Model
creating the possibility of distributed load reduction for Bridge Reliability. University of Nebraska.
with increasing span length. Lenner, R., 2014. Safety Concept and Partial Factors
for Military Assessment of Existing Concrete
Bridges. Universitat der Bundeswehr Munchen.
7 CONCLUSIONS Lenner, R., de Wet, D.P.G., Viljoen, C., 2017. Bridge
Loading and Traffic Characteristics in South
A preliminary traffic load model for single lane traffic Africa. J. South African Inst. Civ. Eng.
has been derived for bridge design in South Africa. Lenner, R., Keuser, M., Sykora, M., 2014. Safety
One year of data from the Roosboom WIM station Concept and Partial Factors for Bridge
Assessment under Military Loading. Adv. Mil.
Technol. 9.
Nowak, A.S., 1994. Load Model for Bridge Design
Code. Can. J. Civ. Eng. 21, 36–49.
doi:10.1139/l94-004
Nowak, A.S., 1993. Load model for highway bridges.
Struct. Saf. 13, 53–66.
Nowak, A.S., Hong, Y., 1991. Bridge Live-Load
Models. J. Struct. Eng. 117, 2757–2767.
Oosthuizen, A.P.C., Meintjies, C.J., Trumpelmann,
V., Peters, D., Ullmann, K.K.A.B., Oppermann,
G.H.P., 1991. TMH7 Part 2: Traffic Loading
(1991) Proposed Substitution of Section 2.6.
Cape Town.
Sedlacek, G., Merzenich, G., Paschen, M., Bruls, A.,
Sanpaolesi, L., Croce, P., Calgaro, J.A., Pratt,
M., 2008. Background document to EN 1991-
Part 2 - Traffic loads for road bridges - and
consequences for the design.
Sivakumar, B., Moses, F., Ghosn, M., 2008. Protocols
for Collecting and Using Traffic Data in Bridge
Design Prepared for NCHRP. Paramus.
doi:10.17226/14521
Van der Spuy, P.F., 2014. A Comparative Study
Between the South African and European Bridge
Design Codes for Bending and Shear. University
of Cape Town.
Zhou, X.Y., 2013. Statistical Analysis of Traffic
Loads and Traffic Load Effects on Bridges.
Universite Paris Est.
Zhou, X.Y., Schmidt, F., Jacob, B., 2012.
Extrapolation of Traffic Data for Development
of Traffic Load Models: Assessment of Methods
Used During Background Works of the
Eurocode, in: Bridge Maintenance, Safety,
Management, Resilience and Sustainability -
Proceedings of the Sixth International
Conference on Bridge Maintenance, Safety and
Management. pp. 1503–1509.

You might also like