See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/237190164

Fatigue strength of prestressed concrete girder bridges

Article in Canadian Journal of Civil Engineering · February 2011

DOI: 10.1139/l88-027

CITATIONS READS

9 1,531

2 authors, including:

Andrzej S. Nowak
Auburn University
291 PUBLICATIONS 6,554 CITATIONS

SEE PROFILE

All content following this page was uploaded by Andrzej S. Nowak on 03 February 2015.

The user has requested enhancement of the downloaded file.

 Fatigue strength of prestressed concrete girder bridges1
RAJEHZ. AL-ZAID'AND ANDRZEJS . NOWAK
Deparrmetlr of Civil Etzgineeritlg, UtziversiQ of Michigan, Atztl Arbor, MI 48109, U.S.A.
Received Janua~y12, 1987
Reviscd manuscript accepted October 27, 1987

A model for evaluating the fatigue life of a prestressed concrete girder bridge is presented. Experimental studies indicate that
fatigue is a random phenomenon. Therefore, the approach is based on probabilistic methods. The fatigue life of the prestressed
concrete girder is estimated from the fatigue properties of its components, namely, the cast-in-place slab, precast beam,
prestressing steel, and, if partially prestressed, the tension reinforcement. The model utilizes the available S-N curves for
Can. J. Civ. Eng. Downloaded from www.nrcresearchpress.com by Renmin University of China on 05/28/13

structural materials and Miner's rule to evaluate the fatigue life under variable amplitude loading. A model for determining the
distribution function of time to first cracking is also presented. To demonstrate the developcd models, a typical prestressed
concrete girder bridge is analyzed. On the basis of this investigation, it is concluded that fatigue in prestrcssed concrete girder
bridges designed according to the American Association of Statc Highway and Transportation Officials specifications is not a
limiting design criterion.
Key words: bridges, fatigue, prestressed concrete, structural rcliability.

Un modele servant a Cvaluer la resistance a la fatigue d'un pont h poutres en bCton prkcontraint est prCsent6. Des Ctudes
expkrimentales indiquent quc la fatigue est un phknomene aleatoire. Par consequent, cette approche est basCe sur des methodes
probabilistiques. La resistance a la fatigue du pont i poutres en beton precontraint est CvaluCe a partir des caracteristiques de
fatigue de ses composantes, soit la dalle coulee sur place, la poutre prCfabriquCe, l'acier de prkcontrainte et les armatures tendues
si elles sont partiellement prkcontraintes. Le modele a recours aux courbes S-N pour les ClCments de construction et a la regle de
Miner pour Cvaluer la rksistance a la fatigue lorsque le chargement est i amplitudes variables. Un modble pour determiner la
fonction de rCpartition du temps jusqu'h la premikre fissuration est propose. Un pont h poutres en beton precontraint type est
analysk dans le cadre d'une demonstration des modeles. Cctte etude conclut que la fatigue dans les ponts a poutres en beton
prkcontraint conqus selon le cahier des charges de 1'AASHTO n'est pas un critbre de conception restrictif.
Mots clis : ponts, fatigue, beton prkcontraint, fiabilite de la structure.
For personal use only.

[Traduit par la revue]

Can J . Civ. Eng. 15, 199-205 (1988)

Introduction can be evaluated by employing reliability theory in the structural

The increasing interest in fatigue of concrete structures is due safety analysis and design. Structural reliability is defined as the
primarily to the use of higher strength materials together with probability that the structure will perform satisfactorily for a
refined design and analysis procedures, widespread use of specified period of time under a stated set of use conditions. In a
concrete structures in different types of applications in which general sense, satisfactory performance is the set of criteria that
the structure is subjected to many repetitions of cyclic loadings, is necessary to accomplish the intended function. Time is that
and the new recognition of the effects of repeated loading on the operating period over which the structure is expected to operate
characteristics of the materials, which may lead to increase in while the desired function is accomplished. Use conditions
crack widths and deflections (ACI Committee 215, 1974). involve the total environment (load, material properties, etc.)
Despite all the existing data, the fatigue process by which under which the structure is expected to perform. Structural
fatigue damage occurs is not completely known yet. In general, reliability has been the subject of extensive research effort
it can be said that fatigue is a process of progressive permanent during the last decade (Kraker et al. 1982; Thoft-Christensen
internal structural changes in a material subjected to repetitive and Baker 1982). Many methods have been proposed consider-
loads. These changes may result in progressive growth of cracks ing the type of problem, the accuracy required, the parameters
and complete fractures of steel or concrete if the load repetitions involved, and the uncertainty associated with these parameters.
and (or) their amplitudes are sufficiently large (Holmen 1979). A recent comprehensive review of the literature on fatigue
Fatigue strength is defined as the maximum stress, expressed in reliability theory is available in Wirsching et al. (1982). A
percentages of the static strength that can be supported reliability model for steel highway bridges has been proposed by
repeatedly for a given number of cycles. Nyman and Moses (1985). Wang and Yao (1983) pointed out
Fatigue analysis involves a considerable degree of uncertain- that the Monte Carlo simulation of Miner's rule (Miner 1945)
ty, mostly due to the unpredictable future loading conditions, may be used to estimate the structural reliability involving
variability of material properties, and the use of simplified fatigue cumulative damage under random loading conditions.
analytical models to predict the structural response under a Only a limited amount of work has been done on the problem of
given loading condition. The risk of unacceptable consequences fatigue reliability of prestressed concrete members. Siriaksorn
(1980) considered the reliability against fatigue failure of
partially prestressed concrete beams. In that study the stress
NOTE:Written discussion of this paper is welcomed and will be range limits specified by the ACI Committee 2 15 (1974) and the
received by the Editor until July 31, 1988 (address inside front cover).
'This paper is a revised version of a paper presented at the 1986
ACI Committee 343 (1977) on bridge structures were used to
Canadian Society for Civil Engineering Bridge Conference and pub- determine the fatigue limit states for concrete, steel, and
lished in the proceedings. prestressing strands.
2~resentaddress: Department of Civil Engineering, King Saud The main objective of this study is to develop a reliability-
University, Riyad, Saudi Arabia. based model for the evaluation of fatigue life of prestressed
 200 CAN. 1. CIV. ENG. VOL. 15. 1988

concrete girder bridges. The model is based on references curves of the structural materials and Miner's rule (Miner 1945)
(Nyman and Moses 1985; Wang and Yao 1983; Wirsching et al. to determine the distribution of the variable amplitude load
1983), taking into consideration the different nature of the cycles to failure, NF, from the statistics of constant amplitude
considered problem. A model for determining the distribution fatigue. In this model, failure is defined as the crushing of
function of time to first cracking is also developed. concrete or rupture of steel.
Let m denote the random Miner sum at failure, Pi be the
Fatigue strength probability of occurrence of a stress cycle of stress level Si, NFi
When a structure is subjected to repeated applications of be the number of cycles to failure at each stress level Si, then it
random loads (e.g., truck traffic load), the time, T, and the can be shown that (Al-Zaid 1986)
number of cycles, N, at which failure occurs are random 1L
variables. Fatigue reliability analysis concerns the prediction of
Can. J. Civ. Eng. Downloaded from www.nrcresearchpress.com by Renmin University of China on 05/28/13

the distribution of lifetime of such a structure and its probability

of failure within a specified period of time. where NF is the number of cycles to failure under random loading
Let Q denote all factors that may give rise to fatigue failure of and k is the number of stress level blocks.
structure, and let R denote all factors that determine the In [6] both m and NFi are random variables, so NF is also a
resistance to this failure. The variables R and Q should be random variable. In this study, the fatigue resistance denoted by
measured in the same units (e.g., R = number of cycles to R in the above formulation is represented by NF and is assumed
failure and Q = number of cycles applied). As for fatigue, either to be independent of time.
R or Q or both are, in general, functions of time. Therefore, the On the other hand, the load effect at any particular point of
probability of failure is also a function of time. There are two time, t, denoted by Q in the preceding formulation, is defined
methods of incorporating the time aspect into the probability of as the number of load cycles encountered by the structure by that
failure analysis based on the way in which R and Q are defined time and is denoted by N(t). The function, N(t), is a nonsta-
(Kraker et al. 1982). The two definitions to be distinguished are tionary process and is related with time as follows:
(a) R(t) and Q(t) are the governing resistance and load,
respectively, for the period (0,t) and (b) R(t) and Q(t) are the [71 N(t) = fqt
instantaneous resistance and load at the point of time t. where fq is the frequency of cyclic loading and t is its duration.
If R(t) and Q(t) are defined as in (a), the probability, For highway bridges, fq is defined as the average daily truck
For personal use only.

P[R(t) < Q(t)] , relates to aperiod of time (0,t) and gives directly traffic (ADTT).
the probability distribution function, FAt), of the random time Consequently, the limit state function is defined as
to failure, T (Kraker et al. 1982). Thus,

On the other hand, if R(t) and Q(t) are defined as in (b), the and the probability of failure at time t conditional on survival
probability, p[R(t) < Q(t)], relates only to a particular point of until time t is given by
time and gives the so-called conditional probability of failure or
hazard function, h(t). The hazard function h(t) is defined such PI h(t)=p[N~<N(t)l
that the value of h(t)dt indicates the probability that failure will Hence, the probability of failure during the time interval (0,t)
occur in (t,t + dt), given that failure has not occurred in (O,t), can be calculated using [5]. The model is schematically illus-
thus: trated in Fig. 1.
[2] h(t) = limp[R < Q in (t,t + At) I R > Q in (0,t)l It is clear that NF and N(t) are independent random variables
and accordingly [9] can be evaluated by using any one of the
Ar+ 0
It can be shown (Al-Zaid 1986; Wirsching et al. 1982) that available reliability analysis methods.
the reliability function, L(t), can be related in a simple way to If the mean-value second-moment method is used, then the
the hazard function, h(t), as follows: instantaneous reliability index, P(t), is given by
[lo1 P(t) = [E(NF) - E{N(~))I/V'ufNF) + u:N(,,]
and h(t) is approximated by
where (0 5 FT(0) 5 1) = probability of failures at time t = 0 and '1 h(t) = - @[P(t)l
L(t) = reliability function defined as the probability of survival where E( ) and u2( ) are the mean and variance of the variable
during the interval (0, t). If is assumed to be zero, then [31 in parentheses, respectively. These parameters can be estimated
becomes as follows:

and the probability distribution function of time to failure, F d t ) , [131 U ( N ~=

) E(NF) V(NF)
is given by

[5] FT(t)=l-LT(t)=I-exp[-[bh(r)dr]
[14] v2(NF)=v2(m)+ -
E(NF)
E(m)
[ 1' 2 [-'I2
i=l E(NF,)
V2(NFi)

[151 E{N(t)) = E(f,)t

For the reliability analysis of fatigue strength of prestressed
concrete girders, a model was developed which utilizes the S-N [16] a[N(t)] = u(f,)t
 AL-ZAID AND NOWAK
Can. J. Civ. Eng. Downloaded from www.nrcresearchpress.com by Renmin University of China on 05/28/13

FIG. 1. Graphical representation of the proposed fatigue strength reliability model

where V( ) is the coefficient of variation of the variable in number of components (= 4 if partially prestressed and 3 if fully
parentheses. Equations [12] and [14] are obtained by expanding prestressed).
NF in [6] in Taylor series about E(m) and E(NF,) and truncating In [20], the lower bound is the probability of failure for the
the series at the linear terms. girder if its, components are perfectly correlated, and the upper
Equation [9] can also be evaluated in a very simple and bound is the probability of failure for the girder if its
efficient way using the Monte Carlo simulation method. The components are independent. Because all components are
steps are summarized as follows: subjected to the same number of cycles and their stresses are
also related to each other, a high correlation between the failure
1. Rewrite [9] in the form
[17] h(t)
where
= p[X < t] - Fx(t)
modes is expected and the probability of failure for the girder
given by [20] will be close to the lower bound. Furthermore,
because the stress levels in the cast-in-place slab and precast
beam are usually low, the probability of failure of these
For personal use only.

components will be small. Accordingly, the range given by [20]

is also expected to be narrow.
2. Use the computer and the probability distribution func- Fatigue cracking
tions of m, f,, and NFi (assumed to be known) to generate n
values of X. Corrosion of reinforcing steel and prestressing strands is
3. Rank the Xis in a descendant manner and construct the mainly affected by the duration of exposure to moisture than by
probability distribution function of X as crack width (Nowak and Grouni 1986). Therefore, knowledge
of frequency of crack opening during the lifetime of the girder is
1
[I91 F ~ ( x i )=----, ;= 1, ...,r~ of practical significance.
n+ 1 This part of the investigation concerns the determination of
the probability distribution function of time to first cracking.
4. Plot Fx(xi) on a normal probability paper and use [17] to
find h(t) at any given lifetime t. Cracking is defined as the event when the cracking moment of
the critical section is being exceeded by the total moment
Thus, for each structural component, the above procedures applied at that section. After the onset of first cracking the
can be used to calculate its probability of failure in a particular tensile strength of concrete is reduced to zero and the crack stays
period of time. In this study, however, the prestressed concrete open whenever the applied moment MQ exceeds the decompres-
composite girder is modeled as a system of four or three com- sion moment Md,,
ponents depending on whether it is partially prestressed or fully Let Mc,(t) be the cracking moment of the critical section of
prestressed, respectively. These components are the cast-in- the prestressed concrete girder at time t. Mc,(t) depends
place slab, precast beam, prestressing strands, and if partially primarily on the initial prestressing force, and on the material
prestressed, the tension reinforcement. properties and section dimensions of the girder. Mc,(t) is also
Because fatigue failure (crushing of concrete or rupture of affected by the cyclic creep and reduction in the flexural rigidity
steel) is brittle, it is reasonable to assume that the redistribution of the concrete caused by the repetitions of live load. For
of stresses within the section after failure of any of its com- simplicity, this effect is assumed to be taken care of by applying
ponents will result in the subsequent failure of other compon- an average live load movement over the life of the girder. The
ents. With this assumption, the behavior of the prestressed validity of this assumption is checked in the numerical example.
concrete composite girder is like a series system. Therefore, its Let be the probability of cracking of girder k at time t
probability of failure within a particular period of time is conditional on survival until time t. If the Monte Carlo
bounded by (Thoft-Christensen and Baker 1982) simulation technique is used to generate n girders (by generating
tl the section dimensions, material properties, and the initial
[20] max FTi(t) 5 PF [in (0,t)l 5 n
1 - .i= 1 [I - F ~ ; ( ~ ) I
-
prestressing force), then Pr,kis given by
[211 P , k = 1 - FM,[MR(~)I
where FTi(t) is the probability distribution function of the time
to failure of component i as obtained from [5], and n is the where FM, is the probability distribution function of the live
 202 CAN. J . CIV. ENG

c
SYMM
I
LANE 1 1
SHOUL~ER

c
.-
4
0
0
c
3
LL
- .

c Stress Range Max. Stress Level

-
.0
Can. J. Civ. Eng. Downloaded from www.nrcresearchpress.com by Renmin University of China on 05/28/13

C
3 .
0
L I
FIG. 2. Cross section of the prestressed concrete girder bridge 4
.-
(I)

n
-
0
E 4
5
z
I

a,
(I)
L
a,
Cj
- I

i
For personal use only.

Stress Range and Maximum Stress Level

FIG. 4. Stress range and maximum stress level spectra in prestressing
strands.

The sample mean and variance of FTnk(t)are determined as

follows:

Also of interest is the probability that a crack opens on the

condition that cracking has occurred. This corresponds to the
64 I probability that the decompression moment of the critical
' 0.0 500 1000 section is being exceeded by the total moment. Such a prob-
Live Load Moments at Midspon Sections ability can be determined using the preceding formulation with
o f the Girders ( kN m ) MCr(t) in [22] substituted with Mdec(t), where Mdec(t) is the
FIG. 3. Girders live load moments spectra. decompression moment of the critical section at time r.

Numerical example
load plus impact moment at the critical section caused by the
daily truck traffic, and MR(t) is given by The developed models are demonstrated by considering a
numerical example. The cross section of a typical prestressed
concrete highway bridge is shown in Fig. 2. The bridge is
where MD is the total dead load moment. Equation [21] gives simply supported with 27.4 m span with 10 AASHTO type IV
the conditional probability of cracking per truck. The rate of girders spaced at 2.4 m. The precast beams were pretensioned
cracking per day is determined as with high relaxation strands. The ages of the precast beam at
release of the prestressing force and at casting the slab are
assumed to be 3 and 30 days, respectively. The parameters of
where ADTT is the average daily truck traffic. Thus, the proba- resistance (dimensions and material properties) are assumed to
bility distribution function of time to first cracking of girder k is be normally distributed. The mean girder web thickness and
given by (see [5]) flange width is 2.54 mm larger than nominal, with the
coefficient of variation V = 0.02; mean girder depth and flange
[241 F , k ( t ) = I - exp [lbh,k
d ~ ] thickness equal to nominal,- V = 0.015; mean effective slab
width is equal to nominal, v = 0.15; mean slab thickness is 0.76
 AL-ZAID AND NOWAK 203
Can. J. Civ. Eng. Downloaded from www.nrcresearchpress.com by Renmin University of China on 05/28/13
For personal use only.

M, Mdec ( k N . m )
Scmax % R %
FIG. 6. Effect of variability in the prestressing force at release on the
FIG. 5. Maximum stress level and ratio of minimum to maximum spectra of cracking and decompression moments,
stresses spectra at top of the precast beam.

mm larger than nominal, standard deviation is 11.94 mm; mean

effective depth of steel in the girder is 1.52 mm larger than
nominal; mean effective depth of steel in the slab is 10.16 mm;
mean area of prestressing steel is 1.O1 x nominal, V = 0.0125; Pi 50 yrs
mean area of reinforcing steel is equal to nominal, v = 0.025;
mean modulus of elasticity of prestressing steel is 1.011 x
nominal, V = 0.01; mean modulus of elasticity for reinforcing
steel is equal to nominal, V = 0.025; mean ultimate strength of
prestressing steel is 1.04 x nominal, V = 0.025; mean yield
stress of prestressing steel is 1.03 x nominal, V = 0.022; mean
Mdec
compressive strength of concrete is 23.46 MPa with V = 0.18
for f: = 27.6 MPa, and the mean is 27.60 MPa with V = 0.15
for = 34.5 MPa. Details of the numerical computations are
presented next.
Stress sDectra
The spectra of the live load moments at the midspan sections
of the girders were taken from Nowak and Zhou (1985) and are I
shown in Fig. 3 on normal probability paper. The critical girder
a,
(girder 2) was analyzed. Unshored construction was assumed. ln
The analytical model developed in Al-Zaid (1986) was used to a,
construct the stress spectra in the prestressing strands and at top 2 I
of the precast beam. only the uppkr tail of the distribution of the
live load moment, which represents 16% of the traffic, was
considered. It was approximated by a normal distribution.
Material properties, section dimensions, and hence the dead r-5
load moments were taken as the average values. The Monte I
Carlo simulation method was used. The results are presented in
Figs.
" 4 and 5.
Available fatigue strength models reviewed in the reference FIG. 7. Effect of the variability in the live load on the spectra of
(Al-Zaid 1986) were based on experimental data obtained for cracking and decompression moments.
 CAN. I. CIV. ENG. VOL. 15, 1988
Can. J. Civ. Eng. Downloaded from www.nrcresearchpress.com by Renmin University of China on 05/28/13

Age of The Precast Beam (Days)

For personal use only.

Age of T h e Precast Beam (Days)

FIG.8. Variation of the cracking and decompression reliability of the girder with time. 0 , shored construction; A, unshored construction.

the materials under higher stress ranges and maximum stress prestressing force. The variability in the live load moment has
levels. Some materials (e.g., reinforcing bars) have been no significant effect. This justified the assumption made
observed to have endurance limits. Hence, extrapolation of previously where the effect of live load on the cracking and
available models to include the observed low stress levels (Figs. decompression moments of the girder was assumed to be
4 and 5) should be based on experimental data. Therefore, the represented by the effect of an average live load moment.
fatigue strength analysis was terminated at this stage. Following the presented procedure, a total of 500 random
sections with random material properties were generated using
Cracking reliability the Monte Carlo technique. The analysis was done for both
The same girder analyzed in the preceding section was shored and unshored precast beam during construction. The
considered. First, the effects of two parameters on the cracking variations of the cracking and decompression reliabilities of the
and decompression moments distribution functions were inves- girder with time are shown in Fig. 8.
tigated. These included the initial prestressing force and the
creep caused by live load. Conclusions
The resulting distribution functions are presented in Figs. 6 The salient conclusions drawn from this investigation are as
and 7. The cracking and decompression moments of the girder follows:
were observed to be very sensitive to the variability in the initial - Based on the available live load models (MTC truck
 AL-ZAID A N D NOWAK 205
survey, Aganval and Wolkowicz 1976) and bridge deck analysis AASHTO. 1983. Standard specifications for highway bridges. 12th ed.
methods (semi-continuum method), the girders load spectra American Association of State Highway and Transportation Officials,
revealed that the AASHTO distribution factor (AASHTO 1983) Washington, DC. \
is very conservative. For example, in the critical girder of the ACI COMMITTEE 215. 1974. Considerations for design of concrete
bridge considered in the numerical example, the maximum structures subjected to fatigue loading. American Concrete Institute
Journal, 71(3): 97-121.
observed midspan moment was 1017 k N . m compared with the
ACI COMMITTEE 343. 1977. Analysis and design of reinforced con-
corresponding AASHTO value of 1627 kN . m (Fig. 3). crete bridge structures. American Concrete Institute, Detroit, MI.,
- Because of the effect of the distribution of the transverse pp. 67-68.
positions of trucks on the highway bridges, the girders were AL-ZAID,R. Z. 1986. Fatigue reliability of prestressed concrete girder
observed to encounter different load spectra (Fig. 3). bridges. Ph.D thesis, University of Michigan, Ann Arbor, MI.
- In prestressed concrete girders designed according to HOLMEN, J . 0 . 1979. Fatigue of concrete by constant and variable
AASHTO specifications, the stress range spectrum in the amplitude loading. Bulletin No. 79-1, Division of Concrete Struc-
Can. J. Civ. Eng. Downloaded from www.nrcresearchpress.com by Renmin University of China on 05/28/13

prestressing strands, calculated for an average section with tures, NTH-Trondheim, Norway.
average material properties, showed a small variation with the KRAKER,A. D. E., TICHLER, J. W., and VROUWENVELDER, A. C. W. M.
variation in the live load level. The maximum stress range 1982. Safety, reliability and service life of structures. Heron Nether-
lands, 27( 1 ).
observed was about 2 % of the ultimate strength of strands (Fig. MINER,M. A. 1945. Cumulative damage in fatigue. Transactions of
4). the American Society of Mechanical Engineers. 67: A 159-A 164.
- Girders shored during construction showed a better NOWAK, A. S., and GROUNI,H. N. 1986. Serviceability criteria in
resistance to cracking and decompression that unshored ones. prestressed concrete bridges. ACI Journal, Proceedings, 83(1):
For instance, in the example considered, the probability of 43-49.
cracking in 50 years life of the girder was found to be 0 . 1 4 and NOWAK,A. S., and ZHOU,J. 1985. Reliability models for bridge
0 . 2 for shored and unshored girders, respectively. analysis. Report No. UMCE 85-3. Department of Civil Engineering,
-In summary, under the assumptions made in this investiga- University of Michigan, Ann Arbor, MI.
tion, it can be concluded that fatigue of prestressed concrete NYMAN, W. E., and MOSES,F. 1985. Calibration of bridge fatigue
girder bridges designed according to the current AASHTO design model. ASCE Journal of Structural Engineering, l l l ( 6 ) :
1251-1266.
specifications does not appear to be a problem.
SIRIAKSORN, A. 1980. Serviceability and reliability analysis of partially
Acknowledgements prestressed concrete beams. Ph.D thesis, University of Illinois at
Chicago, Chicago, IL.
For personal use only.

This research was supported by a fellowship award from the THOFT-CHRISTENSEN, P., and BAKER,M. J. 1982. Structural relia-
King Saud University, Saudi Arabia. It was also partially bility theory and its applications. Springer-Verlag, New York, N.Y.
supported by the National Science Foundation under grant No. WANG,D. F., and YAO,J. T. P. 1983. Fatigue damage and structural
ECE-8413274 to The University of Michigan with Dr. J. Scalzi reliability. ASCE Annual Convention, Philadelphia, Pennysylvan-
ia, May 16-19.
as Program Director. The authors are very grateful for this
WIRSCHING, P. H.,ET AL. 1982. Fatigue and fracture reliability: A
support. state-of-the-art review. Report of Committee on Fatigue and Fracture
AGARWAL, A. C., and WOLKOWICZ, M. 1976. Ontario commercial Reliability, Journal of the Structural Division, ASCE, lOS(ST1):
vehicle survey, 1975. Interim Report, The Ministry of Transportation 1-88.
and Communications, Ontario.

View publication stats