Tailor Made Concrete Structures – Walraven & Stoelhorst (eds)

© 2008 Taylor & Francis Group, London, ISBN 978-0-415-47535-8

Corrosion monitoring for underground and submerged concrete

structures – examples and interpretation issues

R.B. Polder, W.H.A. Peelen & G. Leegwater

TNO Built Environment and Geosciences, Delft, The Netherlands

ABSTRACT: Since about 1980 Corrosion Monitoring Systems have been used in many concrete structures in
aggressive environment worldwide. While these systems work properly in aboveground environment, some ques-
tions have arisen for submerged conditions, e.g. the outer sides of tunnels, piers in seawater or foundations in wet
soil. One question concerns macro-cell formation between reinforcement in submerged concrete and in nearby
aerated concrete, which might lead to severe corrosion for certain types or configurations of structures (“hollow
leg”). In addition, in some cases, unexpected monitoring signals have been measured in submerged structures.
The interpretation of electrochemical methods for monitoring the corrosion activity is not straightforward and
new criteria have to be developed. This paper reports on an example of corrosion monitoring of an underground
structure, the Green Heart Tunnel in The Netherlands. New criteria for interpretation of underground corrosion
monitoring are proposed.

1 INTRODUCTION during the service life with appropriate reliability.

In the Service Life Design (SLD) this performance
The concept of reliability based service life design of criterion was quantified using the DuraCrete models
concrete structures dates back to the 1980s (Siemes for chloride and carbonation induced corrosion.
et al., 1985). DuraCrete (2000) proposed a practi- The CMS was based on sensors to monitor the elec-
cal methodology, which has become internationally trical resistance of concrete (indicating its humidity,
accepted (fib, 2006). Important concrete infrastruc- related to carbonation) and the potential and galvanic
ture is increasingly being designed for long service life current density of steel electrodes in the outer concrete
by modelling of degradations and probabilistic assess- skin (indicating depassivation due to chloride penetra-
ment of uncertainties (DuraCrete, 2000). Service life tion). The choice of locations for monitoring sensors
calculations in the design phase are necessarily based within the tunnel was determined by the pilot character
on modelling of degradation processes and aggressive of the project, i.e. a mix of practical and environmental
loads and on testing of trial mixes for their resistance. (ground water chloride content) considerations.
In reality aggressive loads could be more severe or less Although state-of-the-art sensors were used, the
severe than assumed in the design phase. Thus, mon- application in an underground structure is atypical and
itoring of the performance could provide added value questions exist on the interpretation of the results, as
for the management of the structure. mentioned above. Supporting research was conducted
Corrosion monitoring systems (CMS) for concrete on ‘dummy’ GHT concrete elements with built-in sen-
structures have been used for some time with sat- sors, which were exposed to aggressive conditions
isfactory results. However, in submerged structures in the laboratory. This paper aims at assessing the
unexpected signals have been observed and new crite- performance of the CMS in the tunnel and in particu-
ria need to be developed (Raupach et al., 2007). The lar the criteria to distinguish active/passive corrosion
Green Heart Tunnel (GHT) is used as a test case. behaviour of the sensors.
In the particular case of the GHT, the durabil-
ity performance aspects under consideration are the
serviceability (water tightness) and safety (structural 2 GREEN HEART TUNNEL AND ITS SERVICE
integrity) of the tunnel lining that may be impaired LIFE DESIGN
due to chloride induced corrosion of the outer rein-
forcement and carbonation at the inner reinforcement. The Green Heart Tunnel (GHT) is a bored tunnel in
The performance criterion refers to the initiation of the High Speed train Link (HSL) with a length of
reinforcement corrosion, which should be prevented about 8.6 km and a single tube of 14 m inner diameter

187
 six steel bars (further referred to as steps), embedded
parallel to the concrete surface at different depths of
about 10; 20; 30; 40; 50 and 60 mm. The steps are made
of steel comparable to reinforcing steel and similar
corrosion characteristics are assumed. Due to chlo-
ride ingress a profile will develop with higher chloride
contents at the exterior of the elements. Over time,
the critical chloride content for corrosion initiation
will be exceeded for subsequent steps of the AL; con-
sequently they will depassivate and start to corrode.
A potential difference develops between an actively
corroding step and the passive reinforcing cage. By
monitoring the potential of a step versus the reinforce-
ment, the onset of corrosion can be determined. Upon
connecting the step and the reinforcement through
a low-resistance ampere meter, the galvanic current
between a step and the reinforcement can be mea-
Figure 1. Impression of Green Heart high speed railway sured as an indication of the macro-cell current. A
tunnel. high galvanic current indicates active corrosion. More-
over the electrical resistance between the subsequent
steps can be measured and converted to a resistivity
(Fig. 1). The lining consists of rings of ten precast (cell constant about 0.1 m). Measuring potential (open
reinforced concrete segments; a separation wall is cast circuit) and galvanic current (short circuit) are mutu-
in situ between the two tracks. The precast concrete ally exclusive. Normally the circuit is open and steady
composition was 450 kg/m3 CEM III/A 52.5 N LA, state potential differences are monitored. Upon short
aggregate of 28 mm and a w/c of 0.39. The Client, circuiting, the galvanic current is non-steady state
ProjectOrganisatie HSL-Zuid, required a technical ser- (decreasing with time) and thus the moment of measur-
vice life of 100 years. Deterioration mechanisms to ing after closing the circuit is important. It should be
be considered included reinforcement corrosion due realized that this also means that the galvanic currents
to carbonation and chloride ingress. Chloride expo- observed are a strong overestimation of the real steady
sure should be based on seawater, considering pos- state corrosion current density. Sensortec recommends
sible leakage of ground water. The ground water is as the boundary between active and passive conditions
presently brackish (maximum c. 5 g of chloride ion a potential difference of 100 mV and a current of 15 µA
per litre), but may become more saline in the future (measured 5 s after short circuiting), corresponding to
(supposedly c. 20 g/l). Models and input should be an average current density of about 10 mA/m2 . These
state-of-the-art, based on DuraCrete. Specific require- values are further called “conventional criteria”.
ments were given: a maximum chloride diffusivity of Two tunnel rings were provided with sensors. One
5*10−12 m2 /s, tested using the Rapid Chloride Migra- ring had 20 MRE’s embedded (both interior and exte-
tion (RCM) method (at 28 days). Various uncertainties rior). In the second ring, nine elements were equipped
turned up after the SLD was made. As the most rel- with one AL in the exterior face and one MRE in the
evant here, macro cell corrosion was identified as a interior face, as sketched in Fig. 2. The AL elements
potential threat, with increased corrosion rates due to were produced in September 2003 and installed in the
galvanic coupling between depassivated bars on the tunnel in October 2003; after initial tests, regular in situ
outer side and passivated but well aerated bars on the measurements were taken from February 2005.
inner side. Monitoring was thought a way to cope with
uncertainties; however, a general lack of experience
existed. Thus, a monitoring system was installed as a 4 LABORATORY TESTS
pilot project, which also included studying elements
with sensors in the laboratory. Dummy elements were produced as trial objects, one
with three MRE’s and one with one AL and one
MRE (called AL dummy). After accepting them as
3 MONITORING SYSTEM successfully produced, they were placed outside with-
out shelter. In 2004/2005, a test area of 0.3 m × 0.4 m
Two types of sensors from S + R Sensortec GmbH over the AL sensor in the AL dummy was exposed
were chosen: Multi Ring Electrodes (MRE) and Anode to salt water with 20 g chloride/l for 180 days. Steel
Ladders sensors (AL). MRE sensors measure concrete potentials, galvanic currents and concrete resistivi-
resistance from which moisture content can be derived. ties were measured periodically using the embedded
The present focus is on the AL sensors. They consist of sensors. After 180 days the solution was removed and

188
Interior surface Exterior surface criteria for passive/active behaviour mentioned above,
that steel depassivation had already occurred on a wide
Multiring-Section 1 Anode Ladder scale. In view of the short exposure and the relatively
low chloride content of the ground water, this is highly
unlikely. Critical evaluation of the criteria appeared
necessary, which is discussed below.
Cable

5.2 Dummy element testing

Laboratory testing of the test area in the AL-dummy
Figure 2. Principal cross section of element with Multi Ring element during 180 days exposure to 20 g/l chlo-
Electrode (left, inside) and Anode Ladder (right, outside); salt ride solution and of cores taken before and after the
ground water on right hand side. exposure showed the following.
The Anode Ladder sensors clearly respond to salt
Table 1. Summary of data measured in GHT for all steps in water penetration; steel potentials become more neg-
anode ladder sensors, versus reinforcement, February 2006; ative, galvanic current densities increase and concrete
a ± b indicates mean ± standard deviation; * suspect results. resistivity drops, corresponding to penetration of water
and chloride in space and time.
Potential difference Galvanic current The penetration of water is faster than the pene-
Element (mV) (mA/m2 ) tration of chloride ions; apparently they are separated
during transport. After 180 days of exposure, chloride
S01 −100 ± 40 0 ± 10 ions have penetrated the concrete to depths of about
R1 −330 ± 5 −80 ± 10 15 mm and water has penetrated 30–40 mm.
R2 −330 ± 30 0±5 Cores exposed to salt water in the laboratory show
S04 −390 ± 80 −90 ± 20
similar penetration of water and chloride as the test
S06 −420 ± 80 −70 ± 30
S07 −250 ± 20 −50 ± 20 area. In cores that had been stored in 20◦ C 80% RH
S08 −440 ± 90 −70 ± 20 before exposure, chloride surface contents were 0.65%
S09 −410 ± 80 −70 ± 20 by mass of cement in 5 g/l chloride solution and 1.5%
mean −290 ± 40 −47 ± 13 in 20 g/l solution. These surface contents will prob-
ably increase over time. For various levels of water
saturation and chloride concentration, apparent chlo-
ride diffusion coefficients after 180 days were in the
cores were taken and analysed for chloride penetration. range of 2 to 11*10−12 m2 /s.
Other cores were subjected to salt water exposure in The potential difference between the AL steps and
the laboratory and analysed for chloride penetration. the reinforcement ranged from −60 to −600 mV, with
Before the exposure, concrete resistivity was measured a mean of −325 mV; galvanic current density ranged
on the surface using both Wenner and two-point meth- from 0 to −400 mA/m2 , mean −100 mA/m2 .
ods (Polder, 2001). In 2006, another series of exposure Assuming 0.5% of chloride by mass of cement for
testing was started. Firstly all potentials, resistances the critical content, only the first step of the AL sen-
and galvanic currents were measured after more than sor at a depth of 10 mm may be supposed to have
a year with only normal exposure to outside weather. undergone corrosion initiation during 180 days of salt
Then the surface was exposed to water for 35 days and water exposure. The other steps should not have depas-
subsequently to saturated NaCl solution (360 g/l, still sivated. For these steps, the electrochemical criteria
running). Potentials, resistances and galvanic currents suggest active behaviour, which is highly unlikely;
were measured periodically. consequently, the need revision.
The best criterion to distinguish between passive
and active behaviour of the Anode Ladder steps is
5 RESULTS AND DISCUSSION
based on the galvanic current density, with a pas-
sive/active boundary of about 200 mA/m2 of steel
5.1 Measurements in the tunnel
surface area. The potential difference to reinforcing
In Table 1, potential difference and galvanic current steel does not provide a good criterion.
results measured in the tunnel are summarised. At 20◦ C and in the fog room the concrete resistiv-
Measured values differed statistically significantly ity of drilled cores was about 500
m and in 20◦ C
between elements, but showed an increasingly 80% RH about 700
m, generally similar to surface
negative trend in time. The overall mean values early measurements on the element.
2006 for steps at all depths were about −300 mV and The non-wetted side of the element showed a
about 50 mA/m2 . This suggests by the conventional generally stable resistivity at intermediate levels

189
(1000–3000
m), indicating a moderate and relatively cement concrete sheet piles, without active corrosion
stable moisture content. (Peelen & Polder, 2004). Even in the (lower) splash
Further testing in 2006 on the element surface and zone of slag cement structures, low potentials can be
then exposure to water for 35 days and subsequently to found without actual corrosion (Rooij & Polder, 2005).
saturated NaCl solution (360 g/l) for another 15 days Under normal atmospheric conditions, in which the
showed the following. concrete and the steel are well aerated, normal passive
Before water exposure (so in “atmospheric condi- potentials between +100 and −100 mV (SCE) develop
tion”), the outer step 1 showed a negative potential in slag cement concrete, like in Portland cement con-
difference of −200 mV and a galvanic current of crete. As a working hypothesis, oxygen starvation
−13 mA/m2 , both indicating active corrosion by the may influence potentials and galvanic currents in sub-
conventional criteria; other steps showed much smaller merged concrete in such a way that passive/active
values, suggesting passive behaviour. potential boundaries change. Oxidation of reduced
After 35 days of water exposure, step 1 potential species in slag may also contribute to the galvanic
was −440 mV and its galvanic current increased to current.
−165 mA/m2 ; potential differences for steps 2 to 6 From the previous experiments on the AL-dummy
ranged from 0 to −100 mV, galvanic currents from element it was suggested that the step-reinforcement
+2 to −5 mA/m2 , suggesting passive behaviour. After potential was not a good criterion for distinguishing
another 15 days with saturated NaCl, many poten- passive/active conditions. The conventional galvanic
tials exceeded −100 mV and galvanic currents were current boundary value was also not found suitable.
−700 mA/m2 for step 1 and 0 to −6 mA/m2 for other Instead a galvanic current density of 200 mA/m2 was
steps; proposed (corresponding to 300 µA measured current)
for very wet and/or submerged (slag) concrete. As dis-
cussed above, low oxygen availability may cause the
5.3 Interpretation of signals
potential to drop to negative values without actual
Interpreting the signals from the AL steps using the corrosion initiation. Potentials as low as −600 to
conventional criteria, only the data taken under atmo- −800 mV (versus SCE) could develop without real
spheric conditions appear to produce realistic results: depassivation (“imperfect passivation”). In a tunnel
step 1 corrodes, the other steps are passive. This makes element which is submerged on one side and aerated
sense assuming that the chloride profile measured in on the other side, current is exchanged between con-
April 2005 (at the end of the 180 days exposure) has nected steel bars and the potential of either group of
not significantly changed. bars will assume a mixed potential value between that
After 35 days of water exposure and another 15 of the non-aerated steel on the outer side and the aer-
days of saturated chloride exposure, the same steps ated steel on the inner side. Attempts to calculate such
produce quite different potentials and galvanic cur- potentials have been reported recently (Redaelli et al.,
rents. In particular steps 2, 3 and 6 have potentials 2006). For a free corrosion potential of −600 mV of
that suggest active corrosion. It is possible that step 2 non-aerated steel, an actively corroding bar on the out-
has undergone corrosion initiation, but it is unlikely side has a potential of −539 mV (SCE) and a passive
that the corrosion conditions of steps 3 and 6 have bar on the outside −532 mV, while the (connected)
changed dramatically in such a short time. The pre- steel in the inner side has a potential of −173 mV (free
vious experiments have shown that the penetration of corrosion potential +100 mV). In that study, the resis-
water is so quick that the water content at the steel tivity was rather low (80
m in wet conditions and
changes in a few days to weeks. Apparently the mea- 1000
m in semi-dry conditions). The higher resis-
surements taken under different conditions with regard tivity of the GHT case will have a limited effect on
to wetting should be interpreted using different criteria the calculated potentials, so the results can be used
and/or boundary values. for our present analysis. It appears that both corrod-
There is a mechanistic basis for such a change of ing and non-corroding bars can have very negative
interpretation. The potential of passive steel strongly potentials under water-saturated and non-aerated con-
depends on the presence of oxygen at the steel/concrete ditions. In a tunnel, potentials of −539 mV (active)
interface. Under water saturated (permanently sub- and −532 mV (passive) can hardly be distinguished.
merged) conditions the potential may be as low as This can be taken as an explanation for the similarly
−800 mV (SCE) with a negligible corrosion rate negative potentials of active and passive ladder steps
(Arup, 1983; Bertolini et al., 2004) due to oxygen star- in the tunnel. In our laboratory experiments, oxygen
vation. In our tests, the blast furnace slag cement may access could be somewhat higher than in the tunnel,
also have influenced the potentials. Under limited oxy- resulting in potentials in the range between −400 and
gen access, reduced substances in the slag (sulphides, −700 mV. On the other hand, this explanation sug-
low valence manganese species) may cause negative gests that in the tunnel eventually all steel near the
potentials without actual steel corrosion. Strongly neg- wetted side may assume negative potentials, includ-
ative potentials were also found in submerged slag ing the reinforcement cage, even if true depassivation

190
is absent. The potential difference between steps and monitoring system function well. In November 2006,
reinforcement will then disappear. So far, this has a new system with automated data recording and
not as yet occurred in the AL-dummy element (with wireless readout was installed.
supposedly better oxygen access than in the tunnel). The most critical issue for the GHT CMS is the
The expectation is that in time, step-reinforcement interpretation of signals from Anode Ladder steps in
potential differences will decrease, in particular in terms of corrosion initiation. The conventional cri-
the tunnel. This proves again that potential differences teria (potential difference and galvanic current) and
cannot be a useful criterion in submerged conditions. boundary values (100 mV and 15 µA, respectively)
It should be taken into account that reduced species in produce doubtful results. For many sensor steps in
slag cement have not been considered in this discus- the tunnel these criteria suggest corrosion initiation,
sion. If they are important, their effect will most likely while the local conditions most likely have not initiated
support the analysis given here. corrosion. Experiments on the AL-dummy element
Regarding the absolute value of galvanic current under laboratory conditions have shown that prob-
density, the following observation is made. A current ably passive steps exceed the conventional criteria,
density of 200 mA/m2 theoretically corresponds to a if the concrete surface over the steps is exposed to
steel thickness loss of about 200 µm/year. This would water and/or salt solution for a few weeks. This can be
be an extremely high corrosion rate, which in our expe- explained from the electrochemical behaviour of steel
rience occurs rarely in practice. However, high current in water saturated concrete, in particular from the low
densities may exist for the mechanism of macrocell availability of oxygen. Absence of oxygen depresses
corrosion. It could be speculated that oxidation of the the potential to more negative values, even without cor-
reduced slag components contributes to such high cur- rosion inducing chloride levels; this condition may be
rent densities in our case. Current decrease has been termed imperfect passivation. Consequently, a modifi-
observed over connection times from less than a sec- cation of the criteria and boundary values is proposed.
ond to 30 s. Longer connection times will expectedly First, the potential should be no longer used as a
produce lower galvanic current densities. In any case, criterion for distinction between (truly) active and pas-
to interpret galvanic currents in terms of corrosion sive behaviour. Second, the active/passive boundary
rates, more work is needed. In particular measuring value for the galvanic current is proposed to be set at
galvanic current while maintaining the connection for 200 mA/m2 , corresponding to 300 µA. These values
longer times would be valuable. apply for water saturated (or at least very wet) slag
Based on the results and discussion presented, cement concrete.
we propose that the (conventional) criteria for dis- Based on the conclusions with regard to the elec-
tinguishing active and passive steel in submerged trical performance of the monitoring system and the
concrete need to be modified. In aerated concrete, modified interpretation, the Durability Monitoring
a potential difference between ladder steps and rein- System in the Green Heart Tunnel is concluded to
forcement cage of 100 mV and a galvanic current perform well. It is expected that, using modified cri-
density of 10 mA/m2 may well indicate the pas- teria and boundary values, signals can be interpreted
sive/active boundary. Our experiments support that properly with regard to chloride ingress and corro-
position. In very wet conditions (as in submerged sion initiation. Once corrosion has initiated, macro
concrete), the potential difference is no longer a cell corrosion is a significant mechanism. This allows
useful criterion. Galvanic current density may be use- drawing conclusions on the performance of the tunnel
ful, but the boundary value must be increased. At with regard to durability. Further work is underway to
present it appears that 200 mA/m2 is a proper bound- investigate the significance of monitoring results with
ary between active and passive behaviour of steel regard to updating the original design service life of
electrodes (here AL steps) in submerged slag cement the tunnel.
concrete. Due to the supposed influence of reduced
species in the slag, this value may not necessar-
ily apply to Portland cement concrete. Expectedly, ACKNOWLEDGEMENTS
however, the principle of modifying the criteria is
the same for submerged Portland cement concrete The authors acknowledge the cooperation of HSL-
structures. Zuid with the work reported here and their permission
to use the results. We thank in particular the supervi-
sion and discussions provided by Robert Jan Aartsen,
6 CONCLUSIONS Alain Kooiman, Kees Blom, Sander Lokhorst and
Joost Gulikers who represented HSL-Zuid in various
From the measurements on the durability monitor- stages.
ing system in the GHT taken so far, it appears that This study is part of National Programme Delft
all sensors produce correct signals in an electrical Cluster, project DC02.30 “Smart sustainable man-
sense. Sensors, cabling and data readout of the original agement of concrete structures” as Work Package

191
02.33.21 “Technical integration and interpretation of fib, 2006, Model Code for Service Life Design, Task Group
monitoring results for maintenance management”. 5.6, fib Bulletin 34, Zurich
Finally, the authors acknowledge valuable discus- Peelen, W.H.A., Polder, R.B., 2004, Durability assessment of
sions with European Federation of Corrosion Task the concrete sheet piling for the new Dutch heavy-duty
‘De Betuweroute’ railway line, Corrosion Prevention &
Group members Michael Raupach (ibac), Birit Buhr Control, 11–16
Jensen, Ruth Sørensen, Thomas Frolund (COWI) and Polder, R.B., 2001, Test methods for on site measurement
Oskar Klinghoffer and Peter Nygaard (FORCE) of a of resistivity of concrete – A RILEM TC-154 Technical
previous version of this paper. Recommendation, Construction and Building Materials,
Vol. 15 (2–3), 125–132
Raupach, M., Polder, R.B., Frolund, T., Nygaard, P., 2007,
REFERENCES Corrosion monitoring at submerged concrete structures –
macrocell corrosion due to contact with aerated areas?,
Arup, H., 1983, The mechanism of the protection of steel by Eurocorr07, Freiburg
concrete, Proc. Corrosion of Reinforcement in Concrete, Redaelli, E., Bertolini, L., Peelen, W., Polder, R., 2006,
ed. A.P. Crane, Society of Chemical Industry, 151–157 FEM-models for the propagation of chloride induced rein-
Bertolini, L., Elsener, B., Pedeferri, P., Polder, R.B. 2004. Cor- forcement corrosion, Materials & Corrosion, Vol. 57, (8),
rosion of Steel in Concrete: Prevention, Diagnosis, Repair, 628–635
Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim, ISBN Rooij, M.R. de, Polder, R.B., 2005, Duurzaamheid mariene
3-527-30800-8, 392 p betonconstructies (Durability of marine concrete struc-
DuraCrete, 2000, DuraCrete Final Technical Report R17, tures), Report 215, CUR, Gouda, in Dutch
Document BE95-1347/R17, The European Union – Brite Siemes, T., Vrouwenvelder, T., Beukel, A. van den, 1985,
EuRam III, DuraCrete – Probabilistic Performance based Durability of buildings: a reliability analysis, HERON,
Durability Design of Concrete Structures, CUR, Gouda Vol. 30 (3), Delft, 2–48

192