International Journal of Structural Stability and Dynamics

Vol. 19, No. 1 (2019) 1940015 (17 pages)

#.c World Scienti¯c Publishing Company
DOI: 10.1142/S0219455419400157
by UNIVERSITY OF NEW ENGLAND on 07/10/18. Re-use and distribution is strictly not permitted, except for Open Access articles.

Neutral Axis-Based Health Monitoring and Condition

Assessment Techniques for Concrete Box Girder Bridges

Ye Xia*,‡, Peng Wang*,§ and Limin Sun*,†,¶

*Department of Bridge Engineering
Tongji University, 1239 Siping Rd.
Int. J. Str. Stab. Dyn. Downloaded from www.worldscientific.com

Shanghai 200092, P. R. China

†
State Key Laboratory for Disaster Reduction in Civil Engineering
Tongji University, 1239 Siping Rd.
Shanghai 200092, P. R. China
‡yxia@tongji.edu.cn
§1630669@tongji.edu.cn
¶
lmsun@tongji.edu.cn

Received 15 February 2018

Accepted 12 May 2018
Published 9 July 2018

The neutral axis position is considered as a potential parameter of early abnormal and long term
performance for °exural members of structures. This paper is to propose the Neutral Axis
Indicator (NA) based on strain measurements, to evaluate the behavior of NA, and to investigate
challenges associated with the neutral axis for structural health monitoring purpose. A numerical
example of a simply supported reinforced concrete (RC) beam was ¯rst developed and analyzed
followed by an experimental study for further veri¯cation. With the aim of application to real
bridges, a case study was conducted on Gantao River Bridge, a typical concrete box girder bridge,
for detailed analysis using ¯eld data. The numerical and experimental examples, as well as ¯eld
tests, aim to ¯gure out: (1) the sensitivity of NA to structural damages, especially at early age;
(2) stability and uncertainty of NA using ¯eld data; (3) capability of NA on detecting abrupt and
slowly changing structural degradation. The results show that: NA takes on good index accuracy
and damage sensitivity on numerical and experimental cases; NA tends to keep constant during
linear elastic phase, while responses quickly and sharply to nonlinear behaviors such as concrete
cracking and hardening; and NA is a potentially reliable indicator both to capture nonlinear
performance and to detect minute damage. Additionally, the uncertainty of multiple disturbing
e®ects can be mostly eliminated through statistical analysis. Thus, it can be employed e®ectively
in condition assessment of concrete box girder bridges for long-term health monitoring.

Keywords: Neutral axis; strain gauges; ¯eld data; concrete box girder bridge; structural health
monitoring.

¶ Corresponding author.

1940015-1
 Y. Xia, P. Wang & L. Sun

1. Introduction
Bridges degenerate due to severe environment, over loading, initial structural
defects, disasters or some other factors. How to evaluate and ensure the safety of
by UNIVERSITY OF NEW ENGLAND on 07/10/18. Re-use and distribution is strictly not permitted, except for Open Access articles.

these structures is a question attracting enough attention. So far great progress has
taken place in the ¯eld of structural health monitoring (SHM) all around the world,
but mainly focuses on large bridges or irregular bridges. As for middle and small
spanned bridges, typically concrete box girder bridges, much less e®ort has been
made to improve their traditional management and maintenance pattern. Concrete
box girder bridges always consist of °exural members such as beams. Therefore, the
neutral axis position is considered to be a key structural parameter and serve as a
potential indicator for their condition assessment.
Many researchers have investigated the neutral axis position as an indicator
Int. J. Str. Stab. Dyn. Downloaded from www.worldscientific.com

mainly from two aspects. One aspect focuses on how to determine the neutral axis
position based on monitoring data of a structure, i.e. raw data processing. Strain
measurements obtained in on-site conditions are actually a result of combination of
external loads and environmental e®ects (e.g. tra±c, temperature and humidity).1
Moreover, measurements are always contaminated by a certain extent of noise. Xia2
concluded a methodology based on modern signal processing tools. It is expected that
discrete wavelet transform (DWT) algorithm has the potential to extract strain
components induced by tra±c, and noise can be ¯ltered by Kalman ¯lter (KF) esti-
mator. Soman et al.3 analyzed the neutral axis position of a wind turbine tower using
KF. Results shows that the NA-based tracking is a promising method which is low cost
and sensitive to damage at lower levels. Experiments and numerical examples were
conducted by the same group in literature.4 Tang and Ren5 proposed a new method to
determine the neutral axis position using modal macro-strain. A macro-strain modal
analysis was carried out using the Fourier Transform in a similar way to traditional
displacement modal analysis. The e®ectiveness of the method was veri¯ed through
¯nite element (FE) models and experiments.
The other aspect lays on the correlation among the neutral axis position and
the capacity, performance, condition of a target structure.6,7 A double-composite
steel specimen was studied by Stroh et al.8 The neutral axis moved down along the
cross-section with increasing number of loading cycles due to crack formations in
the top concrete slab. The study presents an example of successful damage de-
tection using the neutral axis position. Plude et al.9 studied an FE model of a
highway bridge with steel girders supporting a concrete slab. The sensitivity of
three damage features were compared, namely natural frequency, strain distri-
bution, and the neutral axis position. Among these three parameters, the neutral
axis showed the best performance in indicating delamination and deck deteriora-
tion. Sigurdardottir and Glisic10 performed a detailed uncertainty analysis of the
neutral axis position based on long-term monitoring and controlled testing of a
concrete footbridge. The main sources of uncertainty in long-term evaluation were
creep and shrinkage e®ects as well as in°uences from nonlinear thermal gradients.

1940015-2
 Neutral Axis-Based Health Monitoring Techniques

The same group of researchers studied a large scale testing model of a composite
bridge by introducing arti¯cial damage into the concrete deck with plastic
sheets.11 The structure was tested under its self-weight and a moving truck load
and the neutral axis position was observed to be sensitive to damage.
by UNIVERSITY OF NEW ENGLAND on 07/10/18. Re-use and distribution is strictly not permitted, except for Open Access articles.

In summary, researchers have made some achievements using numerical models

or in the laboratory, and e®orts have been made through controlled tests. However,
only part of studies presented here have central focus on the neutral axis, which
means it is often a secondary parameter associated with other indicators. No sys-
tematic study has been conducted using the neutral axis for damage detection,
performance evaluation, and condition assessment. Most importantly, available
researches are mainly based on numerical or experimental data without employment
of ¯eld data. Furthermore, most studies have concentrated on composite models or
Int. J. Str. Stab. Dyn. Downloaded from www.worldscientific.com

structures, while few have taken concrete bridges into consideration. A general
framework for condition assessment of real bridges using strain measurements
remains a challenge. This paper aims at applying neutral axis related index as a
universal indicator for concrete box girder bridges. First, an index NA was de¯ned
based on strain measurements associated with speci¯c instrumentations. Then a
numerical example was developed to study the behavior pattern of NA. In succession
an experiment of a simply supported RC beam was carried out to verify the nu-
merical results. Finally, a case study using ¯eld data from the long-term health
monitoring system of Gantao River Bridge was established.

2. Neutral Axis Indicator

2.1. Theoretical derivation
Neutral axis is de¯ned as a set of points in the cross-section of a beam-like structure,
at which normal stress and strain vanishes under bending moment. All ¯bers on one
side of the neutral axis are in a state of tension, and those on the opposite in
compression. According to the plane section assumption, the strain distribution
should be linear over depth of the cross-section. As illustrated in Fig. 1, the °exural
strain on the cross-section can be expressed by

x 0

x ð þ yÞ




y
"ðyÞ ¼ lim ¼ lim ¼ ; ð1Þ

x!0
x

!0 


where "ðyÞ is the longitudinal strain,
x is the length of the undeformed beam ele-
ment,
x 0 is the length of the deformed beam element,

is the bending angle,  is the
curvature radius of the beam, and y is the distance of the point from the neutral axis.
According to Hooke's law, the stress ðyÞ at any point on the cross-section is
proportional to "ðyÞ by E. The force equilibrium under pure bending can be
expressed by
Z Z Z
EðyÞy
ðyÞdA ¼ EðyÞ"ðyÞdA ¼ dA ¼ 0: ð2Þ


1940015-3
 Y. Xia, P. Wang & L. Sun
by UNIVERSITY OF NEW ENGLAND on 07/10/18. Re-use and distribution is strictly not permitted, except for Open Access articles.
Int. J. Str. Stab. Dyn. Downloaded from www.worldscientific.com

Fig. 1. The °exural strain on the cross-section.

Since  is a constant along a given cross-section, EðyÞ and y are the only variables
that change with respect to dA. Therefore, Eq. (2) can be simpli¯ed as
Z
EðyÞydA ¼ 0: ð3Þ

Equation (3) indicates that the essence of neutral axis is a curve dividing the
material sti®ness of the cross-section equally. And EðyÞ here is actually an equivalent
secant modulus. Due to the material nonlinearity of concrete, EðyÞ changes as
concrete cracking and hardening take place. In fact, the neutral axis position will be
a®ected as long as there is loss of elastic modulus or e®ective sectional area.

2.2. De¯nition of NA on bridges

In most cases, bridge monitoring system is installed after bridge construction,
making it impossible to track dead-load strain history. However, the strain increment
caused by tra±c load or other environmental e®ects can be observed and recorded. In
order to extract valuable information from raw strain measurements, an index
named NA is de¯ned in this paper based on live-load strain measurements, which
indicates the neutral axis position due to live load, as illustrated in Fig. 2.
Strains at bottom and top of the cross-section are denoted by "t and "b ; respec-
tively. Then the relation between "t and "b can be expressed by
"t y
¼ t; ð4Þ
" b yb

1940015-4
 Neutral Axis-Based Health Monitoring Techniques
by UNIVERSITY OF NEW ENGLAND on 07/10/18. Re-use and distribution is strictly not permitted, except for Open Access articles.

Fig. 2. NA in de¯nition of strain measurements.

where yt is the distance between the top strain gauge and the neutral axis, and yb is
the distance between the bottom strain gauge and the neutral axis. It is easy to
achieve yb by
"b
Int. J. Str. Stab. Dyn. Downloaded from www.worldscientific.com

yb ¼ h; ð5Þ
"t þ "b
where h denotes the vertical distance between two sensors. Then NA is determined as
"b
NA ¼ yb þ h 0 ¼ h þ h 0; ð6Þ
"t þ "b
where h 0 denotes the height of the bottom gauge in the section.
The de¯nition of NA provides a practicable and promising method by integrating
the measured strain and structural parameters. The response of a concrete box girder
bridge under tra±c loads is critically determined by its bending behavior, again due
to live load. If there are only two strain sensors available with the cross-section, the
line connecting them is uniquely de¯ned, and NA can be calculated by Eq. (6). In
case of more than two strain sensors available, which leads to better understanding of
the strain distribution along the depth of the cross-section, regression method is used
to determine NA best ¯tting the strain distribution. The type of sensors to acquire
strain data is not the focus of this paper. However, with the wide application of
more durable distributed sensing technologies, Fiber Bragg Grating for example,12–14
NA is expected to generate reliable and critical results for long-term structural
evaluation.

3. Numerical Analysis
3.1. FE model
An FE model of a typical RC beam was developed using commercial software
ABAQUS15 in order to study the performance of NA. The concrete and the rein-
forcement were modeled, respectively and then bonded together. Static two point
loading was applied on the up-surface of the beam to create pure bending moment.
Modeling details are illustrated by Fig. 3.
The material constitutive model is important due to its direct in°uence on the
numerical results of NA under material nonlinearity. The parameters are illustrated
by Fig. 4 to de¯ne the mechanical behavior of concrete and steel bars. An

1940015-5
 Y. Xia, P. Wang & L. Sun
by UNIVERSITY OF NEW ENGLAND on 07/10/18. Re-use and distribution is strictly not permitted, except for Open Access articles.

(a) (b)
Int. J. Str. Stab. Dyn. Downloaded from www.worldscientific.com

(c) (d)

Fig. 3. FE model of a typical RC beam: (a) concrete; (b) reinforcement; (c) constraints; (d) boundary and
loading conditions.

(a) (b)

Fig. 4. Material constitutive model: (a) stress–strain curve for concrete under uniaxial compression and
tension; (b) stress–strain curve for longitudinal rebar and stirrup.

1940015-6
 Neutral Axis-Based Health Monitoring Techniques

elastoplastic model with hardening is employed for compressed concrete, while an

elastic model with softening is employed for concrete under tensile stress. In this
study, the stress–strain curve proposed by code for design of concrete structures16 is
used for concrete under uniaxial compression and tension, as illustrated in Fig. 4(a).
by UNIVERSITY OF NEW ENGLAND on 07/10/18. Re-use and distribution is strictly not permitted, except for Open Access articles.

A perfect elastoplastic model is employed for steel. The bilinear stress–strain curves
of longitudinal rebar and stirrup are shown in Fig. 4(b). The x-axis and y-axis
represents the strain and stress of material under uniaxial loading in compression and
tension, respectively.

3.2. Numerical results

The NA-load curve of the midspan section under static load is shown in Fig. 5. There
is a perfect horizontal line in the early loading as expected. When load reaches
Int. J. Str. Stab. Dyn. Downloaded from www.worldscientific.com

34.5 kN, NA begins to abruptly increase while cracks generate from deck bottom.
Then the increasing rate of NA is obviously slowed at the load of 46.9 kN. Finally,
NA reaches a level signi¯cantly higher than before with a slight decline.
The whole process can be divided into three stages: linear elastic state, concrete
cracking, and compressive concrete hardening. During the linear elastic state, the
value of NA equals to approximate 97.7 mm, which agrees with the calculating
results from Eq. (3) by substituting EðyÞ with the elastic modulus Ec . When the
tensile strain at downside of the section exceeds the cracking limit, the loss of sti®ness
at downside leads the neutral axis position to a new balance. Soon after concrete
cracking, the increasing rate of NA slows down obviously, indicating the hardening of
compressive concrete.
The example in Fig. 5 is a RC beam but prestressed RC beams are more widely
used in middle-small spanned bridges. And the concrete cracking always happens
before the compressive concrete hardening. Thus, it is di±cult to study the e®ect of
concrete hardening on NA alone. Concluded from Fig. 5, the mixed in°uence on NA
by cracking and hardening tends to be a slower increasing speed. In other words, it

Fig. 5. NA-load curve of RC beam under static load.

1940015-7
 Y. Xia, P. Wang & L. Sun
by UNIVERSITY OF NEW ENGLAND on 07/10/18. Re-use and distribution is strictly not permitted, except for Open Access articles.
Int. J. Str. Stab. Dyn. Downloaded from www.worldscientific.com

Fig. 6. NA-load curve of prestressed beam under static load.

can be assumed that concrete hardening may cause the neutral axis position to fall.
To con¯rm the assumption, the original beam was ¯rst prestressed along its longi-
tudinal direction, and then static load was exerted. Since there were certain amounts
of initial compressive stress in the cross-section, the stage of concrete hardening was
shifted before cracking. The results of prestressed RC beam are illustrated by Fig. 6.
To di®er from the former case, it is marked with the legend \(PC)".
The assumption is con¯rmed as shown in Fig. 6. Concrete hardening tends to
a®ect the value of NA in a contrast way against concrete cracking. It can be also
explained according to Eq. (3) that the equivalent secant modulus of concrete
decreases as compressive concrete hardening develops in the section. Thus, the
neutral axis moves downwards to achieve a new balance. However, the e®ect of
concrete hardening is less strong than that of concrete cracking on NA. The phe-
nomenon can hardly be found on RC beams because concrete cracking always hap-
pens ¯rst. But it is considered as a common case on prestressed concrete beams, which
provides a reference to assess the cross-section condition using NA as an indicator.

4. Experimental Study
4.1. Experiment setup
A simply supported RC beam was prepared for static test. The pattern and sensitivity
of NA as an indicator were veri¯ed when material nonlinearity and damage developed
in the cross-section. The total beam length was 1800 mm, and the distance between
the two bearings was 1600 mm. The rectangular section was 120 mm wide and 200 mm
high, reinforced by two steel rebars at upside with 10 mm in diameter and two steel
rebars at downside with 14 mm in diameter. Stirrups were made up of steel rebars
with a diameter of 8 mm. Static load was applied on the specimen by a lifting jack
until its rupture. Four strain sensors were installed along the depth of the midspan
section with a gauge of 150 mm. Detailed information can be found in Fig. 7.

1940015-8
 Neutral Axis-Based Health Monitoring Techniques
by UNIVERSITY OF NEW ENGLAND on 07/10/18. Re-use and distribution is strictly not permitted, except for Open Access articles.

(a)
Int. J. Str. Stab. Dyn. Downloaded from www.worldscientific.com

(b)

(c)

(d)

Fig. 7. The reinforced beam for experimental study (unit: mm): (a) the RC beam; (b) loading positions; (c)
loading devices; (d) measuring points of strain sensors.

1940015-9
 Y. Xia, P. Wang & L. Sun
by UNIVERSITY OF NEW ENGLAND on 07/10/18. Re-use and distribution is strictly not permitted, except for Open Access articles.
Int. J. Str. Stab. Dyn. Downloaded from www.worldscientific.com

Fig. 8. Comparison between numerical and experimental results.

4.2. Experimental results

The NA-load curve of the midspan section is shown in Fig. 8. It can be found that in
the early stage of loading, NA remains almost stable as the midspan section is intact
and material is within linear range. Slight °uctuation is caused by the uncertainty for
strain measurements when the absolute value of strain is small.17 Then NA begins to
rise sharply with the ¯rst crack appearing in the specimen. Compressive concrete
shows nonlinear behavior after cracks formation in midspan and bearing regions.
Finally, NA reaches a level which is signi¯cantly higher than before, with growing
speed slowing, cracks developing and concrete hardening. During the whole experi-
ment, NA responses in a pronounced way when a change takes place in structural
performance.
The two solid lines in Fig. 8 stand for two numerical cases respectively, which
have been discussed in Sec. 3.2. A comparison is made between experimental and
numerical results (solid blue line). It is obvious that they agree well with each other
since the NA-load curve drawn from the experimental results shows the same pattern
as that from numerical simulation. The pattern can be divided into three parts as
mentioned above. Nevertheless, there are still observable di®erence between the two
results, because the true mechanical property of material, especially concrete, is
di±cult to simulate precisely. Based on these results, the index accuracy, sensitivity,
and structural condition relativity of NA are veri¯ed.

5. Case Study on Gantao River Bridge

5.1. Engineering background
NA has been proved to respond in a pronounced way when large deformation
(concrete cracking or hardening) takes place through numerical and experimental
examples. When NA is applied on a real concrete bridge with live-load induced strain
magnitude of 10
5 , however, it is doubtful whether NA may work under such small

1940015-10
 Neutral Axis-Based Health Monitoring Techniques
by UNIVERSITY OF NEW ENGLAND on 07/10/18. Re-use and distribution is strictly not permitted, except for Open Access articles.

(a) (b)
Int. J. Str. Stab. Dyn. Downloaded from www.worldscientific.com

Fig. 9. Photographs of the Gantao River Bridge: (a) elevation view; (b) plan view.

deformation. Hence, it is suggested to adopt live tra±c loading for NA determina-

tion, while the in°uence of random moving loads can be evaluated. Therefore, a case
study in a real bridge is conducted.
The Gantao River Bridge, located in north area of China, is a typical concrete box
girder bridge in need of extra focus on health condition, thus equipped with bridge
health monitoring system. The bridge consists of ¯ve spans in one continuous unit,
with the length of 40m per span. The parallel box girders are reinforced by regular
steel rebars and prestressed tendons. Photographs in elevation and plan view are
shown in Fig. 9.

5.2. SHM system

A typical SHM system was designed and installed on Gantao River Bridge, see
Fig. 10. A few types of sensors were deployed in parallel topologies in order to collect
valuable information to monitor the structural condition. The strain monitoring
subsystem is of interest for the research in this paper. Each strain sensor was inte-
grated with a temperature sensor for temperature compensation. A set of two strain
sensors, installed on both sides of a box girder at speci¯c sections (1/4 and 3/4
quarter span), were employed to capture the strain distribution, and to evaluate the
position of neutral axis. Moreover, a set of three strain sensors were deployed on both
sides of the 1/2 span section, since there usually tends to be larger deformation due to
larger bending moments under tra±c loads.
Figure 11 illustrates the dynamic strain response of Gantao River Bridge under
tra±c loads, where 01-S01 denotes the 1st strain sensor on the section 01, and 02-S02
denotes the 2nd strain sensor on section 02. NA at section 01, 02 and 03 are evaluated
based on the strain distribution pattern, as concluded in Table 1. The maximum
strain was used to calculate NA in order to mitigate measurement error. Strain
gauges were connected to rigid steel bar to increase the measuring length so as to
mitigate the e®ect of concrete material uncertainty. Therefore, the strain value here

1940015-11
 Y. Xia, P. Wang & L. Sun
by UNIVERSITY OF NEW ENGLAND on 07/10/18. Re-use and distribution is strictly not permitted, except for Open Access articles.
Int. J. Str. Stab. Dyn. Downloaded from www.worldscientific.com

Fig. 10. Sensor topology in plan, in elevation and in cross-section view.

is not exactly the strain of the measured concrete surface but the average strain of
78 cm in length and is ampli¯ed at a certain ratio. It can be seen that NA at di®erent
sections are not the same because of the varied sizes. Due to the fact that the bridge
consists of variable cross-section girders, NA changes as section properties (Sections
01, 02 and 03) changes. Besides, noise and other environmental e®ects contribute to
the results. The application of NA to real concrete box girder bridges gets veri¯ed
through the interpretation in Fig. 11.

1940015-12
 Neutral Axis-Based Health Monitoring Techniques
by UNIVERSITY OF NEW ENGLAND on 07/10/18. Re-use and distribution is strictly not permitted, except for Open Access articles.
Int. J. Str. Stab. Dyn. Downloaded from www.worldscientific.com

Fig. 11. NA based on dynamic strain distribution at section 01, 02 and 03 under tra±c loads.

Table 1. Evaluation of NA at section 01, 02 and 03.

Section 01 Section 02 Section 03

Sensor location Strain/" NA/mm Strain/" NA/mm Strain/" NA/mm
Top 31.80 20.22 15.22
Middle / 1264 70.08 1300 / 1217
Bottom 84.25 107.63 44.08

5.3. Uncertainty analysis of NA parameters

The neutral axis-based parameter, i.e. neutral axis location, varies under tra±c
loading due to various uncertainty sources,2,5,10,11 such as: (1) uncertainty related to
the load pattern caused by passing vehicles randomly in number, time and spacing;
(2) uncertainty in the location of load especially regarding the transverse location
that may cause di®erent NA results by the exact same load; (3) uncertainty of
measurement errors of strain gauges; (4) uncertainty of time-variable structural
conditions caused by temperature, boundary conditions, etc.
With these uncertainties involved in the raw data as illustrated in Fig. 12, the NA
can be evaluated using a probabilistic approach and statistical analysis.18 A statis-
tical hypothesis testing was then conducted to verify the distribution mode of
samples. Results are given in Table 2. Normal probability plots are illustrated by
Fig. 13 to present an intuitive understanding.
Figure 14 interprets the statistical results of NA at mid-span, and quarter-span
locations from approximately 1100 events. It is observed from Fig. 14 that: (1) the
NA follows the Gaussian distribution at any location, and the Gaussian parameter

1940015-13
 Y. Xia, P. Wang & L. Sun
by UNIVERSITY OF NEW ENGLAND on 07/10/18. Re-use and distribution is strictly not permitted, except for Open Access articles.

Fig. 12. Traditional method on evaluation of NA from in situ tests.

Int. J. Str. Stab. Dyn. Downloaded from www.worldscientific.com

Table 2. Statistical hypothesis testing: NA follows Gaussian distribution.

Location Section 01 Section 02 Section 03

Mean/mm 1194 1288 1212
Standard deviation/mm 43 20 25
Signi¯cance level (0.01) Fail to reject Fail to reject Fail to reject

 can be used to estimate the true NA of a speci¯c section with improved stability
and accuracy; (2) NA at quarter-span locations with only two measurements (at top
and bottom) come out with larger variance or bandwidth while NA at mid-span
location with three measurements (at top, middle, and bottom) results in smaller
variance or bandwidth and hence more stable and accurate; (3) the e®ect of random
tra±c loading on  can be minimized when sampling size is enough; (4) the uncer-
tainty caused by time-variable structural conditions can be avoided by picking out

(a) (b)

Fig. 13. Normal probability plots of samples from di®erent locations: (a) section 01; (b) section 02; (c)
section 03.

1940015-14
 Int. J. Str. Stab. Dyn. Downloaded from www.worldscientific.com
by UNIVERSITY OF NEW ENGLAND on 07/10/18. Re-use and distribution is strictly not permitted, except for Open Access articles.

(c)

1940015-15
Fig. 13. (Continued )

Fig. 14. Statistical analysis based on histogram of NA from in situ tests.

Neutral Axis-Based Health Monitoring Techniques
 Y. Xia, P. Wang & L. Sun

groups of peaks and valleys of strain response when trucks passing by, and evaluating
NA based on these extracted data; (5) uncertainty of measurement errors can also be
minimized when only peaks and valleys of strain response are used for statistical
analysis.
by UNIVERSITY OF NEW ENGLAND on 07/10/18. Re-use and distribution is strictly not permitted, except for Open Access articles.

6. Conclusion and Remarks

In this paper, an index NA is proposed and investigated as an indicator for condition
assessment of concrete box girder bridges using distributed sensors. Systematic
analysis has been conducted using numerical and experimental examples, as well as
¯eld tests. Based on the results, the following conclusions can be drawn:

(1) Behavior pattern: NA tends to keep constant during linear elastic phase, while
Int. J. Str. Stab. Dyn. Downloaded from www.worldscientific.com

responses quickly and sharply to nonlinear behaviors such as concrete cracking

and hardening, whose in°uences on NA are distinguishable. The tendency of NA
caused by cracking is stronger than that caused by hardening. In addition, NA
works well both under static load and dynamic load.
(2) Damage sensitivity: In experimental study, NA behaves sensitively to structural
damage like cracking. In ¯eld tests, NA index accurately shows clear Gaussian
distribution from statistical analysis, and most importantly, potentially provides
quantitative assessment if with long-term monitoring data which is not available
for time being. Basically, NA index is capable of detecting not only signi¯cant
damage but more importantly also small damages at early stage.
(3) Uncertainty and stability: the uncertainty of many e®ects can be mostly elimi-
nated through statistical analysis, so the NA index stays stable in a long-term
manner thus becomes possible to assess structural condition using long-term
monitoring data.
(4) Real application: NA is proved to be a potentially reliable indicator both to capture
nonlinear performance and to detect minute damage. Hence, it can be expected for
e®ective application in condition assessment for concrete box girder bridges.

Acknowledgment
This paper is supported by the National Key Research and Development Program of
China (2017YFC1500605), Ministry of Transport of the People's Republic of China
(2015318J38230), National Natural Science Foundation of China (51508405), and
Science and Technology Commission of Shanghai Municipality (18DZ1201203,
17DZ1204103).

References
1. Y. Q. Ni, H. W. Xia and X. W. Ye, Neutral-axis position based damage detection of
bridge deck using strain measurement: Numerical and experimental veri¯cations, 6th
European Workshop on Structural Health Monitoring, Dresden, Germany (2012).

1940015-16
 Neutral Axis-Based Health Monitoring Techniques

2. H. W. Xia, SHM based condition assessment of in-service bridge structures using

strain measurement. PhD thesis, The Hong Kong Polytechnic University, Hong Kong
(2012).
3. R. N. Soman, P. H. Malinowski and W. M. Ostachowicz, Neutral axis tracking for damage
by UNIVERSITY OF NEW ENGLAND on 07/10/18. Re-use and distribution is strictly not permitted, except for Open Access articles.

detection in wind turbine towers, Wind Energy 98(8) (2015) 2203–2209.

4. R. Soman, K. Majewska, M. Mieloszyk et al., Application of Kalman Filter based neutral
axis tracking for damage detection in composites structures, Compos. Struct. 184 (2018)
66–77.
5. Y. Tang and Z. Ren, Dynamic method of neutral axis position determination and damage
identi¯cation with distributed long-gauge FBG sensors, Sensors 17(2) (2017) 411.
6. S. Chakraborty and J. T. DeWolf, Development and implementation of a continuous
strain monitoring system on a multi-girder composite steel bridge, J. Bridge Eng. 11(6)
(2006) 753–762.
7. A. J. Cardini and J. T. Dewolf, Long-term structural health monitoring of a multi-girder
Int. J. Str. Stab. Dyn. Downloaded from www.worldscientific.com

steel composite bridge using strain data, Struct. Health Monitoring 8(1) (2008) 47–58.
8. S. Stroh, R. Sen and M. Ansley, Load testing a double-composite steel box girder bridge,
Transp. Res. Record J. Transp. Res. Board (2010) 36–42.
9. R. Plude, J. Christenson, J. DeWolf and A. Jamalipour, Quantifying damage measures for a
composite steel girder bridge using ¯nite element analysis, 8th International Workshop on
Structural Health Monitoring, Stanford Univ, Stanford, CA, Sep 13–15, 2011.
10. D. H. Sigurdardottir and B. Glisic, Neutral axis as damage sensitive feature, Smart Mater.
Struct. 22(7) (2013) 2896–2912.
11. D. H. Sigurdardottir and B. Glisic, Detecting minute damage in beam-like structures
using the neutral axis location, Smart Mater. Struct. 23(12) (2014) 125042 (15pp).
12. G. Z. Xue, X. Q. Fang, X. K. Hu et al., Measurement accuracy of FBG used as a surface-
bonded strain sensor installed by adhesive, Appl. Optics 57(11) (2018) 2939–2946.
13. S. A. Neild, M. S. Williams and P. D. McFadden, Development of a vibrating wire strain
gauge for measuring small strains in concrete beams, Strain 41(1) (2005) 3–9.
14. B. Wei, P. Wang, M. G. Yang and L. Z. Jiang, Seismic response of rolling isolation
systems with concave friction distribution, J. Earthq. Eng. 21(2) (2017) 325–342.
15. D. S. Simulia, Abaqus analysis user's manual. Dassault Systemes, Pawtucket, USA
(2010).
16. Code for design of concrete structures, GB 50010-2002. National Standard of the People's
Republic of China. China Structural Science Academe Beijing (2002).
17. E. Gutierrez and S. Primi, Experimental analysis of the shear transfer mechanism in a
composite bridge beam manufactured from advanced polymer composites, European
Commission: Ispra, Italy (2005).
18. Y. Oshima and M. Kado, Long-term monitoring of composite girders using optical ¯ber
sensor. Bridge maintenance, safety, management, resilience and sustainability, in Proc.
Sixth Int. IABMAS Conf., Stresa, Lake Maggiore, Italy (2012).

1940015-17