Professional Documents
Culture Documents
Abstract: The vibration-based tension estimation method is a common approach for health assessment of cable structures. In this meth-
odology, cable natural frequencies along with mechanical and geometrical properties of the cable were used to estimate the cable tension.
In this paper, following the vibration-based tension estimation method, stay cables of new Ironton-Russell Bridge are evaluated and the results
compared against the lift-off test results. It is shown that the difference between estimated tension forces and directly measured tension forces
(lift-off test) is negligible. A numerical analysis is also performed to investigate the impact of errors in measuring cable natural frequencies
and cable axial stiffness on estimated cable tensions. DOI: 10.1061/(ASCE)ST.1943-541X.0002054. © 2018 American Society of Civil
Engineers.
Introduction the theoretical cable natural frequencies (coming from the cable
model) and experimentally measured cable natural frequencies is
Tracking the variation of cable tension over time is a common minimized (Haji Agha Mohammad et al. 2017a). Different combi-
approach for damage diagnosis of cables in cable structures (Yang nations of cable models and cable response measurement tech-
et al. 2015; Nazarian et al. 2016a, b; Scarella et al. 2016; Bao et al. niques have been used to estimate the cable tensions in numerous
2017; Huang et al. 2017; Pacitti et al. 2017). This approach is also cable structures around the world following the vibration-based
used for health monitoring of different parts subjected to tension approach. Cable tension of two stay cables of Hwamyung Bridge
loads in civil structures, like steel eyebars in bridges (Mazurek in Korea was estimated using three methods: lift-off test, electro-
2016). However, ambient parameters like temperature variation can magnetic sensors, and the vibration-based approach. It was shown
also alter dynamic behavior of cables and superstructure of the that the vibration-based approach is an inexpensive method in both
bridge (Norouzi et al. 2016; Chen et al. 2017; Zhao et al. 2017). labor and cost (Cho et al. 2012). Using conventional accelerometers
Knowing cable length, cable mass per unit length, cable axial stiff- to measure the ambient response of stay cables of the Danube
ness, and cable bending stiffness (in cases where bending stiffness Channel bridge in Austria, cable tensions were accurately estimated
is considerable), natural frequencies of cables can be calculated following the vibration-based approach (Geier et al. 2006). In
using a cable model (Yang et al. 2014). In a vibration-based cable another work, the microwave interferometer technique was used
tension estimation method, this procedure is reversed; the cable ten- to measure the ambient response of stay cables of Cesare Cantu
sion is estimated by adjusting the cable tension in the cable model Bridge in Italy. The taut string model was then used along with
so that the error function representative of the difference between mechanical and geometrical cable properties to estimate the cable
tensions (Gentile 2010). Using a similar cable model and using
1
Graduate Research Assistant, Dept. of Mechanical and Materials adaptive sparse time-frequency analysis method to identify the
Engineering, Univ. of Cincinnati, 717 Engineering Research Center, instantaneous natural frequencies of cables from acceleration mea-
ML-30, Cincinnati, OH 45221 (corresponding author). Email: Hajiagsn@ surements (Bao et al. 2017) suggested a method to estimate the
mail.uc.edu instantaneous cable tension. Wireless sensor networks (Cho et al.
2
Research Associate, Dept. of Electrical Engineering and Computing 2010; Sim et al. 2013) and vision-based methods (Kim and Kim
Systems, Univ. of Cincinnati, 717 Engineering Research Center, ML-30,
2011; Feng et al. 2017) have also been used extensively to measure
Cincinnati, OH 45221. Email: Norouzmi@mail.uc.edu
3
Professor and Head of Structural Dynamics Research Laboratory,
cable natural frequencies in the vibration-based cable tension
Dept. of Mechanical and Materials Engineering, College of Engineering estimation methodology.
and Applied Science, Univ. of Cincinnati, Cincinnati, OH 45221. Email: In this paper, the cable model of Mehrabi and Tabatabai
Randall.Allemang@uc.edu (Mehrabi and Tabatabai 1998) is chosen for estimating the cable
4
Associate Research Professor, Dept. of Electrical Engineering and tension of stay cables of the new Ironton-Russell bridge (120 cables).
Computing Systems, Univ. of Cincinnati, 721 Engineering Research The estimated tensions are then compared against lift-off test results.
Center, ML-30, Cincinnati, OH 45221. Email: Victor.Hunt@uc.edu Moreover, reviewing different cable models, five cables with me-
5
Professor, Dept. of Electrical Engineering and Computing Systems, chanical and geometrical characteristics representative of a wide
Univ. of Cincinnati, Rhodes 812J, ML-30, Cincinnati, OH 45221. Email: range of stay cables in cable-stayed bridges around the world are
Arthur.Helmicki@uc.edu
6 selected to investigate the impact of errors in cable natural frequen-
Graduate Research Assistant, Dept. of Electrical Engineering and
Computer Science, Univ. of Cincinnati, 717 Engineering Research Center,
cies and cable axial stiffness on the estimated cable tensions.
ML-30, Cincinnati, OH 45221. Email: venkatcr@mail.uc.edu
Note. This manuscript was submitted on August 3, 2017; approved on
December 8, 2017; published online on April 13, 2018. Discussion period Lift-Off Test
open until September 13, 2018; separate discussions must be submitted for
individual papers. This paper is part of the Journal of Structural Engineer- The lift-off test is usually used to measure the tension of a limited
ing, © ASCE, ISSN 0733-9445. number of strands in a stay cable. The tension of each strand in the
Fig. 2. (a) Schematic of lift-off test; and (b) force-displacement diagram in lift-off test.
the sheathing are free to move. Stay cables away from the towers
(like B30R) are longer, with lower inclinations. As a result, a
larger portion of strands lie on the sheathing and vibrate with
the sheathing as a unit. Consequently, auto-power spectrums
of the corresponding stay cables show distinct frequency con-
tents [Fig. 8(a)]. On the other hand, stay cables that are close
to the towers are shorter, with higher inclination, and as a result,
the stands inside the sheathing move independently and the ac-
quired auto-power spectrums are not as clean as long cables
[Fig. 8(b)]. The natural frequencies of the stay cables of the
new Ironton-Russell Bridge were measured using the peak-
picking method and are presented in Appendixes II and III. Fig. 9
shows the natural frequencies of representative short and long
stay cables of the new Ironton-Russell bridge with respect to
Fig. 7. Test setup. mode number n.
Based on Fig. 9, it is seen that the measured cable natural
frequencies of the new Ironton-Russell bridge fit well to straight
lines. Consequently, the coefficient of the nonlinear term n3 in
resolution of 0.003125 Hz) with a sampling frequency of 64 Hz
Eq. (6) is negligible
(Fig. 7). The auto-power spectrums of the measured ambient
response of stay cables B30R and B16L are shown in Fig. 8.
As shown in Fig. 1, there is no grout used to fill the space π2
ω1s ð1 þ 0.039μÞ ≅0 ð9Þ
between the strands and HDPE sheathing; hence, strands inside 2ζ 2
Fig. 8. Auto-power spectrum of ambient response of stay cable: (a) B30R; and (b) B16L.
Error Analysis
Downloaded from ascelibrary.org by Univ. of Alabama At Birmingham on 04/17/18. Copyright ASCE. For personal use only; all rights reserved.
1s
(a) (b)
Fig. 10. Convergence of ω1s for stay cable: (a) B30R; and (b) B16L.
(a) (b)
Fig. 11. Convergence of ζ for stay cable: (a) B30R; and (b) B16L.
Fig. 12. Lift-off test results and estimated tensions for stay cables on the left side of the bridge.
tension force in comparison with the case where there is no error In Figs. 16 and 17, It is shown that for a fixed value of λ2 ,
in cable properties. However, this trend for the error in cable the error in the estimated tension increases as ζ decreases.
axial stiffness EA is reversed; an increase or decrease in EA Based on Eq. (5), the nondimensional parameter ζ represents
causes an underestimation or overestimation of the cable tension the importance of cable flexural stiffness in the overall dynamic
force in comparison with the case where there is no error in behavior of the cable so that ζ decreases as the diameter of the
cable properties (Fig. 15). Moreover, Fig. 14 shows that for a cable increases (Bouaanani 2006). Consequently, the vibration-
fixed number of natural frequencies used to construct the error based cable tension estimation method used in this paper is less
function N, a specific percentage of the error in the natural accurate for cables with large diameters (higher flexural stiff-
frequency of higher modes affects the estimated tension more ness values). On the other hand, for a fixed value of ζ, the error
than the same level of error in the natural frequencies of lower slightly increases when λ2 increases. The effects of cable sag,
modes. angle of inclination, and axial stiffness are represented in the
Using the first five cable natural frequencies to build the cable tension estimation method through the sag-extensibility
error function (N ¼ 5) and introducing a 10% error to the parameter λ2. Increasing sag increases λ2 , whereas increased
third natural frequency, the cable tension estimation error angle of inclination and axial stiffness decreases λ2 (Johnson
for cables with dimensionless parameters 0.1 < λ2 < 3.1 and et al. 2003). Therefore, the vibration-based cable tension esti-
50 < ζ < 650 are shown in Fig. 16. Similar to Fig. 16, Fig. 17 mation method used in this paper gets less accurate when the
shows the error in the cable tension estimation process caused angle of inclination and axial stiffness of the cable decrease
by a 5% error in calculation of cable axial stiffness EA when and cable sag increases. However, based on Figs. 16 and 17,
there are five cable natural frequencies used to build the error variation in tension estimation accuracy caused by variation of
function. λ2 is negligible in comparison with variation of ζ. It is worth
Fig. 13. Lift-off test results and estimated tensions for stay cables on the right side of the bridge.
mentioning that using different numbers of cable natural properties representative of a substantial portion of stay cables in
frequencies in the tension estimation process and introducing cable-stayed bridges around the world were then picked to numeri-
various error levels to cable natural frequencies and cable axial cally investigate the effects of natural frequency measurement
stiffness resulted in behavior similar to that shown in Figs. 16 errors:
and 17. • Estimated cable tension force converges to the actual cable ten-
sion as the number of cable natural frequencies used to build the
error function increases;
Conclusion • An increase or decrease in cable natural frequencies results in an
overestimation or underestimation of cable tension force;
Using vibration-based tension estimation methodology, cable ten- • An increase or decrease in cable axial stiffness results in an
sions of stay cables of the new Ironton-Russell Bridge were esti- underestimation or overestimation of cable tension force;
mated and the results compared against lift-off test measurements. and
It was shown that the estimated results are in a good agreement with • The error in estimated tension force increases as the diameter of
the lift-off test results. Five cables with mechanical and geometrical the stay cable increases (ζ decreases).
5 5
Error (%)
Error (%)
0 0
-5 -5
-10 -10
2 3 4 5 6 7 2 3 4 5 6 7
(a) N (b) N
10 10
5 5
Downloaded from ascelibrary.org by Univ. of Alabama At Birmingham on 04/17/18. Copyright ASCE. For personal use only; all rights reserved.
Error (%)
Error (%)
0 0
-5 -5
-10 -10
2 3 4 5 6 7 2 3 4 5 6 7
(c) N (d) N
10 5% increase in the 1
st
natural frequency
nd
5% increase in the 2 natural frequency
5 rd
Error (%)
Fig. 14. Tension estimation error due to 5% error in natural frequency of cable types: (a) 1; (b) 2; (c) 3; (d) 4; and (e) 5.
(a) (b)
(c) (d)
1% increase in EA
5% increase in EA
10% increase in EA
1% decrease in EA
5% decrease in EA
10% decrease in EA
No error
(e)
Fig. 15. Tension estimation error due to error in measurement of axial stiffness of cable types: (a) 1; (b) 2; (c) 3; (d) 4; and (e) 5.
3.8
Error (%)
Error (%)
-3.3
3.6
0 -3.4 3.5
3.4
100 3
3.2 200 -3.5 2.5
3.5 300 700 2
Downloaded from ascelibrary.org by Univ. of Alabama At Birmingham on 04/17/18. Copyright ASCE. For personal use only; all rights reserved.
Fig. 16. Tension estimation error due to (a) 10% increase; and (b) 10% decrease in the third natural frequency of the cable when N ¼ 5.
0 0.4
Error (%)
Error (%)
-0.2 0.2
3.5 0
3 100
-0.4 2.5 0 200
2 3.5 300
700 3
600 1.5 400
500 2.5
400 2 500
1 1.5
300 1 600
200 0.5 2 0.5
100 2 0 700
(a) 0 0 (b)
Fig. 17. Tension estimation error due to (a) 5% increase; and (b) 5% decrease in cable axial stiffness EA when N ¼ 5.
Appendix I. Mechanical and Geometrical Properties of Stay Cables of New Ironton-Russell Bridge
Appendix II. Measured Natural Frequencies of Stay Cables of New Ironton-Russell Bridge in Hz
Appendix III. Measured Natural Frequencies of Stay Cables of New Ironton-Russell Bridge in Hz
M20L 1.697 3.378 5.041 6.731 8.306 9.909 11.39 13.44 15.03 — — — — — —
M19L 2.087 4.166 6.222 8.159 10.17 — — — — — — — — — —
M18L 2.05 4.053 6.072 7.984 — — — — — — — — — — —
M17L 2.628 5.197 7.784 10.22 — — — — — — — — — — —
M16L 2.597 5.091 7.753 10.24 — — — — — — — — — — —
B16L 2.294 4.475 6.959 9.156 — — — — — — — — — — —
B17L 2.463 4.897 6.969 — — — — — — — — — — — —
B18L 2.425 4.769 7.006 — — — — — — — — — — — —
B19L 2.072 4.109 6.103 8.113 — — — — — — — — — — —
B20l 1.875 3.741 5.575 7.425 9.022 — — — — — — — — — —
B21L 1.65 3.3 4.9 6.488 8.109 9.594 — — — — — — — — —
B22L 1.497 2.991 4.459 5.934 7.372 8.653 — — — — — — — — —
B23L 1.438 2.869 4.294 5.709 7.138 8.522 9.859 11.07 — — — — — — —
B24L 1.441 2.869 4.297 5.725 7.138 8.525 — — — — — — — — —
B25L 1.222 2.438 3.647 4.859 6.034 7.272 8.425 9.537 10.77 11.8 — — — — —
B26L 1.156 2.3 3.444 4.578 5.713 6.859 7.909 9.131 10.16 — — — — — —
B27L 1.056 2.103 3.15 4.194 5.234 6.263 7.288 8.353 9.356 10.33 11.34 12.33 13.31 14.62 —
B28L 1.034 2.069 3.103 4.131 5.134 6.325 7.234 8.272 9.272 10.16 11.24 — — — —
B29L 0.95 1.888 2.825 3.766 4.709 5.65 6.581 7.516 8.447 9.394 10.28 11.23 12.12 13.03 —
B30L 0.9313 1.85 2.766 3.684 4.603 5.509 6.431 7.356 8.213 9.159 10.06 10.9 — — —
Cable Estimated tension (kN) Lift-off tension (kN) Difference (%) Lift-off standard deviation (kN)
B15R 2,670.5 2,574.7 3.7 26.8
B14R 2,376.7 2,249.5 5.6 23.8
B13R 2,066.6 1,949.6 6 31.2
B12R 1,965.2 1,987.4 −1.1 26.3
B11R 1,951.5 2,078.2 −6.1 81.2
B10R 1,894.6 1,925.3 −1.6 60.6
B9R 1,849.8 1,915.6 −3.4 47.9
B8R 1,489.2 1,480 0.6 71.9
B7R 1,276.7 1,318.5 −3.2 37
B6R 1,266 1,255.3 0.8 37
B5R 1,471.3 1,447.7 1.6 26.4
B4R 1,399 1,346.7 3.9 44.2
B3R 1,222.3 1,257.6 −2.8 29.7
B2R 997.2 1,053.3 −5.3 22.2
B1R 819.3 898.2 −8.8 31.5
M1R 797.1 873.6 −8.7 37.3
M2R 1,293.6 1,278.9 1.1 18.4
M3R 1,094.5 1,069.6 2.3 40.6
M4R 1,222.2 1,169.6 4.5 14.6
M5R 1,182.5 1,196.3 −1.1 52.2
M6R 1,430.6 1,409.8 1.5 22.4
M7R 1,731.9 1,662.8 4.1 101.4
M8R 1,693.3 1,679.9 0.8 61.8
M9R 1,736.1 1,705.9 1.8 24.5
M10R 1800.9 1,789.9 0.6 28.9
M11R 2,236.9 2,164.3 3.3 40.3
M12R 1,985.8 1,935.5 2.6 41.6
M13R 2,124.3 2,263.2 −6.1 71.2
M14R 2,407 2,389 0.7 73.8
M15R 2,562.7 2,499.2 2.5 43.4