Jtam-A4.dvi


JOURNAL OF THEORETICAL

AND APPLIED MECHANICS

52, 2, pp. 289-299, Warsaw 2014

ASSESSMENT OF THE TECHNICAL STATE OF LARGE SIZE STEEL
STRUCTURES UNDER CYCLIC LOAD WITH THE ACOUSTIC

EMISSION METHOD – IADP

Barbara Goszczyńska, Grzegorz Świt, Wiesław Trąmpczyński
Kielce University of Technology, Kielce, Poland; e-mail:bgoszczynska@tu.kielce.pl

In this paper, a global monitoring system based on the measurement of acoustic emission
(AE) due to active deterioration processes within steel structures is presented. This allows
one to locate and identify deterioration processes (initiation and development of cracks)
under cyclic load for large size structures as bridges, gantries, etc. The resulting data can
be used to locate damage zones that are dangerous for construction, to assess the general
condition of the structure and to serve for structural health monitoring.

Keywords: acoustic emission, cyclic load, crack initiation, large size steel structures

1. Introduction

The material fatigue, fatigue cracking and damage caused by corrosion are the main reasons
for failures of steel structures. Yet, the knowledge of the fatigue life of large steel structures
is rather limited. Additionally, computational methods or criteria for assessing the life of such
objects operating under variable loads are not available.
Analyses of fatigue lifewithin the range of high-cycle fatigue (at the stage of crack initiation)

are based on theWöhler diagram, which is constructed using the results of fatigue tests carried
out for sinusoidal load cycles at a constant amplitude. The diagrams are also usedwhenmaking
calculations for elements of structuresunder loadcycles at randomamplitudes inaccordancewith
the Palmgren-Miner rule (Palmgren, 1924). The fatigue crack propagation is usually described
by the Paris law (Paris and Erdogan, 1960). For the sake of the assessment of the service life of
engineering structures, a number of simplifications is introduced into the proceduresmentioned
above. Although those make calculations easier, on a number of occasions even make them
possible, the results obtained for the fatigue life or the crack propagation velocity may differ
several times from the real values.
Recently developed methods for calculating the strength of structure members containing

defects, such as R6 curve method, FITNET or methods based on failure advanced models (e.g.
that of Gurson-Twergaard-Needelman) still have very limited practical applications (Neimitz et
al., 2008).
Hence,when the calculationmethodsprovide only limited information on the fatigue life, the

solution can be experimental investigations to assess the technical state of structures in service.
Recently, non-destructive testing (NDT)methods,which allows one to detect and locate various
damages and defects has been developing dynamically (Hoła and Schabowicz, 2010).
Practical applications of majority of non-destructive testing methods are limited only to a

relatively small measurement area which results from features of a given method. That does
not constitute a major limitation if the structure, or its elements, is relatively small-sized. By
repeating the measurements, it is possible to cover the whole object with the measurement
area rather quickly. For large size structures, e.g. bridges, gantries, wheel excavators in surface
mining, that becomes a serious drawback.



290 B. Goszczyńska et al.

Such limitations do not have the acoustic emission method AE (Ono, 2011; Ranachowski et
al.; 2009, Yu et al., 2011; Ziehl, 2008), which makes it possible to detect and locate destructive
processes, and also to determine how intensive they are. The AE method shows the occurrence
of destructive processes and their location, yet it does not describe the mechanisms generating
such processes, which is a very important aspect of the structure diagnostics.
Therefore, in works by Gołaski et al. (2012), Goszczyńska (2014), Goszczyńska et al. (2012,

2013), the acoustic emission method called Identification of Active Damage Processes (IADP)
was proposed. It does not only record the occurrence and location of destructive processes,
but also makes it possible to identify them. The method has been applied to the analysis of
destructive processes in pre-stressed concrete members and it has been successfully used to
diagnose over seventy real engineering structures, mainly bridges.
The present paper presents the application of the IADP method to the analysis of steel

structures under cyclic load.

2. Acoustic emission method – The IADP method – basics

The acoustic emission (AE) involves generation of transient elastic waves due to local dynamic
change in the structure of thematerial. A signal of the acoustic emission is generated as a result
of sudden release of energy accumulated in the material by propagating micro-damages. The
attenuation of waves results from absorption, a transformation from elastic energy into thermal
one. Thus, the generation of AE signals provides information on degradation of properties of
the material when compared with its properties at an instant preceding the emission. The fact
the processes generating AE signals accompany only the active damage, i.e. that brought about
or developed under conditions prevailing during themeasurement.
Figure 1 presents a diagram of wave generation by destructive processes caused by beam

loading. The waves are recorded by sensors, usually piezoelectric ones, with the frequency ope-
rating range of 0.1-2.0MHz, mounted on the structure surface.

Fig. 1. A diagram of wave generation by destructive processes due to loading

The acoustic emission signal can be characterised by such parameters as: the number of
counts, numberof counts topeak, duration, rise time, amplitude, energy, strength,mean effective
voltage, root mean square, mean frequency, reverberation frequency and initiation frequency.
These parameters (twelve AE parameters called here the reference parameters) are then used
(in the IADPmethod) to create the reference database for destructive processes using statistical
analysis, based on pattern recognition.
A reliable assessment of the structure technical state depends not only on detecting the

occurrence of destructive processes but also on finding the zones where those are generated.
The location is accomplished on the basis of difference in time of the arrival of an AE signal
generated at the area of failure to sensors spaced on the surface of the tested member at the
known velocity of wave propagation. In this manner, it is possible to locate AE sources along a
plane (linear location) or within a specific area (zonal location) or in space (Nair andCai, 2010;



Assessment of the technical state of large size steel structures... 291

Goszczyńska, 2014; Goszczyńska et al., 2012; Aggelis et al., 2011). Using an appropriate number
of AE sensors adjusted to the tested structure, it is possible to cover the whole structure or a
fragment of it with themeasurement areas.
The IADP method is one of acoustic emission testing techniques. It makes it possible not

only to locate but also to identify active destructive processes, which is significant for the struc-
ture diagnostics (Gołaski et al., 2012; Goszczyńska et al., 2012; Świt, 2011). Identification of
destructive processes involves a comparative analysis of acoustic emission signals measured in
the tests of a structure in service against the reference signal database established beforehand
for individual destructive processes. For this method, it is necessary to:

• determine and identify, in laboratory and technical tests, individual destructive processes
characteristic for a given construction,

• establish a reference database for formerly determined destructive processes.

The application of the IADPmethod involves:

• performing measurements, in which acoustic emission signals for individual members of
the structure are recorded, while the structure remains in service,

• carrying out a comparative analysis of the recordedAE signals against the reference signal
database, and thus identifying the occurrence of destructive processes,

• locating places where the destructive processes occurs.

The reference signal database for those processes was compiled by means of conducting a
number of tests on various types of specimens for different loading schemes. The tests were de-
signed to obtain a single dominant destructive process among those that can occur in structures
under investigation, in this case steel ones.
Experiments were conducted for three types of steel: St3s, 18G2A and steel cut out of an

old bridge structure, for different kinds of specimens and types of load:

• simple loads – quasi-static tests on specimens under simple loading (tension, bending of
notched specimens) conducted at the following temperatures: T20 = +20◦C, T0 = 0◦C,
T
−30 =−30◦C, T−60 =−60◦C.

• simple load cycles – specimens with a hole and notches perpendicular to the axis of the
specimen under load cycles at temperature T20 =+20◦C,

• load cycles applied to riveted and welded steel beams with a notch at temperature
T20 =+20◦C

• quasi-static and cyclic loads applied tomodels of welded and bolted nodes at temperature
T20 =+20◦C.

In this way, four destructive processes were differentiated and the reference signal database
was established. The corresponding AE signals were defined as Classes:

Class 0 – signals generated by crack initiation,

Class 1 – signals generated by steel yielding at the tip of the crack,

Class 2 – signals generated by “noise”,

Class 3 – signals generated by crack growth,

Class 4 – signals resulting from the interference of waves generated by more than one
destructive process.

Statistical analysis based on pattern recognition was applied to record acoustic emission
signals usingNOESIS software. The pattern recognition can be categorised into: arbitrary clas-
sification using unsupervised (USPR) learning procedure and classification that employs the



292 B. Goszczyńska et al.

training set in form of reference signals in the supervised (SPR) learning procedure. In order to
compile a database (signal classes), arbitrary pattern analysis was used, whereas for signal class
recognition, supervised analysis was applied (Goszczyńska et al., 2013; Świt, 2011).
As regards statisticalmethods applied to item recognition, it is important to optimally select

recorded parameters of acoustic emission. Because many parameters of acoustic emission show
strong mutual correlation, which makes it possible for them to carry the same information on
an AE source, the degree of correlation between those parameters is defined by the so called
dendrograms. Using them, one can reduce the number of parameters of AE signals in the clas-
sification process, which shortens the duration of the analysis. Twelve parameters of AE signals
with a different level of adjustment were adopted for the analysis (mentioned above as reference
parameters, Goszczyńska et al. (2013)).

3. Application of the IADP method to the evaluation of the technical state of
steel structures

Using the established reference signal database for steel, the destructive processes were analysed
for different types of members and structures and also modes of loading.

3.1. The IADP method applied to the analysis of failure of the standard specimen under
cyclic loading

Exemplary results of tests on steel specimens (with a hole) made of steel taken from an old
bridge under load cycles at 10Hz frequency, are presented below. It is a standard specimen used
to determine the coefficients of the Paris equation (Paris and Erdogan, 1960).
AE sensors, spaced on the specimen as shown inFig. 2, recorded both theAE signals genera-

ted by the fatigue process and by the backgroundwhich consistedmainly of the acoustic signals
from specimen friction in the hydraulic holders and the oil movement in the machine cables.

Fig. 2. Diagram of a specimen with a hole and a notch

The recorded files were subjected to multi-parameter analysis taking into account twelve
(mentioned above) parameters, and the results of the analysis are shown in Fig. 3 in form of a
summation graph of one of the parameters (RMS – rootmean square) against time, for different
signal Classes defined above.
In the first phase of the cyclic loading, a local zone of plastic deformation is created in the

vicinity of partially cut notches, which precedes the crack initiation and the notch tip becomes
rounded. The process is accompanied by Class 1 signals and was observed optically by using
magnifying glass.
During thefirst 30000-70000 cycles, fatigue cracking is initiated. It is clearly formedonly after

approx. 25000 cycles (2500s what was observed optically). The process, i.e. cracking initiation,
corresponds roughly to the first stage of fatigue cracking and is accompanied byClass 0 signals.
The second stage, which is accompanied by Class 3 signals, involves the crack propagation

in a located plastically deformed zone. The growth of the fatigue crack leads to the final stage
- the specimen failure (such a process was observed optically).



Assessment of the technical state of large size steel structures... 293

Fig. 3. Examplary graphs of changes in the RMS (root mean square) with division into classes of
acoustic signals during fatigue crack propagation (steel collected from the bridge)

The end of the diagram shows Class 4 signals, which accompany the final stage of the
specimen failure. It results fromthe interference ofwaves generated bymore thanonedestructive
process and the friction on the crack surface.
Class 2 signals, features signals generatedby the systemthat loads andmonitors the specimen

(noise of the testing machine and specimen grips).
It can be seen that the IADP method makes it possible to trace the process of the crack

propagation up to specimen failure.

3.2. The IADP method applied to the analysis of a beam under cyclic loading – model
tests (18G2A, St3s, bridge steel)

Acoustic emission testing was conducted on beams having T-shape cross-section
(8×120×2000mm) with narrow cut-out in the centre (1.5×40mm) of the beam span. Expe-
riments were carried out on two types of beams:

– beamwith a welded flange (Fig. 4),

– beamwith a riveted flange (Świt et al., 2011)

towhich quasi-staticmonotonic loading and cyclically variable (sinusoidal) loadingwere applied
at two points as shown in Figs. 5 and 6.

Fig. 4. Scheme drawing of welded beams 120mm in height

Fig. 5. Diagram showing points at which the loading force is applied and those at which the tested
beam is supported



294 B. Goszczyńska et al.

Fig. 6. The test stand with a welded beam (18G2A steel) with a narrow cut-out in the centre and
acoustic sensors mounted on it (a) and cracked the beam section (b)

Figure 6 shows the welded beam positioned on the test stand, the spacing of AE sensors
(sixth sensors in total, on both sides of the cut-out) and the cracked due to cyclic loading beam
section.
In the cyclic loading tests, the beam loading was set to be applied through the force P, the

valueofwhichvaried sinusoidally.Thevalueof theminimum-maximumload,both forweldedand
riveted connections, was set to amount up to approx. 0.1-0.46 of the maximum value obtained
from the quasi-static tests. The values for welded beams, the exemplary results for which are
presented in Fig. 7, were as follows: 2.6-26kN, 18G2A steel – failure after 7500 cycles (Fig. 7a),
and 2.2-22kN, St3s steel – failure after 17377 cycles (Fig. 7b) at the frequency of 4Hz.

Fig. 7. Classes of AE signals recorded by two sensors as a function of the sum of the absolute energy
over time; (a) – 18G2A, (b) – St3s

Figure 7 shows the sum of AE signals recorded by two sensorsmounted on both sides of the
cut. The signals were subjected to multi-parameter analysis employing the reference data base.
The result is shown as a summation graph of one parameter (ABEN – absolute energy) as a
function of time with division into signal classes.
In both cases, after approx. 5000 cycles, the processes of steel yielding (Class 1) and cracking

initiation (Class 0) are recorded, which are then followed by the crack growth (Class 3).
For both types of steel (18G2A steel and St3s steel) the moments of crack initiation and its

growth were clearly recognized and could be located using IADPmethod. Themoment of crack
initiation and its growth was confirmed by optical observation.



Assessment of the technical state of large size steel structures... 295

3.3. Application of the IADP method to the analysis of a node under cyclic loading –
model tests

Acoustic emission tests were performed on nodes (models of the bridge nodes) fabricated in
two versions:

• welded one and,

• bolted one (M8 bolts and tightening torque M = 35N), which is a model of a riveted
connection (Fig. 8).

Fig. 8. Diagram of the tested node – bolted and welded

The nodes were first subjected to quasi-static monotonic loading (as in Fig. 9) up to failure
due to welds or bolts cracking. Then, a cyclically variable loading whose value (the force P)
varied sinusoidally at a frequency of 2Hz, was applied. In the cyclic loading, the value of the
maximum load both for welded and bolted connections was set to amount up to approx. 83% of
the maximum value obtained from the quasi-static tests (Pmin =2kN – 0.83Pmax).
Figure 9 shows the test stand and the spacing of AE sensors (eight sensors in total, on both

sides of the node).

Fig. 9. The test stand (with the bolted node under loading) and spacing of AE sensors (on one side of
the node)

Exemplary results of comparative analysis (employing the reference signal database) of AE
signals generated while the bolted node is loaded are presented in Fig. 10. It shows the sum
of AE signals recorded as a sum of the eight sensors (Fig. 10a), and signals recorded only by



296 B. Goszczyńska et al.

one sensor (sensor 5 located next to the bolts – Fig. 10b), which were subjected to the analysis
using the reference data base, as a summation graphof one parameter (ABEN - absolute energy)
against time with division into signal classes.

Fig. 10. Classes of AE signals recorded a sum of all sensors and by sensor 5 as a function of the sum the
absolute energy over time (bolted node)

After approx. 4000 cycles (2000s), the process of steel plastic yield is recorded (Class 1), and
then (following approx. 8000 cycles) signals are generated by the crack initiation and its growth
(Class 0 and 3). Also signals Class 4 are found.
After approx. 12000 cycles, the shear failure of the first bolt is observed visually. For 18000

cycles, two consecutive bolts suffer shear failure. Both graphs are very similar in shape, what
shows that indications of the IADPmethod are analogous for the sensor placed near by the zone
of damage and the sensors placed within the structure.
The results for correspondingwelded structures (bolts replaced with short welds) are shown

in Fig. 11. As before, the sum of AE signals recorded as a sum of all the sensors (Fig. 11a)
and signals recorded only by one sensor (sensor 5 located next to the bolts – Fig. 11b), were
subjected to the analysis using the reference data base, as a summation graph of one parameter
(ABEN - absolute energy) against time.

Fig. 11. Classes of AE signals recorded by all sensors and by sensor 5 as a function of the sum the
absolute energy over time (welded node)

The process of plastic yielding develops in a quite uniform manner since the beginning of
loading (Class 1). In the first stage (200 cycles), a crack is initiated (Class 0), which is followed
by a short-term growth (Class 3). Then, the process becomes inhibited (probably due to stress
redistribution), only to start again (Class 0 and 3 signals) after over 4000 cycles until failure
(Class 4).



Assessment of the technical state of large size steel structures... 297

The initial crack propagation, its stop, re-development and the instant of the weld failure
was recorded visually at the same time.
As before, both graphs are very similar in shape, what shows that indications of the IADP

method are analogous for the sensor placed near the zone of damage and the sensors placed
within the structure.

3.4. Application of the IADP method to the analysis of a structure under service load –
bridge tests

The tests were conducted on the riveted steel bridge (Fig. 12a), where twenty three AE
sensors were spaced along the span (1-6 sensors) and within the bridge nodes. An exemplary
spacing of the sensors in the truss shoe is presented in Fig. 12b (Goszczyńska et al., 2013).

Fig. 12. Side view of the bridge middle span with an exemplary spacing of sensors in the support node

Figures 13 and 14 (Gołaski et al., 2011) show the results of comparative analysis (employing
the reference signal database) of AE signals generated for the unloaded bridge (Fig. 13) and
with a train passing over it (Fig. 14) in form of a diagram of one reference parameter (signal
strength – SSTR) as a function of channels with division into signal classes.

Fig. 13. Diagram of the signal strength as a function of channels (position of the sensors) for the
unloaded bridge

For the unloaded bridge, all sensors record onlyClass 2 signals, i.e. those generated by noise.
When a train passes over the bridge, additionally Class 1 signals, generated by steel yielding,
are found in the measurement zone of sensors no. 10 and 16. Close visual inspection revealed
in sensor zone 16 a plasticized (and loosened) rivet the damage of which was not seen under a
layer of paint.



298 B. Goszczyńska et al.

Fig. 14. Diagram of the signal strength as a function of channels with a train passing over the bridge

Hence, using the IADPmethod, it was possible to find a plasticized rivet within such a large
size structure as a steel bridge.

4. Conclusions

The IADP method made it possible to determine the condition of steel structures under cyclic
loads, i.e. to determinewhere,when andwhat kind of destructive processes appear under service
loads.
The conducted analysis and results of tests indicate that the IADPmethod:

• allows one to detect and identify destructive processes in members of steel structures,

• allows one to detect the moment of cracks formation and track their development,

• makes it possible to locate active destructive processes in large size structuresunder service
loads,

• consequently, the IADP method can be applied to monitor and diagnose steel structures.

Reference

1. Aggelis D.G., Kordatos E.Z., Matikas T.E., 2011, Acoustic emission for fatigue damage
characterization in metal plates,Mechanics Research Communication, 38, 2, 106-110

2. Gołaski L., Goszczyńska B., Świt G., Trąmpczyński W., 2012, System for the global mo-
nitoring and evaluation of damage processes developing within concrete structures under service
loads,The Baltic Journal of Road and Bridge Engineering, 7, 4, 237-245

3. Gołaski L.,GoszczyńskaB., ŚwitG.,TrąmpczyńskiW., 2011,Application ofAE formoni-
toring the damageprocesseswithin steel structures,Proceedings 57-thAnnual ScientificConference
“Scientific Problems of Engineering Rzeszów-Krynica 2011”, 119-126

4. Goszczyńska B., 2014, Analysis of the process of crack initiation and evolution in concrete with
acoustic emission testing,Archives of Civil and Mechanical Engineering, 14, 2, 134-143

5. GoszczyńskaB., ŚwitG., TrąmpczyńskiW.,KrampikowskaA.,Tworzewska J., Two-
rzewski P., 2012, Experimental validation of concrete crack identification and location with aco-
ustic emissionmethod,Archives of Civil and Mechanical Engineering,XII, 1, 23-28

6. GoszczyńskaB., ŚwitG.,TrąmpczyńskiW., 2013,Monitoring of active destructive processes
as a diagnostic tool for the structure technical state evaluation,Bulletin of the Polish Academy of
Sciences, Technical Sciences, 61, 1, 97-108



Assessment of the technical state of large size steel structures... 299

7. Hoła J., Schabowicz K., 2010, State-of-the-art non-destructivemethods for diagnostics testing
of building structures – anticipated development trends,Archives of Civil and Mechanical Engine-
ering, 10, 3, 5-18

8. Nair A., Cai C.S., 2010, Acoustic emission monitoring of bridges: Review and case studies,
Engineering Structures, 32, 6, 1704-1714

9. Neimitz A., Dzioba I., Graba M., Okrajni J., 2008, Assessment of Strength, Service Life
and Structural Reliability of Members Containing Defects (in Polish), Wydawnictwo Politechniki
Świętokrzyskiej, Kielce, 1-438

10. Ono K., 2011, Aplication of acoustic emission for structure diagnostic,Diagnostyka, 2, 58, 3-18

11. Palmgren A., 1924, Die Lebensdauer vonKugellagernVerfahrenstechnik,Zeitschrift des Vereins
Deutscher Ingenieure, 58, 339-341

12. Paris P.C., Erdogan F., 1960, A critical analysis of crack propagation laws, Journal of Basic
Engineering. Transaction ASME, 85, 528-534,

13. Ranachowski P., Rejmund F., Jaroszewski M., Wieczorek K., 2009, Study of structural
degradation of ceramic material of insulators in long term operations,Archives of Metallurgy and
Materials, 54, 1, 205-216

14. Świt G., 2011, Analysis of the Destructive Processes in Bridge Structures of Beams Prestressed
with the Phenomenon of Acoustic Emission (inPolish),WydawnictwoPolitechniki Świętokrzyskiej,
Kielce, M 21, 1-174

15. Świt G., Trąmpczyński W., Płonecki L., 2011, Application of acoustic emission for the mo-
nitoring of large –size steel elements,Przegląd Mechaniczny, 10/11, 27-32

16. YuJ., ZiehlP. , ZárateB.,CaicedoJ., 2011,Predictionof fatigue crackgrowth in steel bridge
components using acoustic emission, Journal of Constructional Steel Research, 67, 8, 1254-1260

17. Ziehl P.H., 2008, Applications of acoustic emission evaluation for civil infrastructure, Society of
Photo-Optical Instrumentation Engineers, Proceedings, 6934, ISSN 0277-786X

Manuscript received April 26, 2013; accepted for print June 18, 2013