Vous êtes sur la page 1sur 6

ARTICLE IN PRESS

International Journal of Adhesion & Adhesives 29 (2009) 234– 239

Contents lists available at ScienceDirect

International Journal of Adhesion & Adhesives


journal homepage: www.elsevier.com/locate/ijadhadh

Adhesive defect density estimation applying the acoustic emission technique


D. Croccolo , R. Cuppini
DIEM, University of Bologna, Viale Risorgimento, 2, 40136 Bologna, Italy

a r t i c l e in f o a b s t r a c t

Article history: The aim of this work is to study in depth the use of the acoustic emissions, as a non-destructive testing
Accepted 18 March 2008 technique, applied to the adhesively bonded joints. In a previous paper, three defect densities of the
Available online 10 June 2008 adhesive, 0%, 50% and 100%, were investigated, while in the present work two more intermediate values
Keywords: have been considered: 25% and 75%. The first goal of this paper is to confirm the good correlation,
Anaerobic highlighted in the previous one, between the adhesive defect density and the cumulative counts of the
Metals acoustic emissions; the second goal is to improve a methodology which might be capable of estimating,
Non-destructive testing with accuracy, the adhesive bonding defect and the final releasing moment of the assembled joints. In
Shaft–hub couplings this paper, the new results have been integrated with the previous ones and a more detailed map of the
adhesive slope variations is presented.
& 2008 Elsevier Ltd. All rights reserved.

1. Introduction of the structure during a single loading test. Hence, the AE


technique can be successfully applied in order to evaluate the
The acoustic emissions (AE) uttered by the stressed materials density of the adhesive defects as a NDT method if the value of the
are the stress waves produced by sudden movement of a crack. maximum load applied during the tests is sufficiently lower than
Hence, the source of AE is a defect-related process caused by a the failure load of the joint [3–5].
crack propagation or by a rapid plastic deformation. The rapid AE inspection is a powerful aid for the study of the failure of
deformation, in fact, produces a stress wave which radiates into adhesively bonded joints because it is strictly related to the
the structure and may excite a sensitive piezoelectric transducer. adhesive strength, and to its damage. Some recent works pointed
As the stress and the crack propagation or the plastic rapid out a strong correlation between the failure load of the joints and
deformation rise in the material, the emissions generated rise too. the cumulative counts of the emissions, and between the defect
The signals acquired by the sensors are, usually, amplified and density of the adhesive and the cumulative counts of the
plotted in order to define the tensile state of the material. emissions [6–12].
Therefore, AE inspection is usually carried out during a controlled Therefore, encouraged by the good results obtained in [12], we
loading test of the structure. decided to study in depth the correlation between the adhesive
AE differ from the other non-destructive testing (NDT) bonding defects and the AE adding some other intermediate levels
methods mainly because the signal is produced by the material of defect density. In the present work, we analyzed and added the
itself, not by an external source, and because the AE detect the correlations of 25% and 75% defect density in order to define the
micro-movements, while the other methods detect the geome- complete map of the cumulative counts slope variations and to
trical discontinuities; the other differences are summarized in forecast, more accurately, the releasing moment of the joints.
Table 1. It is widely demonstrated in the literature that the most
common NDT techniques (for instance, radiography, ultrasounds,
parasitic currents and magnetic particle) are not able to estimate
the density of the adhesive bonding defects produced by a lack of
2. Methodology
glue, unclean surfaces, and an insufficient degree of polymeriza-
tion once the parts are assembled together [1,2]. On the contrary,
The test specimen, shown in the drawing of Fig. 1 and in the
one of the benefits of the AE inspection is that it is possible to
pictures of Figs. 2 and 3, was designed and manufactured by the
perform a non-intrusive inspection applied to the whole volume
authors with the aim and the scope presented in [12]. After each
test the specimen was disengaged, cleaned using the Loctite 7063
 Corresponding author. Tel./fax: +39 0512093413. cleaner and a cutter, and reused for a new test. The female part of
E-mail addresses: dario.croccolo@unibo.it (D. Croccolo), the specimen, Fig. 2, and the male one, Fig. 3, are in 18NiCrMo3
rossano.cuppini@mail.ing.unibo.it (R. Cuppini). face-hardened steel and are held together with a screw and a nut.

0143-7496/$ - see front matter & 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijadhadh.2008.06.001
ARTICLE IN PRESS

D. Croccolo, R. Cuppini / International Journal of Adhesion & Adhesives 29 (2009) 234–239 235

The fundamental dimensions of the specimen, as reported in the When the adhesive defect was simulated on the specimen,
drawing of Fig. 1, are summarized in the following list: there was another step (spreading the contact surfaces with a film
of lube oil) between #2 and #3.
 mean coupling diameter Dm ¼ 99.13 mm;
 coupling axial length L ¼ 25 mm;
2.2. Tests run
 taper angle a ¼ 21;
 external diameter of the female part DAa ¼ 141.6 mm;
 QA ¼ Dm/DAa ¼ 0.7. The new additional tests were performed with two defect
densities (25% and 75%) and the same three interference levels,
already used for the previous tests with adhesive defect density
2.1. Specimen setting-up cycle
equal to 0%, 50% and 100%. The new tests were performed with the
aim to complete the previous design of experiment (DOE) data set
The specimen setting-up cycle is realized by the following
and to obtain a more detailed response surface for the failure load
steps:
(releasing moment) of the adhesive (Mad), which has been
calculated in each test by subtracting the moment due to the
1. disengaging the specimen;
interference (Mint) from the total releasing moment (Mtot)
2. cleaning the contact surfaces using the Loctite 7063 cleaner
measured by the torque machine, following the same methodol-
and a cutter;
ogy presented in [12].
3. spreading the contact surfaces with the Loctite 638 adhesive;
The minimum polymerization time was established at 72 h,
4. generating the contact pressure using a hydraulic press (the
which is the inferior limit needed for a good polymerization of the
pressure is measured using a strain gauge applied on the
adhesive. The load speed variation was set at 15 Nm/s in order to
external diameter of the female part of the specimen);
obtain a quasi-static loading test.
5. tightening the nut in order to maintain the axial load;
Hence, the global factorial plane (three interference levels and
6. waiting for the adhesive polymerization in an oven at the
five adhesive defect density levels) has 15 points and the total
constant temperature of 40 1C for 72 h.
number of tests is equal to 60, since we performed 4 replications
for each point. The number of additional tests is 24. The complete
Table 1 ranges of the factors are the following:
Characteristic of acoustic emission inspection compared with other methods

Acoustic emission Other methods  interference:


level 0: 0.002 mm (H7-m6 cylindrical coupling equivalent)
Detects movement of defects Detect geometric forms of defects level 1: 0.036 mm (H7-r6 cylindrical coupling equivalent)
Requires stress Do not requires stress
level 2: 0.077 mm (H7-t5 cylindrical coupling equivalent)
Each loading is unique Inspection is directly repeatable
More material sensitive Less material sensitive  adhesive defect presence:
Less geometry sensitive More geometry sensitive level 0: surfaces completely cleaned
Less intrusive on plant/process More intrusive on plant/process level 0.5: 25% of the surfaces completely oiled
Requires access only at sensor Requires access to whole area of inspection level 1: 50% of the surfaces completely oiled
Tests whole structure at once Scan local region in sequence
Main problems: noise related Main problems: geometry related
level 1.5: 75% of the surfaces completely oiled
level 2: 100% of the surfaces completely oiled

L=25mm
Dm =99.13mm
DAa =141.6mm

92°

Fig. 1. Drawing of the specimen.


ARTICLE IN PRESS

236 D. Croccolo, R. Cuppini / International Journal of Adhesion & Adhesives 29 (2009) 234–239

Fig. 4. The DOE response surface [13].

Fig. 2. The female part of the specimen.

Fig. 5. The PAC transducer applied on the external surface of the female part.

The correlation coefficients between the releasing moments


and the total cumulative counts and between the releasing
moments and the cumulative counts at 4 kN m have been
calculated. The results analysis confirmed what was discussed
Fig. 3. The male part of the specimen. and presented in [12]: the correlation coefficients, computed by
grouping the results with the same adhesive defect density, are
The defect density (0%, 25%, 50%, 75% and 100%) was simulated always higher than 0.75, whereas no correlation exists if the
by spreading first a film of lube oil and then the adhesive on the results are grouped with the same interference level. The value of
two contact surfaces of the specimens. The percentage of oiled 4 kN m, which is about 30% of the maximum releasing moment of
surface was equal to the desired defect density (four marks were the coupling (13.16 kN m calculated in [13] with no defect and no
put on the female and on the male parts in order to obtain four interference), is an appropriate value in order to acquire a
sectors exactly equal to a quarter of a circle). significant number of cumulative counts and to consider the tests
The numerical results of the new analysis, gauged by applying as non-destructive ones.
the DOE and the ANalysis Of VAriance (ANOVA), were presented in The diagrams of the cumulative counts related to the applied
[13], while in the present work they are summarized with the new load have been cut at the beginning of the slope variations and
response surface shown in Fig. 4. During each test we have also were interpolated with the minimum mean square error method
acquired the AE uttered by the specimens. following the same methodology presented in [12]. The slope
variations indicate the transition from the adhesive micro-fracture
zone to the coalescence one [8], which means an imminent
3. Results
disengaging. For each group of interpolation lines (0%, 25%, 50%,
75% and 100% of the defect density), the average and the standard
The AE parameters acquired and analysed with the Physical
deviation values of the slopes have been calculated and the results
Acoustics Corporation1 (PAC) control unit installed on a personal
are reported in Table 2 and plotted in Fig. 6.
computer are the cumulative counts and the energy uttered as
acoustic waves. The transducer location, placed on the external
diameter of the female part, is reported in Fig. 5. 4. Discussion

1
The specific PAC model employed for the experimental tests was DISP-4, All the tests, also those with defect density equal to 25% and
with a. NANO 30 transducer, 70 dB as sensitivity and 298 kHz as peak frequency. 75%, confirmed the strong correlation between the failure load
ARTICLE IN PRESS

D. Croccolo, R. Cuppini / International Journal of Adhesion & Adhesives 29 (2009) 234–239 237

and the number of cumulative counts and the good relationship 4 kN m, are always higher than 0.75 for all the five adhesive defect
between the AE slopes and the defect density of the adhesive. As densities.
already observed, the correlation coefficients, between the The five areas between each slope average minus and plus its
maximum applied load and the total cumulative counts and standard deviation, corresponding to each defect density, are
between the maximum applied load and the cumulative counts at plotted in Fig. 6. While the area corresponding to 75% of the defect
density has no superposition zones with the two contiguous areas
(50% and 100%), the one corresponding to 25% of the defect
density is partially superimposed on the 50% area. This occurrence
Table 2
Characteristics of the slopes for each group of interpolation lines (0%, 50% and confirms that the AE correlation decreases when the defect
100% of adhesive defect density) density decreases. Nevertheless, it is possible to consider the
overlapping area between 25% and 50% of the defect density as
Slope (cumulative counts/kN m) Defect density
belonging to (a part of) 50% area. It means that a tested joint with
0% 25% 50% 75% 100% a defect density equal to 25% but with an AE slope included in the
overlapping area would be considered as a 50% defected joint and,
Average (x̄) 221 3801 4969 18,070 63,278 therefore, with a lower Mtot than the actual one. Furthermore, the
Standard deviation (s) 206 2633 2962 9229 34,571 maximum error computed in the evaluation of Mtot is lower than
x̄s 15 1168 1984 8841 28,707
x̄+s 427 6434 7908 27,299 97,849
10% and, anyway, in a safety condition. Hence, the actual
cumulative counts slope map is plotted and presented in Fig. 7.

Fig. 6. Areas between each slope average minus and plus its standard deviation: the sketched area is the overlapping zone between 25% and 50% of adhesive defect density.

Fig. 7. Areas between each slope average minus and plus its standard deviation: final slope map.
ARTICLE IN PRESS

238 D. Croccolo, R. Cuppini / International Journal of Adhesion & Adhesives 29 (2009) 234–239

Table 3
Characteristics and test values of the four additional tests

Test D% I Mtot (kN m) Mint (kN m) Mad (kN m) Mtest (kN m) Cumulative counts at 4 kN m Slope (cumulative counts/kN m)

V11 1 1 10.04 1.82 8.22 4 22,732 5683


V1.50 1.5 0 10.22 0.09 10.13 4 94,740 23,685
V21 2 1 8.85 1.55 7.30 4 293,888 73,472
V1.52 1.5 2 8.49 3.12 5.37 4 78,524 19,631

Fig. 8. Areas between each slope average minus and plus its standard deviation containing the four additional test slopes.

Table 4
Comparison between the test values and the estimated values of the four additional tests

Test Mtot-real (kN m) Mint (kN m) Mad-real (kN m) Mad-estimateda (kN m) Mtot-estimated (kN m) e% (Mtot)

V11 10.04 1.82 8.22 8.69 10.51 4.4


V1.50 10.22 0.09 10.13 9.88 9.97 2.5
V21 8.85 1.55 7.30 6.89 8.44 4.9
V1.52 8.49 3.12 5.37 5.9 9.02 5.9

a
According to the mathematical model defined in [13].

The AE technique is, therefore, suitable to forecast the defect 5. Conclusion


density of the adhesive and the failure load of the joint with a
good degree of confidence. In this paper, the use of AE technique as an NDT one in the field
In order to verify the non-destructive methodology discussed of adhesively bonded joints is studied in depth. The defect
above, we performed four additional tests in the same boundary densities considered are 0%, 25%, 50%, 75% and 100%, while the
conditions. All the specimens were loaded until failure in interference levels are 0.002, 0.036 and 0.077 mm. The tests have
acquiring the AE and measuring the final releasing moments. been performed using some special conical specimens and
The specimen characteristics, the values of the test results, and according to the DOE method in order to evaluate the presence
the computed data are presented in Table 3; all the computations and the amount of the adhesive defect density and in order to
were made in accordance with the methodologies presented in define, with accuracy, the map of the cumulative counts slope
this paper. The interpolation lines of the four new specimens, as variations. The analysis of the AE confirmed the strong correlation
shown in Fig. 8, stay inside their own area with regard to the between the maximum load applied and the total cumulative
defect density and confirm that the AE technique can be used to counts, as well as between the maximum load applied and the
forecast the defect density of the adhesive with a good degree of cumulative counts at 4 kN m. The correlation coefficients range
confidence. from 0.75 up to 0.96. This strong correlation was verified only for
Finally, the estimated total releasing moments (Mtot) computed the same defect density of the adhesive, whereas it does not occur
according to [12] were compared to the real ones. The comparison when the interference level is the same.
results reported in Table 4 pointed out a small difference between The complete map of the slope variations referring to the
the actual values and the estimated ones, with errors always cumulative counts of the five adhesive defect densities is reported
lower than 6%. in Fig. 6. The area of 25% defect density partially overlaps the 50%
ARTICLE IN PRESS

D. Croccolo, R. Cuppini / International Journal of Adhesion & Adhesives 29 (2009) 234–239 239

one; as discussed before, this superposition zone is not repre- References


sented in Fig. 7 as it was considered as a part of the 50% defect
density area, in a safety condition. [1] Adams RD, Cawley P, Guyott CCH. Int J Adhes Adhes 1987;21:279–90.
Finally, it is possible to state that the AE technique may be used [2] Munns IJ, Georgiou GA. Insight 1995;37(12).
[3] Hill R. NDT Int 1977:63–71.
as an NDT one during the production phase in order to foresee the [4] Muravin GB, Leksovskii AM, Zimting VN, Kubasov NI, Rozumovich EE.
actual failure load of the joint. The present work was developed Defectoskopiya 1991:56–63.
applying the method to some conical steel joints but it may be [5] Solomos GP, Lucia AC, Santulli C, Caretta A. In: Proceedings of the 6eme
conference Europeenne sur les controles non destructifs, Nice, 1994.
extended to other adhesively bonding couplings because there is, [6] Dal Re V. Int J Mater Prod Technol 1999;14(5/6):456–66.
in general, no dependence on the joint’s shape. Obviously, it is [7] Dal Re V. Lamiera 2002:2.
necessary to know a priori the defect density slope map of the [8] Croccolo D, Cuppini R, Dal Re V. In: Proceedings of the XXXI Convegno
Nazionale AIAS, Parma, 2002, p. 196.
cumulative counts and the mathematical model of the failure load
[9] Croccolo D, Cuppini R, Dal Re V. In: Proceedings of the XXXI Convegno
which may be obtained, with a good accuracy, by applying an Nazionale AIAS, Parma, 2002, p. 211.
appropriate DOE. After that, it is possible to apply a moment [10] Dal Re V, Dragoni E, Guarini G. In: Proceedings of the twelfth Danubria-Adria
equal, for instance, to 25–30% of the estimated model failure load symposium, Sopron, 1995. p. 57.
[11] Dal Re V, Dragoni E. Öiaz 1997;6:446.
and to forecast the value of the releasing moment by comparing [12] Croccolo D, Cuppini R. Int J Adhes Adhes 2006;26(7):490.
the AE acquired to those uttered by the test specimen; finally, it is [13] Croccolo D, Cuppini R. In: Proceedings of the fifth international conference on
also possible, during the production phase, to decide whether to mechanics and materials in design, Porto, 2006.

discard or to accept the product.

Vous aimerez peut-être aussi