Characterisation of thermally sprayed near net shape oxide ceramic and cermet coatings by acoustic emission analysis Fr.-W. Bach, K. Möhwald, M. Erne, T. Bause, C. Scheer, Hannover / D Near net shape coating is a trend in thermal spraying being aimed at for several years to lower costs through shortened spraytime and reduced after treatment work. Because of better microstructure compared to conventional coatings, the thickness is often also reduced. To characterise the quality of those thin layer systems the standardized tensile adhesive test is not suitable, as the adhesion outweighs the cohesion by far due to reduced quantities of coating defects. To characterize the coatings behaviour under tensile stress, three-point bending tests were performed. As especially thin cermet coatings do not show abrupt catastrophic failure in bending tests, ultrasonic signals being emitted from cracks propagating through the coatings were taken during the tests. Three coating systems (Cr2O3, Cr3C2-NiCr and WCCoCr) were investigated in this work using fine grained feedstock powders with three different size distributions per system. The tests showed the negative influence of stress moments in case of the oxide ceramic and the embrittlement of the cermets, especially the Cr3C2-NiCr, with increasing spraying temperature regime due to increased formation of solid solutions and decarburization. It can be said, that the analysis of acoustic emission during bending tests gives valuable information about how to achieve thin wear and corrosive protecting coatings being qualified for high operational demands. 1 Introduction Bending tests on brittle materials were established in the field of structural ceramics with the objective to correlate the defects in the microstructure of free standing ceramic bars with their fracture toughness. The test method was adopted by thermal sprayers and combined with the analysis of acoustic emission (AE) since the end of the 1970´s [1]. As the specimen is loaded, strain is increased until cracking occurs in the coating. During loading acoustic events in the ultrasonic range are emitted by the specimen, which relate to plastic deformation, phase transformations as well as crack initiation and growth [2]. Those acoustic events can be recorded by transducers attached to the sample, which transform the high frequency stress waves into a voltage output (see Fig. 1). This burst signals exceeding background noise, which is masked out by a determined threshold level, can be analysed regarding the amount of total counts, the cumulative energy, the rise time and the amplitude distribution [3]. This criteria related to cracking can be correlated with the applied force onto the specimens and therefore their displacement. The goal is to get information about cracking mechanisms which drive the failure of coatings in operation until delamination occurs.

Fig. 1: Burst output of an ultrasonic transducer [2]

AE were performed on oxide ceramics, especially alumina coatings and thermally cycled TBCs [2-14] as well as on cermet coatings [15-17] and one metallic system [18]. There were several ways discussed to achieve the aim of correlating AE events with cracking mechanisms in the examined coating systems. This is due to the fact, that the microstructure of thermally sprayed coatings is inhomogeneous compared to sintered bulk materials and therefore conclusions about the cracking mechanisms often could not be validated afterwards. The distribution of following AE criteria were tried to correlate predominantly with the approach of two different cracking mechanisms (microcracking between single lamellae and macro- or transveral cracking as origin for delamination) and will

be discussed later on: Cumulative events [2, 8, 9, 11, 14, 15, 17], cumulative counts [4-6, 10, 13, 16, 18], cumulative energy [3, 8, 9, 11-12, 14-15, 17-18] and distributions of amplitudes of events [7, 14-16, 18]. 2 Experimental Spraying processes and Equipment

The spraying experiments were carried out with a F4 plasma gun (Sulzer Metco Inc.) and a K2 HVOF gun (GTV GmbH). As substrates specimens of structural steel S235JR with the dimensions 120 mm x 20 mm x 4 mm were chosen. Sample preparation was done with corundum EKF 54 at 0,3 MPa pressure. As said, for each powder feedstock system three size distributions (nominal -15+5, -20+5 and -25+5 µm) came into operation. Before spraying three different types of powder feeding systems (Powder Feed Dynamics Mark XV, Sulzer Metco TWIN 120 and Thermico CPF2-VD) were investigated in respect of their ability to feed the different feedstock powders with minimal variance as this is the first step in near net shape coating.

The experiments were designed as one L9-fields per investigated feedstock investigating three other parameters besides the feedstock grain size distribution (amperage in case of APS- and λ in case of HVOF-spraying, standoff distance and powder feed gas rate) on three levels. Besides the behaviour in three point bending tests the layer criteria roughness characterised by the parameters Ra and RZ, applied coating thickness per gun traverse, porosity and hardness according to Superrockwell HR15N were taken to assure the goal of achieving near net shape coatings with minimised necessity of aftertreatment work. The results of this preliminary experiments were analyzed and parameter settings being optimal for the layers criteria in a weighted order (minimal variance for deposited coating thickness, roughness, hardness and bending behaviour at least) were chosen using nominal is best, minimising, maximising and signed target statistical evaluation. (For a very comprehensive introduction to Mr. Genichi Taguchi´s method of Robust Design see [19].) The confirmatory experiments for the optimal parameter sets forecasted by two software packages (Statistica and Minitab) were done on structural steel S235JR and stainless steel X5CrNi18-10. For comparison samples were sprayed using standard fractionated powders -45+25 µm. Besides the named criteria on this sets ball on disk and corrosion tests were done. All samples (preliminary and confirmatory experiments) were examined by means of XRD to correlate the coatings characteristic hardness and bending behaviour with their phase composition. Bending tests and AE equipment

The bending tests were carried out on an universal testing machine Zwick Z250. The specimens dimensions were 100 mm long, 20 mm wide and 4 mm thick, the span between the supporting bars was 80 millimetres. The specimens were put in the holder with the coating downside being loaded from back. The AE system used was an AMSY-5 by Vallen Systeme GmbH achieving 10 MHz sampling rate. Pre-amplification of the signals was done with a Vallen AEP-4 with a gain of 30,6 dB in the bandwidth range from 2,5 kHz to 3.8 MHz. A high-pass filter of 160 kHz was used to eliminate low frequency noise. Two transducers from Dunegan Engineering Company Inc. (DECI) named SE650-P were attached by magnets on the backside of the samples opposite the coating with silicon grease, having their highest sensibility in the range from 400 to 800 kHz. The following AE characteristics were taken: Number of events, counts, amplitude, energy, event duration and ringdown time. The threshold for the interpretation of the data set was set to an amplitude of 40 dB to suppress sound emitted by the friction occurring between the specimen and the loading fin as well as the supporting bars. 3 Results & Discussion Investigation of energy and amplitude distributions

For interpretation of the taken AE signals first the energy and amplitude distributions were examined with the goal to correlate trends deviating from normal distribution with different cracking mechanisms like it was done in [2, 6 and 9]. The named works pointed out a trimodal distribution of the energy and the amplitude of AE-events from grey alumina coatings due to three different cracking mechanisms, distinguished after the range of released energy into micro-, transitional and macro-cracks. The distributions recorded in the course of this work where somewhat different: For all three coating systems the amplitude and energy distribution could be at most divided in two ranges. The energies in the lower range are lognormal distributed (see the fit for sample V1 in Fig. 2 in the range up to 500 aJ) and in the upper range they are normal distributed (see second fit for V1) In some samples showing a lower cumulative number of events and lower maximum released energy per event, the second trend in the high-energy region cannot be detected at all as it is not reached (see sample V4 in Fig. 2). Another result is gained for the amplitude distributions. The fitting of the histograms of the natural logarithm of the amplitudes like it is done e.g. in [2] show a Gaussian distribution connoting that the amplitudes are lognormal distributed over the hole range. There is only a little skewness onto the left side as here resides the background noise from friction in the experimental setup being not totally masked out (see the lower region in the diagrams of Fig. 3). To verify the nature of this distributions Kolmogorov-Smirnov- and Shapiro-Wilk-tests were performed. If the test is limited to the ranges distinguished through different fits, the tests resulted in a correspondence with the given distributions to a confidence interval of > 99 %. One explanation for this difference between previous done and this work might be the advancement of the hardware used for recording the AE events. The equipment used in [2] is able to record more than 350 events per second steady-state, the Vallen system used in this work writes up to over 30.000 events per second to harddisk. This also might be the reason for the differing total number of events between this work and e.g. [2] by a factor of 30 up to 150 for oxide ceramic coatings taken into account, that the same threshold for the recording of the signals was used. This discussion seems inevitable to succeed in correlating the AE characteristics energy and amplitude with cracking behavior of thermally sprayed coatings. Deflected from this results and the named literature the distribution of energy emitted by each sample was manually divided in two ranges if possible, and the upper range was interpreted as to be caused by macrocracks causing delamination of the coatings during the bending tests. The cumulative released energy was used to quantify the damage potential of the used parameter settings for the tensile stress conditions during the tests. This criteria was analyzed using the smaller-is-better type to achieve coatings with higher fracture toughness realized in the

5 50 500 5000

Observed Value of released Energy [aJ]

0,0005

0,0050

0,0500

0,5000

5,0000

Exp

ecte

d N

orm

al V

alue

E [aJ] V1 E [aJ] V4

Fig. 2: Normal probability plot of energy distributions of two chromia coatings

confirmatory experiments. The results of these experiments will be discussed later on. Correlation of AE signals with coatings fracture behaviour Besides the interpretation of the distribution of recorded AE signals, the fracture behaviour of coatings observed in bending tests can be correlated to the negative influences of residual stress moments in case of oxide ceramics and the embrittlement of cermets due to decarburization.

Fig. 3 on the following page shows the amplitude distribution over time and sample deflection of three differently sprayed chromia coatings using a feedstock powder in the fraction -5+15 µm. After an deflection of 8 mm coating a) shows the worst damage of all three samples with delamination of the coating over a width of about 1 cm. Sample b) contrarily does not show end-to-end delamination and in c) just a few chunking occurred on one side of the specimen. This differing observations can be related to

1. the total number of events decreasing from a) to c) and

2. the abnormal cumulation of events with high amplitude after a deflection of about 1 mm in case of a) and b)

Furthermore the number of cumulated events at the beginning of the experiments in case of a) and b) correlate to the number of events occurring over the whole duration. In fact, the most of these events in case a) could not be recorded due to overmodulation

of the transducers. The origin of this events are tensile residual stress moments in the coatings. As known the heat transferred during the plasma spraying process into the substrate cause this stresses through the mismatch in thermal expansion coefficients of the steel substrate and especially oxide ceramic coatings [20, 21]. The strongest effect on the formation of residual stress has hereby the deposition rate of feedstock material per time and substrate area as the outermost heat is transferred from the molten material into the substrate. The powder feedstock size has in turn the highest positive effect on coating deposition of the parameters investigated in the chromia system being eliminated but this effect is flattened as in Fig. 2 samples prepared with the same feedstock powder are compared. Therefore the parameter sets resulting in formation of high or low residual stress moments correspondent predominantly to a high deposition rate. The deposited coating thickness per gun traverse is on sample b) with 11,1 µm 50 % higher than on c) with 7,4 µm, as the parameter set is optimal for high deposition rates with the used feedstock. For a) the deposition rate lies with 10,6 µm lower than for b), but here the short spraying distance of 80 mm causes additional stress moments besides the impact of the molten feedstock.

Resumptive it can be said, that with the correlation of AE signals and the bending behaviour of oxide ceramic coatings valuable information about the stress moments residing in the coating can be gained. Especially for thin coatings it can be estimated, that low total stress moments in the coatings like it is achieved in coating c) should result in good

Fig. 3: Effects of residual stress moments on bending behaviour of plasma sprayed -15+5 µm chromia coatings

performance in operation. The total number of cracking events is significantly lower in this coating system compared to the other ones.

In case of HVOF spraying of cermet powders an embrittlement of the coatings can occur due to decarburization of the carbides of the feedstock powder and the mixing of carbides with the metal matrix forming solid solutions [22]. This processes are esteemed to lower resistance against tribological wear and the fracture toughness of cermet coating systems.

Fig. 4 shows two X-ray diffractograms of Cr3C2-NiCr coatings sprayed using one fine (sample E2) and one coarser grained feedstock powder (sample N2) with different parameter sets. Though spraying of the narrow fractionated powder was done with cool parameter sets (λ = 1,4, spraying distance = 270 mm) in comparison to the coarse grained feedstock (λ = 1,0, spraying distance = 300 mm), the fine powder is significantly more oxidised during spraying (see the lower diffractogram). Besides the obvious amounts of chromia in the coating presumably descending from the Cr content of the feedstock matrix, decarburization

Fig. 4: Negative effects on the fracture behaviour of Cr3C2-NiCr coatings due to decarburization and oxidation

of the chromium carbide is clarified by the quantitative amounts of Cr23C6 in the coating. Decarburization takes also place when spraying the coarse powder, but the degree of non stoichiometry with the Cr7C3-phase occurring is lower and the quantitative amounts are significant lower.

The effect of the stronger oxidation and decarburization of the fine powder can clearly be seen by the damage symptoms of the coatings on the right side of Fig. 4. Unlike N2 the specimen E2 shows delamination of parts of the coating throughout the whole cross section of the substrate. This result is in accordance with the slightly higher hardness measured on E2 (88 HR15N) in comparison to N2 (86 HR15N) and the higher wear rate for E2 in ball on disk tests It can be said, that the fine grained powder feedstock tends more to oxidation and decarburization than coarser grained powders. But as this effect was forecasted for the parameter set which first and foremost was chosen to reduce the aftertreatment work of the coating system, another parameter set for the fine powder was realized promising better fracture toughness. Basing on the results of the preliminary tests it was deduced, that a strong interaction between the narrow and fine powder fraction and λ exists. For the investigated levels in the preliminary tests (1,4; 1,5 and 1,6) the worst delamination was forecasted for the hot level of 1,4 being the set used for E2, but much better results for spraying with λ of 1,5 or 1,6. For comparison one specimen was sprayed with a λ of 1,6 (named E2_2 in Fig. 5) and better results to N2 were achieved. Comparing the cumulative released energies of all three samples in Fig. 5 the strong interaction between the fine grained

powder feedstock and the level of λ used can clearly be seen.

4 Conclusion Recapitulatory it can be stated, that the combination of bending tests and the investigation by means of Acoustic Emission gives the ability to correlate the fraction behaviour of thermally sprayed coatings with the causing layer criteria phase composition in case of cermets and residual stress moments especially in the case of oxide ceramic coatings. Examination of the fracture mechanisms occurring under load gives valuable information about how to spray coatings that should show good stress-strain-behaviour due to increased fracture toughness. To achieve this, further work should be done regarding

- the understanding which AE event correlates to what type of fracture mechanism, delamination etc. which is discussed until now on contraversly in literature and

- correlation of more layer criteria with the results of bending tests to get a better insight into the dependencies between coatings characteristics and the mechanisms causing failure in operation.

As the focal point of this work laid in achieving near net shape coating it can be said, that using fine grained powder especially in the case of cermets demands for highly adjusted spraying equipment and process to avoid the embrittlement of the coatings by changes in the feedstocks phase composition. On the other hand reducing the thickness of oxide ceramic coatings, which is often aimed for because of the low deposition efficiency of these powders, should be

E 2_1 N 2 E 2_20 1 2 3 4 5 6 7 8 9

Defelection [mm]

-5E5

0

5E5

1E6

1,5E6

2E6

2,5E6

3E6

3,5E6

4E6

4,5E6

5E6

E cu

mul

ativ

e [a

J]

Fig. 5: Cumulative energies of AE events recorded on Cr3C2-NiCr coatings sprayed with –15+5µm (samples E 2_1 and E 2_2) and -45+25 µm (sample N 2) feedstock powders

accompanied by investigation of residual stress distributions in the coatings to avoid the coatings from precipitate failure under operational conditions. 5 Acknowledgements This work was granted by the consortium “Otto von Guericke” e. V. (AiF) under the reference number 14.509 N. This support is greatfully acknowledged by the authors. 6 References [1] Steffens, H. D. and H.-A. Crostack: Non destructive testing of thermally sprayed coatings. Proceedings of the 9´th ITSC, The Hague, 19. – 23. May 1980, pp. 120/128 [2] Lin, C. K., Berndt, C. C: et al.: Acoustic emission studies of alumina-13 % titania free-standing forms during four-point bend tests. Journal of the American Ceramic Society 80 (1997), Issue 9, pp. 2382/2394 [3] C. C. Berndt: Acoustic Emission Evaluation of Plasma-Sprayed Thermal Barrier Coatings. Journal of Engineering for Gas Turbines and Power 107 (1985), pp. 142/146. [4] Shankar, N. R., Berndt, C. C. et al. (1983). "Acoustic emission from thermally cycled plasma- sprayed oxides. American Ceramic Society Bulletin 62 (1983), Issue 5, pp. 614/619 [5] Berndt, C. C. and H. Herman: Failure during thermal cycling of plasma-sprayed thermal barrier coatings. Thin Solid Films 108 (1983), pp. 427/437 [6] C.C. Berndt: Discrimination Of Micro And Macrocracking Processes In Plasma Sprayed Ceramic Coatings. Proceedings of the 11th. Inter. Thermal Spraying Conf., Montreal, 7-12 September 1986, pp. 585/594 [7[ Richard, C. S., Lu, J. et al.: Study of Cr2O3 coatings. Part 1. Microstructures and modulus. Journal of Thermal Spray Technology 4 (1995), Issue 4, pp. 342/346 [8] Lin, C. K., Lei, S.-H. and C. C. Berndt: Acoustic emission response of plasma-sprayed alumina-3% titania deposits. Thin Solid Films 310 (1997), pp. 108/114 [9] Lin, C.-K. and C. C. Berndt: Acoustic emission studies on thermal spray materials. Surface and Coatings Technology 102 (1998), pp. 1/7 [10] Berndt, C. C., Lin, C. K. and S.-H. Lei: Characterization of Cracking within Thermal Spray Deposits by an Acoustic Emission Method – Extended

Abstract. Journal of Thermal Spray Technology 7 (1998), Issue 3, pp. 334/336 [11] Kucuk, A., Berndt, C. C. et al.: Influence of plasma spray parameters on mechanical properties of yttria stabilized zirconia coatings. II: acoustic emission response. Materials Science and Engineering Part A 284 (2000), Issue 1, pp. 41/50 [12] Kucuk, A., Berndt, C. C. et al.: Deformation characteristics of plasma sprayed thermal barrier coatings. Ceramic Engineering and Science Proceedings 21 (2000), Issue 4, pp. 681/689 [13] Ma, X. Q., Cho, S. et al.: Acoustic emission source analysis of plasma sprayed thermal barrier coating during four-point bend tests. Surface and Coatings Technology 139 (2001), Issue 1, pp. 55/62 [14] Park, J. H., Kim, J. S. et al.: Acoustic emission characteristics for diagnosis of TBC damaged by high- temperature thermal fatigue. Journal of Materials Processing Technology 187/188 (2007), pp. 537/541 [15] Dalmas; D., Benmedakhene, S. et al.: Characterization of Cracking Within WC-Co Coated Materials by an Acoustic Emission Method During Four Points Bending Test. In: Proceedings of the 1st International Thermal Spray Conference, 9.-11. May 2000, Montreal, pp. 1335/1340 [16] Miguel, J. M., Guilemany, J. M. et al.: Acoustic emission study on WC-Co thermal sprayed coatings. Materials Science and Engineering, Part A 352 (2003), Issue 1-2, pp. 55/63 [17] Bouaricha, S., Legoux, J. G., et al.: Bending behavior of HVOF produced WC-17Co coating: Investigated by acoustic emission. Journal of Thermal Spray Technology 13 (2004), Issue 3, pp. 405/414 [18] Voyer, J. and H. Kreye: Determination of cracking resistance of thermal spray coatings during four-point bend testing using an acoustic emission technique. Journal of Thermal Spray Technology 12 (2003), Issue 3, pp. 416/426 [19] Madhav S. Phadke: Quality Engineering Using Robust Design. Prentice Hall PTR, 1989, ISBN: 0137451679 [20] Mateijka, D., Benko, B.: Plasma spraying of metallic and ceramic materials. John Wiley and Sons, 1989, ISBN: 0471918768 [21] Heimann, R.B.: Plasma-Spray Coating. VCH Verlagsgesesellschaft mbH, 1996, ISBN: 3527294309 [22] Zimmermann, S. et al.: Improved coating properties by optimised carbide powders for modern HVOF systems. Proceedings of the 6´th Kolloquium

Hochgeschwindigkeitsflammspritzen, 27.-28.11.2003, Erding, pp. 31/38