Introduction

The application of metallic foam core sandwich structures in engineering components has grown in recent years due to their attractive properties such as high strength to weight ratio, high surface to volume ratio, high flexural rigidity, interconnected porosity, high permeability, and special thermal properties (Ref 1-4).

Various manufacturing processes such as cladding, brazing, and welding of the metal sheets have been employed to place the skins on the surface of foam structures to manufacture metallic foam core sandwich structures (Ref 5). These methods have problems in forming and shaping the skins, especially for complex shapes (Ref 6). A method under development by researchers in our laboratory is thermal spraying of coatings on the metallic foam structures to produce the skins. Thermal spray processes potentially offer an economical and technically viable way of depositing skin layers on complex components that cannot otherwise be easily fabricated. Different spray methods including APS, HVOF, and Wire Arc spraying have been employed to deposit super alloy skins on the foam surfaces (Ref 6-9). To be able to deposit thermal spray coatings on the open pore metallic foam structures, it is essential to prevent the penetration of the in-flight molten droplets into the foam holes. This issue has been solved in previous studies by filling the surface of the foam using a paste consisting of metallic powders and a curable resin (Ref 10, 7, 9). This process is followed by curing the paste in an oven at the proper temperature, and subsequently surface machining and grit-blasting to remove the paste on top of the foam struts and expose the tips of the struts slightly above the surface. Molten droplets deposited on the surface of the prepared foam by thermal spraying techniques bond to the strut tips and the other droplets to form the skin. Finally, the temporary surface made of metallic powder and epoxy resin is removed by a burn-out process at the temperature of epoxy degradation, while the coating remains bonded on top of the struts to make the sandwich structure skins.

Adhesion strength, porosity, and microhardness are known as basic and key quality characteristics of thermally sprayed coatings (Ref 11). It is commonly believed that the adhesion of thermally sprayed coatings strongly depends on the condition of the substrate surface prior to the spraying process (Ref 12, 13). The cleanliness and roughness of the surface are the most significant factors affecting the adhesion strength of thermal spray coatings. Grit-blasting of the substrate is required as a surface roughening technique before deposition of almost all the thermal spray coatings to get desirable bonding to the substrate through a mechanical keying mechanism (Ref 14). Grit-blasting also removes the surface contamination and oxides, and provides an active metallic surface that can improve the metallurgical bonding between the substrate and the coating (Ref 15). It has been understood that increasing surface roughness leads to better bonding strength between coating and substrate, but that it increases the porosity content at the same time, affecting the mechanical strength and hardness of the coating (Ref 10, 16). Grit-blasting is controlled by adjusting the following parameters to optimize the outcome: blasting air pressure, grit size, blasting distance and angle, number of passes, and the geometry of the nozzle. The investigation of the effects of each parameter and their interactions on the roughness of the paste-filled foam substrate surface and consequent influence on the coating quality is a complex task that has to be done for fabrication of quality metallic foam core sandwich structures. There are some studies on the surface roughness of bulk substrates in response to variations in the grit-blasting process (Ref 17-23). The bond strength between the thermal spray coating and substrate has also been investigated by several researchers (Ref 20, 24-27). However, surface preparation of the foam surfaces prior to thermal spraying, and its effect on the adhesion strength and porosity content of the skin have not been the focus of any previous study. In fact, a very high percentage of the prepared foam substrate surface consists of the filled paste and only the strut tips exposed above this surface are responsible for bonding between the foam and the skin. Hence, the roughness of the paste-filled foam substrate surface can be viewed as the average height of the exposed struts tips above the temporary surface of the cured paste. It is expected that the temporary surface will exhibit the same droplet impact behavior as bulk surfaces, but there is no connection between temporary surface and coating after the burn-out process.

Therefore, the objective of this study was to optimize the grit-blasting process parameters for surface roughening of the temporary layer on the metallic foam substrates prior to twin-wire arc spraying of alloy 625 to obtain a dense and high adhesion strength coating. This work presents the application of RSM in developing empirical relationships between significant grit-blasting input factors and adhesion strength and between those grit-blasting input factors and the porosity content of the coating.

Materials and Experiments

Preparation of the Metallic Foam Substrates

A 10-mm-thick sheet of open cell nickel foam (supplied by Inco Special Products, Mississauga, ON, Canada) with a pore density of 40 pores per inch (PPI) was used as the substrate (Fig. 1). The Mechanical properties of the as-received foam sheet was measured through the compression test on five cubic samples of 10 mm in edge length that were cut from the nickel foam sheet. Compression test was done by tensile/compression test machine (Autograph AG-I Universal tester, 50 KN, Shimadzu) under constant cross-head moving rate of 1 mm per minute. Under uniaxial compressive loading, the yield strength of the nickel foam was measured as 0.9 ± 0.1 MPa and the Elastic Modulus as 0.020 ± 0.005 GPa. A paste was spread over and into the surface of the foam to obtain a smooth and temporary near-surface skin on both faces of the foam sheet to prevent penetration of the sprayed droplets into the foam internal spaces during Arc Spraying of the coating. The filling paste was prepared by mixing 60 volume percent of Inconel 625 powder (AMDRY 625 Sulzer Metco) with 40 volume percent of a resin (Acrodur 950 L, BASF) to make a high-viscosity mixture. The paste layer extended 1-2 mm into the foam and was cured using a multi-step curing procedure which finally led to a stiff and stable thermoset temporary surface on the foam structure. Curing of the paste was conducted by step-wise heating to 75°, 95°, 125°, and 195° centigrade in a laboratory oven, holding for 2 h at each temperature.

Fig. 1
figure 1

40 PPI Nickel foam (a) sheet (b) SEM micrograph of the pore structure

Full size image

Design of Experiments for Grit-Blasting

A statistical design of experiments based on response surface methodology, the Box-Behnken method, was employed to study the effect of grit-blasting parameters on the prepared foam surface roughness, coating porosity content, and also to indicate the optimum levels for the process parameters. Box-Behnken designs are response surface designs based on the construction of a balanced incomplete block design (Ref 28, 29). This design consists of the midpoints of all edges of a k-factor cube and includes replicated center points for orthogonality and estimation of the global standard deviation (Fig. 2) (Ref 30).

Fig. 2
figure 2

The geometry of Box-Behnken design. Design points are at the midpoints of the edges of the design space and at the center

Full size image

The effects of four variables/factors, namely pressure (A), blasting angle (B), blasting distance (C), and grit size (D), each with three levels (Table 1) were studied using Design Expert version 9 software (http://www.statease.com) in 29 experiments, including five replicates for the center point. The number of blasting gun passes (E) was kept constant for all the experiments. A second-order polynomial model was used to fit the response surfaces with the four varying grit-blasting factors. The proposed response function is represented by Eq 1:

Table 1 Grit-blasting factors with their levels
Full size table
$$ Y = \upbeta_{0} + \mathop \sum \limits_{i = 1}^{k} \upbeta_{i} X_{i} + \mathop \sum \limits_{i = 1}^{k} \upbeta_{ii} X_{i}^{2} + \mathop \sum \nolimits \mathop \sum \limits_{i < j} \upbeta_{ij} X_{i} X_{j} + \epsilon $$
(1)

where Y is the response, β0 is the mean of the response, β i represents the effect of variable X i , β ii , and β ij are the coefficients of regression which present the effect of interactions of variables X i and X j , and ε is the random error (Ref 31).

Grit-Blasting

Crushed brown Al2O3 grit (McMaster-Carr Company) was used for grit-blasting in three different US standard grit sizes including 16 (1180 Microns), 30 (600 Microns), and 46 (335 Microns). Based on the RSM design, the middle grit size was supposed to be 31, but because of availability a grit size of 30 was utilized instead of 31 in the experiments (the effect is negligible).

Wire Arc Spraying

A twin-wire-arc spraying system (ValuArc, Sulzer Metco Inc., Westbury, NY) was utilized to spray Inconel 625 coating (an oxidation resistant Nickel base alloy of Ni 58%, Cr 20-23%, Mo 8-10%, Fe 5%, Nb 3-4% provided by Sulzer Metco, Canada) on the surface of grit-blasted samples with different surface roughnesses. The spraying parameters employed in this study (Table 2) had been previously optimized for the Inconel 625 in our laboratories (Ref 2).

Table 2 Wire Arc thermal spraying parameters
Full size table

After deposition of the coating, a burn-out process was conducted at 400 °C for 3 h to degrade the resin in the filling paste mixture and remove the remaining powder from the sandwich structure. A typical schematic of the nickel foam core sandwich structure prepared by twin-wire arc spraying of Inconel 625 and the SEM image of its cross section are shown in Fig. 3(a) and (b), respectively.

Fig. 3
figure 3

Cross section of deposited Inconel 625 on foam surface using wire arc spraying (a) Macro-scale (b) SEM image

Full size image

Recording the Responses

Roughness Measurements

The average roughness (R a, the arithmetic average of the height of the irregularities from the mean line measured within the sampling length) of the prepared foam surfaces after grit-blasting process was measured using a tracer head Surfometer (Precision Devices Inc. Milan, Michigan) with five replicates for each sample.

Microstructural Characterization

Microstructural characterization of the coating cross sections was conducted by scanning electron microscopy (TM3000, Hitachi High-Technologies Canada Incorporated, Toronto, ON) to investigate the bonding between strut tips and deposited skin, and also to measure the porosity content of the thermally sprayed skins. The ImageJ software (free access for public, http://imagej.nih.gov/ij/) was employed for quantitative measurement of the coatings porosity content. Also, the strut tips’ exposure in the grit-blasted samples was studied using scanning electron microscopy (SEM).

Adhesion Strength

The adhesion strength of the coated samples was measured using a PosiTest portable adhesion tester (Stone Tucker Instruments Inc.) based on the ASTM D4541-09 standard. E-6000 Craft Carded Adhesive was utilized to bond the adhesion test handles to the coating surface. The effect of different substrate roughness and various spraying distances on the adhesion strength of the coating skins to the foam core was examined.

The results obtained from the above experiments were analyzed using the experimental design to fit a model to the data. The factor and level combinations, along with the measurement data, are given in the test matrix (Table 3).

Table 3 Box-Behnken experiment design for grit-blasting of the surface with various levels of process parameters and resulting response values: surface roughness, porosity, and adhesion
Full size table

Predictive Model for Responses

In this study a response surface method (RSM) was employed to predict the effect of grit-blasting process parameters on the responses. Initially, the effects of the grit-blasting parameters on three separate responses were investigated: surface roughness, porosity content, and adhesion strength. However, it was not possible to fit a model to the adhesion strength results because the measured values did not always represent the adhesion strength of the coating to the foam substrate. For sufficiently high adhesion strengths, failure occurred in the foam instead of at the interface between the foam and coating. In these cases, the measured values represented the tensile strength of the foam, not the adhesion strength. Therefore, only the roughness of the prepared surface of the filled foam after grit-blasting and the coating porosity content were chosen as the responses for further investigation.

The responses are expressed below as a function of pressure (A), blasting angle (B), blasting distance (C), and grit size (D):

$$ {\text{Response}} = {\text{f}}\left( {A, B, C, D} \right). $$
(2)

For the four factors, the previously mentioned polynomial model (Eq 1) can be written as follows (Eq 3):

$$ {\text{Y}} = \upbeta_{0} + \upbeta_{1} \left( A \right) + \upbeta_{2} \left( B \right) + \upbeta_{3} \left( C \right) + \upbeta_{4} \left( D \right) + \upbeta_{11} \left( {A^{2} } \right) + \upbeta_{22} \left( {B^{2} } \right) + \upbeta_{33} \left( {C^{2} } \right) + \upbeta_{44} \left( {D^{2} } \right) + \upbeta_{12} \left( {AB} \right) + \upbeta_{13} \left( {AC} \right) + \upbeta_{14} \left( {AD} \right) + \upbeta_{23} \left( {BC} \right) + \upbeta_{24} \left( {BD} \right) + \upbeta_{34} \left( {CD} \right). $$
(3)

After determining the coefficients (at a 95% confidence level) based on the Box-Behnken RSM, an empirical relationship between the responses and the parameters were developed using these coefficients. The final statistical models to predict the responses based on the actual values of the grit-blasting parameters are given below:

$$ \begin{aligned} {\text{Roughness}}\;(\upmu {\text{m}})& = - 8.56 + 1.72(A) + 0.23(B) + 0.016(C) + 0.49(D) - 0.011(A^{2} ) - 0.0014(B^{2} ) \hfill \\ & \quad \,- 0.0004(C^{2} ) - 0.007(D^{2} ) + 0.001(AB) + 0.0029(AC) - 0.0015(AD) \hfill \\&\quad \, - 0.00038(BC) - 0.0017(BD) + 0.00096(CD) \hfill \\ \end{aligned} $$
$$ \begin{aligned} {\text{Porosity}}\;(\% ) \,& = \, - 6.042 \, + \, 0.368(A) \, + \, 0.063(B) \, + \, 0.031(C) \, + \, 0.19(D) \, - \, 0.0046(A^{2} ) \\& \quad\, - 0.0004(B^{2} ) - 0.0002(C^{2} ) - 0.002(D^{2} ) + 0.00078(AB) - 0.00002(AC) \\& \,\quad- 0.0002(AD) - 0.0002(BC) - 0.0006(BD) - 0.0003(CD) \hfill \\ \end{aligned} $$

Checking the Adequacy of the Model

ANOVA analysis was employed to check the adequacy of the models (Ref 32). The model P value and probability were checked to confirm the significance of the empirical relationships. Based on the F value, the most significant factors affecting surface roughness and coating porosity are blasting pressure and blasting distance. The determination coefficient (R-squared) in the ANOVA analysis confirmed the proper fit of the model. According to the R-squared value, less than 1% of the total variation is not explained by the empirical relationship for the surface roughness. This number for the porosity content is 2%. The adjusted R squared for both responses is also high, indicating a high significance for the empirical relationships. It also can be seen that the predicted R squared value is in a good agreement with the adjusted R squared value. The coefficient of variation (CV%) shows a low value in the ANOVA analysis for both responses, confirming the reliability and precision of the fitted models. The significance of the model is reflected by the P value of less than 0.05 for the empirical relationships. The lack of fit P value was 0.8071 and 0.2106 for the porosity content and surface roughness model, respectively, which is insignificant (Ref 33). The normal probability results showed that the residuals fall on the straight line that indicates the errors are distributed normally (Ref 34). Figure 4 and 5 show the graphs for predicted values versus actual values of roughness and porosity. It can be seen that the predicted values are in a very good agreement with the actual values, again demonstrating the adequacy of the fitted model.

Fig. 4
figure 4

Predicted vs. actual roughness

Full size image
Fig. 5
figure 5

Predicted vs. actual porosity

Full size image

Results and Discussion

Effect of Grit-Blasting Parameters on the Responses

The developed models can be used effectively to predict the surface roughness of filled foam surfaces and also porosity content of the coating after spraying as the functions of grit-blasting process parameters. The main parameters and their interactions that affect the surface roughness before spraying and coating porosity content are shown in the form of perturbation plots in Fig. 6 and 7, respectively. In RSM, the effect of all factors at a particular point can be compared by employing perturbation plots (Ref 35). The perturbation plots show the variation of the response by changing deviation of each factor from a specific reference point, while the other factors are kept constant at the reference point value. The slope of each line is representative of the effectiveness of the related factor (Ref 28). From the perturbation plots, it can be concluded that the most significant factor affecting the responses is the air pressure of the grit-blasting process (curve A). The next important factor is the blasting distance (curve B). The other two parameters, blasting angle and grit size, have very small effect on the responses. Their curves are approximately straight and symmetric around the reference line (curves C and D).

Fig. 6
figure 6

Effect of grit-blasting parameters on the surface roughness (perturbation plot)

Full size image
Fig. 7
figure 7

Effect of grit-blasting parameters on the coating porosity content (perturbation plot)

Full size image

The effects of the two most significant grit-blasting factors, namely pressure and blasting distance, on the surface roughness of prepared foam and porosity content of the sprayed skin were plotted using statistical results in surface plots (Fig. 8 and 9). It can be seen that variations in grit-blasting pressure significantly affects both surface roughness and porosity content.

Fig. 8
figure 8

The effect of blasting distance and pressure on the filled foam surface roughness

Full size image
Fig. 9
figure 9

The effect of blasting distance and pressure on the coating porosity content

Full size image

Process Optimization

Coating adhesion has been known as a factor of mechanical strength for metallic foam core sandwich structures. It has been observed that increasing in the surface roughness prior to the coating deposition leads to significant improvement in adhesion strength of the deposits. For the samples with surface roughness of less than 30 µm, failure occurred at the interface of the coating and the foam struts. This illustrates that adhesion strength of these samples was not high enough to make a strong connection between foam and the skin. For the samples with roughness higher than 31 µm, failure occurred in the structure of the foam. It can be concluded that surface roughness around 31 µm and higher is appropriate to produce strong enough contacts between strut tips and the coating. The failure mode of samples prepared using grit-blasting parameters of run 5 and 15 with alloy 625 coating after adhesion strength have been shown in Fig. 10(a) and (b), respectively. It should be mentioned that the foam preparation procedure, including grit-basting, does not affect the mechanical properties of the foam. In fact, the temporary surface protects the foam struts from any damage that can be caused by grit-blasting process.

Fig. 10
figure 10

Failure mode of the coating adhesion test sample for (a) sample 5 and (b) sample 15

Full size image

A non-grit-blasted surface of the filled foam was coated by arc spraying of alloy 625 to illustrate the role of grit-blasting in this application. As it has been shown in Fig. 11, there was no connection between struts and coating for a non-grit-blasted filled foam substrate. As a result, coating was delaminated from the surface of sample after burn-out process. This indicates significance of surface roughening process to get the strut tips sufficiently bonded to the skin.

Fig. 11
figure 11

Sample prepared by spraying on non-grit-blasted surface

Full size image

Cross section of a sandwich structure produced by deposition of alloy 625 on the sample 15 in Table 3 has been shown in Fig. 12. It can be seen that the nickel foam struts have been nicely covered by the molten alloy 625 droplets, and the strut tips are embedded in the coating structure.

Fig. 12
figure 12

SEM micrograph of cross section of strut tips and coatings for grit-blasted sample of run 15

Full size image

Therefore, higher value of surface roughness leads to stronger adhesion of the coating to the foam. Cross sections of the grit-blasted samples were studied by SEM microscopy to investigate distribution and exposed height of nickel foam struts. SEM images of the cross section of the non-grit-blasted sample and the grit-blasted samples for the non-grit-blasted sample and experimental runs numbered, 5, 15, and 25, have been shown in Fig. 13(a-d), respectively. It can be observed that a layer of the filler paste has covered the struts tips for non-grit-blasted sample, which prevents any connection between coating and the foam surface. By increasing the surface roughness, struts start to rise from the filled surface of the foam (Fig. 13b and c). As the adhesion test results show in Table 3, the surface roughness of the prepared foam surface shown in Fig. 13(b) was not enough to make a strong bond with the skin and failure happened at the coating/foam interface. Surface roughness for the sample shown in Fig. 13(c) was high enough to make an adhesion strength higher than foam cohesion strength. While surface roughening is essential for sufficient adhesion strength, but over blasting can lead to deformation of the foam struts and reduce their ability to make mechanical bonds with the skins. For the sample with surface roughness of 42.4, a bent strut is shown in Fig. 13(d). This illustrates that after getting a specific height of struts exposed, more grit-blasting of the surface leads to deformation of struts and the increase in roughness is insignificant. Deformation of struts from the straight forms to the bent shapes can also lead to reducing the covering capability of the in-flight droplets when they hit the strut tips, and finally increase in the coating porosity content.

Fig. 13
figure 13

SEM Micrograph of (a) non-grit-blasted surface, (b) run 5, (c) run 15, (d) run 25

Full size image

Based on the adhesion test results (Table 3), a specific minimum roughness value is needed to provide the desirable adhesion strength while a maximum limit should be considered to avoid bending of the foam struts. This surface roughness range was considered as the boundary for the following optimization step.

Numerical optimization was conducted on the developed RSM model using Design Expert software by limiting the adhesion strength in the range of 164-173 and then defining the target of minimum porosity content for the coating. Hundred different solutions were suggested by the model that the best one is shown in Table 4.

Table 4 The selected set of parameters for grit-blasting process for minimum porosity content and acceptable adhesion strength
Full size table

Three validation runs were performed for the suggested optimization to confirm the reliability and repeatability of the model. The value of the porosity content and adhesion strength are shown in Table 4.

Summary and Conclusion

Statistical design of experiments was employed to predict the effect of grit-blasting parameters on the surface roughness of the prepared foam and coating porosity content. RSM (Box-Behnken method) was used to fit response models in Design Expert version 9 software based on blasting pressure, blasting angle, blasting distance, and grit size. Pressure and blasting distance were found to be the most significant factors in grit-blasting process. Higher surface roughness led to higher adhesion strength while affected the porosity content. Optimization of the process was conducted by statistical software to obtain good enough adhesion strength while porosity content is the minimum. The suggested solution for optimized condition was verified by three experimental replications. Base on the results, the minimum porosity content for alloy 625 coating on 40 ppi nickel foam to have a high enough adhesion strength (in the range of 164-173) was 3.9%.