Abstract
Extensive efforts have been devoted in both the engineering and scientific domains to seek new designs and processing techniques capable of making stronger and tougher materials. One such method for enhancing such damagetolerance in metallic alloys is a surface nanocrystallization technology that involves the use of hundreds of small hard balls which are vibrated using highpower ultrasound so that they impact onto the surface of a material at high speed (termed Surface Mechanical Attrition Treatment or SMAT). However, few studies have been devoted to the precise underlying mechanical mechanisms associated with this technology and the effect of processing parameters. As SMAT is dynamic plastic deformation process, here we use random impact deformation as a means to investigate the relationship between impact deformation and the parameters involved in the processing, specifically ball size, impact velocity, ball density and kinetic energy. Using analytical and numerical solutions, we examine the size of the indents and the depths of the associated plastic zones induced by random impacts, with results verified by experiment in austenitic stainless steels. In addition, global random impact and local impact frequency models are developed to analyze the statistical characteristics of random impact coverage, together with a description of the effect of random multiple impacts, which are more reflective of SMAT. We believe that these models will serve as a necessary foundation for further, and more energyefficient, development of such surface nanocrystalline processing technologies for the strengthening of metallic materials.
Introduction
Deformation processes based on the impact of particles have been utilized extensively to harden surfaces, by imparting residual compressive stresses and/or by creating a workhardened layer for the purpose of improving surface properties, such as resistance to fatigue or wear.^{1,2} More recently, these techniques have also been used to make stronger engineering metal alloys,^{3,4,5} through the use of techniques such as air blast and ultrasonic shot peening,^{6} surface mechanical attrition treatment,^{7} particle impact processing,^{8} and surface nanocrystallization and hardening,^{9} specifically to develop surface hardness and structural gradients. Here we focus on the technique of surface mechanical attrition treatment (or SMAT) which has been widely used in the generation of nanostructures and subsurface gradients to improve the damagetolerance of structural materials.^{10,11,12,13} The technique was first introduced by Lu and Lu,^{14} who exploited an ultrasonic setup to vibrate spherical balls actuated by highpower ultrasound.^{15} During this process, hard spherical balls are randomly impacted on surface of target materials, specifically to establish nanosized crystalline layers on a variety of metallic materials,^{16,17,18,19,20} including steels, nickelbased alloys and Mgbased alloys.^{21} However, as numerous indents are generated on the sample surface layer by such impacts, the successful use of SMAT requires a number of parameters to be effectively controlled, including ball size, ball density, impact velocity and impact kinetic energy.
One good example of the role of SMAT in enhancing the properties of metallic alloys is illustrated in Fig. 1 for an AISI 316 L austenitic stainless steel, where we show how the process can induce superior mechanical properties through surface modification.^{3,4} Specifically, we compare the microstructure and the mechanical behavior of a SMATtreated AISI 316 L stainless steel with that of an asannealed AISI 316 L stainless steel. Following SMAT treatments, the hardness distribution along the crosssection of the SMATtreated steel is a factor of ~3–4 times higher than that of the asannealed steel (Fig. 1a), which is consistent with the uniaxial tensile test results (for both engineering and true stress/strains) which show corresponding factors of 4 and 3fold increases in the yield and tensile strengths, respectively (Fig. 1b). Crosssection scanning electron microscopy (SEM) and electron backscatter diffraction (EBSD) images and related schematic illustrations show that this results from a gradient microstructure in grain size in the SMATtreated steel from surface to core (from ~80 nm to ~20 μm), with dualphase crystal phases comprising alternating nanoband austenite and nanolamellar martensite within each grain^{3,4} (Fig. 1c, e, g), as compared to a nominally equiaxed singlephase austenitic grain structure in the annealed steel (Fig. 1d, f, h). A comparison in Fig. 1i, j with other highstrength steels, namely dualphase steels and TRIPsteels,^{22,23} of the tensile and yield strength properties of the SMATtreated stainless steel as a function of their tensile ductility, clearly reveals that SMAT treatments have the capacity to radically enhance the damagetolerant properties of medium to highstrength steels in terms of much improved combinations of strength and ductility.
The purpose of the current work is to provide a relaible mechanistic and modeling basis for the successful adoption of surface nanocrystallization for the purpose of inducing these exceptional mechanical properties. Specifically, we use analytical modeling and numerical analysis of the indent size induced by random impacts, as a function of the parameters such as the ball size, ball density, impact velocity and kinetic energy, and of the resulting depths of the impact plastic zones; theoretical results are then validated by experiment. Based on this, a global random impact model and a local impact frequency model are developed to analyze the statistical characteristics of the impact coverage on the material under treatment. We further provide an analogous model to describe the phenomenon of full coverage by random multiple impacts, which is most representative of SMAT. Given that the salient mechanisms responsible for the nanostructuring are associated with surface nanorecrystallization induced by the plastic deformation and residual stresses associated with the impact events, we believe that our numerical and experimental observations, coupled with systematically developed modeling, can afford the necessary foundation for further development and optimization of superior surface mechanical attrition procedures.
The SMAT treatment process is a dynamic plastic deformation process, where plates of the material to be treated are placed on the top side of a chamber containing hundreds of 3mm diameter hard balls, which are vibrated using highpower ultrasound so that they impact onto the plate surfaces at high speed. Plastic deformation in the surface layers, coupled with the large strains and high strain rates, result in a progressive reduction of the surface microstructure into nanograins. Indentation size is important here to achieving 100% coverage; moreover, the depth of plastic zone beneath the intents is also critical as it determines the thickness of the workhardened surface layer as well as depth below the surface where beneficial residual compressive stresses are created. To fully understand and optimize the SMAT process, in current study quantitative relationships between SMAT parameters, indent size and plasticzone size are determined and validated by experiment.
Previous studies^{24} on WC/Co ball impacts on Nibase C2000 Hastelloy plate revealed that the plasticzone depth increases with the kinetic energy of balls and, for a given kinetic energy, the ball size. Associated analytical models^{25,26} to predict the surface deformation in a component induced by highenergy impacts have been mainly based on Hertzian theory for the elastic contact between a sphere and a semiinfinite solid,^{27} where depth of the plastic zone (r_{y}) induced by highenergy ball impacts has the following dependencies on the density ρ_{B}, normal impact velocity v_{n} and diameter d_{B} of the impacting ball:^{24}
These analytical approaches though neglect factors such as work hardening, strainrate sensitivity, friction and thermal effects. The work of Dai^{31,32} offers an analytical model to predict the surface roughness (indent depth) of aluminum 5052 plates treated with a surface nanocrystallization and hardening process, but this specific approach^{32} only takes into consideration the strainhardening effect during the impact process and merely provides the means to calculate the peaktovalley size of the impactdamaged zones. Moreover, Dai’s analyses are only strictly applicable to deformation induced by normal impacts, i.e., perpendicular to the surface of the target solid. During SMAT, the impact processes are random with a high strain rate under impact deformation. Accordingly, to accurately predict the response of a sample submitted to such surface treatments, the effects of high strain rates on the flow stress must be included in the constitutive models. To address this, we apply here the JohnsonCook (JC) plasticity hardening model^{28} to estimate the plastic zones induced by high strainrate random impacts involving elasticplastic dynamic deformation. This model represents an empirical constitutive equation that was developed from the experimental study of the fracture characteristics of metals subjected to various strains, strain rates, temperatures and pressures.^{28} The equivalent flow stress, σ_{f}, in J–C model is given by:
where σ_{y} is the initial yield strength of the material at reference temperature (T_{r}), \(\bar \varepsilon\) is the equivalent strain, and B, n, C and m are material constants. Specifically, the constant B and strainhardening exponent n represent the constitutive law at low homologous temperatures, C and m are the corresponding ratedependent constants which respectively represent the coefficient of strainrate hardening and the thermalsoftening exponent, T and T_{m} are respectively the current and material’s melting temperature, and \({\dot{\bar{\varepsilon}}}\) is the equivalent plastic strain rate which is normalized in Eq. (2) by a reference strain rate \(\dot \varepsilon _0\) set to unity.
In addition, we used finite element analysis (FEA), which can provide a powerful method for simulating single and multiple ball impacts onto a target, as the dynamic, highvelocity impact of a ball and the resulting elasticplastic behavior of the target can be both readily analyzed. Figure 2a, b illustrates the FEA model for the threedimensional (3D) dynamic analysis that was performed in this study using the commercial code Abaqus/Explicit mode. Eightnode linear brick elements with reduced integration (C3D8R) were used with the mesh modeled to be very fine close to the impact zone, to achieve good resolution of the plastic strain distribution beneath the surface, and to be coarser away from the impact zone. The ball was also fine meshed to ensure that it possessed a smooth spherial surface condition on impact during the SMAT process. Specifically, we used 194, 840 elements in the single impact model. The size of elements in the impact zone was chosen to be 1/24^{th} of dimple size; this is considerably smaller than previous mesh convergence studies^{33}^{–}^{35}, which utilized elements that were 1/15^{th} and 1/10^{th} of the dimple size. The implication of this is that we have taken a smaller mesh size parameter than all previous methods to ensure convergence of our calculations. The target material plate was restrained against all displacements and rotations on the top end.
In this work, a widelyused AISI 316 L austenitic stainless steel was chosen as the target material; its nominal chemical composition and material properties (for the JC model) are given, respectively, in Tables 1 and 2.^{29} Our experimental studies^{3,4,30} showed that both the hardness and yield/tensile strengths can be significantly improved in SMATtreated materials (Fig. 1). Our numerical study is consistent with the experimental results; details are presented below.
Results
Numerical analysis
To explore the salient micromechanisms which underpin these nanostructuring procedures, we designed the numerical simulation system depicted in Fig. 2. Schematic illustrations of the indentation process are shown in Fig. 2c, together with simulated contours of the indent created by one impact of a rigid ball (Fig. 2d); d_{B} is the ball diameter, v and v′ are, respectively, the impingement and rebound velocities, D is the indent diameter and h is the indent depth. For a 3mm ball impacting on an AISI 316 L stainless steel surface at a speed of 8 m/s, the depth of the indent increases with change in time from t = 0 to ~ 5 μs, whereupon the depth subsequently is reduced due to springback (Fig. 2d). In addition, a small pileup is generated near the edge of indent. The impact angle θ, defined as the angle between the impact direction and the horizontal plane in 3D space, has an important effect on the indent size. Indeed, the indent shape will be elliptical if θ is less than 90° (Fig. 2e, f). The indent diameters along two crossover directions, D_{xy} and D_{zy}, are measured for the impacts at θ = 15° and θ = 30°, respectively (Fig. 2g, h). As the ellipticities, D_{xy}/D_{zy}, are not large, for simplification the indent shape can be regarded as being circular in our study. Note that the model developed here is a macroscopic model without emphasis on nano/microscale phenomena such as dislocation activity. However, we consider the subsequent recovery following impact, as shown in Fig. 2d; after the ball bounces back off the surface, a permanent indent is generated, the final indent depth of which is determined by the plastic deformation during contact and elastic recovery from springback to an equilibrium position where all internal stresses are relatively in balance.
Numerical solutions for indent size under normal impact
The indent size generated by normal impacts is significantly influenced by the ball diameter, impact velocity and ball density based on simulation results of numerous impact cases (Fig. 3a–d). The following numerical relations are found: (a) the indent diameter D has a linear relationship with ball diameter d_{B}, and a powerlaw relationship with both the normal impact velocity ν_{n} and ball density ρ_{B}, as described in Eq. (3); (b) the indent depth h, which describes the roughness of the component surface, has a linear relationship with the ball diameter d_{B} and normal impact velocity ν_{n} and squareroot law with the ball density ρ_{B}, as described in Eq. (4):
The indent depth dependencies in Eq. (4) are consistent with the peaktovalley value predicted for aluminum treated with a surface nanocrystalline and hardening process from the analytical model proposed by Dai et al.^{31} Additionally, the indent size can be found to have a relationship with the impact kinetic energy E_{k} of the impact ball:
where:
The value of the indent diameter and the indent depth can be further numerically expressed as:
or
where d_{B}, v_{n}, ρ_{B} and E_{k} are the ball diameter, normal impact velocity, ball density and impact kinetic energy, respectively, with units listed in Table 3. k_{1}, k_{2}, k_{3} and k_{4} are material constants. Based on our study, we find that k_{1}, k_{2}, k_{3} and k_{4} remain almost constant; their average values for AISI 316 L stainless steel, obtained from the simulation results, were k_{1} = k_{1average} = 40, k_{2} = k_{2average} = 0.35, k_{3} = 56, and k_{4} = 0.68. Figure 3a–d illustrates the simulated points and fitting curves that are fit well by Eqs. (7, 8).
The newly developed model can also be applied to other structural materials, such as AISI 304 L stainless steels, where the indent size generated by single normal impact again satisfies the proposed function in Eqs. (7, 8), with the constants found to be k_{1} = 36.8, k_{2} = 0.29, k_{3} = 51.5, and k_{4} = 0.57. The model was further verified by the equivalence of the numerical solutions with experiments for the indent depth, specifically the peaktovalley values, for surface nanocrystallized and hardened 5052 aluminum plates^{31} where a value of k_{2} ≈ 0.71 was found.^{32}
Numerical solutions for indent size under random impact
The normal impact model described above can be readily extended to impacts at oblique angles, where the oblique impact angle θ is defined between the impact direction and the horizontal plane in 3D space, as shown in Fig. 3e. First, the material parameters k_{1} and k_{2} in Eqs. (7, 8) must be verified for impacts at oblique angles. Here, the value of the indent diameter and the indent depth can be expressed as:
With the same ball size, d_{B} = 3 mm, same ball density, ρ_{B} = 8.0 g/cm^{3} (equivalent to 8,000 kg/m^{3}), and same magnitude of impact velocity v = 10 m/s, the profiles of indentations induced by impacts at different oblique angles are shown in Fig. 3f, where h and h′ are two values of the indent depth due to the nonsymmetrical shape of indents resulting from oblique impacts. As seen in Fig. 3g, h, the values of the corresponding constants k_{1}, k_{2} and k_{2}′ for impacts at different oblique angles are almost the same with those values for the normal impact case. Different indent sizes with different oblique impact angles can now be predicted both for SMAT and, by comparison, for shot peening. As the impact angle in shot peening tends to be uniform, Eq. (3) can be used directly; for SMAT, conversely, the impact angle θ is random, θ ∈ [0°, 90°], such the average indent size \(\bar D_{{\mathrm{SMAT}}}\) and \(\bar h_{{\mathrm{SMAT}}}\) can be evaluated as:
where \(\gamma _D = \frac{{\mathop {\int }\nolimits_0^{{\mathrm{90}}} ({\mathrm{sin}}\theta )^{1/2}{\mathrm{d}}\theta }}{{{\mathrm{90}}}} \approx {\mathrm{0}}{\mathrm{.76}}\), and \(\gamma _h = \frac{{\mathop {\int }\nolimits_0^{{\mathrm{90}}} ({\mathrm{sin}}\theta ){\mathrm{d}}\theta }}{{{\mathrm{90}}}} \approx {\mathrm{0}}{\mathrm{.64}}\). Eq. (14) can then be used to give a quantitative prediction of surface roughness of a material following SMAT treatments.
Experimental measurements
The ball size and density, used in Eqs. (7, 8), are easily measured and controlled, whereas the impact velocity cannot be controlled directly in SMAT experiments because the motion of balls is propelled by the vibration of the ultrasonic horn surface in random fashion (Fig. 4a). To validate these solutions (Eqs. (7, 8)), the following experiment was designed to determine whether such numerical simulations using finite element methods are capable of predicting accurate indent sizes.
The velocities of the flying balls were measured by taking consecutive frame photographs with a highspeed camera (Phantom VR706) at the rate of 5,000 frames per second. The delay time of each frame was 0.2 ms so that the impact velocity could be calculated by counting the number of frames a ball travels from the bottom of the chamber to the sample surface for a fixed distance of 25 mm. Impingements were recorded by the highspeed camera. As an example, 16 frames were captured during the flight of a 3mm steel ball onto the sample surface; four typical images are shown in Fig. 4b. The resulting vertical velocity distribution of the selected balls for multiple impacts, shown in Fig. 4c, gives an average velocity of 7.0 m/s.
The measured diameters of the indents on the surface of the SMATtreated AISI 316 L stainless steel are shown in Fig. 4d. Choosing the larger circular dimples because they were more likely induced by normal impacts, the average diameter from measurements on 100 indents was 515 μm, consistent with an average indent diameter of 525 μm predicted by the FEM simulation and 531 μm from the analytical model of Eq. (7). There is approximately a 3% difference between the experimental value and that calculated by the proposed numerical solutions.
Plastic zone under oblique impacts
Considering the oblique impact model with the same ball size d_{B} = 3 mm, ball density ρ_{B} = 8,000 kg/m^{3}, and initial normal velocity v_{n} = 10 m/s, the contours of the displacement along the thickness (ydirection) for different oblique angles are shown in Fig. 5b–f. The area under each impact is extruded to form peaks (designated in blue) and valleys (in red). The percentage of the pileup area at the rim of the indent increases with increasing oblique angle; by contract, normal impacts lead to a symmetrical shape of indent.
Figure 5g shows the distribution of equivalent plastic (Mises) strain, \(\bar \varepsilon\), along the depth of the indent for different oblique impact angles, where \(\bar \varepsilon = \frac{1}{{3\sqrt 2 }}( {( {\varepsilon _{{\it{xx}}}  \varepsilon _{{\it{yy}}}} )^2 +\, ( {\varepsilon _{{\it{yy}}}  \varepsilon _{{\it{zz}}}} )^2 +\, ( {\varepsilon _{zz}  \varepsilon _{{\it{xx}}}} )^2 +\, 6( {\varepsilon _{{\it{xy}}}^2 + \varepsilon _{{\it{yz}}}^2 + \varepsilon _{{\it{zx}}}^2} )} )^{{\mathrm{1/2}}}\). These numerical results in Fig. 5 show that the location of the maximum equivalent plastic strain is not at the actual surface of the impact, but slightly below it. They also clearly reveal that impacts generated with the same normal impact velocity, v_{n}, will induce the same degree of plastic deformation beneath the impact point. The plastic strains are nonsymmetric for impacts at an oblique angle. The equivalent plastic strain, shown in Fig. 5, represents the equivalent plastic strain distribution along the depth beneath the impact point. Specifically, the depth of the material influenced by plastic deformation under such impacts, r_{y}, can be computed to be approximately equal to the indent diameter D, if the onset of plastic deformation is defined to occur when the equivalent plastic strain reaches a certain critical value of \(\bar \varepsilon\) = 0.2%. Accordingly:
For impacts at random oblique angle θ, as in SMAT, the average depth of plastic zone can be calculated as \(\bar r_y^{\,{\mathrm{SMAT}}} \approx \gamma _l \cdot k_1 \cdot (d_B) \cdot (v)^{{\mathrm{1/2}}} \cdot (\rho _B)^{{\mathrm{1/4}}}\) with γ_{l} ≈ 0.76. This result has significant importance in estimating the plastic zone influenced by impact deformation.
Analytical evaluation of SMAT coverage
In order to control the process of the surface modification treatment accurately so that sufficient indent coverage can be attained, it is important to know the relationship between impact ball parameters and the indent coverage on a sample surface. This section focuses on how to develop such relationships using the threedimensional impact model depicted in Fig. 3e. To do this, the velocity components are first redefined in Eq. (16) as:
Global coverage rate of impacts
The impact process associated with SMAT is random, which implies that some regions of the sample surface may receive more rigid ball impacts while others less. However, by increasing the duration of the treatment, it can be presumed that the entire component surface will eventually be covered by indents. To analyze this, a random impact model was developed to investigate the statistical characteristics of impacts using MATLAB. We used a sample to be SMATtreated with a surface diameter of 70 mm, and made the following assumptions in the model: (a) the impact events are independent from each other, which means (i) any interaction of flying balls before they impact the surface is not considered, and (ii) an impact event is modeled to take place only after the preceding impact event is complete, i.e., from plastic deformation to elastic recovery; (b) the impact velocity for each flying ball is identical, (c) the oblique impact angles are random; (d) the indent size created by each impact is different but can be calculated by the proposed numerical solution in Eq. (11).
The simulation results are presented in Fig. 6. After 1000 impacts by 8mm balls, some parts of the surface still have not been covered by indents (Fig. 6a–c), while after 20,000 impacts, the entire plate surface is covered (Fig. 6d–f). To investigate, as a function of ball parameters, the percentage of indent coverage, specifically the ratio of indented area to total sample surface area, impact processes with identical (Fig. 6g) and random oblique angles (Fig. 6h) were respectively modeled. The scatter points in Fig. 6g, h show the percentage of area covered by indents, the coverage ratio, CR, vs. the total number of impacts N. Seven fitting curves in Fig. 6g, h all satisfy the following type of Avrami equation:
where k_{C} is a fitting parameter related to the indent area (S_{indent}) for a single impact and the sample surface area (S_{plate}), given in terms of a constant α by:
By substituting Eqs. (11, 18) into Eq. (17), the relationship between the indent coverage and impact ball parameters can be obtained as:
The simulation results in Fig. 6g give a value of α of unity provided the impact angle θ is uniform, e.g., as in shot peening. Figure 6h shows the corresponding predictions for the coverage rate as a function of the number of random oblique angle impacts, i.e., for SMAT. The coverage ratio for SMAT can be expressed as:
where γ_{C} was obtained from the fitting curves to be equal to 0.76, and γ_{C} = γ_{D} because the coverage rate is directly related to the average indent size.
Local evaluation of the SMAT impact frequency
To construct a method for calculating the effective coverage, we envisage a model where randomly distributed impacts are assumed to occur on a twodimensional circular area S_{plate} with radius R_{plate} (based on the size of our SMAT fixture), as shown in Fig. 6i. Each circle here represents an indent. The diameter D of the indent can be determined by Eq. (11) for a given set of ball parameters. We take the center points of the indents (P = [P_{1}, P_{2}, P_{3},…, P_{i}, …, P_{N}]) to have coordinates (x_{i}, y_{i}), i = 1, 2, 3,…, N.
Assuming that a location P_{S} with a coordinate (x_{s}, y_{s}) on the sample surface will be impacted by the ith impingement at a random oblique angle if the distance d_{si} between P_{s} and the center point P_{i} of the indent is less than or equal to D_{i}/2, where D_{i} is the diameter of the indent generated by the i^{th} impingement and calculated by Eq. (11), it follows that:
Accordingly, the Nsequenced field variable \(\eta _{P_S} = \left( {\eta _{P_{\rm{S}}}^1,\,\eta _{P_{\rm{S}}}^2,\,\eta _{P_{\rm{S}}}^3, \ldots ,\,\eta _{P_{\rm{S}}}^i, \ldots,\,\eta _{P_{\rm{S}}}^N} \right)\) can be generated depending on whether the position P_{S} is impacted by the flying balls or not. Similar to the diffuseinterface model (Fig. 6j) in classical phasefield theory,^{33} the field variables \(\eta _{P_S}\) should be continuous functions between 0 and 1 depending on the percentage of the area covered.
Since each indent size is rather small compared with the whole sample surface, by a factor of between ~10^{−5} and 10^{–4}, the field variable \(\eta _{P_S}\) can be assumed to be a step function, i.e., with a discontinuous sharp interface, as shown in Fig. 6k for only two values, 1 or 0, for simplification:
The number of impacts n_{s} received by a randomly selected position P_{S} can then be defined as:
We further find that the probability (P(x)) of one surface point receiving x number of impacts can be characterized by a normal distribution, as shown in Fig. 6l, m and expressed as:
where a and c are real constants and \(\bar n\) is the average number of impacts. For a 3mm ball (v = 8.0 m/s, ρ_{B} = 8,000 kg/m^{3}) impacting an AISI 316L stainless steel 70mm diameter plate, the indent diameter can be calculated by Eq. (11), with the number of impacts obtained from Eqs. (19, 20).
The statistical characteristics of the impact frequency for different types of impact, i.e., at uniform impact and random oblique impact angles, are shown respectively in Fig. 6l, m. If full coverage is defined to be equivalent to 99% of the actual coverage, it is clear that points on the plate must be impacted on the order of 4 to 5 times to achieve full coverage under normal impacts, whereas for impacts at random oblique angles, which is the situation during SMAT treatments, each point would need around 11 to 12 hitting impacts.
Fullcoverage multipleimpact model
Based on the analysis described above, clearly the process of SMAT demands multiple impacts to attain full coverage. Accordingly, here we develop a fullcoverage multipleimpact model to describe this scenario. Figure 7a, b presents the modeling considerations for multiple impingements which consider simultaneously full coverage, the location of random impacts and oblique angle impacts. Each circle in Fig. 7a represents an indent induced by one impact, with the circle diameters calculated using Eq. (11). For the random impacts during SMAT, the global full coverage is generated from Eq. (19), as illustrated in Fig. 7a. It is apparent that the indent sizes are different from each other due to the differing oblique angles of impact. For computational efficiency, only the local fullcoverage model is transferred into the finite element formulation (Fig. 7b), where we consider a total of 64 balls, all with a diameter of 3 mm. Using this local fullcoverage, random multipleimpact model, the impact location and impact angle of each ball are obtained from the corresponding global coverage using Eqs. (17–24). For such random, multiple impacts, we consider the overlap of indents by explicitly analyzing the first and second impacts. Specifically, the second ball is considered to impact the concave region of the first indent. Multiple impacts involving complex mechanical and multiaxial loading conditions result in strain hardening and high strainrate hardening under the plastic deformation. To explain how we handle this, we have developed the flow chart to illustrate how our model for full coverage multiple impacts works and how it is coupled to the finite element analysis. This is described in the Methods section with the flow chart shown in Fig. 8.
The contours of the resulting surface roughness were derived (Fig. 7c) and show clear indents with a spherical base generated by the impacts of the balls. The sizes of the indents are different, however, and exhibit some sharp ridges on the side of impacting surface.
We also investigated the states of accumulated plastic strain using the random multipleimpact model. Specifically, by consideration of the concave geometry caused by the previous impact, the equivalent plastic strain distribution on the impact surface induced by four successive random impacts during the SMAT process was calculated; results are shown in Fig. 7d. It is apparent from this figure that the sharp ridges of the indent induced by the first impact are changed by the second impact. The random multiple impact model described in this paper attempts to simulate the real random impact process in threedimensional space; it therefore provides predictions of the strain as well as the geometry of the indents on the impact plane. The oblique angles of impact and the impact locations are both modeled as random; however, the impact angle and location of each impact are recorded for one fullcoverage random impact process and such data are then used to buildup the local finite element model to investigate and calculate the material behavior under such random impacts. To illustrate the geometry of the deformation process during multiple impacts, the contours of displacement on the impact surface are shown in Fig. 7c, where the final geometry is represented by the contours of the resulting surface roughness. The contours of equivalent plastic strains in the cross section of plate during the impact process are illustrated in Fig. 7e, f. Whereas the maximum equivalent plastic strain is calculated to be ~0.1 beneath one single normal impact (Fig. 7e), after random multiple impacts, the maximum equivalent plastic strain is increased by more than a factor of three to 0.36 (Fig. 7f). The distribution of such equivalent plastic strain in the depth direction beneath the center of the local impact zone is plotted in Fig. 7g. It is interesting to note here that if the onset of plastic deformation is defined to occur when the equivalent plastic strain reaches a certain critical value of \(\bar \varepsilon\) = 0.2%, the depth of the deformation surface layer under a single impact at one location and that after random multiple impacts stays essentially unchanged. In fact, for such random multiple impacts, each point would receive some 11 to 12 impacts, as shown in Fig. 6m. The fundamental reason why the equivalent strains caused by a single impact vs. multiple impacts are almost identical below a depth of 280 μm is that the material displays strong strain hardening. As the incident energy of the ball is limited, the maximum strain levels saturate after only a small number of impacts. Our analysis also predicts that the depth of the material influenced by plastic deformation is approximately equal to the indent diameter D, as given in Eq. (15). Thus, from a practical perspective, if a larger depth of the deformed layer is required, one approach would be to utilize higherdensity balls with a larger diameter to impact the target surface at high velocity. For example, in some treatment cases, 8mm diameter balls with a larger density have been employed precisely for this purpose. However, the magnitude of the plastic strains and the width and total volume of the zone of plastic deformation induced by random multiple impacts will be naturally larger than that induced by a single impact at one point.
The effect of random multiple collisions on the equivalent flow stress is shown in Fig. 7h. The flow stress calculated on the assumption of full coverage and multiple impacts is clearly much larger than the corresponding stress calculated assuming successive multiple impacts at one location. We have also determined the influence of the number of impacts as well as the random impact angles on the value of the equivalent flow stress; the flow stress shown in Fig. 7h is the steadystate value. Our results indicate that adjacent random impacts markedly influence the resulting stress distribution. Thus, a multipleimpact model has been developed that can be utilized to simulate the evolution of mechanical properties as a function of the processing parameters associated with SMAT. Currently, in our analysis, the fullcoverage random multipleimpact model for the SMAT process is statistical in nature. The random impact location and random impact angle for each impact are recorded until the impact surface is fully covered by indents. We believe that these models can be used as an efficient simulation tool to provide a highfidelity understanding of the driving force for nanocrystallization. Moreover, the approach can be readily extended to any metallic material, even with complex shapes.
Discussion
The surface modification technique of surface mechanical attrition, or SMAT, involves the highspeed impact of small hard balls onto a metal surface to induce surface nanocrystallization and consequent graded microstructures and hardening in structural steels and other metals. To aid the successful use of this technique, in this work analytical and numerical solutions for the indent size, indent depth, extent of local plasticity and impact coverage, as a function of the process parameters of ball size, impact velocity, ball density and kinetic energy, have been proposed and developed. Indeed, we derive both single impact and fullcoverage random multipleimpact models for these scenarios. Our analyses reveal that during such impactinduced surface nanocrystallization, the indent diameter generated by a single impact displays a linear relationship with ball diameter and a powerlaw relationship with the normal component of impact velocity and ball density; conversely, the indent depth, which describes the roughness of the component surface, displays a linear relationship with ball diameter and impact velocity and a squareroot relationship with ball density. We also derive the relationship between the extent of plasticity surrounding the indents and the parameters of ball impact process for random oblique impact angles that are characteristic of surface nanocrystallization techniques.
In general, we report here a newly developed model which we believe represents a viable approach for characterizing the mechanical effect of SMAT for a wide range of metallic materials. As indicated in the numerical solution, the material constants can be obtained for other materials using these same analytical formulas. As such, our analytical formulations can be readily applied to other metallic materials by employing any impact parameters to obtain the material constants. We believe that our results, which we verify by experiment, can provide a necessary foundation for an improved understanding of the SMAT procedure for metals, as well as for the further development of microstructural evolution in metallic alloys for improved strength, ductility, hardness and toughness.
Methods
The procedures describing the general framework of how we implement our proposed full coverage multiple impacts model is described below, in particular to address strain hardening and high strain rate hardening under the plastic deformation, which we solve using finite element methods. Since the elasticplastic material behavior is nonlinear, the solutions are obtained using an iterative procedure which is shown schematically in Fig. 8. Each iteration of our procedure is comprised of two steps: i) a global step which, using the equilibrium equations, the elasticity law and the boundary conditions, consists of calculating the displacement at the nodes of the structure and the total strain and stresses, ε and σ, at the Gauss points, assuming the increment of plastic strain Δε^{p} between time increments of t and t + Δt are known at all points, and ii) a local step involving a consideration of the total strain ε (specifically its increment Δε) resulting from the global step at each point, i.e., calculation of the increments of elastic and plastic strains Δε^{e}, Δε^{p} and the stresses σ using the equations of plasticity. Programming the global step is a complex task, as it is independent of the applied plastic model and is totally performed by the commercial FEA package. However, the local step represents finite increments with the integration algorithm required to satisfy the material equations. The stress, elastic and plastic strains and hardening parameters must be calculated, stored and updated.
The first operation of the material behavior calculation is focused on the formulation of the elastic constant matrix, zero temporary variables, and setting the convergence tolerance. Starting from an equilibrium state at time t, the material equations are all satisfied and all quantities (stress, strain, etc.) at the integration point are known from the global FEA analysis. The total strain increment dɛ is decomposed into an elastic dɛ^{e} and plastic part dɛ^{p}:
and the elastic predictor \(\sigma _{\it{ij}}^{\it{pr}}\) is calculated as:
where \(\sigma _{{\it{ij}}}^{\it{0}}\) denotes the stressstate at instant t, and δ is the secondorder identity tensor. μ and λ are the shear modulus and Lamé constant which are defined as the functions of the Young’s modulus E, and Poisson’s ratio ν for isotropic materials as:
The von Mises (equivalent) stress, \(\bar \sigma\), based on purely elastic behavior (elastic predictor), can be calculated as:
where \(S_{\it{ij}}^{\it{pr}}\) is the deviatoric stress and
where σ_{m} = (σ_{xx} + σ_{yy} + σ_{zz})/3 is the hydrostatic stress. Assuming isotropic plastic behavior, the von Mises yield function can be applied as:
where σ_{f} is the flow stress calculated by the constitutive equation in Eq. (2), as:
The yield function (Eq. (30)) indicates that plastic deformation begins when the sum of the squares of the principal components of the deviatoric stress reaches a certain critical value. If the value of the yield function, F(σ), is less than, or equal to, zero, the elastic predictor becomes the new state of stress at time t + Δt; Otherwise, the yield surface is extended and plastic flow occurs. The direction of plastic flow, r, is given by:
The equivalent plastic strain rate can be calculated by:
where \(\dot e_{{\mathrm{ij}}}\) = de_{ij}/dt, dt is the current time step, and de_{ij} is the increment of deviatoric strain defined as:
where dε_{m} = (dε_{xx} + dε_{yy} + dε_{zz})/3. The increment of equivalent plastic strain is approximately obtained by:
where h stands for the hardening at the beginning of the increment which can be calculated by the tangent of the flow stress equation as:
With respect to the implicit algorithm that we use with the Backward Euler integration method, the increment of the equivalent plastic strain at t + Δt is determined by the information at t + Δt whose value is unknown at time t, such that the iteration method needs to be applied to solve the incremental plastic consistency parameter, \(\delta \bar \varepsilon\), until the convergence is satisfied,
where n denotes the nth iteration. The equivalent plastic strain rate used in the iteration loops is obtained from:
At this point, the iteration is completed and the elasticplastic solution is finalized. Then plastic strain, elastic strain, stress components and equivalent plastic strain are then updated as:
The dissipated inelastic energy, W_{d}, is also calculated as,
where V is the geometrical domain occupied by the sample. The increment in plastic specific energy is defined as:
The thermomechanical coupling is treated as local adiabatic heating with the temperature taken as an inner variable; the new temperature of the material T^{new} is thus calculated using the temperature resulting from the plastic work, viz.:
where T^{int} is the initial temperature, ρ_{P} and C_{v} are density and specific heat of the workpiece, respectively.
Data availability
Experimental and numerical data from this study are available, respectively from Dr. Shan Cecilia Cao of the University of California Berkeley (email: shancao@berkeley.edu) and from Dr. Xiaochun Zhang of the Shanghai Institute of Applied Physics (email: zhangxiaochun@sinap.ac.cn) upon request.
References
 1.
Unal, O. & Varol, R. Surface severe plastic deformation of AISI 304 via conventional shot peening, severe shot peening and repeening. Appl. Surf. Sci. 351, 289–295 (2015).
 2.
Majzoobi, G. H., Azizi, R. & Alavi Nia, A. A threedimensional simulation of shot peening process using multiple shot impacts. J. Mater. Process. Technol. 164, 1226–1234 (2005).
 3.
Cao, S. C. et al. Natureinspired hierarchical steels. Sci. Rep. 8, 5088 (2018).
 4.
Cao, S. C. et al. Lightweight isometricphase steels with superior strengthhardnessductility combination. Scr. Mater. 154, 230–235 (2018).
 5.
Wu, G., Chan, K.C., Zhu, L., Sun, L. & Lu, J. Dualphase nanostructuring as a route to highstrength magnesium alloys. Nature 545, 80 (2017).
 6.
Unal, O., Maleki, E. & Varol, R. Effect of severe shot peening and ultralow temperature plasma nitriding on Ti6Al4V alloy. Vacuum 150, 69–78 (2018).
 7.
Tao, N. R. et al. An investigation of surface nanocrystallization mechanism in iron induced by surface mechanical attrition treatment. Acta Mater. 50, 4603 (2002).
 8.
Sun, L., He, X. & Lu, J. Nanotwinned and hierarchical nanotwinned metals: a review of experimental, computational and theoretical efforts. npj Computational. Materials 4, 6 (2018).
 9.
Xu, S., Xiong, L., Chen, Y. & McDowell, D. L. Sequential slip transfer of mixedcharacter dislocations across Σ3 coherent twin boundary in FCC metals: a concurrent atomisticcontinuum study.npj Comput. Mater. 2, 15016 (2016).
 10.
Jung, J. et al. Continuum understanding of twin formation near grain boundaries of FCC metals with low stacking fault energy. npj Computational. Materials 3, 21 (2017).
 11.
Wegst, U. G. K., Bai, H., Saiz, E., Tomsia, A. P. & Ritchie, R. O. Bioinspired structural materials. Nat. Mater. 14, 23–36 (2015).
 12.
Gludovatz, B. et al. A fractureresistant highentropy alloy for cryogenic applications. Science 345, 1153–1158 (2014).
 13.
Shekhawat, A. & Ritchie, R. O. Toughness and strength of nanocrystalline graphene. Nat. Commun. 7, 10546 (2016).
 14.
Lu, K. & Lu, J. Surface nanocrystallization (SNC) of metallic materialspresentation of the concept behind a new approach. J. Mater. Sci. & Technol. 15, 193 (1999).
 15.
Ritchie, R. O. The conflicts between strength and toughness. Nat. Mater. 10, 817–822 (2011).
 16.
Sun, J. et al. Simultaneously improving surface mechanical properties and in vitro biocompatibility of pure titanium via surface mechanical attrition treatment combined with lowtemperature plasma nitriding. Surf. Coat. Technol. 309, 382–389 (2017).
 17.
Guo, H.Y. et al. Nanostructured laminar tungsten alloy with improved ductility by surface mechanical attrition treatment. Sci. Rep. 7, 1351 (2017).
 18.
Huang, L., Lu, J. & Troyon, M. Nanomechanical properties of nanostructured titanium prepared by SMAT. Surf. & Coat. Technol. 201, 208 (2006).
 19.
Lu, A. Q., Liu, G. & Liu, C. M. Microstructural evolution of the surface layer of 316L stainless steel induced by mechanical attrition. Acta Metall. Sin. 40, 943 (2004).
 20.
Wang, K., Tao, N. R., Liu, G., Lu, J. & Lu, K. Plastic straininduced grain refinement at the nanometer scale in copper. Acta Mater. 54, 5281 (2006).
 21.
Pei, Z. et al. Atomic structures of twin boundaries in hexagonal closepacked metallic crystals with particular focus on Mg. npj Comput. Mater. 3, 6 (2017).
 22.
Zhang, H. W., Hei, Z. K., Liu, G., Lu, J. & Lu, K. Formation of nanostructured surface layer on AISI 304 stainless steel by means of surface mechanical attrition treatment. Acta Mater. 51, 1871 (2003).
 23.
Tao, N. R. et al. Mechanical and wear properties of nanostructured surface layer in iron induced by surface mechanical attrition treatment. J. Mater. Sci. & Technol. 19, 563 (2003).
 24.
Shaw, L. L. & Zhu, Y. T. in Materials Processing Handbook Ch. 31 (eds Groza, J. R., Shackelford, J. F., Lavernia, L. J. & Powers M. I.) (CRC Press, Boca Raton, USA, 2007).
 25.
Alobaid, Y. F. Shot peening mechanics  experimental and theoreticalanalysis. Mech. Mater. 19, 251 (1995).
 26.
Guagliano, M. Relating Almen intensity to residual stresses induced by shot peening: A numerical approach. J. Mater. Process. Technol. 110, 277 (2001).
 27.
Hertz, H. On the Contact of Elastic Solids. (Macmillan, Jones & Schott, London, 1896).
 28.
Johnson, G. R. & Cook, W. H. Fracture characteristics of three metals subjected to various strains, strain rates, temperatures and pressures. Eng. Fract. Mech. 21, 31 (1985).
 29.
Nasr, M. N. A., Ng, E. G. & Elbestawi, M. A. Effects of strain hardening and initial yield strength on machininginduced residual stresses. J. Eng. Mater. & Technol. 129, 567 (2007).
 30.
Chen, X. H., Lu, J., Lu, L. & Lu, K. Tensile properties of a nanocrystalline 316L austenitic stainless steel. Scr. Mater. 52, 1039 (2005).
 31.
Dai, K., Villegas, J. & Shaw, L. An analytical model of the surface roughness of an aluminum alloy treated with a surface nanocrystallization and hardening process. Scr. Mater. 52, 259–263 (2005).
 32.
Dai, K., Villegas, J., Stone, Z. & Shaw, L. Finite element modeling of the surface roughness of 5052 Al alloy subjected to a surface severe plastic deformation process. Acta Mater. 52, 5771–5782 (2004).
 33.
Moelans, N., Blanpain, B. & Wollants, P. An introduction to phasefield modeling of microstructure evolution. CalphadComput. Coupling Phase Diagr. & Thermochem. 32, 268 (2008).
 34.
Zimmermann, M., Schulze, V., Baron, H., Löhe, D., A novel 3D finite element simulation model for the prediction of the residual stress state after shot peening, (ed. Tosha, K.) Proceedings of the 10th International Conference on Shot Peening, 63–68 (Academy Common Meiji University, Tokyo Japan, 2008).
 35.
Klemenz, M., Schulze, V., Vöhringer, O. & Löhe, D. Finite element simulation of the residual stress states after shot peening. Mater. Sci. Forum 524, (349–354 (2006).
Acknowledgements
The authors express their sincere gratitude to the National Key R&D Program of China (Project No. 2017YFA0204403), the Major Program of National Natural Science Foundation of China: NSFC 51590892, the Shenzhen Science Technology and Innovation Commission under the Technology Research Grant no. JCYJ20160229165310679. They further thank the Research Center of AREVA NP for financial support, and to Mrs. E. Aublant, Dr. G. Kermouche and Prof. J.M. Bergheau there for helpful discussions. J.L. acknowledges support from the Guangdong Provincial Department of Science and Technology under the Grant no. JSGG20141020103826038 and from the Shenzhen Science Technology and Innovation Commission under the Technology Research Grant no. 2014B050504003. R.O.R was supported by the Mechanical Behavior of Materials Program (K13) at the Lawrence Berkeley National Laboratory, funded by the U.S. Department of Energy, Office of Science, Office of Basic Energy Sciences, Materials Sciences and Engineering Division, under Contract No. DEAC0205CH11231. The authors also wish to thank Profs. C.T. Liu, T. F. Hung and Ming Dao for helpful discussions, and Yifan Zheng, Haokun Yang, Haojie Jiang, Yan Bao, Hui Yang and Fansheng Cheng for technical assistance.
Author information
Affiliations
Contributions
S.C.C., X.Z., SQ.S., J.L. and R.O.R. conceived the research; S.C.C. and X.Z., supervised by R.O.R., performed the numerical simulations and experiments, and analyzed the data with the help from SQ.S., J.L. and R.O.R; S.C.C., X.Z. and R.O.R. wrote the manuscript, which all authors discussed and commented on.
Corresponding authors
Ethics declarations
Competing interests
The authors declare no competing interests.
Additional information
Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Cao, S.C., Zhang, X., Lu, J. et al. Predicting surface deformation during mechanical attrition of metallic alloys. npj Comput Mater 5, 36 (2019). https://doi.org/10.1038/s4152401901716
Received:
Accepted:
Published:
Further reading

Predicting mechanical properties of 316L stainless steel subjected to SMAT: A sequential DEMFEM investigation
International Journal of Mechanical Sciences (2021)

A constitutive model incorporating grain refinement strengthening on metallic alloys
Journal of Materials Science & Technology (2021)

Literature review on the mechanical properties of materials after surface mechanical attrition treatment (SMAT)
Nano Materials Science (2020)

Exploring the strength and ductility improvement of Cu–Al alloys
Materials Science and Engineering: A (2020)

A dislocation density–based comparative study of grain refinement, residual stresses, and surface roughness induced by shot peening and surface mechanical attrition treatment
The International Journal of Advanced Manufacturing Technology (2020)