1 Introduction

Nowadays, titanium alloys have been widely used in medical field due to their low density, low elastic modulus, biological inertia, non-toxic, non-magnetic, and good mechanical properties compared with traditional metals (Ref 1,2,3,4,5,6,7,8). It is known that interfacial friction largely determines the longevity of titanium orthopedic implants. Generally, the wear resistance of original titanium alloys is not up to clinical expectations. Therefore, surface modification is usually adopted as an effective way to improve the wear resistance of titanium implants (Ref 9). Most studies focused on the overall influence of surface modification on tribological properties. Few studies considered the relevance between mechanical properties and wear resistance. The existing researches on the relevance between them remain controversial (Ref 10,11,12,13). Mechanical properties such as strength, toughness, and hardness are changed after surface modification, and the amount of change depends on the specific surface modification method to a great extent. To deeply analyze the influence of surface modification on tribological properties and effectively optimize the surface modification process for obtaining better wear resistance, the relevance between mechanical properties and tribological properties must be thoroughly studied.

It is proved that the reciprocal of a certain mechanical parameter is correlated with tribological properties. For instance, a reciprocal relationship between wear rate and hardness was recorded by Sharma et al. when studying the correlation between hardness and fretting wear resistance of ultrafine-grained Cu-TiB2-Pb composites (Ref 14). They found the results commensurate well with Archard’s relationship of abrasive wear. Su et al. also found the linearly proportional relationship between wear rate to the inverse hardness in nanocrystalline Co-W alloy coatings (Ref 15). Actually, the relevance between mechanical properties and wear rate is not very obvious when all the mechanical parameters are considered separately (Ref 16, 17). However, a good correlation can be achieved between the combination of mechanical parameters and wear resistance (Ref 18, 19). Therefore, attempts have been made to combine more than one mechanical parameter to determine a better relevance (Ref 20). When the influence factors were combined, the relevance and balance mechanism between these mechanical parameters and tribological properties were further explored based on the results of individual relevance analysis (Ref 21). Some studies associated mechanical properties with specific wear rates or friction coefficients. For instance, Harsha et al. investigated the relevance between mechanical properties and tribody abrasive wear behavior. The wear loss was found only related to certain mechanical properties, such as (SE)−1 and (HSE)−1 (Ref 22). Lancaster et al. drew similar conclusions when studying the relevance between mechanical properties and tribological behavior of different metal alloys (Ref 23).

As an effective surface modification method, the ultrasonic surface rolling process (USRP) can generate gradient nanostructure in material surface, accompanied by the increase in hardness and decrease in surface roughness. The significant effect of USRP on improving the wear resistance of titanium alloys has been verified by experiments. Yet, the particular bio-tribological behavior in humanoid body fluid and the relevance of various mechanical parameters on tribological properties have not been studied. In order to better understand the USRP treatment on the tribological behavior of titanium alloy, material transfer, tribofilms, and wear mechanism during friction and wear process as well as the relevance between mechanical properties and specific wear rate (SWR)/friction coefficient (COF) were analyzed in depth in the present work. Relevant results are helpful to understand the performance of USRP-treated titanium alloy as important joint material in near real human environment.

2 Experimental

The commercial TB8 titanium alloy was adopted in this study to stand for orthopedic implant material because of its low elastic modulus which can reduce stress shielding to a large extent, to meet the clinical requirements for implant material. TB8 plates with the chemical composition (in wt.%) of 14.0–14.9 Mo, 2.4–3.2 Nb, 2.5–3.5 Al, 0.15–0.25 Si, 0.0018 Fe, 0.10–0.15 O, 0.012 C and balanced Ti were treated by USRP before mechanical properties tests and friction-wear tests. USRP parameters included a spindle speed of 20 rpm, a feed rate of 0.08 mm/r, an amplitude of 8 μm, and a static load of 200 N, 400 N, and 600 N.

D/Max2500PC x-ray diffractometer with Cu-Kα radiation, measurement angle of 20°—90° and scanning speed of 4°/min was used for analyzing the phases of TB8 after USRP. VEGA3 scanning electron microscope was used for observing the surface topographies, and the composition of surface layer was measured with its included Inca Energy 350 energy spectrometer. ESCALAB 250 x-ray photoelectron spectroscopy with Al-Kα (1486.6 eV) monochromatic light source was used for analyzing the composition of sample surface after friction-wear tests.

All tensile tests were performed on 100 KN MTS 810 material testing system. The geometry and dimension of tensile samples were designed according to the standard of ASTM E8M-09 and the allowance range of experimental conditions. As schematically shown in Fig. 1, the holding length was 15 mm. The stretching speed was 0.6 mm/min.

Fig. 1
figure 1

Schematic diagram of tensile sample (mm)

Full size image

According to ASTM E384-08, microhardness was measured by MHVD-1000 MPa Vickers microhardness tester. Surface roughness (Ra) was measured by OLS3000 optical confocal profilometer. According to ASTM G99-95A, the same tests were repeated three times for each group of samples and the average values were finally adopted. Table 1 lists all the measurements, including the mean values and standard deviations (SD) obtained from five test samples.

Table 1 The mechanical properties of TB8 titanium alloy
Full size table

Friction-wear tests were carried out on MFT-R4000 reciprocating friction and wear tester at 25 °C under humidity of 50 ± 5 RH. Samples were cut into 20 × 20 × 10 mm. ZrO2 ceramic balls with hardness of 1600 HV and surface roughness of 0.1 μm (Ra) were selected as counterparts. The applied load was set as 20 N, and the sliding speed was set as 0.1 m/s with a duration time of 15 min. The friction region was surrounded by 15% hyaluronic acid (HA) solution in the groove of working platform, as shown in Fig. 2. For preparing the compound synovial fluid, HA was weighed and hydrogenated soybean phospholipids (HSPC) was counted accordingly. The correlative ingredients were dissolved in deionized water and magnetically stirred for 180 min with a stirring speed of 500–900 r/min. Then, the stock solution was stirred overnight at a rate of 600 r/min. Temperature was maintained at 37.1 °C during the preparation process to simulate human body environment. The worn loss (V) was measured from the profile of the worn cross section obtained by an optical confocal profilometer (see Fig. 2). Results were averaged from three tests under the same conditions. Specific wear rate and friction coefficient were chosen to represent tribological properties. Specific wear rate (SWR/mm3/N·m) is the ratio of worn loss to sliding distance, which can be calculated through:

$$SWR = \frac{V}{F \cdot L}$$
(1)

where V/mm3 is worn loss, F/N is external load and L/m is sliding distance.

Fig. 2
figure 2

Schematic extraction process of typical image sequence for post-processing analysis

Full size image

The friction coefficient is the ratio of the friction (F) between two surfaces to the vertical force (N) acting on one of them. The corresponding formula is shown as:

$$\mu = \frac{F}{N}$$
(2)

Linear regression was used to evaluate the relevance between mechanical properties and tribological properties of USRP-treated TB8 titanium alloy by finding the determining coefficient R2, as shown in the following formula:

$$R^{2} = \left( {\frac{{n\sum {xy} - \left( {\sum x } \right)\left( {\sum y } \right)}}{{\sqrt {n\left( {\sum {x^{2} } } \right) - \left( {\sum x } \right)^{2} } \sqrt {n\left( {\sum {y^{2} } } \right) - \left( {\sum y } \right)^{2} } }}} \right)^{2}$$
(3)

where x and y are variables, and n represents the number of values or elements. R2 gives the proportion of change for one variable. That is the proportion of data closest to the best fit line.

3 Results and Discussion

3.1 Surface Characteristics After USRP

The microstructure of TB8 matrix mainly contains β phase and a small amount of α’ phase, as shown in Fig. 2. For a better understanding of the effects of USRP on the bio-tribological properties of TB8, the morphology and element distribution of sample surfaces treated by different USRP loads were investigated and are shown in Fig. 3. The entire surfaces were observed and specific regions were intercepted for element analysis. Metal debris formed by machining can be seen on the surface of the original sample. As the USRP load increases, surface defects diminish gradually. For the original sample and 200 N USRP sample, the selected detection regions contain visible surface defects like the oxide of titanium, indicating the relatively heterogeneous distribution of Ti, O. Yet, this kind of unevenness disappears obviously with the increase in USRP load. Clearly, the smoothest surface is found in 600 N USRP sample. EDS results further verify the fact that USRP treatment makes elements more evenly distributed.

Fig. 3
figure 3

The surface morphology and corresponding EDS image of (a) original sample, (b) 200 N USRP sample, (c) 400 N USRP sample, and (d) 600 N USRP sample

Full size image

The XRD patterns of samples prepared with different USRP loads are shown in Fig. 4. After 200 N USRP, most diffraction peaks recede, indicating that grains in the surface layer have been refined. The diffraction peak of (102) is further weakened by continuously increasing the USRP load, but the diffraction peak of (002) has an obvious orientation growth effect. This is because the impact energy transferred to surface grains is aggrandized with the increase in USRP load, which raises the free energy of atoms in the β phase. As a result, the energy of atoms deposited on the substrate surface gets higher, and the surface diffusion migration activation energy also increases, making the crystallization process easier. Therefore, it can be concluded that a rising USRP load contributes to the formation of nanolayers, intensifies the phase transition process, and enhances the optimal growth ability of nanolayers.

Fig. 4
figure 4

XRD patterns of sample surfaces prepared with different USRP loads

Full size image

The grain size was calculated according to the Debye–Scherrer formula. As can be seen from Table 2, grains in the outermost surface have been refined to nanoscale after USRP and become smaller with the increase in USRP load.

Table 2 Grain size for different USRP samples (nm)
Full size table

3.2 Friction and Wear Properties After USRP

The worn cross-sectional profiles of samples treated with different USRP loads are comparatively shown in Fig. 5. Generally, all profiles show the same pattern, but the profile of the wear track becomes shallower and narrower after USRP, indicating that better wear resistance can be obtained by introducing USRP treatment on TB8 surface. Meanwhile, the width and depth of the wear track gradually decrease with the increase in USRP load, and this trend seems to be more obvious as USRP load increases. Also, the fluctuation gradually decreases due to the hard and smooth USRP-treated surface.

Fig. 5
figure 5

Comparison of cross-sectional wear scars

Full size image

Table 3 lists the measured and calculated friction and wear data of TB8. As USRP load increases, the worn loss, specific wear rate, and friction coefficient all decrease, which is consistent with the varying trend of observed wear track. The reduction slightly shrinks when a comparison is taken between different USRP samples. The variation of COF with USRP load is more even.

Table 3 Friction and wear properties of TB8 titanium alloy
Full size table

The worn surface morphologies of samples after friction and wear test in 15% HA solution are shown in Fig. 6. Furrows and severe material spalling can be observed in the original sample surface, reflecting the severity of wear. After USRP, the wear state improved significantly. For the original sample and 200 N USRP sample, the wear mechanism is characterized by adhesive wear and abrasive wear. For 400 N USRP sample, the main wear mechanisms are adhesive wear and oxidation wear. Although furrows can still be found on the sample surface, the wear is relatively light due to the shallow abrasion marks. For 600 N USRP sample, the wear mechanism is mainly oxidation wear. Some minor furrows took away part of the substrate and caused a small amount of wear.

Fig. 6
figure 6

Worn surface morphologies of TB8 titanium alloy in 15% HA solution for (a) original sample, (b) 200 N USRP sample, (c) 400 N USRP sample, and (d) 600 N USRP sample

Full size image

Figure 7 shows the XPS measurements of TB8 alloy treated with different USRP loads after friction and wear test in 15% HA solution. As shown in Fig. 7(a), the binding energy of the signal emitted by Na 1 s, O 1 s, Ti 2p1/2, and Ti 2p2/3 electrons was 1070 eV, 531 eV, 464 eV, and 458 eV, respectively.

Fig. 7
figure 7

(a) The full XPS of TB8 after friction and wear test in 15% HA solution, (b) Na 1 s peaks, (c) O 1 s peaks, (d) Ti 2p peaks, and (e) C 1 s peaks

Full size image

The high-resolution narrow scan spectra in Fig. 7(b) show that Na 1 s peak maintains at 1070.6 eV. Its intensity gradually decreases with the increase in USRP load, indicating the fact that Na mostly exists on the surface. This is because Na comes from 15% HA solution and deposits on the worn surface after friction and wear test. Since wear gets lighter as USRP load increase, less Na remains on flatter worn surface. The high-intensity peak at 284.101 eV in Fig. 7(e) is mainly related to the C=C chemical bond, while the low-intensity peak with higher binding energy may be the characteristic of TiCxOy. Ti 2p spectrum has a complex pattern, as shown in Fig. 7(d). In the initial stage of USRP, the curve fluctuates gently without protruding peak. As USRP loads up to 400 N, a relatively stable state is obtained and two obvious peaks emerge. Take 600 N USRP sample for detailed analysis. The Ti 2p peak can be deconvolved into Ti-C bond and Ti-O-C bond, indicating that titanium oxide is partially transformed into more compact titanium carbide (TiC) and titanium carboxide (TiCxOy). TiC can be obtained at 457.393 eV of Ti 2p2/3, and TiCxOy can be obtained at 463.091 eV of Ti 2p2/3. As for Fig. 7(c), one main peak of O 1 s at 531.582 eV is related to the chemical bond of O and Ti. The other peak at 528.824 eV belongs to TiCxOy. Hence, the friction surface of TB8 in 15% HA solution is constituted by TiC, TiO2, and TiCxOy.

More Ti 2p was found in the spectrum diagram of the original sample, indicating that more metallic matrix is exposed under adhesion wear compared with USRP sample. This is also confirmed by the spalling pit observed in Fig. 6(a). Meanwhile, the Ti–O bond occupies a large proportion in Fig. 7(c) and (d), which means the worn surface is almost covered by TiO2. This also verifies the occurrence of oxidation wear.

A compound conversion occurs when USRP load increases from 200 to 400 N. As shown in Fig. 7(d), in addition to the presence of some expected elements such as Ti and O, C was also distinctly detected, revealing the occurrence of material transfer in friction interface due to severe oxidation. The C element was derived from 15% HA solution. Besides, it can be seen from the fine spectrum of C 1 s that except for the C=C double bond, two bonds of C element (O-C and Ti-C) were also found after peak-differentiating and imitating, whose orbital binding energy located near 285.651 and 287.31 eV, respectively, corresponding to TiC and TiCxOy.

Both the chemical composition analysis and the worn surface morphology demonstrate that a smoother and denser oxide layer composed of TiO2, TiC, and TiCxOy is formed on the friction interface of 600 N USRP sample, which protects TB8 substrate from further oxidation. Hence, the 600 N USRP sample presents the best wear resistance in 15% HA solution.

3.3 Relevance between Specific Wear Rate and Mechanical Properties

To investigate the correlation between mechanical properties and wear resistance of USRP-treated TB8 titanium alloy, statistical data based on mechanical parameters (tensile strength, elongation, hardness, surface roughness) and specific wear rate were extracted and plotted using Pearson’s coefficient and coefficient of determination at 95% confidence intervals. The statistical analysis of the correlation between single mechanical parameter and specific wear rate at steady state can be found in Fig. 8(a).

Fig. 8
figure 8

Statistical analysis between single mechanical parameter of USRP-treated TB8 titanium alloy and specific wear rate: (a) first-order positive correlation and (b) first-order negative correlation

Full size image

As observed, hardness plays an absolutely important role in wear process among these parameters. Tensile strength, elongation, and surface roughness have a slightly weaker effect on SWR. A negative correlation between hardness and SWR can be seen, which is in agreement with the Acadian law. Therefore, the correlation between the reciprocal of single mechanical parameter and SWR was further investigated. The statistical analysis data are shown in Fig. 8(b).

After taking the inverse, hardness still shows the highest correlation with SWR. The coefficients of determination for tensile strength and elongation are somewhat higher, and the influence of surface roughness comes to the fore. When relative sliding occurs, interfacial micro-convex presses into the opposite surface and plows on it, leading to local plastic deformation until material detaches from surface. Higher hardness leads to higher resistance of the material to impurity compression, which proves the absolutely important role of hardness in retarding wear process. Plastic properties, including tensile strength and elongation, are also important because they determine the plastic deformation capacity of material. Higher tensile strength and elongation delay the fracture of interfacial material and its removal from substrate. Surface roughness determines the number and morphology of micro-convex, which also affects the wear process, especially in earlier stage. As described in Table 1, tensile strength, elongation, and hardness all increase significantly with the increase in USRP load. Meanwhile, surface roughness was greatly reduced after USRP. Thus, the correlation of these parameters with the SWR of USRP sample is outstanding.

The mechanical parameters themselves have a certain correlation. For example, there is generally an approximately proportional relationship between indentation hardness and tensile strength. From this point of view, the influence of mechanical parameters on wear resistance is not independent. For USRP treatment, tensile strength, elongation, and hardness are simultaneously increased, and surface roughness is reduced in the meantime. Clearly, there is a close relationship between mechanical parameters when material is treated by USRP. In order to get insight into the influence and balance mechanism of these mechanical parameters on tribological properties, the relationship between combined mechanical parameters and SWR was investigated. For simplicity, the combined mechanical parameters are expressed as the reciprocal of one parameter multiplied by another. For example, (SH)−1 is the reciprocal of tensile strength multiplied by hardness. Figure 9(a) shows the statistical analysis results of two combined mechanical parameters versus specific wear rate at steady state.

Fig. 9
figure 9

Statistical analysis between combined mechanical parameters of USRP-treated TB8 titanium alloy and specific wear rate: (a) second-order negative correlation and (b) third-order negative correlation

Full size image

A strong correlation is found between all the two combined mechanical parameters and specific wear rate, demonstrating the interdependence of mechanical parameters for USRP-treated metals. As can be seen from Fig. 9(a), the largest coefficient of determination appears in the (SH)−1 case, followed closely by the (SR)−1, (EH)−1, and (HR)−1. Since there is an inherent relationship between hardness and tensile strength, the combination of these two mechanical parameters shows the strongest correlation with SWR. Due to the characteristics of USRP, other combinations also show strong correlations. SWR increases with the decrease in (SR)−1 and (HR)−1. Clearly, in both cases, surface roughness plays a dominant role. This is because the data points in Fig. 9(a) correspond to samples treated with different USRP loads. For lower loading, the tensile strength and hardness are relatively small, while the surface roughness is relatively high. Thus, the limited improvement in wear resistance caused by tensile strength and hardness will be partially offset by large surface roughness, which further deteriorates the wear resistance. In addition, the coefficients of determination for combinations involving hardness are all high, indicating the importance of hardness in the two-factor analysis.

The statistical analysis of three combined mechanical parameters vs. SWR is shown in Fig. 9(b). A pretty evident relevance is obtained when the combination of tensile strength, elongation, and hardness is considered. This once again confirms the fact that plastic mechanical properties and hardness supplement each other, together determining the wear resistance. (SHR)−1 also shows a good relevance with SWR, and the linear fitting curve trend changes when surface roughness is involved in. For (EHR)−1 case, certain relevance can still be found, where hardness and surface roughness plays a leading role. But the determining coefficient is lower with the participation of elongation, compared with the case of just hardness and surface roughness. For (SER)−1 case, very little dependence is obtained.

Based on the above analysis, it can be concluded that hardness plays a major role in controlling the wear resistance of TB8 titanium alloy. Tensile strength shows better relevance when combined with hardness or surface roughness. Elongation is significantly correlated only when combined with hardness. Yet, elongation reveals its irrelevant characteristics when combined with more factors. It is worth noting that though plays a crucial role, hardness is not the only factor (Ref 24).

3.4 Relevance Between Friction Coefficient and Mechanical Properties

The correlation between single mechanical parameter and COF at steady state was analyzed. When single factor is considered, tensile strength and surface roughness are found to be closely correlated with COF. Elongation also shows a good correlation. However, hardness does not present particularly strong correlation with COF. This is because friction coefficient mainly reflects the ratio of the frictional force between two surfaces to the vertical force acting on one of them, which is influenced by the material of contact surface, the lubricant, the surface roughness, and the relative sliding speed. In the present work, the effect of surface roughness is more obvious under the same experimental conditions. As reported in the literature, COF increases with the increasing surface roughness (Ref 25). A linear fit of COF versus surface roughness is shown in Fig. 10.

Fig. 10
figure 10

Relevance between the reciprocal of surface roughness and COF

Full size image

When considering two factors, all the combinations show good relevance except (ER)−1. High relevance can be found in the case of (SE)−1 and (HR)−1. Corresponding linear fit charts are shown in Fig. 11. The apparent relevance between combined mechanical properties and tribological properties can be understood from the intrinsic relevance of these mechanical properties. According to the Support Vector Regression (SVR) algorithm, to achieve a high tensile strength, strain rate must be less than 0.01 s−1 (Ref 26). Lower strain is also conducive to achieving high elongation. It is thus clear that strain rate is the key influential factor both for tensile strength and elongation. Since plastic deformation controls the whole friction and wear process, strain rate thus has a great correlation with tribological properties, which leads to the apparent relevance between (SE)−1 and COF. As to the combination of hardness and surface roughness, it is known that surface roughness has a certain influence on the measurement of hardness. When they are considered together, the role of surface roughness becomes even more prominent. Meanwhile, with the decrease in surface roughness, the inhibition effect of lubrication on adhesion friction is more significant. This is why the combinations with high relevance all involve surface roughness. It should be pointed out that the relevance between the two-factor combinations and COF is not so high and close to each other compared with SWR.

Fig. 11
figure 11

Relevance between combined mechanical properties and COF

Full size image

3.5 Correlation Prediction Model

Due to the lack of understanding on the functional dependence between mechanical properties and wear resistance, traditional regression models assume a single linear dependence between output and input. However, the variables of mechanical properties are highly interdependent and intricate, creating a tendency for dispersion when multiple mechanical parameters are intertwined. Such trends make it difficult to build a straightforward correlation model using traditional regression analysis techniques. To overcome these limitations, a special correlation model using artificial neural network (BP) was first developed in this work, as shown in Fig. 12. It can handle the highly complex relationship between mechanical properties and wear resistance, and thus easily and accurately predict the wear resistance of material by testing its mechanical properties. The model, exemplified by the correlation data of TB8 titanium alloy, can be applied to other metals and alloys to explain and predict the observed wear resistance variations and largely reduce friction and wear test time.

Fig. 12
figure 12

Diagram of the neural network structure for correlation model developed in this study

Full size image

The current model draws on the experience of Balasundar et al. when they used a neural network to establish the relationship between microstructure and tensile properties as well as creep properties (Ref 27, 28). They adopted the Levenberg–Marquardt algorithm and Bayesian regularization to train the neural network. For the sake of adapting to the relationship between multiple mechanical properties and wear resistance, two hidden layers were feed forward installed in a 4–3–2 architecture, namely 4 neurons in the input layer, 3 neurons in the hidden layer, and 2 neurons in the output layer. Considering the complexity of mixed mechanical properties, only the mean value of each mechanical parameter was used in correlation analysis. The optimized neural network established here was then used to conduct the simulation experiment to study the influence of different mechanical properties on wear resistance by changing a specific mechanical parameter within the experimental range while keeping other characteristics at a constant value (mean value). In order to further understand the simulation experiment, the following formula was used to normalize the trend graph of simulation experiment between 0 and 1:

$$x_{N} = \frac{{x - x_{\min } }}{{x_{\max } - x_{\min } }}$$
(4)

where xN is the normalized value of a parameter and x is the predicted value of this parameter. xmin and xmax are the minimum and maximum values of parameters concerned in the database.

The simulation experiments also verify that all the mechanical parameters have a relatively uniform trend in bionic body fluid for USRP-treated TB8 titanium alloy. The characteristic of increasing wear resistance with increasing hardness predicted by the BP neural network, especially when other mechanical properties are considered in combination with hardness can also show a good correlation, but the little correlation with elongation is confirmed by the theory reported in the literature. Meanwhile, the wear resistance predicted by the artificial neural network is in good agreement with the experimental observation.

4 Conclusions

The bio-tribological properties of TB8 titanium alloy treated by USRP were studied in this work. Surface characteristics were investigated to help analyze the effect of USRP on wear resistance. The relevance between mechanical properties and tribological properties was also studied to make a deep analysis. The conclusions are as follows:

  1. 1.

    The highest tensile strength, hardness, and the smallest surface roughness are obtained under the 600 N ultrasonic surface rolling process (the highest load within the selected parameter range).

  2. 2.

    The best wear resistance is obtained under 600 N USRP. The smoothest and densest oxide layers composed of TiO2, TiC, and TiCxOy are formed on friction interfaces to protect the TB8 substrate from further oxidation.

  3. 3.

    For SWR, when a single factor is considered, hardness plays a vital role during the wear process. When combined factors are considered, strong relevance is found in all the two-factor cases. Also, a pretty evident relevance is obtained in (SEH)−1 case as well as (SHR)−1 case.

  4. 4.

    For COF, when considering a single factor, tensile strength and surface roughness are closely related. Elongation also shows a good correlation. When considering two-factor cases, all the combinations show good relevance except (ER)−1. Groups with the best relevance are (SE)−1 and (HR)−1.

  5. 5.

    A particular correlation model using an artificial neural network is first established to handle the highly complex relationships between mechanical properties and wear resistance. The wear resistance predicted by the artificial neural network is in good agreement with the experimental observation.