This study investigated the effect of heat treatment processes on the dry sliding wear resistance of the TC21 Ti-alloy at several levels of normal load and sliding speed. Response Surface Methodology (RSM) has been used as a design of the experiment procedure. OM and FESEM besides XRD analysis were used for results justification. Highest hardness of 49 HRC was recorded for WQ + Aging specimens due to the plenty of α″ which decomposed to αs and the more αs, while the lowest hardness of 36 HRC was reported for WQ specimens. The results revealed that specimens subjected to water quenching and aging (WQ + Aging) under extreme load and speed conditions (50 N and 3 m/s), possessed the poorest wear resistance although they had the highest hardness. While those left in the annealed condition revealed the highest wear resistance although they had much lower hardness when compared to other conditions. A mathematical polynomial model for wear resistance expressed in wear rate was developed, validated then used to get the optimum parameters.
Several engineering applications require engineers to get materials with high strength, stiffness, fracture toughness, and extreme service temperatures with low weight[
The wear behavior of TC21 has been investigated from both sliding wear and fretting wear perspectives. Elshear et al.[
In order to get valid, reliable conclusions along with keeping costs and time of experimental runs as minimum as possible Design of Experiments (DOE) is used extensively in tribological field as wear test is classified as a destructive test. One of the most used designs in either industry or research work is Taguchi designs or Taguchi orthogonal arrays, which can be used in both the process design and product stage to enhance manufacturability and reliability of the product[
Several researchers utilized RSM as DOE technique to investigate wear behavior of Ti alloys. El-Tayeb et al.[
Literature reporting on using traditional heat treatment processes to control dry sliding wear behavior of the newly developed TC21 alloy is limited. This could be attributed to two reasons, the first one, Ti64 alloy is still the preferred Ti-alloy. The second reason, most of research related to wear behavior of TC21 alloy focuses on surface modification techniques although they have a lot of disadvantages. Those disadvantages include large expenses, complicated procedures, high energy consumption, and environmental hazards[
The alloy under investigation is TC21 Ti-alloy supplied by Baoji Hanz Material Technology Co., Ltd., China with chemical composition shown in Table 1. With a diameter of 7 mm and a length of 140 mm, the alloy used in this study was annealed in rod form. β transus temperature of this alloy is 950 ± 5 °C[
Table 1 Average chemical composition of studied TC21 Ti-alloy, wt. %.
Al Mo Zr Sn Nb Cr Si Fe Ti 6.09 2.94 2.12 2.06 1.96 1.46 0.09 0.06 Balance
There are 4 different heat treatment cycles used in this work, see Fig. 1. Table 2 summarizes the details of the heat treatment cycles. An electric programmable furnace (Muffle furnace/model HTC03/1) with a controlled atmosphere was used for all heat treatment cycles. To get specimens suitable for next different tests, the TC21 rods were cut into small specimens of 7 mm dimeter and 12 mm length by means of electrical discharge machining (EDM) wire cut machine (NOVICUT 350M model 2015). These small specimens were ground up to 1000 grit. For metallography examination purposes, specimen from every group were selected and embedded in cold mounting resin, ground, polished, and finally etched using 3% HF, 30% HNO
Graph: Figure 1 Cycles of different heat treatments.
Table 2 Heat treatment processes details.
Cycle Heat treatment cycles 1 1000 °C/15 min + AC, for short (AC) 2 1000 °C/15 min + WQ, for short (WQ) 3 1000 °C/15 min + AC + Aging (575 °C/4 h), for short (AC + Aging) 4 1000 °C/15 min + WQ + (Aging 575 °C/4 h), for short (WQ + Aging)
Rockwell hardness (scale C) test was carried out using Rockwell hardness tester (United True-Blue II model U-2004) according to ASTM E18 standards. Seven readings have been recorded for each specimen. Dry sliding wear test for 15 min at ambient temperature was carried out using pin on disk tester for selected specimens based on a design of experiments procedure. Wear specimens (Φ7 and 12 length) were fixed against high-speed steel (HSS) disk with hardness of 64 HRC. Before each individual run, the disk was ground with emery paper of 1000 grit, both the disk and specimen were cleaned with acetone, and then an air blower was used for drying and blowing any contaminations. To get the mass loss due to wear, an electronic balance with a resolution of 0.0001 g was used to weigh the specimen before and after the test. The wear resistance expressed by the wear rate (WR) is given by:
Graph
where, Δm: mass loss in grams (g), t: time in minutes (min).
The test was repeated three times at the same levels of normal load and sliding speed then the average was determined and recorded. At the beginning of each run, each specimen was left for period of time until the surface is fully upset to the desk surface to get a uniform wear rate and avoid the effect of run-in period.
To identify and assess the wear mechanisms, Field emission scanning electron microscopy (FESEM) of worn-out surfaces was performed for some specimens under conditions of (10 N; 1.5 m/s) and (50 N; 3 m/s) which represent low and severe wear conditions, respectively. Also, some collected debris was optically examined.
The output response that is interesting in this investigation is the wear resistance of the TC21, expressed in wear rate (WR). RSM is used to model the WR as a function of input parameters. According to Refs.[
The low and high levels of the input factors had been assigned based on the literature survey, considering the technical capabilities of the wear testing machine available. Table 3 illustrates the levels of the input factors. In this study, the face-centered central composite design (CCD), Fig. 2, was used to construct the design matrix. The face-centered CCD consists of a total of 11 points, detailed as 4 factorial, 4 axial, and 3 center points. These 11 points were used for each level of the categorical factor (heat treatment). So, we get 55 runs in total in the design matrix (Table 4). The Design Expert 13 software was used for purposes of DOE and subsequent statistical analysis.
Table 3 Input factors and levels.
Factor Type Code Level – Low (− 1) Medium (0) High (+ 1) – Normal load (N) Numeric A – 10 30 50 – Sliding speed (m/s) Numeric B – 1.5 2.25 3 – Heat treatment Categoric C Annealed AC WQ AC + Aging WQ + Aging
Graph: Figure 2 Face-centered central composite design.
Table 4 Design matrix including input factors and corresponding responses.
Run Factor 1 A: Load (N) Factor 2 B: Speed (m/s) Factor 3 C: Heat treatment Response WR (g/min) Run Factor 1 A: Load (N) Factor 2 B: Speed (m/s) Factor 3 C: Heat treatment Response WR (g/min) 1 10 1.5 Annealed 1.89 29 30 1.5 WQ 3.17 2 50 1.5 8.05 30 30 3 6.27 3 10 3 2.62 31 30 2.25 4.63 4 50 3 10.92 32 30 2.25 4.81 5 10 2.25 2.31 33 30 2.25 4.49 6 50 2.25 10.27 34 10 1.5 AC + Aging 2.07 7 30 1.5 3.54 35 50 1.5 8.77 8 30 3 5.46 36 10 3 2.5 9 30 2.25 4.81 37 50 3 29.72 10 30 2.25 5.33 38 10 2.25 2.32 11 30 2.25 5.22 39 50 2.25 17.82 12 10 1.5 AC 2.08 40 30 1.5 3.43 13 50 1.5 7.76 41 30 3 6.42 14 10 3 2.16 42 30 2.25 5.7 15 50 3 37.26 43 30 2.25 5.81 16 10 2.25 2.14 44 30 2.25 5.98 17 50 2.25 13.77 45 10 1.5 WQ + Aging 1.91 18 30 1.5 3.87 46 50 1.5 9.86 19 30 3 5.72 47 10 3 2.31 20 30 2.25 5.11 48 50 3 58.43 21 30 2.25 5.09 49 10 2.25 2.47 22 30 2.25 5.39 50 50 2.25 42.68 23 10 1.5 WQ 1.77 51 30 1.5 3.67 24 50 1.5 6.79 52 30 3 5.37 25 10 3 2.47 53 30 2.25 6.41 26 50 3 23.25 54 30 2.25 6.02 27 10 2.25 2.03 55 30 2.25 6.34 28 50 2.25 10.28
Figure 3 shows the microstructure of annealed and different heat treatment conditions. The microstructure of the annealed consists of equiaxed α-phase that is uniformly distributed within a matrix of β-phase (Fig. 3a). According to phase volume fraction analysis based on image processing, α-phase which is soft phase[
Graph: Figure 3 OM images of microstructure: (a) Annealed, (b) AC, (c) WQ, (d) AC + Aging and (e) WQ + Aging.
The different heat treatment processes resulted in a variety of microstructures. This induced a remarkable variation in the hardness of the treated specimens as illustrated in Fig. 4. Annealed specimens showed a hardness value of 38 HRC. WQ specimens revealed the lowest hardness value of 36 HRC. While WQ + Aging specimens obtained the high hardness of 49 HRC. This reflects about 36% increase in hardness as compared to WQ and WQ + Aging specimens. Therefore, specimens after WQ + Aging had the highest hardness due to the plenty of α″ which decomposed to α
Graph: Figure 4 Hardness of TC21-Ti alloy at different conditions.
Figure 5 illustrates the wear rate for all treatment conditions of the TC21 Ti-alloy under all tested levels of both the normal load and sliding speed. One can conclude that the effect of sliding speed is limited for all treatment conditions under low and medium normal loads of 10 and 30 N, respectively. While under same speeds the effect of normal load was significant. For WQ + Aging specimens under 10 and 30 N, WR was increased then decreased when the speed increased from 1.5 to 2.25 m/s and then increased from 2.25 to 3 m/s, respectively. This may be attributed to the occurrence of adhesive wear which decreases by increasing the sliding speed which in turn reduces the time and opportunity of material diffusion between the two friction mates especially when no lubricant film is used. On the other hand, under severe normal load of 50 N, all treatment conditions showed a dramatic increase in WR when speed was increased from 1.5 to 3 m/s. This increase was as minimum as possible for the Annealed condition and as maximum as possible for WQ + Aging treatment.
Graph: Figure 5 Wear rate for different conditions.
Although the hardness of the annealed is much lower than that of WQ + Aging, the wear resistance of annealed samples is higher than that of WQ + Aging under the same combination of high load and speed. This seems to be intuitive specially when compared to competitive materials such as steels, but the microstructure variation resulting from different heat treatments and the frictional thermal effect occurring upon these extreme conditions of testing play an important role in this unfamiliar behavior.
Figure 6 reveals an inverse relationship between the surface hardness and wear resistance expressed by WR, where wear rate increase (wear resistance decrease) is associated with hardness increase. By comparison of wear debris collected during testing of the annealed and WQ + Aging samples, it was noticed that the size of debris from WQ + Aging was much larger than that from annealed specimens as shown in Fig. 7. This suggests that the TC21 Ti-alloy undergoes a change in wear behavior from plastic deformation in the annealed condition to more brittle fracture of surfaces in WQ + Aging condition. This suggestion is supported by FESEM results of WQ + Aging worn surface which revealed existing of smoothed compacted layers which are generally damaged in a brittle manner[
Graph: Figure 6 Correlation between hardness and Wear rate at extreme conditions of load and speed.
Graph: Figure 7 Wear debris of WQ + Aging and annealed specimens at 50 N and 3 m/s.
Graph: Figure 8 FESEM of worn surfaces under 50 N normal load and 3 m/s sliding speed for (a) Annealed, (b) AC, (c) WQ, (d) AC + Aging, and (e) WQ + Aging.
Furthermore, as the normal load increases, the real area of contact between the two friction mates increases leading to an increase in temperature due to high friction force which is the frictional thermal effect. As a result of low thermal conductivity and high chemical affinity of the titanium especially at high temperatures, a chemical reaction with the ambient oxygen occurred and titanium oxides formed as revealed by the XRD spectrum analysis of wear debris of the annealed specimens as shown in Fig. 9. The presence of the titanium oxide is thought to specimens some protection for the tribolayer of the annealed samples and hence they had a better wear resistance under the extreme conditions of load and speed. On the other hand, the absence of oxides in WQ + aging, Fig. 10 is attributed to a very high rate of removing the tribolayer, and hence no chance for a chemical reaction occurs.
Graph: Figure 9 XRD spectrum of the annealed specimen's debris.
Graph: Figure 10 XRD spectrum of the WQ + Aging specimen's debris.
The morphologies of some worn surfaces obtained under several conditions of load and speed for all different conditions are shown in Figs. 8 and 11. Under low load of 10 N and low speed of 1.5 m/s, the worn surfaces, Fig. 11, showed ploughing remarks caused by debris or asperities on the counter face of the HSS disk with excessive plastic deformation, especially for the annealed condition which also showed some small adhesion marks which may be attributed to its low hardness. Therefore, under those low conditions, the predominant wear mechanism is abrasive wear mechanism. When the testing conditions reach extreme levels i.e., 50 N and 3 m/s, a severe rapture can be observed as a result of delamination and spalling occurring, Fig. 8, due to brittle fracture, especially in WQ + Aging specimen because of its high hardness, Fig. 8d.
Graph: Figure 11 FESEM of worn surfaces under 10 N normal load and 1.5 m/s sliding speed for (a) Annealed, (b) AC, (c) WQ, (d) AC + Aging, and (e) WQ + Aging.
The following chart, Fig. 12, illustrates and summarizes the sequence of statistical analysis used in this study. It involves the analysis of the response variations obtained from the experimental work by a well-established statistical method known as the Analysis of Variance (ANOVA). This was in addition to utilizing the response transformation (Box–Cox power transformation). This transformation is an efficient way to develop an equation for a mathematical model that could have a good fit for the experimental data.
Graph: Figure 12 Statistical analysis sequence.
Table 5 shows the ANOVA results of the final improved model based on a 95% confidence level. The results show that the reduced quartic model after transformation is significant (p = 0.0001) with a model F-value of 433.12, which means there is only a 0.01% chance that an F-value this large could occur due to noise. The F-value for lack of fit is 2.05, indicating that it is insignificant (p = 0.1327) when compared to the pure error. Furthermore, all terms with p less than 0.05 are statistically significant. It is obvious that the normal load has been identified as the most significant input factor, followed by the sliding speed and the type of heat treatment. Also, the interaction effect between the load and the speed has been identified as the most significant interaction.
Table 5 ANOVA results for reduced quartic model after inverse Sqrt transform.
Source Sum of squares df Mean square F-value p-value Model 1.43 32 0.0447 433.12 < 0.0001 Significant A-Load 1.25 1 1.25 12,123.08 < 0.0001 B-Speed 0.0987 1 0.0987 955.93 < 0.0001 C-Heat treatment 0.0202 4 0.0050 48.87 < 0.0001 AB 0.0079 1 0.0079 76.34 < 0.0001 AC 0.0125 4 0.0031 30.27 < 0.0001 BC 0.0035 4 0.0009 8.51 0.0003 A2 0.0034 1 0.0034 33.36 < 0.0001 B2 0.0100 1 0.0100 96.42 < 0.0001 ABC 0.0081 4 0.0020 19.50 < 0.0001 A2B 0.0000 1 0.0000 0.4735 0.4986 A2C 0.0031 4 0.0008 7.61 0.0005 AB2 0.0000 1 0.0000 0.1371 0.7147 B2C 0.0071 4 0.0018 17.08 < 0.0001 A2B2 0.0021 1 0.0021 20.39 0.0002 Residual 0.0023 22 0.0001 Lack of fit 0.0016 12 0.0001 2.05 0.1327 not significant Pure error 0.0007 10 0.0001 Cor total 1.43 54
The fit statistics, Table 6, show that the coefficient of determination R-squared (R
2
Graph
3
Graph
4
Graph
5
Graph
6
Graph
where, L = load in (N) and S = sliding speed (m/s).
Table 6 Fit statistics for reduced quartic model after inverse Sqrt transform.
Std. dev. 0.0102 R2 0.9984 Mean 0.4616 Adjusted R2 0.9961 C.V. % 2.20 Predicted R2 0.9813 Adeq precision 78.5628
Graph: Figure 13 Internally residuals for the final regression model.
To illustrate the combined effect of independent parameters on the response (WR), 3D response surface plots and 2D contour plots are constructed for all heat treatment conditions, as shown in Figs. 14 and 15, respectively. According to those plots, the wear rate increases with the increase in normal load and sliding speed, especially at high levels. In addition, this increase in wear rate is most dramatic in the WQ + Aging condition, Fig. 14e, while it is too small in the annealed condition, Fig. 14a.
Graph: Figure 14 3D response surface for WR of (a) annealed, (b) AC, (c) WQ, (d) AC + Aging, and (e) WQ + Aging.
Graph: Figure 15 2D contour plots for WR of (a) annealed, (b) AC, (c) WQ, (d) AC + Aging, and (e) WQ + Aging.
To validate the obtained regression model, confirmation tests were carried out. The input parameters chosen within the design space constraints. Table 7 summarizes the input parameters levels applied, the corresponding experimental WR, and predicted WR. From the results, the model has a good prediction ability with average absolute error equal to 3.91%. In addition, all predicted values are within the 95% prediction interval (PI) limits of the model.
Table 7 Comparison between experimental and predicted wear rate.
No. Load (N) Speed (m/s) HT Expt. WR (g/min) × 10–3 Predict. WR (g/min) × 10–3 95% PI low 95% PI High % of abs. error 1 20 2.5 Annealed 3.47 3.62 3.31 3.97 4.32 2 30 2.75 Annealed 5.54 5.45 4.88 6.11 1.62 3 40 2 Annealed 6.79 6.86 6.08 7.78 1.03 4 20 2.5 AC 3.23 3.35 3.07 3.65 3.72 5 30 2.75 AC 5.42 5.81 5.18 6.53 7.2 6 40 2 AC 7.11 7.36 6.49 8.39 3.52 7 20 2.5 WQ 3.49 3.28 3.02 3.59 6.02 8 30 2.75 WQ 5.53 5.69 5.07 6.39 2.89 9 40 2 WQ 5.97 6.11 5.45 6.88 2.35 10 20 2.5 AC + aging 3.68 3.79 3.46 4.16 2.99 11 30 2.75 AC + aging 5.94 6.39 5.66 7.23 7.58 12 40 2 AC + aging 7.81 8.21 7.18 9.42 5.12 13 20 2.5 WQ + aging 3.97 3.85 3.51 4.22 3.02 14 30 2.75 WQ + aging 6.19 6.42 5.7 7.27 3.72 15 40 2 WQ + aging 9.93 10.28 8.86 11.99 3.52 Average 3.91
The best treatment is considered that one giving a microstructure withstand extreme operating conditions i.e., maximum normal load and maximum sliding speed but shows the minimum wear rate. According to this optimization criteria and by using the regression model equations, the optimum solution is shown in Fig. 16. The optimum set of input parameters are 42.75 N normal load, 3 m/s sliding speed and annealed condition (Equiaxed microstructure) which give optimum wear rate of 8.49 g/min with maximum desirability of 0.655 which means the gool of optimization is achieved by 65.5%. Table 8 summarizes confirmation test results at optimum conditions, the experimental WR within the 95% prediction interval (PI) limits of the model with average absolute error of 6.04%.
Graph: Figure 16 Optimum set of input parameters necessary to get the minimum WR under extreme conditions.
Table 8 Confirmation test results at optimum conditions.
No. Load (N) Speed (m/s) HT Expt. WR (g/min) × 10−3 Predicted WR (g/min) × 10−3 95% PI low 95% PI high % Absolute error 1 42.75 3 Ann 8.93 8.49 7.31 9.93 4.93 2 42.75 3 Ann 9.15 8.49 7.31 9.93 7.21 3 42.75 3 Ann 9.03 8.49 7.31 9.93 5.98 Average 6.04
- With respect to annealed specimens (38 HRC), the minimum hardness achieved by WQ specimens of 36 HRC, while the maximum hardness achieved by WQ + Aging specimens of 49 HRC.
- Under extreme wear condition (50 N, 3 m/s), although WQ + aging specimens had the maximum hardness, they showed the worst wear resistance. While, the annealed ones showed the best wear resistance regardless having much lower hardness.
- Abrasive wear mechanism is predominant under low wear conditions (10 N, 1.5 m/s) while delamination wear mechanism is predominant under extreme conditions.
- Using RSM, regression model for wear resistance expressed in wear rate has been developed as a function in normal load, sliding and type of heat treatment. Based on ANOVA, the normal load has been identified as the most significant input factor followed by the sliding speed and the type of heat treatment. Also, the interaction effect between the load and the speed has been identified as the most significant interaction.
- Model validation results revealed the experimental results are within 95% prediction interval of the model with average absolute error of 3.91% hence, the developed model is valid to predict WR within the design space.
- The obtained model was used to predict the optimum levels of input factors required to get the minimum wear rate under severe conditions of load and speed. Experimental results showed that the actual WR under those optimum levels is close to the predicted one with average absolute error of 6.04%.
Conceptualization, R.N.E., A.A. and M.E.-S.; methodology, R.N.E., A.A., and M.E.-S.; validation, R.N.E. and A.A.; formal analysis, R.N.E.; resources, R.N.E.; data curation, A.A. and R.N.E.; writing-original draft preparation, A.A.; writing—review and editing, R.N.E., M.E.-S. and A.S.S.; visualization, R.N.E. and A.A., M.E.-S., A.S.S. and R.N.E. supervision. All authors have read and agreed to the published version of the manuscript.
Open access funding provided by The Science, Technology & Innovation Funding Authority (STDF) in cooperation with The Egyptian Knowledge Bank (EKB). The authors would like to acknowledge the fund from Science Technology and Development Fund-Egypt, Grant No. 43215.
All data generated or analyzed during this study are included in this published article.
The author declares no competing interests.
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
By Ali Abdelmoneim; Ramadan N. Elshaer; M. El-Shennawy and Arafa S. Sobh
Reported by Author; Author; Author; Author