Measurement 49 (2014) 256–265

Contents lists available at ScienceDirect

Measurement

journal homepage: www.elsevier .com/ locate /measurement

A fuzzy logic predictive model for better surface roughnessof Ti–TiN coating on AL7075-T6 alloy for longer fretting fatiguelife

0263-2241/$ - see front matter � 2013 Elsevier Ltd. All rights reserved.http://dx.doi.org/10.1016/j.measurement.2013.11.042

⇑ Corresponding author at: Department of Mechanical Engineering,Faculty of Engineering, Assiut University, Assiut 71516, Egypt.

E-mail addresses: erfan_zalnezhad@yahoo.com (E. Zalnezhad), ah_sarhan@um.edu.my (A.A.D. Sarhan).

E. Zalnezhad a,b, Ahmed A.D. Sarhan a,c,⇑a Center of Advanced Manufacturing and Material Processing, Department of Mechanical Engineering, Faculty of Engineering,University of Malaya, Kuala Lumpur 50603, Malaysiab Faculty of Engineering, Islamic Azad University, Chalous Branch, Iranc Department of Mechanical Engineering, Faculty of Engineering, Assiut University, Assiut 71516, Egypt

a r t i c l e i n f o a b s t r a c t

Article history:Received 5 April 2013Received in revised form 30 June 2013Accepted 25 November 2013Available online 4 December 2013

Keywords:AL7075-T6 alloyTi–TiN coatingPVD magnetron sputteringSurface roughnessFuzzy logicFretting fatigue

In this study, the fretting fatigue resistance of AL7075-T6 alloy is investigated using surfacetreatment Ti–TiN multilayer coating by physical vapor deposition (PVD) magnetron sput-tering technique. A fuzzy logic model was established to forecast the surface roughness ofTi–TiN coating on AL7075-T6 with respect to changes in the input process parameters ofDC power, temperature, DC bias voltage, and nitrogen flow rate. The results indicate anagreement between the fuzzy model and experimental results with 95.349% accuracy.The fretting fatigue lives of Ti–TiN-coated specimens with the lowest surface roughnessresulting from fuzzy logic were enhanced by 18% at low cyclic fatigue, while at high cyclicfatigue the result was reversed.

� 2013 Elsevier Ltd. All rights reserved.

1. Introduction

Light-weight and high-strength aluminum alloys suchas 7075 are widely utilized in aircraft engines, fuselageand automobile parts. Aluminum alloys themselves donot provide adequate mechanical strength for structuralparts. Therefore, enhancing surface properties is vital inpractical applications, particularly when aluminum is incontact with other components [1]. Ceramic coatings likeTiN, CrN, and SiC are possibly effective in raising surfaceshardness, corrosion and wear resistance. However, coat-ings used for aluminum alloys via old-fashioned tech-niques such as hard anodizing, glazing and thermalspaying, are provided with little support from the underly-

ing material besides the inadequate adhesion strength,decreasing their durability [2,3]. Fretting fatigue is a phe-nomenon which occurs when two components are in con-tact with each other and one or both are subjected to acyclic load, often resulting in damage which may causepremature component failure.

Ti–TiN coating applied by PVD technique is a method ofimproving the hardness of Al7075-T6 surface by reducingwear and extending fretting fatigue service life [4–6]. Theformation of TiN coating on the surface of materials isamong the most effective techniques of improving mate-rial wear resistance. However, the characterization of sur-face topography is imperative in applications includingabrasion, wear, and lubrication. Surface hardness androughness are important when identifying the perfor-mance of a coated surface [7,8]. High surface quality isessential in achieving lower surface roughness, which af-fects the practical characteristics such as friction, frettingfatigue, wear, and light reflection [9]. The surface rough-

E. Zalnezhad, A.A.D. Sarhan / Measurement 49 (2014) 256–265 257

ness of hard materials coated using PVD magnetron sput-tering technique is mostly affected by the selecting of var-ious coating parameters including DC bias voltage,nitrogen percentage, temperature, and DC power. Hence,for investigating and predicting the best surface roughnessat different Ti–TiN coating parameters, a reliable system-atic approach is required. Soft computing techniques areuseful when exact mathematical information is not avail-able. The techniques differ from customary computing inthat they are tolerant of uncertainty, imprecision, approx-imation, partial truth and met heuristics. Fuzzy logic is asoft computing method that plays a significant role in in-put–output parameter relationship modeling [10–12]. Fuz-zy modeling is called upon when subjective knowledgeand expert suggestions may significantly define the objec-tive function and decision variables. Fuzzy logic is ideal inpredicting the coating characterization of surface topogra-phy based on the input variables due to the nonlinear con-dition in the coating process [13]. Fuzzy logic wasestablished to develop the rule model for predicting thesurface roughness value in Ti–TiN coating on AL7075-T6alloy based on parameter and performance interaction.

The goal of this study is to evaluate the fretting fatiguelife of Ti–TiN coated Al7075-T6 alloy with different surfaceroughness.

In this present work, multilayer Ti–TiN was coated onAl7075-T6 substrate at different parameter conditions.Each parameter comprises four levels, namely nitrogenpercentage, power (DC), substrate temperature, and DCbias voltage. A fuzzy rule-based method was proposed topredict the surface roughness of Ti–TiN coating onAL7075-T6 alloy, after which the fretting fatigue lives ofcoated specimens with better surface roughness wereinvestigated.

2. Experimental details

2.1. Surface treatment and measurement

Aluminum 7075-T6 alloy was employed in this researchwork. The material’s composition was attained by Energy-dispersive X-ray spectroscopy (EDX) apparatus (Table 1).The ultimate strength and yield stress of Al7075-T6 were

Table 1Chemical composition of AL 7075-T6.

Al Cu Si Mg Cr Zn Mn

91 1.85 0.47 1.8 0.28 4.6 0.06

Table 2Parameters and levels used in the experiment.

Parameters Experime

1

A DC power (w) 300B Temperature (�C) 150C Nitrogen low rate (%) 3D Substrate biases voltage (v) 25

obtained by performing a number of tensile tests(rut = 590 MPa and ry = 520MPa, respectively). All sampleswere polished with SiC paper of 800–2000 grit, and weresurface-mirrored by diamond liquid. The substrate wascleaned with acetone in an ultrasonic bath for 14minutes,and then thoroughly rinsed with distilled water. An SGControl Engineering Pte Ltd. series PVD magnetron sputter-ing machine was used to experimentally deposit thin filmsof metal. DC generators were designed to facilitate metalsputtering. The chamber was evacuated to below2.67 � 10�3 Pa (2 � 10�5 Torr) before the argon gas forsputtering was introduced. A constant sputtering pressureof 6.9 � 10�1 Pa (5.2 � 10�3 Torr) was maintained. Table 2shows the parameters and levels throughout the experi-ment. The DC bias voltage, substrate temperature, nitrogenpercentage and DC power as coating parameters, were ar-ranged according to the experimental array provided in Ta-ble 3, to learn how to develop the sputtered Ti–TiN thinfilm roughness.

The TiN coating’s surface roughness was determinedusing micro-roughness equipment (MAHR). The layerswere characterized by scanning electron microscopy(SEM), focused ion beam technique (Quanta FEG250) andatomic force microscopy (AFM-NANOSCOPE DIMENSIOND13000).

2.2. Fretting fatigue

All specimens for the fretting fatigue test were ma-chined at an initial surface roughness of Ra = 0.6 ± 0.1 lmby lathe turning (CNC LATHE MACHINE, Miyano, BNC-42C5). The round shape samples utilized in this experi-mental work were prepared in accordance with ISO stan-dard [14]. Fretting fatigue pads were fabricated from AISI4140 steel plate with Vickers hardness of 346HV. Substratematerial (179HV) is softer than the pads but Ti–TiN coating(540HV) is harder. Illustrations of the fretting fatigue sam-ples and friction pads can be found in Fig. 1. A bridge-typefriction pad and ring-type load cell were designed andmanufactured to simulate the fretting fatigue conditions.Fig. 2 shows a schematic view of the fretting fatigue testsetup used in this study.

The samples were loaded rotationally in a rotatingbending fretting fatigue test apparatus (Fig. 2). The frettingfatigue tests were performed at a constant contact pres-sure of 100 MPa, while the samples comprised uncoatedand Ti–TiN-coated AL7075-T6. Plain and fretting fatiguetesting were carried out at laboratory temperature(25 �C) in a two-point loading rotating bending machine

ntal condition levels

2 3 4

350 400 500180 200 220

4 5 650 75 100

Table 3Standard L16 (44) experimental array.

Experiment Parameters combination

A B C D

1 1 1 1 12 1 2 2 23 1 3 3 34 1 4 4 45 2 1 2 26 2 2 1 47 2 3 3 28 2 4 3 29 3 1 3 410 3 2 4 311 3 3 1 212 3 4 2 113 4 1 4 214 4 2 3 115 4 3 2 416 4 2 1 3

(a) Fretting fatigue specimen

(b) Drawing of fretting pad

Fig. 1. Drawings of the fretting fatigue specimen and the fretting pad.

Fig. 2. Schematic of fretting fatigue test rig.

Table 4The measured surface roughness.

Experiment Measured surface roughness (lm) Average surfaceroughness (lm)

1st 2nd 3rd

1 0.048 0.069 0.054 0.0572 0.065 0.082 0.072 0.0733 0.054 0.063 0.059 0.0594 0.097 0.115 0.095 0.1025 0.042 0.055 0.071 0.0566 0.098 0.114 0.119 0.1107 0.022 0.038 0.045 0.0358 0.083 0.071 0.065 0.0739 0.091 0.108 0.156 0.118

10 0.108 0.121 0.111 0.11311 0.117 0.131 0.12 0.12312 0.072 0.069 0.086 0.07613 0.072 0.106 0.102 0.09314 0.084 0.071 0.089 0.08115 0.111 0.121 0.119 0.11716 0.241 0.252 0.374 0.289

258 E. Zalnezhad, A.A.D. Sarhan / Measurement 49 (2014) 256–265

(R = �1) at a rotational speed of 3000 rpm. The tests werestopped if the samples did not fail after 1 � 107 cycles.

The amount of friction coefficient (l) between pads(AISI 4140 steel) and AL7075-T6 was calculated to bearound 0.607. Friction force could be found from the equa-tion F = l�N, in which N is the contact load calculated by aring-shaped load cell (Fig. 2(a)) and F is the friction forcemeasured from the friction test machine.

3. Ti–TiN multilayer coating results

The coated specimens’ surface roughness was measuredwith micro-roughness equipment. Each roughness mea-surement was repeated three times and the calculatedaverages are tabulated in Table 4. A typical Ti–TiN multi-layer coating on Al7075 is depicted in Fig. 3(a), which re-veals under SEM that the coat’s structure is columnar. Asdemonstrated in Fig. 3, there are two portions to the coat-

ing, pure titanium as an interfacial layer and titanium ni-tride. Fig. 3(b) presents the diffusion rate of titanium andnitrogen, chemical composition of AL 7075-T6 and theinterfacial layer of titanium and TiN and aluminum. Thethicknesses of TiN coating at different parameter condi-tions are illustrated in Table 5.

TiN coating surface morphology was also studied usingatomic force microscopy (AFM). Fig. 4(a) and (b) show typ-ical examples of high and low surface roughness of Ti–TiNcoating under different parameter conditions.

4. Fuzzy logic-based model to investigate the effects ofTi–TiN multilayer parameters on substrate surfaceroughness

The relationship between the input parameters (sput-tering DC bias voltage, substrate temperature, nitrogenpercentage and DC power) with the output parameter(surface roughness of Ti–TiN coating on AL7075-T6) was

Inte

nsit

y%

(a)

(b)Fig. 3. (a) SEM micrograph and (b) EDX of the Ti–TiN coated AL7075-T6.

Table 5The thickness of Ti–TiN coating on different samples.

Sample Ti–TiN thickness (lm)

1 3.12 2.83 1.94 1.15 1.46 3.87 1.98 1.79 3.5

10 2.611 7.112 5.313 3.714 5.415 7.516 8.6

(a) Typically high surface roughness at DC power 200 (w),temperature 180 (˚C), nitrogen flow rate 4%,

and DC bias voltage 25(V)

(b) Surface roughness test at DC power 350 (w),temperature 200 (˚C), nitrogen flow rate 5%, and

DC bias voltage 50 (V)

Fig. 4. AFM surface morphologies of Ti–TiN thin film at differentparameters condition.

E. Zalnezhad, A.A.D. Sarhan / Measurement 49 (2014) 256–265 259

referred to for rule construction. Fuzzy linguistic variablesand fuzzy expression for input and output parameters arepresented in Table 6. Four membership functions were uti-lized for each variable as input: Low (L), Medium (M), High(H), and Very High (VH). The output variable (Roughness)also comprises four membership functions, Excellent,Good, Average, and Bad. The Input and Output variablecharacteristics are given in Table 6.

4.1. Fuzzy logic variables membership function for input andoutput

The event and type of membership function are mainlydependent on the relevant event in which the membership

functions are chosen for fuzzification. There are four mem-bership functions for each input and output parameter. Forinput and output variables, the Gaussian and triangularforms of membership function were used to describe thefuzzy sets. The triangular form of membership functionwith only one definite value has progressively increasingand decreasing characteristics [10,11]. The input variableswere partitioned according to the experimental parameterranges. Membership functions for fuzzy set input DC biasvoltage, nitrogen percentage, temperature, and DC powervariable are found in Fig. 5(a)–(d) respectively, while amembership function for a roughness fuzzy set is illus-trated in Fig. 5(e).

Table 6Fuzzy linguistic and abbreviation of variables for each parameter.

Parameters Linguistic variables Range

InputsA – DC power (W) Low (L), medium (M), high (H),

very high (VH)300–500

B – Temperature( C)

Low, medium, high, very high 150–220

C – Nitrogen flowrate (%)

Low, medium, high, very high 3–6

D – DC biasvoltage (V)

Low, medium, high, very high 25–100

OutputRoughness (lm) Excellent, Good, Average, Bad 0.035–

0.289

(a) Input variable “A” (DC power (w))

(c) Input variable “C” (nitrogen flow rate (%))

(e) Output variable “surfac

Deg

ree

of m

embe

rshi

p

Deg

ree

of m

embe

rshi

p D

egre

e of

mem

bers

hip

Fig. 5. Membership functions for the

260 E. Zalnezhad, A.A.D. Sarhan / Measurement 49 (2014) 256–265

4.2. Structure of fuzzy rules

A set of 16 rules was constructed based on the actualexperimental surface roughness of Ti–TiN coating onAL7075-T6, as shown in Tables 3 and 5. Experimental re-sults were simulated in Matlab software according toMamdani Fuzzy Logic as follows:

1. IF (A is L) and (B is L) and (C is L) and (D is L) then(Roughness is Good)

2. IF (A is L) and (B is M) and (C is M) and (D is M) then(Roughness is Good)

3. IF (A is L) and (B is H) and (C is H) and (D is H) then(Roughness is Good)

(b) Input variable “B” (temperature (˚C))

(d) Input variable “D” (DC bias voltage (v))

e roughness”

Deg

ree

of m

embe

rshi

p D

egre

e of

mem

bers

hip

input and output parameters.

(a) Surface roughness in relation to changeof DC power (A) and nitrogen flow rate (C)

(b) Surface roughness in relation to change ofDC bias voltage (D) and temperature (B)

Surf

ace

Rou

ghne

ss (

µm

)

Nitrogen flow rate%

Surf

ace

Rou

ghne

ss (

µm

)

Temperature (˚C)

DC bias voltage (V)

DC power (W)

Fig. 6. The predicted surface roughness by fuzzy logic in relation toparameters change.

E. Zalnezhad, A.A.D. Sarhan / Measurement 49 (2014) 256–265 261

4. IF (A is L) and (B is VH) and (C is VH) and (D is VH)then (Roughness is Bad)

5. IF (A is M) and (B is L) and (C is M) and (D is M) then(Roughness is Good)

6. IF (A is M) and (B is M) and (C is L) and (D is VH) then(Roughness is Bad)

7. IF (A is M) and (B is H) and (C is H) and (D is M) then(Roughness is Excellent)

8. IF (A is M) and (B is VH) and (C is H) and (D is M)then (Roughness is Good)

9. IF (A is H) and (B is L) and (C is H) and (D is VH) then(Roughness is Bad)

10. IF (A is H) and (B is M) and (C is VH) and (D is H) then(Roughness is Bad)

11. IF (A is H) and (B is H) and (C is L) and (D is M) then(Roughness is Bad)

12. IF (A is H) and (B is VH) and (C is M) and (D is L) then(Roughness is Average)

13. IF (A is VH) and (B is L) and (C is VH) and (D is M)then (Roughness is Average)

14. IF (A is VH) and (B is M) and (C is H) and (D is L) then(Roughness is Average)

15. IF (A is VH) and (B is H) and (C is M) and (D is VH)then (Roughness is Bad)

16. IF (A is VH) and (B is M) and (C is L) and (D is H) then(Roughness is Bad)

4.3. Defuzzification

Defuzzification is the conversion of a fuzzy quantity to aprecise value, just as fuzzification is the conversion of aprecise value to a fuzzy quantity [15]. There are severalmethods that researchers utilize for defuzzifying, includingcenter of largest area, first (or last) of maximum method,center of sum, centroid, weight average, mean of max,etc. [16]. Selecting the method is essential as it greatlyinfluences the model’s speed and accuracy. In the currentmodel, the centroid of area (COA) defuzzification methodis used due to its wide acceptance and capability to pro-duce more accurate results compared to the others. Theresultant membership functions in COA are developed byconsidering the union of each rule’s output, which meansthat the fuzzy output set’s overlapping area, is counted asone, providing more results [17,18].

Fig. 6(a) and (b) are examples that demonstrate thesuitable assent between parameter change and Ti–TiNcoating surface roughness values predicted by the fuzzy-based model. Fig. 6(a) shows that with increasing the DCpower from 300 to 400 W there is no remarkable increasein surface roughness, but with further increase from 400 to500 W, surface roughness drastically mounts. Fig. 6(a)indicates that increasing the nitrogen flow rate from 3%to 5.5% significantly lowers surface roughness, while byagain increasing the amount of nitrogen, surface roughnessis slightly increased slightly goes up. Temperature is yetanother element with effects on the surface roughnessquality of thin film Ti–TiN. Surface roughness diminisheswith increasing the temperature from 150 to 220 �C, as re-vealed in Fig. 6(b). The figure is manifests that the idealsubstrate DC bias voltage to achieve excellent roughnessis amplitude of 60–70 V; when raising DC bias up to

100 V, surface roughness is extremely increased. After cre-ating the fuzzy rules, five additional, new, separate exper-iments were performed while the proposed fuzzy modelwas used to predict surface roughness under the same con-ditions (Table 7) to explore fuzzy model accuracy and er-ror. The comparison between the fuzzy model predictionvalues and experimental results is depicted in Fig. 7(a)and (b). According to these figures, the experimental andfuzzy values are in very close assent to each other; hencethe fuzzy logic technique can be efficiently utilized for pre-dicting Ti–TiN coating surface roughness from magnetronsputter coating technique. Fig. 7(b) and (c) portray theaccuracy and error graphs, also indicating that the pro-posed model can satisfactorily predict the surface rough-ness of Ti–TiN multilayer-coated AL7075-T6.

5. Fretting fatigue test results

In order to examine the fretting fatigue life of Ti–TiN-coated specimens with low surface roughness investigatedin the fuzzy logic section, some experiments were exe-cuted and the results are illustrated in Fig. 8. Fig. 8(a)shows a comparison of the number of cycles to failure ver-sus bending stress for fretting fatigue and plain fatigue ofuncoated specimens. Clearly, the fatigue life of uncoatedspecimens lessened with increased bending stress. More-over, it can be deduced that fretting has a destructive effect

Table 7The accuracy and error of the fuzzy logic model prediction.

No. of Exp. Parameters (Inputs) Surface roughness result (output) (lm) Error Accuracy

Measured surface roughness Predicted surface roughness (Fuzzy)

A B C D 1st 2nd 3rd Average % %

1 425 210 3.3 80 0.285 0.266 0.273 0.275 0.288 4.727 95.2732 475 215 3.5 95 0.198 0.241 0.258 0.232 0.238 2.586 97.4143 380 160 5.5 30 0.079 0.096 0.085 0.087 0.092 5.747 94.2534 235 185 5.8 35 0.059 0.068 0.085 0.071 0.074 4.225 95.7755 375 190 4.7 40 0.072 0.071 0.059 0.067 0.071 5.970 94.03

Accuracy of model = 95.349

(a) Comparison of fuzzy logic model prediction with themeasured surface roughness in Ti-TiN coating on AL7075-T6

(b) The accuracy

(c) Error percentage

Fig. 7. Comparison of fuzzy logic model prediction with the measuredsurface roughness in Ti–TiN coating on AL7075-T6, the accuracy and errorpercentage.

262 E. Zalnezhad, A.A.D. Sarhan / Measurement 49 (2014) 256–265

on fatigue life. The reduction in percentage of the frettingfatigue life of uncoated specimens vs. stress compared toplain fatigue is illustrated in Fig. 8(b). This figure proposesthat a drop in fretting fatigue life takes place when the ap-plied bending stress is increased. For example, the reduc-tion percentage is nearly 57% at a stress of around150 MPa, while at 300 MPa it is roughly 32%. Fig. 8(c) pre-sents the fretting fatigue S/N curve of uncoated and coatedspecimens with lower surface roughness. By studyingFig. 8(c) and (b), it can be understood that the fretting fati-

gue lives of Ti–TiN-coated specimens with low surfaceroughness improved by only 18% at high bending stress,while at low bending stress, the results are reversed.Fig. 9 shows the tested specimens’ fracture surfaces exam-ined using optical microscopy, suggesting that the fracturesurface contains two different areas: a fatigue region gen-erated by crack propagation and a tensile zone which trig-gers specimen fracture when it is sufficiently weakened bydistinct crack development. Fig. 10 shows the cracked andshattered thin film Ti–TiN coating at a bending stress of300 MPa and 5.7E104 cycles caused by fretting pads. Thecracks occurred far from the fretting pads. Fig. 11 demon-strates the fractured Ti–TiN-coated specimen at a bendingstress of 250 Mpa and 1.2E105 cycles. The SEM image indi-cates that the fracture of the coated specimen containscracks and pits inside the specimen and around the coat-ing. A crashed region can be seen in the image as well,which is a result of the friction pads’ contact during the fa-tigue test. Fig. 10 shows the cracked and shattered thinfilm Ti–TiN coating at bending stress of 300 MPa and5.7E104 cycles caused by fretting pads. The cracks occurredat some distance from the fretting pads. Fig. 11 demon-strates the fracture of Ti–TiN coated specimen at a bendingstress of 250 Mpa and 1.2E105 cycles. The SEM image indi-cates that the fracture of coated specimen has cracks andpits inside the specimen and at the region of the coating.A crashed region can be seen in the image as well whichis resulted from friction pads during fatigue test.

6. Discussion

6.1. The effect of Ti–TiN coating parameters on surfaceroughness

The experimental and fuzzy model prediction resultsindicate that the surface roughness of Ti–TiN-coated spec-imens does not enormously increase when DC power israised from 300 to 400 W; however, surface roughness sig-nificantly increases if DC power is further raised to 500 W(Fig. 6(a) and (b)). When DC power gets increased up to400 W, the sputtered and ionized elements become moreenergized while sputtering rate improves, a phenomenonthat may cause the distance between energized atoms todecrease and makes the surface denser. When power is in-creased more from 400 to 500 W, the surface becomes den-ser and rougher. Besides, nitrogen is also a key parameterin Ti–TiN coating in helping to achieve good surface rough-

(a) S/N curve of plain fatigueand fretting fatigue for uncoated specimens

(b) Reduction percentage of fretting fatigue life vs. stress

Ben

din

g st

ress

(MP

a)

No. of cycle to failure

(c) S/N curve of fretting fatigue for uncoatedand Ti-TiN coated with lowestsurface roughness specimens

(d) Improvement percentage of fretting fatigue life vs. stress

Impr

ovem

ent p

erce

ntag

e of l

ife

%

Bending Stress (MPa)R

edu

ctio

n p

erce

nta

ge o

f lif

e %

Bending stress (MPa)

Fig. 8. S/N curve of plain fatigue and fretting fatigue of uncoated and coated specimen with lower roughness and improvement percentage of frettingfatigue life vs. stress.

(a) Fracture of uncoated specimen atbending stress 200MPa and 1.5E106 cycles

(b) Fracture of Ti-TiN coated specimenwith low surface roughness at bending

stress 200MPa and 6E105 cycles

Fig. 9. Optical images of fractured surfaces.

E. Zalnezhad, A.A.D. Sarhan / Measurement 49 (2014) 256–265 263

ness. The Ti–TiN multilayer coating roughness was alsomeasured as a function of nitrogen content in the argon-nitrogen gas mixture [19,20]. It is note worthy that with-out nitrogen doping, the pure Ti coating is rough(0.37 lm), while the nitrogen mix is effective in reducingsurface roughness. The coating’s surface roughness dimin-ishes when nitrogen is raised from 3% to 5.5%, and whennitrogen is further increased to 6%, surface roughness esca-lates (Fig. 6(a)). This may be due to the fact that the surfacebecomes more brittle. Fig. 6(a) also shows the best surfaceroughness value of Ti–TiN coating, at which point DC

power is between 300 and 400 W and nitrogen is at 5.5%.DC power has an important role in coating thickness. Ta-ble 6 shows the list of coatings thickness at differentparameter conditions. From the study of this table can beunderstood that the coatings thickness are increased withincreasing of DC power which this phenomenon is due toincreasing in the deposition rate. It means that more DCpower can separate more atoms from the target. Thereforemore atoms or molecules can stick to the surface duringsputtering. Beside that vacuum pressure also has a signifi-cant effect on coatings thickness. The film thickness im-

Fig. 10. The crack propagation on the surface of Ti–TiN coated specimenunder fretting pad (100 MPa) at bending stress 300 MPa and 5.7E104

cycles.

Fig. 11. Fracture of Ti–TiN coated specimen at bending stress 250 MPaand 1.2E105 cycles.

264 E. Zalnezhad, A.A.D. Sarhan / Measurement 49 (2014) 256–265

proves at high vacuum pressure. In this study, the vacuumpressure was kept constant. With inserting the nitrogenflow rate the film thickness changes due to change in vac-uum pressure. With increasing the nitrogen flow rate, thevacuum pressure decreases and therefore the coatingsthickness decrease. The best thickness was obtained atlow nitrogen flow rate and high DC power.

The DC bias voltage also plays a role in Ti–TiN coatingregarding attaining lower surface roughness. Coated spec-imen roughness diminishes when raising DC bias voltagefrom 25 to 62 V, while it begins to increase tremendouslyfrom 62 to 100 V (Fig. 6(b)). Thus, by biasing the substrateto an optimal value (62 V) the thin film Ti–TiN coating’suniformity is enhanced, and it was found that renucleationis initiated beyond the critical bias voltage value. The in-creased adatoms’ mobility along with higher nucleation

density is the result of raising substrate bias voltage duringmagnetron sputter coating. The best substrate bias voltagevalue to attain the best surface roughness seems to be62 V. When temperature goes up from 150 to 220 �C sur-face roughness declines, something attributed to the fasterand more energized atoms which create a more homoge-neous surface. Substrate temperature controls the delu-sions of atoms during Ti–TiN development [20]. Theeffect of deposition temperature on surface roughness ispresented in Fig. 6(b). Finally, differences in roughnessare evident where roughness declines from 0.18 to0.085 lm at arising deposition temperature.

6.2. S/N curves and fretting fatigue life

The corresponding fretting fatigue and plain fatigue S/Ncurves at 100 MPa contact pressure are shown in Fig. 8(a).The figure obviously demonstrates a significant decrease infretting fatigue life compared to the plain fatigue life, andparticularly at lower bending stresses. Fig. 8(b) revealsthe reduction percentage of fretting fatigue life versusstress. The lower fretting fatigue life of uncoated speci-mens in comparison to plain fatigue is due to the existenceof stress concentration in the interface area between thefriction pads and substrate. Because of locally concentratedfrictional shear stress at the contact surface, fretting fati-gue cracks get generated in this area. Thus, the decreasedfatigue life by fretting destruction is believed to be dueto the increase in crack initiation life produced by the localstress concentration from fretting as well as the accelera-tion of initial crack propagation from fretting [21].

Fig. 5(c) displays the S/N curve of fretting fatigue for un-coated and Ti–TiN-coated specimens with the lowest sur-face roughness. The fretting fatigue life of coatedspecimens with the lowest roughness improved 18% atlow cyclic fatigue, while at high cyclic fatigue, the resultsare reversed. Surface roughness has very complex and evencontradictory effects on fretting resistance. The plasticityindex in surfaces with high roughness is higher than insmooth surfaces [22], so at the asperities’ tips on a roughsurface some plastic deformation will occur. Work harden-ing is likely to prevent these deformed asperities frombeing completely flattened, so that the sharper asperitieson a rough surface can accumulate more of the tangentialmovement by elastic deformation. During fretting, anabundance of hard oxide debris accumulates on the con-tact surface and can cause severe abrasion, but on a roughsurface, there is a greater chance that wear-debris will es-cape from the contact areas into the adjacent hollows ordepressions instead of ploughing the worn surfaces. How-ever, in some cases, the increased surface roughness canresult in araising coefficient of friction, which is not bene-ficial for fretting fatigue resistance. A rough surface con-taining potential stress raisers is very dangerous,especially under fatigue conditions [23,24].

7. Conclusion

In this research work, AL7075-T6 samples were coatedwith Ti–TiN at varying parameter conditions. Fuzzy logic

E. Zalnezhad, A.A.D. Sarhan / Measurement 49 (2014) 256–265 265

model was used to predict the surface roughness of thecoated specimens. The fretting fatigue lives of uncoatedand coated specimens with low surface roughness wereinvestigated. From the experimental and computationalresults the following conclusions are obtained:

1. For AL7075-T6 alloy coated with Ti–TiN by sputter-ing machine, 350WDC power, 200 �C, 5.5% nitrogenflow rate, and 50 V substrate DC bias voltage arerecommended in order to obtain the lowest surfaceroughness for the specific test range of 0.035 lm.

2. The fuzzy model percentages of error and accuracywere found to be 5.97% and 95.349%, respectively.This indicates that the fuzzy logic prediction modelpredicted the AL7075-T6’s thin Ti–TiN film surfaceroughness with great accuracy.

3. Fretting can decrease the fatigue life of uncoatedAL7075-T6 by 57% in high cycle regions and 32%in low cycle regions (Fig. 8(b)).

4. The fretting fatigue life of Ti–TiN-coated specimenswith the lowest surface roughness was enhancedby 18% in low-cyclic fatigue, while in high-cyclicfatigue, the results are reversed.

Acknowledgements

The authors acknowledge the financial support underthe University Malaya Research Grant (Grant No.UM.TNC2/RC/AET/GERAN (UMRG) RP001C-13AET) fromthe University of Malaya, Malaysia.

References

[1] R. Sadeler, S. Atasoy, A. Arici, Y. Totic, The fretting fatigue ofcommercial hard anodized aluminum alloys, J. Mater. Eng. Perform.18 (2009) 1280–1284.

[2] E. Zalnezhad, Ahmed A.D. Sarhan, M. Hamdi, Investigating thefretting fatigue life of thin film titanium nitride coated aerospaceAl7075-T6 alloy, Mater. Sci. Eng., A 559 (2013) 436–446.

[3] W. Ren, S. Mall, J.H. Sanders, S.K. Sharma, Evaluation of coatings onTi–6Al–4V substrate under fretting fatigue, Surf. Coat. Technol. 192(2005) 177–188.

[4] I. Milosev, H.H. Strehbtow, B. Navinsek, Comparison of TiN, ZrN andCrN hard nitride coatings: electrochemical and thermal oxidation,Thin Solid Films 303 (1997) 246–254.

[5] B. Subramanian, M. Jayachandran, Characterization of reactivemagnetron sputtered nanocrystalline titanium nitride (TiN) thin

films with brush plated Ni interlayer, J. Appl. Electrochem. 37 (2007)1069–1075.

[6] E. Zalnezhad, Ahmed A.D. Sarhan, M. Hamdi, Optimizing the PVD TiNthin film coating’s parameters on aerospace AL7075-T6 alloy forhigher coating hardness and adhesion with better tribologicalproperties of the coating surface, Int J Adv Manuf Technol, doi:10.1007/s00170-012-4022-6.

[7] M. Diesselberg, H.R. Stock, P. Mayr, Corrosion protection ofmagnetron sputtered TiN coatings deposited on high strengthaluminium alloys, Surf. Coat. Technol. 177–178 (2004) 399–403.

[8] I. Iordanova, P.J. Kelly, R. Mirchev, V. Antonov, Crystallography ofmagnetron sputtered TiN coatings on steel substrates, Vacuum 81(2007) 830–842.

[9] W.J. Chou, C.H. Sun, G.P. Yu, J.H. Huang, Optimization of thedeposition process of ZrN and TiN thin films on Si (100) usingdesign of experiment method, Mater. Chem. Phys. 82 (2003) 228–236.

[10] V. Hajipour, A. Kazemi, S.M. Mousavi, A fuzzy expert system toincrease accuracy and precision in measurement system analysis,Measurement 46 (2013) 2770–2780.

[11] A.K. Nandi, J.P. Davim, A study of drilling performances withminimum quantity of lubricant using fuzzy logic rules,Mechatronics 19 (2009) 218–232.

[12] A. Gupta, H. Singh, A. Aggarwal, Taguchi-fuzzy multi outputoptimization (MOO) in high speed CNC turning of AISI P-20 toolsteel, Expert Syst. Appl. 38 (2011) 6822–6828.

[13] E. Zalnezhad, A.D. Ahmed, M. Sarhan, Hamdi, Prediction of TiNCoating Adhesion Strength on Aerospace AL7075-T6 Alloy UsingFuzzy Rule Based System, IJPEM 13 (2012) 1453–1459.

[14] ISO Standard, Metallic materials-rotating bar bending fatiguetesting, ISO International, 2010.

[15] J. Jang Sr., C.T. Sun, E. Mizutani, Neuro-Fuzzy and Soft Computing-AComputational Approach to Learning and Machine Intelligence,Pearson Education, India, 2005.

[16] J.L. Lin, K.S. Wang, B.H. Yan, Y.S. Tarng, Optimization of the electricaldischarge machining process based on the Taguchi method withfuzzy logic, J. Mater. Process. Technol. 102 (2000) 48–55.

[17] E. Cox, Fuzzy fundamentals, IEEE Spectrum, Oct. Issue, 1992.[18] E. Zalnezhad, A.D. Ahmed, M. Sarhan, Hamdi, Surface Hardness

Prediction of CrN Thin Film Coating on AL7075-T6 Alloy Using FuzzyLogic System, IJPEM 14 (2013) 467–473.

[19] E. Posti, I. Nieminen, Coating thickness effects on the life of titaniumnitride PVD coated tools, Mater. Manuf. Processes (2007) 363–382.

[20] J.H. Huang, K.W. Lau, G.P. Yu, Effect of nitrogen flow rate on structureand properties of nanocrystalline TiN thin films produced byunbalanced magnetron sputtering, Surface Coat. Technol. 191(2005) 17–24.

[21] W. Rena, S. Mall, J.H. Sanders, S.K. Sharma, Evaluation of coatings onTi–6Al–4V substrate under fretting fatigue, Surf. Coat. Technol. 192(2005) 177–188.

[22] A. Volchok, G. Halperin, I. Etsion, The effect of surface regularmicrotopography on fretting fatigue life, Wear 253 (2002) 509–515.

[23] X.H. Zhang, D.X. Liu, H.B. Tan, X.F. Wang, Effect of TiN/Ti compositecoating and shot peening on fretting fatigue behavior of TC17 alloyat 350 �C, Surf. Coat. Technol. 203 (2009) 2315–2321.

[24] Ahmed A.D. Sarhan, E. Zalnezhad, M. Hamdi, Hamdi, The influence ofHigher Surface Hardness on Fretting Fatigue Life of Hard Anodizedaerospace AL7075-T6 Alloy, Int. J, Mater. Sci. Eng., A 560 (2013) 377–387.