Iaetsd experimental investigation and predictive modelling for surface roughness of

Experimental Investigation and predictive modelling for Surface Roughness of
drilling on GFRP composites
Prasanna Ragothaman
Department of Mechanical Engineering, Sri Ramakrishna Engineering College, Coimbatore, India
prasse1994@gmail.com
Abstract: Glass Fiber Reinforced Plastics composites have an increased application in recent days, due to its enhanced
structural properties, Mechanical and thermal properties. Drilling of holes in GFRP becomes almost unavoidable in
fabrication. The heterogeneous nature of this kind of materials makes complications during machining operation.
However, drilling is a common machining practice for assembly of components. The quality of holes produced in the
GFRP material is severely affected by surface roughness, Circularity and Delamination . The objective of the study is
to apply the full factorial design, ANOVA and Fuzzy logic model to achieve an improved hole quality considering the
minimum surface roughness through proper selection of drilling parameters. The regression method is employed in the
Experimental investigation and Mathematical modelling of drilling of GFRP material using HSS drill bits and the fuzzy
logic model for the validation of the mathematical model.
Keywords: GFRP, ANOVA, Fuzzy logic, Full factorial Method, Drilling, Surface Roughness.
1. Introduction
Glass fiber Reinforced Plastics (GFRP) are widely being used in the automotive, machine tool industry, aerospace
components, sporting equipment’s [1] because of their particular mechanical and physical properties such as specific
strength and high specific stiffness. An aircraft fuselage structure around 100,000 holes is required for joining purpose
[2, 3]. About 60% of the rejections are happening in aircraft industry due to the defects in the holes [4]. Many of these
problems are due to the use of non-optimal cutting tool designs, rapid tool wear and cutting parameters [5, 6]. Among
the defects caused by drilling with tool wear, Delamination appears to be the most critical [7].The surface finish of the
work piece is an important attribute of hole quality in any drilling operation. During machining many factors affect the
surface finish. Many theoretical models have concluded that the effect of spindle speed on surface roughness is minimal
[8]. In practice, however spindle speed has been found to be an important factor [9]. The quality of drilling surfaces
depends on the cutting parameters and tool wear, while changing the cutting parameters causes to tool wear
[10].Researchers have attempted to model the surface roughness prediction using multiple regression, mathematical
modeling based on physics of process, fuzzy logic [11].Machining operation being highly nonlinear in nature, soft
computing techniques have been found to be very effective for modeling [12].The influence of process parameters such
as spindle speed, lubrication and feed rate on surface finish were investigated during the experimentation of Metal
matrix composites. The experiments were conducted according to the full factorial design [13] .The percentage of
contribution of highest influential factors can be determined using analysis of variance(ANOVA) a statistical tool, used
in design of experiments[14,15]. Fuzzy logic is a mathematical formalism for representing human knowledge involving
vague concepts and a natural but effective method for systematically formulating cost effective solutions of complex
problems [16]. A model was developed for surface roughness on drilling of GFRP composites using fuzzy logic [17].
The primary objective of this study is to quantify the influence of process input parameters on surface roughness by
formulating a mathematical model and validating using Fuzzy logic model.
2. Design of experiment
Design of experiment is the design of all information -gathering exercises where variation is present, whether under the
full control experimenter or not. The cutting speed, feed rate and thickness of GFRP plate are the three parameters
under investigation in the present study. A full factorial experimental design with a total number of 27 holes drilled into
the GFRP specimen to investigate the hole quality on Surface Roughness. The full factorial design is the most efficient
way of conducting the experiment for that three factors and each factor at three levels of experiments is used. Hence as
per Levelsfactor
(factors to power of levels) formula = Levelsfactors
,N = 33
= 27, N- number of experiments.
INTERNATIONAL ASSOCIATION OF ENGINEERING & TECHNOLOGY
International Conference on Advancements in Engineering Research
ISBN NO : 378 - 26 - 138420 - 8
www.iaetsd.in
77

Table 1Assignment levels of process parameters
Fig 1: Fabricated GFRP plate
3. Specimen Preparation
The Glass fiber reinforced composite used is fabricated using hand lay-up technique [12]. The composition type is
Glass fibers (fiber length=20-30mm) reinforced with isopathalic resin with 30% reinforcement .The material was
fabricated and then cut into pieces of 22cm x 11cm for all the three thicknesses of plate (Fig.1).
3.1 Methodology
Fig 2.Surface Roughness tester Fig 3.Experimental setup
Experiments were carried out in high speed radial drilling machine using HSS drill of 10mm diameter. Experiments
were carried out according to full factorial design. It provide a powerful and efficient method for designing processes
that operate consistently and optimally over a variety of conditions. The selected levels of process parameters were
given in Table 1.Fig. 3 shows the photographic view of the experimental setup. Further, the hole quality characteristics
surface roughness measured using roughness tester [Mitutoyo TR-200].Fig. 2 shows the measurement of hole quality
characteristics using roughness tester. Point angle was measured before every drill for 27 experiments using Digital
Profile Projector [OPTOMECH, 10x magnification].
4. Results and Discussion
4.1. ANOVA
The Analysis of variance is extensively used to analyze the experimental results. ANOVA tests the significance of
group difference between two or more groups. The normal probability plot represents that all the points on the normal
plot lie close to the straight line (main line) or not. Versus fits plots represents that how far deviation occur from the
normal distribution. An interaction plot is occurs when the change in response from the one level of a factor to another
level differs from change in response at the same two level second factor. A main effect plot is present when different
levels of an input affect the responses directly.
4.2. ANOVA for Surface Roughness
Fig.4 Represent that all the points lie closer to the regression line, this implies that the data are fairly normal and there is
a no deviation from the normal. Histogram graph shows the skewness. The Equation No. 1 represents that feed has
much effect on Ra. The main effect plot for Surface Roughness has been shown in the Fig 5. The plot shows that Ra
decrease with low cutting speed and low feed rate for 15 mm plate, as well as the initial (without wear in drill bit )
point angle has less effect on Ra. Table 3. Shows that the analysis of variance of second order model with 95%
confidence interval for the Surface roughness experiments. Parameter A gives 44.2% contribution to the Ra.
Factors Levels
1 2 3
Speed, s(rpm) 280 900 1800
Feed, f(mm/rev) 0.18 0.71 1.40
GFRP Plate
thickness, t(mm)
5 10 15
ISBN NO : 378 - 26 - 138420 - 8
www.iaetsd.in
78

Fig.4 Residual plot for Ra
Fig.5 Main Effect Plot for Ra
Fig.6 Interaction Plot for Ra
4.3 Mathematical model for Surface Roughness
The models were based on the Box-Behnkn design method. The developed second order mathematical model for
surface roughness.
Surface Roughness = 4.87 - 0.00086 (s) + 2.49 (f) + 0.249 (t) (1)
Table 3Analysis of variance
Predictors Coef SE Coef T P
Constant 4.873 2.352 2.07 0.05
S 8.56e-4 1.09e-3 -0.78 0.442
F 2.494 1.367 1.49 0.151
T 0.2487 0.1673 2.07 0.017
S = 3.54808 R-Sq = 21.1% R-Sq(adj) = 10.8%
Source DF SS MS F P
Regression 3 77.44 25.81 2.05 0.135
Residual Error 23 289.54 12.59
Total 26 366.99
Fig.6 Represent that high feed rate and low speed have less effect on Ra while drilling on 5mm thickness plate. When
drilling on 10mm thickness of plate with cutting parameters of low speed and feed rate shows surface roughness is
ISBN NO : 378 - 26 - 138420 - 8
www.iaetsd.in
79

minimized. For 15mm plate high speed and high feed rate has less effect on Ra. From Fig.5 shows that when decreasing
the point angle, the effect of surface roughness is increased. Decreasing the point angle is causes tool wear.
Fig.7 shows the predicted and measured hole characteristics at different drilling process parameter conditions. The
result significantly shows that the values relatively follow the similar trend pattern of the measured value and predicted
values from the developed regression model.
4.4 Fuzzy logic model
Fuzzy logic refers to a logical system that generalizes the classical two-value logic for reasoning under uncertainty. It is
a system of computing and approximate reasoning based on a collection of theories and technologies that employ fuzzy
sets, which are classes of objects without sharp boundaries. Fuzzy logic is the best captures the ambiguity in the input.
Fuzzy logic has become popular in the recent years, due to the fact that it is possible to add human expertise to the
process. Nevertheless, in the case where the nonlinear model and all the parameters of a process are known, a fuzzy
system may be used.
4.4.1.Devolopment of fuzzy logic model
The surface roughness in drilling of GFRP is assumed as a function of three input variables viz.plate thickness, spindle
speed, and feed rate. The Fuzzy logic prediction model is developed using Fuzzy Logic Toolbox available in Matlab
version 7.10(R2010a).In this work Mamdani type Fuzzy Inference Systems(FIS) is used for modeling. The steps
followed in developing The fuzzy logic model are described below.
4.4.1.A.Fuzzification of I/O variables:
The input and output variables are fuzzified into different fuzzy sets. The triangular membership function is
used for simplicity yet computationally efficient. It is easy to use and requires only three parameters to define.The input
variables plate thickness [5-15 mm] ,spindle speed [280-1800 rpm] and feed rate [0.18-1.4 mm/rev] are fuzzified into
three fuzzy sets viz.Low (L),Medium(M),and High(H) as shown in the Fig.8 (a,b,c).The output variable i.e. The surface
roughness is divided into nine fuzzy sets as Very Very Low(VVL),Very Low(VL),Low(L),Medium1 (M1),Medium2
(M2), Medium3 (M3),High (H),Very High (VH),Very Very High (VVH) as shown in Fig.8 (d) to increase the
resolution and accuracy of prediction.
Fig.8 Fuzzification of I/O variables
4.4.1.B. Evaluation of IF-THEN rule
The three input variables are fuzzified into three fuzzy sets each, the size of rule base becomes 27(3x3x3).For
generating the Fuzzy rules, the level of the variable having more membership grade on a particular fuzzy set is
considered. With appropriate level of all the input variables representing the corresponding fuzzy set, the surface
roughness values are used for 27 data sets of fuzzy rule base. Since all the parts in the antecedents are compulsory for
getting the response value, the AND (min) operator is used to combine antecedents parts of each rule.
The implication method min is used to correlate the rule consequent with its antecedent. For example, the first
rule of the FIS can be written as
ISBN NO : 378 - 26 - 138420 - 8
www.iaetsd.in
80

Rule 1: ‘if Thickness is Low and Speed is Low and Feed rate is Low then surface roughness is Very Very low
(VVL)’.
4.4.1.C.Aggregation of Rules
The aggregation of all the rule outputs is implemented using max method, the commonly used method for
combining the effect of all the rules. In this method the output of each rule is combined into single fuzzy set whose
membership function value is used to clip the output membership function. It returns the highest value of the
membership functions of all the rules.
4.4.1.D.Defuzzification
The aggregate output of all the rules which is in the form of fuzzy set is converted into a numerical value (crisp
number) that represents the response variable for the given data sets. In the present work, the centroid defuzzification
method is used for this purpose. It is the most popular method used in most of the fuzzy logic applications. It is based
on the centroid calculation and returns center of area under the curve.
The predicted values of surface roughness are compared with the experimental output, prediction model output
and fuzzy output. The comparison of prediction performance in fuzzy logic output, prediction model output with the
experimental results is given in the Table 4.
Table 4.Surface roughness values for Experimental output, Predicted output and Fuzzy output
S.
No
Plate
thickness
t (mm)
Speed
s(rpm)
Feed
f(mm/rev)
Point angle
θ (◦)
Surface Roughness, Ra (µm)
Experimental
output
Predicted
output
Fuzzy
output
1
5 280 0.18 107º32'07" 3.19 6.3224 2.15
2
5 280 0.71 107º29'47" 11.98 7.6421 10.00
3
5 280 1.40 107º25'32" 6.49 9.3602 5.75
4
5 900 0.18 107º20'44" 4.09 5.7892 3.62
5
5 900 0.71 107º17'58" 9.16 7.1089 7.87
6
5 900 1.40 107º15'20" 8.79 8.8270 7.87
7
5 1800 0.18 107º19'31" 12.96 5.0152 12.1
8
5 1800 0.71 107º12'25" 7.27 6.3349 5.75
9
5 1800 1.40 107º08'28" 3.33 8.0530 2.50
10
10 280 0.18 107º02'57" 5.64 7.5674 3.62
11
10 280 0.71 106º47'11" 8.98 8.8871 7.87
12
10 280 1.40 106º39'47" 11.42 10.6052 10.00
13
10 900 0.18 106º30'09" 1.76 7.0342 2.19
14
10 900 0.71 106º28'42" 7.12 8.3539 5.75
15
10 900 1.40 106º24'50" 8.58 10.0720 7.87
16
10 1800 0.18 106º21'20" 7.72 6.2602 5.75
17
10 1800 0.71 106º19'26" 10.07 7.5799 10.00
18
10 1800 1.40 106º16'32" 8.97 9.2980 7.87
19
15 280 0.18 106º32'07" 8.37 8.8124 7.87
20
15 280 0.71 105º58'10" 10.48 10.1321 10.00
21
15 280 1.40 105º52'37" 15.75 11.8502 14.30
22
15 900 0.18 105º44'12" 5.43 8.2792 3.62
ISBN NO : 378 - 26 - 138420 - 8
www.iaetsd.in
81

S.
No
Plate
thickness
t (mm)
Speed
s(rpm)
Feed
f(mm/rev)
Point angle
θ (◦)
Surface Roughness, Ra (µm)
Experimental
output
Predicted
output
Fuzzy
output
23
15 900 0.71 105º36'04" 18.25 9.5989 16.4
24
15 900 1.40 105º14'39" 11.43 11.3170 10.00
25
15 1800 0.18 105º22'42" 4.74 7.5052 3.62
26
15 1800 0.71 105º08'35" 6.64 8.8249 5.75
27
15 1800 1.40 104º58'49" 8.55 10.5430 7.87
Fig.10 Correlation between Experimental Ra with
predicted Ra and Fuzzified Ra
The variation of surface roughness with different combinations of input variables is studied using the output
surface FIS.Fig.14,Fig.15,Fig 16,and Fig 17 shows the Functional dependence of surface roughness(Ra) with Plate
thickness,feed rate and Spindle speed.It can be observed that the surface roughness increase with increase in plate
thickness or increse in spindle speed or increase in feed rate.And it is also observed that surface roughness is decreases
with small plate thickness, medium spindle speed and small feed rate. Fig.10 indicates that the outputs from the
experiments, Prediction model, Fuzzy are in good correlation with each other.
Fig.11 Rule viewer Fig .12 Surface Viewer
ISBN NO : 378 - 26 - 138420 - 8
www.iaetsd.in
82

Fig .13 Surface roughness vs Thickness Fig.14 Surface roughness vs Speed
Fig .15 Surface roughness vs Feed rate
5. Conclusion
This experimental investigation presents the surface roughness characteristics of drilling on GFRP composites. A
simple regression prediction model was developed based on the function of process variables and the following
conclusions were made
 Surface Roughness was analyzed as a function of process input variables. Validation was done with a
developed fuzzy –rule based model. The results obtained from experiments, Prediction model and the fuzzy
model are in good correlation with each other.
 From analysis of variance and from the fuzzy model, the results indicated that low feed rate, high spindle
speed and 5mm thickness of GFRP plate gives better Surface Roughness.
 It was observed that the surface roughness increases with the decreasing of point angle.
 Further investigations are needed to enhance the hole quality characteristics considering different tool
materials and tool diameters, considering machine vibration, etc during drilling of GFRP composites.
6. References
[1] Park, J. N., Cho, G. J.A Study of the Cutting Characteristics of the Glass Fiber Reinforced Plastics by Drill Tools,
International Journal of Precision Engineering and Manufacturing, vol. 8 (2007) 11-15
[2] VijayanKrishnaraj, Member, IAENG, Effects of Drill Points on Glass Fiber Reinforced Plastic Composite While
Drilling at High Speed, Proceedings of the World Congress on Engineering 2008 Vol II, WCEE 2008, July 2-4(2008)
London, U.K.
[3] Sonbatry El, Khashaba U.A, Machaly T, Factors affecting the machinability of GFRP/epoxy composites, Comp
Structures, 63 (2004) 329-338.
[4] Montgomery, D.C.,. Design and Analysis of Experiments: Response Surface Method and Designs. John Wiley and
Sons. New York, USA, 2005.
ISBN NO : 378 - 26 - 138420 - 8
www.iaetsd.in
83

[5] Konig W, WulfCh, Graß P and Willercheid H, Machining of fiber reinforced plastics, Annals CIRP, 34 (2) (1985)
537-548.
[6] Komaduri R, Machining of fiber-reinforced Composites, Mechanical Engineering, 115 (4), (1993) 58-66.
[7] A.M. Abrao et.al,. Drilling of fiber reinforced plastics: A review, Journal of Materials Processing Technology 186
(2007) 1–7.
[8] Abrao A M, Faria PE, Campus Rubio J., C., Reis P, PauloDavim J Drilling of fiber reinforced plastics: A Review. J
Materl. Process Technology 186 (2007) 1-7.
[9] CaprinoG, Tagliaferi V Damage development in drilling glass fiber reinforced plastics. Int J Mach tools Manuf
(6): (1995) 817-829.
[10]Hocheng, H.and H. Puw. On drilling characteristic fiber reinforced Thermoset &Thermoplastics. Int J Mach tools
Manuf ,32 (1992)583-592.
[11]M.Chandrasekaran, M.Muralidhar, C.M.Krishna and U.S.Dixit, Application of soft
computing techniques in machining performance prediction and optimization:a literature
review,Int J Adv Manuf Technol,Vol.46(2010) 445-464.
[12]M.Chandrasekaran and D.Devarasiddappa ,Development of Predictive Model For Surface Roughness in End
Milling of Al-SiC Metal matrix Composites using Fuzzy logic, Engineering and Technology 68 (2012) 1271-1276
[13]Sureshkumar ManickamShanmugasundram et.al, Experimental Investigation of Prediction
of Hole quality Characteristics of Aluminum Matrix Composite (AMC225xe). Advanced
Materials Research Vols.622-623 (2013) 1305-1309.
[14] C.Y.Hsu,C.S.Chen,C.C.Tsao,Free abrasive wire saw machining of ceramics, Int J Adv
Manuf Technology 40 (2009) 503-511.
[15] Bala Murugan Gopalsamy,Biswanath Mondal,Sukamal Ghosh,Optimisation of machining
parameters for hard machining:grey relational theory approach and ANOVA,The International
journal of Advanced Manufacturing Technology 45 (2009) 1068-1086.
[16]Vikram Banerjee et.al,Design space exploration of mamdani and sugeno inference systems
for fuzzy logic based illumination controller, International journal of VLSI and Embedded
system-IJVES (2012) 97-101.
[17]B.latha and B.S.Senthilkumar, Modeling and Analysis of Surface Roughness Parameters
in Drilling GFRP Composites Using Fuzzy Logic, Materials and Manufacturing Processes
25(8) (2010) 817-827.
ISBN NO : 378 - 26 - 138420 - 8
www.iaetsd.in
84

Iaetsd experimental investigation and predictive modelling for surface roughness of

More Related Content

What's hot (16)

Similar to Iaetsd experimental investigation and predictive modelling for surface roughness of (20)

More from Iaetsd Iaetsd (20)

Recently uploaded (20)

Iaetsd experimental investigation and predictive modelling for surface roughness of