文档库 最新最全的文档下载
当前位置:文档库 › Experimental investigation and empirical modeling of the dry electric discharge machining process

Experimental investigation and empirical modeling of the dry electric discharge machining process

Experimental investigation and empirical modeling of the dry electric discharge machining process

Sourabh K.Saha?,S.K.Choudhury

Department of Mechanical Engineering,Indian Institute of Technology Kanpur,Kanpur208016,India

a r t i c l e i n f o

Article history:

Received15July2008

Received in revised form

1October2008

Accepted30October2008

Available online24November2008

Keywords:

Dry EDM

EDM in gas

Design of experiments(DOE)

Central composite design(CCD)

Analysis of variance(ANOVA)

Response surface analysis

a b s t r a c t

Dry electric discharge machining(EDM)is an environment-friendly modi?cation of the oil EDM process

in which liquid dielectric is replaced by a gaseous medium.In the current work,parametric analysis of

the process has been performed with tubular copper tool electrode and mild steel workpiece.

Experiments have been conducted using air as the dielectric medium to study the effect of gap voltage,

discharge current,pulse-on time,duty factor,air pressure and spindle speed on material removal rate

(MRR),surface roughness(R a)and tool wear rate(TWR).First,a set of exploratory experiments has been

performed to identify the optimum tool design and to select input parameters and their levels for later

stage experiments.Empirical models for MRR,R a and TWR have then been developed by performing a

designed experiment based on the central composite design of experiments.Response surface analysis

has been done using the developed models.Analysis of variance(ANOVA)tests were performed to

identify the signi?cant parameters.Current,duty factor,air pressure and spindle speed were found to

have signi?cant effects on MRR and R a.However,TWR was found to be very small and independent of

the input parameters.

&2008Elsevier Ltd.All rights reserved.

1.Introduction

Electric discharge machining(EDM)is one of the most popular

non-traditional machining processes being used https://www.wendangku.net/doc/a411631056.html,e of

mineral oil-based dielectric liquids is the major cause of

environmental concerns associated with the EDM process[1].

Dry EDM is an environment-friendly modi?cation of the oil EDM

process in which the liquid dielectric is replaced by a gaseous

medium.Dielectric wastes generated during the oil EDM process

are very toxic and cannot be recycled.Also,toxic fumes are

generated during machining due to high temperature chemical

breakdown of mineral oils.The use of oil as the dielectric?uid also

makes it necessary to take extra precaution to prevent?re

hazards.Replacing liquid dielectric by gases is an emerging?eld

in the environment-friendly EDM technology[2–7].High velocity

gas?owing through the tool electrode into the inter-electrode gap

substitutes the liquid dielectric.The?ow of high velocity gas into

the gap facilitates removal of debris and prevents excessive

heating of the tool and workpiece at the discharge spots.Providing

rotation or planetary motion to the tool has been found to be

essential for maintaining the stability of the dry EDM process[4].

Tubular tools are used and as the tool rotates,high velocity gas is

supplied through it into the discharge gap.Tool rotation during

machining not only improves the process stability by reducing arcing

between the electrodes but also facilitates in the?ushing of debris.

The?rst reference to dry EDM can be found in a1985NASA

Technical report[2].It was brie?y reported that argon and helium

gas were used as dielectric medium to drill holes using tubular

copper electrode.Further details are however not https://www.wendangku.net/doc/a411631056.html,ter

in1991,Kunieda et al.[3]showed that introducing oxygen gas into

the discharge gap improves the material removal rate(MRR)in a

water-based dielectric medium.It was in1997that the feasibility

of using air as the dielectric medium was?rst demonstrated by

Kunieda et al.[4].High velocity gas jet through a thin walled

tubular electrode was used as dielectric.Further research in this

?eld has brought out some of the essential features of the process.

It is now known that some of the advantages of the dry EDM

process are:low tool wear,lower discharge gap,lower residual

stresses,smaller white layer and smaller heat affected zone[4–7].

Also,several studies have been made to improve the performance

of the process such as by using a piezoelectric gap control

mechanism[8]and by introducing ultrasonic vibrations to the

workpiece[9,10].Dry EDM has also been successfully implemen-

ted in wire EDM operations[11,12].Tao et al.[13]have recently

reported that oxygen gas and copper tool combination leads to a

high MRR in dry EDM and nitrogen gas–water mixture dielectric

and graphite tool combination leads to high surface?nish in near-

dry EDM.However,the current literature in the?eld is insuf?cient

Contents lists available at ScienceDirect

journal homepage:https://www.wendangku.net/doc/a411631056.html,/locate/ijmactool

International Journal of Machine Tools&Manufacture

0890-6955/$-see front matter&2008Elsevier Ltd.All rights reserved.

doi:10.1016/j.ijmachtools.2008.10.012

?Corresponding author.Tel.:+16177155326.

E-mail address:sourabh.k.saha@https://www.wendangku.net/doc/a411631056.html,(S.K.Saha).

International Journal of Machine Tools&Manufacture49(2009)297–308

in order to make dry EDM a commercially viable process.Suitable process models for accurately predicting the process performance (such as MRR,surface?nish and tool wear rate(TWR))for a given set of input parameters are still not available.A limited knowledge base in parametric analysis makes it dif?cult to choose input process parameter values for obtaining a high performance.

In the current work,a machining unit has been designed and developed for implementing the dry EDM process on an existing oil EDM machine.Parametric analysis and empirical modeling of the process has been done by conducting experiments in two stages:exploratory experiments and designed experiments. Exploratory experiments have been conducted to identify suitable tool dimensions and machining time.Effect of the input parameters such as gap voltage,discharge current,pulse-on time, duty factor,inlet air pressure and spindle speed on MRR and surface roughness(R a)has also been studied one variable at a time in the exploratory stage.In the next stage,central composite design(CCD)of experiments has been used to develop empirical models of MRR,R a and TWR.Experimental data obtained from the CCD runs has?nally been used to perform a response surface analysis of the process.

2.Experimental set-up and procedure

2.1.Experimental set-up

Experiments have been conducted on a Z numerically controlled(NC)oil die-sinking EDM machine in which the Z-axis is servo controlled and the X and Y axes are manually controlled. All three axes have an accuracy of5m m.A dry EDM unit attachment has been designed and developed to enable perform-ing the dry EDM process on this machine[14].A schematic representation of the set-up is shown in Fig.1.The attachment unit has been designed to ful?ll the basic requirements of dry EDM:rotating tool and high velocity gas?ow through tubular tool.The unit comprises a hollow spindle shaft supported on the ?ange of a cylindrical support through a pair of bearings.The shaft can rotate relative to the support cylinder.A collet is mounted at the lower end of the shaft for holding tubular tools.A line of high-pressure gas from an external source such as an air compressor is connected to the support cylinder.Channel for gas?ow is made in the support cylinder.A tube transfers the gas from the support cylinder channel into the shaft-bore through an O-ring seal.The tubular tool mounted at the shaft end receives this high pressure gas while rotating relative to the support cylinder.The attachment unit is connected to the main machine spindle through the support cylinder and it borrows the feed motion(Z direction tool movement)from the Z-NC machine control.

2.2.Experimental procedure

Blind holes were drilled on the workpiece surface by the dry EDM process.Each drilling operation was performed for a?xed amount of time.MRR was calculated by measuring the loss in weight of the workpiece after machining.Initial and?nal weights of the workpiece were measured on Afcoset electronic balance (FX-400)having a resolution of0.001g.The weight MRR was then converted into volumetric MRR by knowing the density of the workpiece material.Similarly,loss in weight of the tool after machining was used to obtain the volumetric TWR.Surface roughness of the end surface of the machined holes on the workpiece was measured.Considering the size of the holes (diameter of the order of10mm and depth of the order of 100m m)and the relative ease of measurement,roughness of only the bottom surface was measured.The center line average(CLA) surface roughness parameter,R a was used to quantify the surface roughness.R a(JIS2001standard)was measured using a contact type stylus-based surface roughness tester,Mitutoyo Surftest SJ-301.Gaussian digital?lter with a cut-off length of0.8mm and an evaluation length of4mm was used.R a was measured along three different lines on the surface and the average value was considered for further analysis.

2.3.Workpiece,tool and dielectric

Experiments were conducted on EN32mild steel(density7.8g/ cm3)workpiece using a copper(density8.9g/cm3)tool.Work-piece was in the form of a strip of dimensions75mm?20mm?5 mm.Tool electrode was in the form of a tube so that high velocity gas could?ow through it.Experiments were performed with air as the dielectric medium.High-pressure air obtained from an air compressor was supplied through the tool into the discharge gap.

2.3.1.Tool design

When experiments were conducted using a thin walled tubular electrode with a single centrally located hole,it was found that a central core was formed on the machine surface after machining. This is because machining takes place only along the tube walls and no material is removed from the workpiece regions corresponding to the tube central hole.Due to the presence of central core during blind hole-drilling,it is not possible to take surface roughness measurements on the bottom surface of the hole.In order to prevent the formation of a central core,a tube with non-central holes was used.Due to tube rotation the entire surface is exposed to sparks and no central core is formed.To overcome the dif?culty in manufacturing tubes with high aspect ratio drilled holes,the tool design as shown in Fig.2has been used.Several small diameter non-central holes are drilled in a solid cylinder from one end.These holes open into a larger central hole,which is drilled into the cylinder from the other end.The top portion of the tool is made to the size of the tool holding collet where as the outer diameter of the lower portion is determined by the diameter of the hole to be drilled in the workpiece.

Nomenclature

a d distance of axial points from the center point in

central composite design(CCD)of experiments

D duty factor(%)

I d discharge current(A)

MRR material removal rate(mm3/min)

N spindle rotational speed(rpm)n number of axial holes for air?ow in the tool electrode P air inlet pressure(kPa)

R a center line average surface roughness(m m)

T on pulse-on time(m s)

T off pulse-off time(m s)

TWR tool wear rate(mm3/min)

V g gap voltage(V)

S.K.Saha,S.K.Choudhury/International Journal of Machine Tools&Manufacture49(2009)297–308 298

3.Plan of experiments

Experiments were conducted in two stages:?rst a series of exploratory experiments followed by a set of designed experi-ments.Exploratory experiments were conducted to select suitable parameters for the later stage experiments and to select appro-priate values of other controllable variables.Experiments based on central composite design (CCD)were then performed to systematically study the effect of various process parameters and their interaction effects on MRR,surface roughness (R a )and TWR.The data obtained from the CCD runs was used to ?t the models of MRR,R a and TWR in terms of the six input parameters:gap voltage (V g ),discharge current (I d ),pulse-on time (T on ),duty factor (D ),air inlet pressure (P )and spindle rotational speed (N ).

Details of the experiments performed in the two stages are discussed in this section.

3.1.Exploratory experiments

Out of the several controllable parameters available on the control panel of the EDM machine,V g ,I d ,T on and D were chosen as the input parameters.Other controllable parameters were either ?xed at average values (such as anti-arc sensitivity to 50%)or at most favorable values (such as servo motor sensitivity).It was found that material removal from workpiece occurred only when reverse polarity was used (i.e.workpiece as negative terminal and tool as positive terminal).Hence,all experiments were conducted with reverse polarity.In addition to the electrical parameters,air inlet pressure (P )and spindle rotational speed (N )were also considered among the input variables.The operating range of the input variables was determined by identifying the feasible conditions for existence of sparks.

Each dry EDM experiment was performed for a pre-determined machining time.The machining time was ?xed by performing a set of experiments with the same input parameters but for different machining times.It was observed that experiments must be conducted for at least 8–10min before stable machining conditions can be achieved.MRR was found to increase steeply with machining time in the beginning,but the rate of increase in MRR dropped as machining time was increased beyond 10min.The machining time for subsequent experiments was thus ?xed at 10min.The corresponding machined depth was observed to be of the order of 100m m,the exact value depending on the MRR.

Initial parametric studies of MRR and R a have been made by considering one variable at a time (OVAT).By keeping all other variables at ?xed average values,one variable at a time was varied and its effect on MRR and R a was studied.TWR was found to be small relative to the MRR (less than 1%of MRR)and exhibited minor variations on changing the input parameters.Hence,OVAT analysis of TWR was not performed during the exploratory stage.For each input variable,experiments were conducted for ?ve different levels and a single run was performed at each level.Although the OVAT analysis does not give a clear picture of the phenomena over the entire range of the input parameters,it can highlight some of the important characteristics.This may be

Fig.1.Schematic diagram of the experimental

set-up.

Fig.2.Machining with tool electrode having multiple non-central holes for high-velocity air ?ow.

S.K.Saha,S.K.Choudhury /International Journal of Machine Tools &Manufacture 49(2009)297–308

299

helpful in reducing the number of variables or restricting the range of the variables in later stage experiments.

The effect of tool geometry on MRR and R a was investigated and the optimum values were used for later stage experiments. Two geometrical parameters have been studied here:the tool outer diameter(OD)and the number of holes(n)for air?ow.The radial position and size of the holes were kept constant.Each hole was of diameter2mm and made on a pitch circle diameter of 5mm.Experiments were?rst conducted for identical input parameters using two tools with different outer diameters:11.5 and15mm.The tool with a smaller outer diameter(11.5mm OD) had a higher performance in terms of MRR,TWR and R a.Hence, tool OD of11.5mm was chosen and experiments were then conducted with tools having different number of air-?ow holes. The results are shown in Fig.3and suggest the existence of an optimum number of holes.When the number of holes(n)was increased,the?ow rate through the tools had to be increased to maintain a constant air inlet pressure.Hence the?ushing ef?ciency and heat carrying capacity increased,which led to an improvement in the MRR and the R a values.However,the tool area over which sparks can occur reduces when n increases.This leads to a reduction in the number of sparks per unit time,thereby lowering the MRR.Also,low frequency sparks lead to a higher R a value due to more energy contained in each spark as compared to high-frequency sparks.Thus,an optimum value of n exists for which the MRR is highest and the R a value is lowest.Hence the optimum three-hole tool was selected for the CCD experiments.

3.2.Central composite design of experiments

In order to develop empirical models for MRR,R a and TWR,a designed experiment was conducted based on the central composite design(CCD).The CCD design is capable of?tting up to the second order including the quadratic terms.In order to conduct the CCD experiments it is necessary to represent all the input parameters into the same range[à11],such that the high level of a parameter is represented as+1and the low level is represented asà1.Apart from the high and low levels,zero level (center point)and7a d levels(axial points)are also included in CCD[15].Value of a d depends on the number of factors.For six factors,a small practical value of a d?60.25(1.56508)was chosen. The high and low level values of the factors were chosen such that the7a d values were within the operating range of the variables. In case of unavailability of the parameter values corresponding to the coded levels in CCD design(due to the machine constraints),the nearest available values were used for conducting the experiments.The actual parameter values used for generating the design matrix are shown in Table1.Parameter values corresponding to the71level were the true values.For the axial and center points,the nearest feasible parameter setting was https://www.wendangku.net/doc/a411631056.html,ing the parameter settings shown in Table1,the design table for a6factor full-factorial CCD with10central runs was generated.Experiment run orders were randomized to remove the effect of systematic errors.The total number of runs was86,with 64factorial points,12axial points and10center points.

The MRR,R a and TWR values obtained for each CCD run are shown in Table2.Analysis of the experimental data was done using the Design Expert7.0.0software.Several models such as linear,linear with?rst-order interaction terms and quadratic models could be?tted using the CCD experimental data.Analysis of variance(ANOVA)-based sequential sum of squares test was done to select the most appropriate model to be?tted.Values of various regression statistics were also compared to identify the best?t model.Additionally,the?tting was improved by eliminating the insigni?cant terms through a backward step-wise model?tting with a term elimination probability value of10% [15].To study the effect of various input parameters and their interactions,a response surface analysis was then done using the developed empirical models.

4.Exploratory experiments:results and discussion

Preliminary parametric studies were made during the ex-ploratory stage by separately studying the effect of each parameter taken one at a time.The results of the OVAT analysis are discussed in this section.

4.1.Effect of spindle speed

The effect of spindle speed(N)on MRR and R a is shown in Fig.4.MRR initially increased with an increase in N but then saturated to a level and did not increase with further increase in N. R a values also followed a similar trend and decreased with increase in N up to a level and then almost saturated.This behavior can be explained by the effect of spindle rotation on the discharge phenomenon.Spindle rotation leads to an improvement in the?ushing ef?ciency.Due to rotation of air as it?ows through the tube,air is forced outward away from the center when it comes out of the tool carrying the debris particles away from the discharge gap.The outward?ow velocity of air is directly proportional to N.This effect leads to an improvement in the ?ushing ef?ciency up to a maximum limit.After this limit,since almost all the debris particles are effectively removed from the discharge gap,there is no scope for further improvement in the ?ushing ef?ciency.Hence,the MRR and R a values saturate at very high N values.Apart from the?ushing ef?ciency,tool rotation has an impact on the spark frequency.At high spindle speeds,a

Fig.3.Effect of number of holes for air?ow in the tool electrode on MRR and R a.Table1

List of actual parameter values used corresponding to the coded levels in CCD. Parameter Coded value

Modi?edàa dà1Modi?ed0+1Modi?ed+a d

Gap voltage,V g(V)5563779199 Discharge current,I d(A)916294249

Pulse-on time,T on(m s)502005007501000

Duty factor,D(%)824487288

Air inlet pressure,P(kPa)58.888.2147205.8245

Spindle speed,N(rpm)300650127519002250

S.K.Saha,S.K.Choudhury/International Journal of Machine Tools&Manufacture49(2009)297–308 300

S.K.Saha,S.K.Choudhury/International Journal of Machine Tools&Manufacture49(2009)297–308301

Table2

Experimental data obtained from the CCD runs.

Run order Input parameters Output response

V g(V)I d(A)T on(m s)D(%)P(kpa)N(rpm)MRR(mm3/min)R a(m m)TWR(mm3/min)

1911620072205.86500.71 3.040.04 2634220072205.81900 5.68 3.14à0.03 363162002488.26500.33 3.220.01 4914275024205.8650 2.28 3.1à0.02 591162007288.26500.55 3.46à0.02 6631675072205.86500.69 2.980.03 763167507288.26500.58 3.76à0.06 87729500881471275 2.67 3.310.02 991427507288.2650 2.31 3.980.01 10631620024205.86500.14 2.660.03 1191167507288.21900 1.03 3.25à0.02 1255295004814712750.87 3.530.02 137729500481471275 1.6 3.260.01 1491427502488.2650 1.33 4.29à0.01 1563162002488.219000.17 2.860.01 16631620024205.819000.68 2.4à0.03 17634220024205.81900 2.31 2.46à0.03 1863162007288.219000.99 2.960.03 19911620024205.86500.47 2.630.04 2063422002488.21900 1.85 3.330.04 217729500481471275 1.58 3.250.01 22634275072205.81900 4.67 2.980.03 2363422007288.21900 5.37 3.110.02 2491422007288.2650 3.19 4.270.03 257749500481471275 4.22 3.940.04 26914220024205.81900 2.44 3.50.02 2791162002488.26500.26 3.420.01 2891162002488.219000.64 2.95à0.02 29634275024205.81900 4.27 2.840.03 3091427502488.21900 2.86 4.080.01 317729504814712750.86 2.87à0.01 3263162007288.26500.36 3.360.02 33911675024205.86500.53 3.010.02 34914275024205.81900 3.6 3.37à0.02 3577295004858.81275 1.19 3.330.01 367729500481471275 1.82 3.210.04 37631620072205.86500.47 2.86à0.07 3863427507288.21900 3.31 4.20.04 39634220072205.8650 4.67 3.3à0.01 40911675072205.8650 1.03 2.920.03 417729500481471275 2.04 3.35à0.02 4263167507288.219000.88 3.70.07 4377291000481471275 2.69 3.130.02 447729500481472250 3.19 3.380.01 45634275072205.8650 2.51 3.290.01 4691167507288.26500.51 3.77à0.02 4791422002488.2650 1.22 3.40.02 4891167502488.219000.9 3.370.02 49911620072205.81900 1.14 2.480.02 5063167502488.26500.49 3.780.02 517729500481471275 1.92 2.990.01 52914220024205.8650 1.53 2.640.03 537795004814712750.51 2.850.03 547729500481471275 1.79 3.180.01 55634275024205.8650 2.04 3.210.01 567729500481471275 1.79 3.30.02 57914220072205.8650 2.69 3.570.04 587729500481471275 1.59 3.79à0.01 5991167502488.26500.55 3.680.02 6063427502488.2650 1.55 3.89à0.01 61914220072205.81900 4.19 3.340.02 6263427507288.2650 1.94 4.280.01 637729500481471275 1.85 3.60.01 64911675072205.81900 1.06 3.1à0.01 65772950081471275 1.1 3.490.03 6691427507288.21900 4.06 3.40.02 67914275072205.81900 4.42 3.280.01 68634220024205.8650 1.81 3.60.03 699929500481471275 1.95 4.28à0.03 70914275072205.8650 2.88 3.65à0.01 71911620024205.819000.68 2.490.02 727729500482451275 2.14 3.460 73911675024205.819000.51 2.750

discharge may be interrupted even during the pulse-on time due to movement of the tool.Thus,several short sparks occur over a single pulse-on time and the spark frequency increases with the spindle rpm.Since the same pulse energy is now distributed over a number of sparks,the crater depth is lower.Hence,lower R a values were observed when N was increased.

4.2.Effect of air inlet pressure

The effect of air inlet pressure(P)on MRR and R a is shown in Fig.5.The?gure suggests that higher air pressure provides a better performance in terms of both the MRR and R a values. Flushing ef?ciency improves with an increase in the pressure.A better removal of debris particles from the gap lowers the arcing probability.Arcing leads to surface damage,hence lower values of R a were obtained with low arcing probability at high pressures.

Also,when arcing is sensed by the EDM machine during a pulse, the pulse-on time is reduced during the duration of the pulse.This increases the non-cutting time during machining leading to a lower MRR at low values of P.

4.3.Effect of current

The effect of discharge current(I d)on MRR and R a is shown in Fig.6.MRR was found to increase with an increase in I d.Spark energy increases with current which leads to higher crater volumes.Thus MRR increases with current.However,R a values remain almost constant with increase in current.For higher currents,the crater volume may increase either due to an increase in the depth of the crater or increase in the diameter of the crater or due to an increase in both of these.The R a value is more sensitive to the crater depth as compared to the crater diameter. Hence,if with an increase in current,the diameter of crater is affected more than the depth of crater then the effect of increased current would not be observed clearly on the R a values.This can further be clari?ed if studies are made on the effect of parameters such as current and voltage on the discharge plasma channel diameter.

Table2(continued)

Run order Input parameters Output response

V g(V)I d(A)T on(m s)D(%)P(kpa)N(rpm)MRR(mm3/min)R a(m m)TWR(mm3/min)

747729500481471275 1.9930.02 7563422007288.2650 4.19 3.69à0.03 7663422002488.2650 1.21 3.070.02 7791162007288.21900 1.41 3.2à0.02 787729500481473001 3.490.01 7963427502488.21900 2.27 3.420.03 8091422007288.21900 4.19 3.810.01 81631675024205.819000.27 3.240.03 8291422002488.21900 1.81 3.080 8363167502488.219000.44 4.030.01 84631675024205.86500.44 2.86à0.02 85631620072205.81900 1.19 2.840.02

86631675072205.819000.79 3.5

0.01

Fig.4.Effect of spindle speed on MRR and R a

.Fig.5.Effect of air inlet pressure on MRR and R a

.

Fig.6.Effect of discharge current on MRR and R a.

S.K.Saha,S.K.Choudhury/International Journal of Machine Tools&Manufacture49(2009)297–308 302

4.4.Effect of voltage

The effect of gap voltage (V g )on MRR and R a is shown in Fig.7.Initially MRR increases with an increase in voltage but an optimum exists and the MRR drops with further increase in voltage.A similar trend was observed for R a values which decrease up to an optimum point and then increase with voltage.Since spark energy is proportional to the gap voltage,increase in voltage would lead to a higher MRR.The decrease in R a values with an increase in voltage suggests that larger but shallower craters are formed at higher voltage values due to expansion of the plasma channel in the discharge gap.Discharge takes place when the effective electric ?eld (?gap voltage/inter-electrode distance)between the electrodes exceeds the dielectric strength of the medium.Hence,with an increase in the gap voltage the discharge gap distance increases and the breakdown electric ?eld can now be achieved even at a larger gap distance.The effective gas velocity at the workpiece surface is lower when the gap distance is high.Thus,?ushing ef?ciency reduces and the probability of arcing increases due to the presence of debris in the tool-workpiece gap.Due to partial removal of debris from the discharge gap,low MRR and a high R a value was obtained at very high voltages.Thus an optimum value of voltage exists at which high MRR and low R a value is obtained.This optimum may depend on the levels of other parameters and can be studied in the later stage designed experiments.

4.5.Effect of pulse-on time

The effect of pulse-on time (T on )on MRR and R a is shown in Fig.8.It can be seen that MRR and R a values both increase with an increase in T on .For a higher T on ,deeper discharge craters are formed and more material is removed per spark since spark energy is directly proportional to T on .This leads to higher R a values.Also,for a rotating tool spark frequency may not be determined by the duty factor alone.Hence,in spite of a constant value of duty factor the MRR increases with T on .For very large values of T on ,the drop in MRR can be explained by the high values of pulse-off time.Since the duty factor was held constant during the experiment,a higher T off value was obtained corresponding to a higher T on value.No material removal occurs during the T off .Hence,a high value of T off increases non-cutting time and reduces the MRR.This non-cutting time however does not have a signi?cant effect on the R a value.

4.6.Effect of pulse-off time

The effect of pulse-off time (T off )on MRR and R a is shown in Fig.9.Since direct control of T off was not available on the EDM machine,duty factor (D )was changed while T on was held constant to change the T off .For a very small value of T off ,the MRR was low but then increased drastically when T off was increased.The MRR then dropped slowly with an increase in T off .For very short pulse-off time,the probability of arcing is high because the dielectric in the gap may not have completely recovered its dielectric strength.Also,debris particles may still remain in the discharge gap.This would lead to a low MRR and high R a value.When T off is suf?ciently high,the dielectric regains its dielectric strength and the debris particles are also ?ushed away from the gap.Thus,a drastic increase in MRR and decrease in R a value is achieved.With further increase in T off ,MRR decreased slowly because machining does not take place during the pulse-off time and it only adds to the non-cutting time.Further advantage due to dielectric strength recovery was not available;hence increase in T off lead to a decrease in MRR.The effect of increased T off on R a values cannot be explained by the above phenomenon as according to it the R a values should have remained unaffected for large values of T off .One of the reasons for the observed trend in R a values could be the effect of T off on the re-solidi?cation of molten workpiece at the discharge crater.Molten material may re-solidify when long pulse-off times are used due to continuous supply of high velocity air through the tool.Also,air ?ow may distort the distribution of molten material in the crater leading to a poorer surface ?nish when the material solidi?es.Further experiments are necessary to clarify this.

Fig.7.Effect of gap voltage on MRR and R a

.

Fig.8.Effect of pulse-on time on MRR and R a

.

Fig.9.Effect of pulse-off time on MRR and R a .

S.K.Saha,S.K.Choudhury /International Journal of Machine Tools &Manufacture 49(2009)297–308

303

https://www.wendangku.net/doc/a411631056.html,D experiments:results and discussion 5.1.Regression analysis

On analysis of the experimental data obtained from the central composite design (CCD)runs (Table 2),models with signi?cant factor effects were obtained for MRR and R a .However,for TWR no suitable model could be obtained which had a signi?cant effect.Regression statistics for the ?tted models of MRR,R a and TWR are

shown in Table 3.Regression analysis for each response is discussed in this section.

5.1.1.Regression analysis for MRR

A two-factor interaction model was found to be the most suitable model based on the analysis of variance (ANOVA)sequential sum of squares test.The ANOVA table for the reduced two-factor interaction MRR model is shown in Table 4.The Model F -Value of 59.39implies the model is signi?cant.There is only a 0.01%chance that a ‘‘Model F -Value’’this large could occur due to noise.Values of ‘‘Prob 4F ’’less than 0.0500indicate model terms are signi?cant.In this case I d ,D ,P ,N ,V g I d ,V g T on ,I d D ,I d P,I d N and T on D were signi?cant model terms.Values greater than 0.1000indicate the model terms are not signi?cant.The ‘‘Lack of Fit F -value’’of 8.09implies the ‘‘Lack of Fit’’is signi?cant.A signi?cant lack of ?t is bad because it is desirable that the model ?ts.Where as a whole model test checks if anything in the model is signi?cant,a lack-of-?t test checks whether anything left out of the model is signi?cant.If the lack-of-?t test is signi?cant,there is

Table 4

Analysis of variance table for response surface reduced two-factor interaction model of MRR.Source Sum of squares (SS)Degrees of freedom Mean square F value p -Value Prob 4F Test result Model

138.6591

12

11.55492

59.39111

o 0.0001

Signi?cant

V g 0.00011910.0001190.0006120.9803I d 91.85232191.85232472.1114o 0.0001T on 0.04009910.0400990.2061030.6512D 17.17029117.1702988.25352o 0.0001P 1.92981 1.92989.9189740.0024N 11.42346111.4234658.71541o

0.0001V g I d 0.92147710.921477 4.7362960.0328V g T on 0.80758610.807586 4.1509070.0452I d D 6.2580151 6.25801532.16555o 0.0001I d P 1.1563141 1.156314 5.9433360.0172I d N 3.642961 3.6429618.72444o 0.0001T on D

3.5737461 3.57374618.36868

o

0.0001

Residual SS 14.20262730.194556Lack of ?t 13.96008640.2181268.0941470.0010Signi?cant

Pure error

0.24253890.026949

Corrected total SS

152.861785

Table 5

Analysis of variance table for response surface reduced quadratic model of R a .Source Sum of squares (SS)Degrees of freedom Mean square F value p -Value Prob 4F Test result Model

10.500

10

1.049973

13.332

o 0.0001

Signi?cant

V g 0.10110.100989 1.2820.2611I d 1.9971 1.99736425.361o 0.0001T on 1.9681 1.96793124.987o 0.0001D 0.44110.441253 5.6030.0205P 4.0851 4.08480651.866o 0.0001N 0.58610.5858067.4380.0079V g I d 0.30810.308025 3.9110.0516T on P 0.30610.306311 3.8890.0523V g 20.47210.472464 5.9990.0166T on

20.64010.6400058.1260.0056

Residual SS 5.907750.078758Lack of ?t 5.362660.081243 1.3420.3336Not signi?cant

Pure error

0.54590.060534

Corrected total SS

16.40785

Table 3

Regression statistics for ?tted models of MRR,Ra and TWR.Statistics

Model

MRR reduced 2FI

R a reduced quadratic TWR linear R 2

0.9070.6400.016Adjusted R 20.8920.592à0.058Predicted R 2

0.869

0.523

à0.187

S.K.Saha,S.K.Choudhury /International Journal of Machine Tools &Manufacture 49(2009)297–308

304

some signi?cant effect that has been left out of the model,and that effect is a function of the factors already in the model.It could be a higher order power of a factor or some form of interaction among the factors.In this case,one of the reasons for the signi?cant lack of?t could be due to the inability to represent the interaction terms which include the T off variable.Since T off could not be controlled independently(but only in terms of T on and duty factor) on the EDM machine,it was not possible to include the interaction terms of the form V g T off,y,NT off in the models.If any of these interaction terms are important,then the?tted model would not be able to capture it and would show a signi?cant lack of?t.

The?nal regression equation for MRR in terms of the actual parameter values is

MRR?à1:04901t0:00536V gt0:02829I dà0:00044T on t0:00864Dà0:00226Pà0:0002Nà0:00066V g I d

t0:00003V g T ont0:001I d Dt0:00018I d P

t0:00003I d Nà0:00004T on D(1) where MRR is in mm3/min and V g in V,I d in A,T on in m s,P in kPa,N in rpm and D is dimensionless duty factor in%.

5.1.2.Regression analysis for R a

The ANOVA table for the reduced quadratic model for R a is shown in Table5.The Model F-Value of13.33implies the model is signi?cant.There is only a0.01%chance that a‘‘Model F-Value’’this large could occur due to noise.Values of‘‘Prob4F’’less than 0.0500indicate model terms are signi?cant.In this case I d,T on,D, P,N,V g2and T on2were signi?cant model terms.The‘‘Lack of Fit F-value’’of1.34implies the Lack of Fit is not signi?cant relative to the pure error.

Fig.10.Response surface of MRR versus:(a)gap voltage and discharge current,(b)discharge current and duty factor,(c)discharge current and air pressure and(d) discharge current and spindle speed.

S.K.Saha,S.K.Choudhury/International Journal of Machine Tools&Manufacture49(2009)297–308305

The?nal regression equation for R a in terms of the actual parameter values is

R a?7:95767à0:13375V gà0:01624I dt0:00334T ont0:00332D à0:00211Pà0:00015Nt0:00038V g I dà0:000004T on P

t0:00081V2

g à0:000002T2

on

(2)

where R a is in m m and V g in V,I d in A,T on in m s,P in kPa,N in rpm and D is dimensionless duty factor in%.

5.1.3.Regression analysis for TWR

Two factor interaction model was suggested on performing an ANOVA-based sequential sum of squares test.However,the predicted R2was found to be negative(Table3).A negative value of predicted R2implies that the overall mean is a better predictor of response than the model.Various models were chosen for ?tting and it was found that the predicted R2was negative in all the cases.Hence,instead of using a regression equation,TWR has been estimated by the mean value:

TWR?0:01mm3=min(3) The near-constant value of TWR can be explained by the deposition of debris material on the tool face after machining.

When high MRR conditions exist,large amount of debris is formed and the amount of material deposited on the tool increases.Under such conditions,the amount of actual material removed from the tool also increases.Similarly,for low MRR conditions the amount of debris deposited on tool and amount of actual tool material removed is also low.Since TWR was calculated by measuring the difference in tool weight before and after machining,the combined effect of deposition and removal of material from tool was included in the measured TWR.The constant value of TWR indicates that the difference between material removed from tool and material deposited on tool remains constant even when the process parameters are changed.

5.2.Response surface analysis

5.2.1.MRR response surface

The response surfaces of MRR were obtained for the interaction terms in the reduced two-factor interaction model.Response surface of MRR versus gap voltage and discharge current is shown in Fig.10(a).From the?gure it can be seen that a high current and voltage combination leads to high MRR due to an increase in the spark energy under such conditions.As seen in the?gure,at low current values MRR increases with an increase in voltage due to an increase in the spark energy.However,at high-current levels, an increase in voltage leads to a slight decrease in the MRR.One of the reasons for this could be the higher amount of debris formation and higher?ushing requirements at high-current levels.Since,discharge gap increases with an increase in voltage (Section4.4),?ushing ef?ciency is reduced at high voltages.The increase in spark energy is dominated by the reduction in?ushing ef?ciency as voltage is increased.This leads to a reduction in MRR as voltage is increased at high-current levels.

Response surface of MRR versus discharge current and duty factor is shown in Fig.10(b).From the?gure it can be observed that MRR increases with an increase in duty factor and current. Similarly,from Fig.10(c)and(d)it can be observed that high MRR is obtained at high current and high air inlet pressure combina-tion and high current and high spindle speed combination.High values of air pressure and spindle speed lead to a better?ushing ef?ciency which improves the MRR.MRR increases on increasing any of the four factors(I d,D,P,N),but it can be seen that I d has the highest effect on MRR.The effect of gap voltage depends on the current level(i.e.on the amount of debris formed).

The effect of pulse-on time is not straightforward as can be seen from the response surface of MRR versus T on and D(Fig.11). For high values of duty factor,an increase in T on leads to a decrease in the MRR where as for low values of duty factor,an increase in T on leads to an increase in MRR.A possible reason for this could be that for high-duty factors,the pulse-off time is low.When a high T on is used,the amount of material which melts during the spark increases due to higher spark energy.However,due to extremely low pulse-off times,a substantial part of the material remains in the spark gap instead of being?ushed away from the gap.With an increase in T on,the amount of debris particles in the spark gap is expected to increase.This would lead to a higher arcing probability and reduce the MRR.On the other hand,high pulse-off time is obtained at low duty factor and suf?cient time for ?ushing of debris from the gap is available in between the sparks. Hence,when a high T on is used more material is melted and the material is also removed due to?ushing.Thus MRR increases on increasing T on at low duty factors.

5.2.2.R a response surface

The response surfaces of R a were obtained for the interaction terms in the reduced quadratic model.Response surface of R a versus gap voltage and discharge current is shown in Fig.12(a). From the?gure it can be seen that the R a value decreases with a decrease in the current.However,an optimum voltage exists for minimum R a.Also,this optimum point depends on the value of the current.The existence of optimum voltage can be explained as in the case of OVAT analysis(Section4.4).The optimum voltage shifts towards higher values as the current decreases.At low currents,the amount of debris formed is lower because of the lower spark energy.Thus the need for?ushing is not as signi?cant at lower currents as in the case of higher currents.At lower voltage values the?ushing ef?ciency is improved due to lower spark gap.Hence for obtaining the same R a value at higher current levels,lower voltage values are required.Thus,the optimum voltage value decreases with an increase in the current value.

Response surface of R a versus pulse-on time and air inlet pressure is shown in Fig.12(b).Corresponding to each air pressure level,a T on value exists at which the R a value is highest and R a decreases on either side of this T on value.For very low values of T on,high-frequency sparks take place leading to shallower crater formation and subsequently a lower R a value.For very large values

Fig.11.Response surface of MRR versus pulse-on time and duty factor.

S.K.Saha,S.K.Choudhury/International Journal of Machine Tools&Manufacture49(2009)297–308 306

of T on,the T off values also increase(since the duty factor is held constant).Thus,more time is available for debris removal leading to an improvement in?ushing ef?ciency.This leads to a lower R a value.The T on corresponding to highest R a value depends on the value of air pressure.At high pressure values,better?ushing conditions exist.Hence even for low T off(corresponding to low T on at constant D)suf?cient?ushing ef?ciency is available.Thus,the T on for highest R a reduces with an increase in air pressure(as seen on the contour plot in Fig.12(b)).

6.Conclusion

In the present work,parametric analysis of the dry EDM process has been done based on experimental results.Experi-ments based on the central composite design(CCD)were conducted to develop empirical models of the process.Following conclusions can be drawn from the analysis of the results:

(a)From the preliminary experiments it was found that EDM

with air as the dielectric is feasible with reverse polarity.

However,high velocity air?ow into the inter-electrode gap through a hollow tubular tool electrode and rotation of the tool are necessary conditions for obtaining a reasonable MRR.

(b)Flow characteristic of air in the inter-electrode gap affects the

MRR and the surface roughness(R a).There exists an optimum number of air-?ow holes(in the tool)for which the MRR is highest and the R a is lowest.

(c)From the designed set of experiments based on CCD it was

found that discharge current,duty factor,air pressure and spindle speed are the signi?cant factors which affect MRR and MRR increases with an increase in any of these factors.On an average,there was an increase by350%in MRR when current was increased from the low level(9A)to the high level(42A).

Similarly,MRR increased by about76%when duty factor was increased from24%to72%,by about21%when air pressure was increased from88.2kPa to205.8kPa and by about59% when spindle speed was increased from650rpm to1900rpm.

For MRR,most signi?cant two-factor interaction effects were present among current and duty factor,current and spindle speed and pulse-on time and duty factor.(d)From CCD experiments it was found that except for gap

voltage all other input parameters(discharge current,pulse-on time,duty factor,air pressure and spindle speed)have signi?cant effect on R a.R a values decrease with a decrease in the values of current and duty factor.Also,R a values decrease with an increase in the values of air pressure and spindle speed.On an average,there was a decrease by about10%in R a when the current was decreased from42A to9A and a decrease by about5%when duty factor was decreased from 72%to24%.Similarly,R a decreased by about13%when air pressure was increased from88.2kPa to205.8kPa and by about5%when spindle speed was increased from650rpm to 1900rpm.No signi?cant two-factor interactions were found for R a.However,R a was found to have a quadratic dependence on gap voltage and pulse-on time.

(e)During the CCD experiments,the tool wear rate(TWR)was

found to be very small(less than about1%of the MRR).It was found that the TWR was independent of the input parameters. References

[1]F.N.Lea?o,I.R.Pashby,A review on the use of environmentally-friendly

dielectric?uids in electrical discharge machining,Journal of Materials Processing Technology149(1–3)(2004)341–346.

[2]V.Ramani,M.L.Cassidenti,Inert-Gas Electrical Discharge Machining,NASA

Technical Brief Number NPO15660,1985.

[3]M.Kunieda,S.Furuoya,N.Taniguchi,Improvement of EDM ef?ciency by supplying

oxygen gas into gap,CIRP Annals—Manufacturing Technology40(1991)215–218.

[4]M.Kunieda,M.Yoshida,N.Taniguchi,Electrical discharge machining in gas,

CIRP Annals—Manufacturing Technology46(1997)143–146.

[5]M.Kunieda,Y.Miyoshi,T.Takaya,N.Nakajima,Y.Z.Bo,M.Yoshida,High speed

3D milling by dry EDM,CIRP Annals—Manufacturing Technology52(2003) 147–150.

[6]Z.Yu,T.Jun,K.Masanori,Dry electrical discharge machining of cemented

carbide,in:14th International Symposium on Electromachining(ISEM XIV), Journal of Materials Processing Technology149(1–3)(2004)353–357.

[7]Y.Zhanbo,J.Takahashi,N.Nakajima,S.Sano,K.Karato,M.Kunieda,Feasibility

of3-D surface machining by dry EDM,International Journal of Electrical Machining10(2005)15–20.

[8]M.Kunieda,T.Takaya,S.Nakano,Improvement of dry EDM characteristics

using piezoelectric actuator,CIRP Annals—Manufacturing Technology53(1) (2004)183–186.

[9]Q.H.Zhang,J.H.Zhang,S.F.Ren,J.X.Deng,X.Ai,Study on technology of

ultrasonic vibration aided electrical discharge machining in gas,in:14th International Symposium on Electromachining(ISEM XIV),Journal of Materials Processing Technology149(1–3)(2004)640–644.

Fig.12.Response surface of R a versus:(a)gap voltage and discharge current and(b)pulse-on time and air pressure.

S.K.Saha,S.K.Choudhury/International Journal of Machine Tools&Manufacture49(2009)297–308307

[10]Q.H.Zhang,R.Du,J.H.Zhang,Q.B.Zhang,An investigation of

ultrasonic-assisted electrical discharge machining in gas,Inter-national Journal of Machine Tools and Manufacture46(12–13)(2006) 1582–1588.

[11]M.Kunieda,C.Furudate,High precision?nish cutting by dry WEDM,CIRP

Annals—Manufacturing Technology50(1)(2001)121–124.

[12]C.C.Kao,J.Tao,S.Lee,A.J.Shih,Dry wire electrical discharge machining of

thin workpiece,in:Transactions of the North American Manufacturing Research Institute of SME34,2006,pp.253–260.[13]J.Tao,A.J.Shih,J.Ni,Experimental study of the dry and near-dry electrical

discharge milling processes,Journal of Manufacturing Science and Engineer-ing130011002(2008),DOI:10.1115/1.2784276.

[14]S.K.Saha,S.K.Choudhury,Modular unit attachment for performing dry and

near-dry electric discharge machining(EDM),Indian Patent Pending, Application no.2032/DEL/2008,2008.

[15]R.H.Myers, D.C.Montgomery,Response Surface Methodology:Process and

Product Optimization Using Designed Experiments,second ed.,Wiley,New York, 2002.

S.K.Saha,S.K.Choudhury/International Journal of Machine Tools&Manufacture49(2009)297–308 308

相关文档