Vous êtes sur la page 1sur 24

1021 O. B. Abouelatta, Prediction of machining operations and surface roughness using artificial neural network, pp.

1021- 10

PREDICTION OF MACHINING OPERATIONS AND SURFACE ROUGHNESS USING ARTIFICIAL NEURAL NETWORK
O. B. Abouelatta
Production "ngineering and +echanical /esign /epartment, 'acult& of "ngineering, +ansoura $ni%ersit&, *0011 +ansoura, "g&pt.

Received 11 February 2013, accepted 27 March 2013 ABSTRACT


Surface roughness is considered as one of the most specified customer requirements in machining processes For efficient use of machine too!s, se!ection of machining process and determination of optima! cutting parameters "speed, feed and depth of cut# are required $herefore, it is necessary to find a suitab!e %ay to se!ect and to find optima! machining process and cutting parameters for a specified surface roughness va!ues &n this %or', machining process %as carried out on (&S& 10)0 stee! in dry cutting condition in a !athe, mi!!ing and grinding machines and surface roughness %as measured Forty five e*periments have been conducted using varying speed, feed, and depth of cut in order to find the surface roughness parameters $his data has been divided into t%o sets on a random basis+ 3, training data set and - testing data set $he training data set has been used to train different artificia! neura! net%or' "(..# mode!s in order to predict machining processes and surface roughness parameter va!ues through bac' propagation net%or' /*perimenta! data co!!ected from tests %ere used as input parameters of a neura! net%or' to identify the sensitivity among machining operations, cutting parameters and surface roughness Se!ected inde*es %ere used to design a suitab!e a!gorithm for the prediction of machining processes ( soft%are %as deve!oped and imp!emented to predict the machining processes and surface roughness va!ues $he resu!ts sho%ed that the proposed mode!s are capab!e of predicting machining operations, cutting parameters and surface roughness 2e&words, Machining operations, Surface roughness, .eura! net%or', 0rediction

1. Int o!u"t#on
&ncreased demands for higher product qua!ity, re!iabi!ity, and manufacturing efficiency !eve!s have imposed stringent requirements on automated product measurement and eva!uation $he progress in the deve!opment of predictive mode!s, based on cutting theory, has not yet met the ob1ective of maintaining high qua!ity production at !o% cost $he most essentia! cutting performance measures, such as, too! !ife, cutting force, roughness of the machined surface, energy consumption, 2 etc , shou!d be defined using e*perimenta! studies $herefore, further improvement and optimi3ation for the techno!ogica! and economic performance of machining operations depend on a %e!!4based e*perimenta! methodo!ogy 5aya1neh et a! 617, aimed to deve!op a better understanding of the effects of cutting parameters on the surface roughness and to bui!d a mu!tip!e regression mode! Suhai! et a! 627, proposed a method for cutting parameters identification using mu!ti adaptive net%or' based fu33y inference system "M(.F&S# $heir resu!ts sho%ed that the M(.F&S mode! can be used successfu!!y for machinabi!ity data se!ection
!ournal of "ngineering #ciences, Assiut $ni%ersit&, 'acult& of "ngineering, (ol. 1, )o. *, +a&, 201*, "-mail address, -es.aun.edu.eg

1022 O. B. Abouelatta, Prediction of machining operations and surface roughness using artificial neural network, pp. 1021- 10

$o dea! %ith a non!inear optimi3ation prob!em %hich aims to minimi3e the unit production cost in mu!ti4pass turnings, 8ie and 9uo 637 proposed a ne% approach, %hich combines genetic a!gorithms %ith a pass enumerating method :omputer simu!ation resu!ts sho%ed that the optimi3ation approach can find the better resu!ts than other a!gorithms proposed previous!y to significant!y reduce the unit production cost 9r3esi' and ;anat 6)7 provided a comprehensive ana!ysis of part surface finish in continuous dry turning of hardened construction stee! %hen using mi*ed a!umina cutting too!s $heir resu!ts sho%ed that hard turning %ith mi*ed a!umina too!s produces specific surface profi!es and microstructures, a!though the Ra roughness va!ues of about 0 2< =m can be comparab!e to those produced by finish grinding Surface roughness has been mode!ed by 0atri'ar 6<7, based on po%er series and Fast Fourier $ransform methods ( method based on neura! net%or's has been used to mode! these surfaces to map the process parameters to roughness parameters &n recent years, the trends %ere to%ards mode!ing of machining using artificia! inte!!igence (rtificia! neura! net%or's are considered one of the important methods of artificia! inte!!igence in the mode!ing of non!inear prob!ems !i'e machining processes (.. sho% good capabi!ity in prediction and optimi3ation of machining processes compared %ith traditiona! methods &n turning operations, >uong and Spedding 6,7 described the app!ication of neura!4net%or' techno!ogy to the se!ection of machining parameters and to the prediction of machining performance in meta! cutting via an on4!ine imp!ementation of the trained net%or' using the : programming !anguage ( neura!4net%or'4based methodo!ogy %as proposed by ?oh!i and @i*it 677, for predicting the surface roughness in a turning process by ta'ing the acce!eration of the radia! vibration of the too! ho!der as feedbac' $hey stated that their methodo!ogy %as ab!e to ma'e accurate prediction of surface roughness by uti!i3ing sma!! si3ed training and testing datasets (.. and mu!tip!e regression approaches %ere used by (si!tAr' and Bun'aC 6D7, to mode! the surface roughness of (&S& 10)0 stee! $he (.. mode! %as used to estimate the surface roughness %ith high accuracy compared to the mu!tip!e regression mode! Mode!s for predicting surface roughness of (&S& 10)0 stee! materia! using artificia! neura! net%or's and mu!tip!e regression %ere deve!oped by (si!tAr' 6-7 Mean squared error of 0 002-2E achieved using the deve!oped (.. outperforms error rates reported in ear!ier studies and cou!d a!so be considered admissib!e for rea!4time dep!oyment of the deve!oped (.. a!gorithm for robust prediction of the surface roughness in industria! settings &n mi!!ing operations, mu!tip!e regression and (.. techniques %ere app!ied by Murthy and Ra1endran 6107, to predict the surface roughness $he resu!ts of the prediction mode!s %ere quite c!ose %ith e*periment va!ues $hey reported that the feed rate %as the most dominant factor in inf!uencing surface roughness $he resu!ts a!so sho%ed that the highest cutting speed, medium feed rate and medium depth of cut produces !o%est surface roughness (n empirica! approach using a statistica! ana!ysis %as emp!oyed by 5uang and :hen 6117, to discover the proper cutting force in order to represent the uncontro!!ab!e

!ournal of "ngineering #ciences, Assiut $ni%ersit&, 'acult& of "ngineering, (ol. 1, )o. *, +a&, 201*, "-mail address, -es.aun.edu.eg

1023 O. B. Abouelatta, Prediction of machining operations and surface roughness using artificial neural network, pp. 1021- 10

factors in end mi!!ing operations using an in4process neura! net%or'4based surface roughness prediction system Response surface mode! and an artificia! neura! net%or' %ere deve!oped by /r3urum!u and F'tem 6127, to predict surface roughness va!ues error on mo!d surfaces $he response surface mode! and an artificia! neura! net%or' %ere compared %ith manufacturing prob!ems such as computationa! cost, cutting forces, too! !ife, dimensiona! accuracy, etc ( ne% technique %as deve!oped by ;ibo%o and @esa 6137, using hybridi3ation of 'erne! principa! component ana!ysis based on non!inear regression and 9(s to estimate the optimum va!ues of surface roughness $heir resu!ts sho%ed that the presented technique gave more accurate prediction mode! than the ordinary !inear regressionGs approach $he need for precision components and parts in manufacturing industries has bought an increase in the need for finishing operations that can satisfy this demand 9rinding has the potentia! to meet these critica! needs for accurate and economic means of finishing parts, and generate the required surface topography (n approach suggesting the combination of design of e*periment method and (.. %as deve!oped by Fred1 and (mamou 61)7 $he bui!t (..s sho%ed !o% deviation from the training data, !o% deviation from the testing data and high sensibi!ity to the inputs !eve!s $he high prediction accuracy of the deve!oped (..s %as confirmed by the good agreement %ith the resu!ts of empirica! mode!s deve!oped by previous investigations Mu'her1ee and ?umar Ray 61<7 attempted to provide a systematic methodo!ogy to deve!op a mu!tivariate !inear regression mode!, hypothesis testing for the inf!uence of non!inear terms to !inear mode!, and according!y se!ection of a suitab!e artificia! neura! net%or'4based inferentia! mode! %ith improved prediction accuracy and contro! of grinding behavior $he deve!opment of neura! mode!4based contro! strategies for the optimi3ation of an industria! a!uminum substrate dis' grinding process %as described by 9ovindhasamy et a! 61,7, using a non!inear autoregressive e*ogenous based neura! net%or' mode! 0re!iminary p!ant investigations sho% that thic'ness defects can be reduced by <0E or more, compared to other schemes emp!oyed ?umar and :houdhury 6177 focused on the prediction of %hee! %ear and surface roughness using t%o techniques, name!y design of e*periments and neura! net%or' /ffect of process parameters, such as pu!se current, duty ratio, %hee! speed and grain si3e on output responses, name!y, %hee! %ear and surface roughness of high speed stee! %ere investigated e*perimenta!!y $his paper aims to reduce the huge number of t%o dimensiona! surface roughness parameters, to propose some inde*es suitab!e for the prediction of machining processes and machining parameters based on surface roughness signature ana!ysis using neura! net%or'

$. E%&e #'ental (etu& an! & o"e!u e 2.1. /esign of e3periment


/*periments have been performed in order to investigate the effects of machining parameters "speed, feed and depth of cut# on the surface finish of the machined surface
!ournal of "ngineering #ciences, Assiut $ni%ersit&, 'acult& of "ngineering, (ol. 1, )o. *, +a&, 201*, "-mail address, -es.aun.edu.eg

102) O. B. Abouelatta, Prediction of machining operations and surface roughness using artificial neural network, pp. 1021- 10

Forty five specimens made of (&S& 10)0 stee!, )< mm diameter %ith 10 mm in height, %ere used for e*perimentation using a turning "facing#, mi!!ing and grinding machines :hemica! composition and mechanica! properties of (&S& 10)0 stee! are !isted in $ab!e 1 "a# and "b#, respective!y $he specimens %ere divided into three groups according to the machining operation+ each has 1< specimens "12 for training and 3 for eva!uation# Hefore emp!oying machining process, each specimen %as po!ished using sandpapers grade 1200 to e!iminate the effect of previous cuts (!! e*periments have been done under dry machining environment Severa! variab!es %ere put under c!ose contro! inc!uding the machining condition "the same machine %as used for a!! e*perimenta! %or' for each group#, and the operator "the same operator machined a!! specimens for each machine#

Table 1. )a* :hemica! composition of (&S& 10)0 stee!


:hemica! composition ";t E# :arbon 0 37 &ron -D D Manganese 07 0hosphorus 0 03< Su!phur 0 0)< Fthers 0 0<

Table 1. )b* Mechanica! properties of (&S& 10)0 stee!


Iie!d tensi!e strength 4&ield "M0a# )<0 Mechanica! properties J!timate tensi!e strength Kic'ers 5ardness 4ultimate "M0a# "5K# <1< 1<< Modu!us of /!asticity "90a# 200

$he first group %as produced by !athe turning machine "Facing# under different cutting parameters ( cemented carbide cutting too! equipped %ith thro%a%ay inserts %as used in the e*periments $he insert type %as ?09. 1,0)12 and the shan' type %as MS?.RL> 2<2<M1, for approach ang!e M3<N $he second group %as produced by a hori3onta! mi!!ing machine ( 34inch " D0 mm outer diameter#, 1)4f!ute high4speed stee! cutters and a D0 mm %idth %as used in the e*periments $he third group %as machined by a surface grinder machine %ith e!ectromagnetic chuc' $he used cutting too! for grinding operations is a!uminum o*ide grinding %hee!s "3D(,0??<KH/#

2.2. #urface roughness measurements


$here is an industria! need for the measurement and c!assification of the topography of engineering surfaces $%o4dimensiona! "2@# ana!ysis is fast, but !imited in the usefu!ness of the resu!ts obtainab!e %hi!st the three4dimensiona! "3@# approach offers greater scope Jnfortunate!y, the data ana!ysis step for 3@ data characteri3ation can be time4consuming, and often requires a s'i!!ed metro!ogist 61D7

!ournal of "ngineering #ciences, Assiut $ni%ersit&, 'acult& of "ngineering, (ol. 1, )o. *, +a&, 201*, "-mail address, -es.aun.edu.eg

102< O. B. Abouelatta, Prediction of machining operations and surface roughness using artificial neural network, pp. 1021- 10

&n this investigation, a set of 2@ surface roughness parameters, as %e!! as profi!e and surface characteristics, such as the amp!itude distribution functions, bearing area curves and symmetrica! curves of geometrica! contact obtained for the machined surface, %ere determined and ana!y3ed $he surface roughness data %ere co!!ected for each of the )< samp!es for the machining conditions as defined in $ab!es 24) ( surface roughness measuring system consists of t%o ma1or parts, hard%are and soft%are %as used $he hard%are consists of t%o main items as sho%n in Fig 1 Mitutoyo Surf$est4SO201 is a surface measuring instrument, used in contact method surface assessment Figures 2 and 3 sho% t%o specia! programs %ritten in Mat!ab$M, Prediction of +achining Process and 5utting Parameters 6P+P5P7 and #urface 8oughness 5alculation Program 6#85P7 %ere used to give a fu!! assessment of surface roughness parameters from the resu!ted surface profi!e, respective!y $he P+P5P program is used to measure, import and e*port surface profi!e and surface roughness data from Mitutoyo Surf$est4SO201 Hy c!ic'ing the P8un #85PP button, the #85P program can be invo'ed to ca!cu!ate more than ,, standard and non4standard surface roughness parameters as sho%n in $ab!e < 61-7 $hree measurements for %or'piece surface roughness %ere made and averaged for each test Figure ) sho%s surface te*ture and surface profi!e for some machining operations

Table $. :utting and measured roughness parameters turning operation


.o 1 2 3 ) < , 7 D 10 11 12 13 1) 1<
9 99

:utting parameters9 f s dc mm:re% rpm mm 0 1, 1)0 03 0 1, 22) 03 0 1, <,0 03 0 1, -00 03 0 1, <,0 01 0 1, <,0 0 7< 0 1, <,0 1< 0 0D <,0 03 0 23 <,0 03 0 32 <,0 03 0 ), <,0 03 0 <<,0 03 0 2, 710 1 2< 0 32 1D0 1 2< 0 <2 )<0 1 7<

Surface roughness parameters99 8a # Pc ;m mm 5ount 7 0< 112 2 72 ) ,7- 3 101 , 77 11< ) 70 3 2< 10< ) D3 D 01 11, 0 71 < 00 121 0 7, 2 -) 7< 107 , 00 111 7) ) )1 D- 0 -) , <113 , 71 , 3D -< 1 D2 - <10, D 7< < 22 120 0 ,D < -1 -0 3 D, 03 -< 2 D)

Surface inde*es #:8a Pc:8a 5ount:;m mm:;m 10 2 1< 21 < 1, 10 3 17 1 2< ) 32 ) D1) < 1< 3 2) 2 3, 3 2< D 12 3 1D 7 21 3 20 2 10 D 17 2 12 1) 711 1 13 0 23 0 1< 0 1< 3 13 1< D

f, s, and dc are feed, rotational cutting speed and depth of cut, respecti%el&. 8a, #, and Pc are arithmetic roughness height, mean spacing of ad-acent local peaks, and peak count, respecti%el&.

!ournal of "ngineering #ciences, Assiut $ni%ersit&, 'acult& of "ngineering, (ol. 1, )o. *, +a&, 201*, "-mail address, -es.aun.edu.eg

102, O. B. Abouelatta, Prediction of machining operations and surface roughness using artificial neural network, pp. 1021- 10

Table +. :utting and measured roughness parameters mi!!ing operation


.o 1 2 3 ) < , 7 D 10 11 12 13 1) 1<
9 99

:utting parameters9 f s dc mm:min rpm mm 2D 1<7 02 << 1<7 02 1<7 1<7 02 110 1<7 02 1)0 1<7 02 110 D0 02 )1 1<7 0 )< 110 D0 02 110 D0 01 110 1<7 0 3< 110 1<7 0 2< 110 1<7 03 2D0 )30) 20 )0 03 110 1<7 0 1<

Surface roughness parameters99 8a # Pc ;m mm 5ount 3 )1 120 7 ,7 3 73 127 3 ,) ) 1< 13< , ,3 < 23 1,< ) <3 3 <D 12, 0 ,7 3 01)- D <7 3 ,, 17- 3 )7 3 07 1)2 , <2 -121 2 ,7 < 11 1,< < <0 < 0, 1)7 1 ,2 3 71 1<0 2 <) , 120< , )0 , 3D 1<D ,0 3 -3 123 < ,<

Surface inde*es Pc:8a #:8a 5ount:;m mm:;m 1- D 3< ) 17 2 3) 1 1< 1 32 7 10 2 31 , 1D , 3< 2 ,< 33 2 1D ) )D , -) 2) 12 D )- 0 1- 2 ), ) 22 3 )0 , -7 32 ) 12 2 2- 1 1) < )0 < 1, < 31 <

f, s, and dc are feed, rotational cutting speed and depth of cut, respecti%el&. 8a, #, and Pc are arithmetic roughness height, mean spacing of ad-acent local peaks, and peak count, respecti%el&.

Table ,. :utting and measured roughness parameters grinding operation


.o 1 2 3 ) < , 7 D 10 11 12 13 1) 1<
9 99

:utting parameters9 (w (c ae m:min m:s ;m , 3< 20 , 3< 30 , 3< )0 , 3< <0 , 20 20 , 20 30 , 20 )0 , 20 <0 D) 20 20 72 3< 20 )D 3< 20 3, 20 20 10 20 20 )0 3< ,0 < 3< 10

Surface roughness parameters99 8a # Pc ;m mm 5ount 0 <7 <3 D 1,) 0 D3 )- 3 1,, 0 7, ), 1 17, 0 D) <1 < 1<7 0 -0 <- 7 13, 1 00 <3 1 1<< 0 70 )- 1,3 0 ,<0 < 1,D 0 ,) ), 7 17D 0 ,, )7 D 1,0 73 )2 7 1-2 0 ,< )D 3 1,7 0 D2 <7 ) 1)1 0 7D ), 3 17, 0 -) <, < 1)<

Surface inde*es Pc:8a #:8a 5ount:;m mm:;m 2D7 3 -) 3 200 2 <- < 231 , ,0 , 1D< ,1 0 1<1 ) ,, , 1<< 2 <3 0 231 , 70 D 2)1 1 72 D 27D < 73 3 2<7 ) 72 7 2,) 2 <D D 2<, < 7) 3 173 0 70 3 22< ) <- ) 1<3 2 <- -

(w, (c, and ae are grinding wheel surface speed, workpiece speed, depth of cut, respecti%el&. 8a, #, and Pc are arithmetic roughness height, mean spacing of ad-acent local peaks, and peak count, respecti%el&.

!ournal of "ngineering #ciences, Assiut $ni%ersit&, 'acult& of "ngineering, (ol. 1, )o. *, +a&, 201*, "-mail address, -es.aun.edu.eg

1027 O. B. Abouelatta, Prediction of machining operations and surface roughness using artificial neural network, pp. 1021- 10

1 0ersona! computer

2 Mitutoyo interface program , 5o!der

3 Specimen

) Sty!us

< Surf$est4SO201 driving unit

7 Surf$est4SO201 contro! unit

F#-. 1. Surf$est4SO201 and persona! computer "0:# +. Anal.(#( & o"e!u e *.1. Anal&sis of surface roughness parameters
&n recent years there has been a pro!iferation of parameters %ith %hich to specify surface te*ture Some of these parameters are usefu!, but most are not 6207 $he resu!t of this rash is confusion and e*pense So, a measure of the strength of the !inear re!ationship bet%een roughness parameters %ere app!ied using MS Fffice /*ce! to reduce the number of surface roughness parameters used in surface ana!ysis $ab!e , sho%s a part of corre!ation coefficient bet%een surface roughness parameter $hen, the surface roughness parameters regrouped %ith that having a strong corre!ation coefficient "8Q0 -0# $he resu!ts of this ana!ysis are grouped in si* categories "&4K&# and !isted in $ab!e 7 $he first group, for e*amp!e, inc!udes 8a, 8<, 8p, 8%, 8t, )Peaks, A5'10=, A5', >' roughness parameters $his means that any roughness parameter %ithin group "&# is enough to study the effect of cutting parameter, for e*amp!e, on the resu!ting surface roughness $he same cou!d be app!ied for other groups "&&4K&#, i e any roughness parameter %ithin the same group cou!d be used to represent the effect of cutting parameter
!ournal of "ngineering #ciences, Assiut $ni%ersit&, 'acult& of "ngineering, (ol. 1, )o. *, +a&, 201*, "-mail address, -es.aun.edu.eg

102D O. B. Abouelatta, Prediction of machining operations and surface roughness using artificial neural network, pp. 1021- 10

F#-. $. P+P5P, 0rediction of Machining Fperations and :utting 0arameters 0rogram

F#-. +. #85P, Surface Roughness :a!cu!ation 0rogram


!ournal of "ngineering #ciences, Assiut $ni%ersit&, 'acult& of "ngineering, (ol. 1, )o. *, +a&, 201*, "-mail address, -es.aun.edu.eg

102O. B. Abouelatta, Prediction of machining operations and surface roughness using artificial neural network, pp. 1021- 10

6a7

?urned workpiece )o. 617

6b7

+illed workpiece )o. 607

6c7 @round workpiece )o. 617 F#-. ,. Samp!es of surface te*ture and surface profi!e for some machining processes *.2. Anal&sis of machining operation signature
$he va!ues of amp!itude, spacing and hybrid parameters of surfaces manufactured by different machining operations are over!apped significant!y /ven if one machining operation resu!ts in a simi!ar surface pattern, it produces surfaces %hose amp!itude, spacing and hybrid parameters vary %ide!y depending on the machining conditions $hus, it is difficu!t to identify surface signature of different machining operations by investigating the magnitudes of the parameters 6217 &t is considered here that a good method of c!assifying machining processes is to use inde*es rather than abso!ute physica! quantities $his section provides evidence of the re!ationships bet%een machining
!ournal of "ngineering #ciences, Assiut $ni%ersit&, 'acult& of "ngineering, (ol. 1, )o. *, +a&, 201*, "-mail address, -es.aun.edu.eg

1030 O. B. Abouelatta, Prediction of machining operations and surface roughness using artificial neural network, pp. 1021- 10

operations and some derived inde*es 5o%ever by e*amining these inde*es, it is possib!e to identify different c!asses of machining operations

Table /. )a* Standard and non4standard surface roughness parameters


Surface roughness parameter Arithmetic a%erage roughness, 8a @eometric a%erage roughness, 8< #kewness, 8sk 2urtosis, 8ku +ean of 8pi %alues, 8pm +ean of 8%i %alues, 8%m +ean of 8ti %alues, 8tm +a3imum height, 8p +a3imum depth, 8% +a3imum peak to %alle& height, 8t Aargest 8ti %alue, 8& ?en-point height, 8BC#O ?en-point height, 8B/C) +ean of 8*&, 8*B 'ullness grade, 2m "mptiness grade, 2p +edian #tandard /e%iation, #/ Gigh spot count at 10=, G#510= Gigh spot count at *0=, G#5*0= Gigh spot count at 00=, G#500= Peak count at 10=, Pc10= Peak count at *0=, Pc*0= Peak count at 00=, Pc00= +ean spacing of ad-acent local peaks, # +ean spacing between profile peaks, #m10= +ean spacing between profile peaks, #m*0= +ean spacing between profile peaks, #m00= )o. of intersections of the profile at 10=, n60710= )o. of intersections of the profile at *0=, n60710= )o. of intersections of the profile at 00=, n60700= )umber of inflection points, g A%erage slope, a 8oot mean slope, < A%erage wa%elength, a 8oot mean wa%elength, < Bearing ratio at 10=, tp10= Bearing ratio at *0=, tp*0= Bearing ratio at 00=, tp00= $urning < D, 7 1< 0 10 2 D2 1) 77 12 32 27 01D )1 17 13< ,0 33 -0 27 013 <) 7 )2 0 )D 0 <2 40 <7 D.10 3 D7 1< 220 11 33 )) 21 << 12 72 103 1 )33 1<, ) 12 7D 31 0 )1 3 )70, 7 - 33 13 7D < ,7 ) ,0 1 -) 1, 1) )< -2 Mi!!ing ) 22 < 3) 40 <3 3 <, D <, 10 <3 1- 011 -1 1, 27 2D 1D 2, 3) 1- 0- <) 3 1) 0 <D 0 )2 0 <2 0.* < 33 1- )0 17 0) 2D -) 1< ,1 D 17 1)7 ,01 7 132 1 0, 10 3- < 3< ) ),,7 3 D )3 13 0D ) <0 3 2) 3 )1 2D 1,, 02 9rinding 0 7D 0 -D 40 27 3 1< 1 -< 2 <D ) <3 2 <2 3 )< < -D < <1 ) <3 2 27 0 32 0 <D 0 )2 0 0< 0.EF 10 -) 72 -1 D) )2 12< 2, )2 <1 17 31 <1 1 <D- 0 1,D 1)) 21 1), 3 170 0 3<0, ) D )0 1) ,2 0 77 0 )D 1 77 22 -3 ,, 2< Jnits =m =m 44444 44444 =m =m =m =m =m =m =m =m =m =m 44444 44444 =m ;m 0ea' 0ea' 0ea' 0ea'Lcm 0ea'Lcm 0ea'Lcm mm mm mm mm :ount :ount :ount :ount degree degree =m =m E E E

!ournal of "ngineering #ciences, Assiut $ni%ersit&, 'acult& of "ngineering, (ol. 1, )o. *, +a&, 201*, "-mail address, -es.aun.edu.eg

1031 O. B. Abouelatta, Prediction of machining operations and surface roughness using artificial neural network, pp. 1021- 10

Table /. )b* Standard and non4standard surface roughness parameters ":ontinued#


Surface roughness parameter Profile slope at line 10=, 10= Profile slope at line *0=, *0= Profile slope at line 00=, 00= 8elati%e length of the profile, lo Peak count, rp Amplitude /ensit& 5ur%e at 10=, A/510= Amplitude /ensit& 5ur%e at *0=, A/5*0= Amplitude /ensit& 5ur%e at 00=, A/500= Autocorrelation function, A5' 5orrelation Aength, 5A Power spectrum densit& function, P#/' 8oughness height uniformit&, Gu 8oughness height skewness, Gs 8oughness pitch uniformit&, Pu 8oughness pitch skewness, Ps #teepness factor of the profile, #' >a%iness factor of the profile, >' $urning 1D- 0 1-3 , 1-3 2 1 0) D1 0 2D 1 10 1 ,1 <, 17 211 , 1D3 0 , D, 40 <) )<1 3 0 713 ,2 0 20 Mi!!ing 1D< , 1-< < 1-3 3 1 02 <D 0 << 1 D0 1 <, 31 22 3<< 0 D2 2 ) 77 0 ))<0 3 0 -< - )< 0 27 9rinding 20) 1 1D2 D 1D1 < 1 01 1,2 0 33 1 D1 1 -0 -33) 0 ,< , 0 -2 0 03 <02 2 21 <2 0 <D 1 37 Jnits degree degree degree =m :ount E E E 44444 =m 44444 44444 44444 44444 44444 44444 44444

First of a!!, a paired of surface roughness parameters %ere se!ected to study its re!ation %ith the machining operations $he se!ected paired of surface roughness parameters, %hich ca!!ed inde*, is based on some trai!s e*cept for the condition of their poor corre!ation coefficient %ith each other $hen, by p!otting this pair of surface roughness parameters, the machining operations can be partia!!y or comp!ete!y separated &f an appro*imate 3one of a machining operations can be got based on the proposed inde*es, the efficiency of the identification can be assessed &t %as found that Pc:8a and #:8a inde*es satisfy the condition of determining an appro*imate 3one $he machining operation 3ones %hich enc!ose the va!ues of the inde*es obtained under different cutting parameters are evident from Figs < and , &t %as found that turning "Facing#, mi!!ing and ground surfaces are distinctive!y separated from other machining operations &n genera!, turned surfaces have the !argest Pc:8a inde* and #:8a inde*, %hereas grinding surfaces have the !o%est Pc:8a inde* and #:8a inde* Mi!!ing surfaces have the second !argest Pc:8a inde* and #:8a inde*

!ournal of "ngineering #ciences, Assiut $ni%ersit&, 'acult& of "ngineering, (ol. 1, )o. *, +a&, 201*, "-mail address, -es.aun.edu.eg

1032 O. B. Abouelatta, Prediction of machining operations and surface roughness using artificial neural network, pp. 1021- 10

Table 0. :orre!ation coefficient bet%een surface roughness parameters ":orre!ation coefficients greater than 0 - are high!ighted %ith bo!d#
Parameters 8B/C) 8< 8sk 8pm 8%m 8tm 8p 8% 8t 8& 8a 2m 8pm 2.3,$ 0 ,<2 0 <-< 8tm 1.222 0 7<0 0 )21 2.3,$ 0 D7) 8p 0 7D0 2.3,4 0 <<, 0 713 0 712 0 7D0 8t 0 71< 2.3/$ 0 2,0 <<0 0 D03 0 71< 2.3+$ 2.313 8& 2.31/ 0 772 0 2,< 0 7,2 2.3,+ 2.31/ 0 D20 0 737 0 D)3 8BC#O 1.222 0 7,0 )2) 2.3+3 0 D7D 1.222 0 7-< 0 <<0 0 732 2.314 8a 0 7<0 2.334 0 )3< 0 ,,< 0 71) 0 7<0 2.3,0 0 77< 2.3++ 0 7<1 2m 0 )3, 0 31) 2.3+/ 0 ,<7 0 02< 0 )3, 0 )DD 0 2)3 0 1<0 0 22D 0 3<) 2p 0 )3, 0 31) 2.3+/ 0 ,<7 0 02< 0 )3, 0 )DD 0 2)3 0 1<0 0 22D 0 3<) +ed 0 )<1 0 ,33 0 770 0 ,21 0 110 0 )<1 0 ,)7 0 1<0 )), 0 27, 0 ,,) 0 72D #/ 0 7<0 1.222 0 3-D 0 ,<2 0 73< 0 7<0 2.3,4 0 D0D 2.3/$ 0 772 2.334 0 31)

Table 1. 9rouping of a high!y corre!ated surface roughness parameters


G ou& No. & && &&& &K Tu n#n- )Fa"#n-* Pa a'ete ( No. 8a, 8<, 8p, 8%, 8t, )Peaks, A5'10=, A5', >' 8sk, 2m, 2p, tp 0=, , tp00=, A/5 0= 8B/C), 8pm, 8tm, 8&, 7 8BC#O, Gu, 8%m Pc10=, Pc20=, Pc*0=, < Pc 0=, Pc00= G#510=, )Cntr10=, G#520=, )Cntr20=, G#5*0=, )Cntr*0=, 10 G#5 0=, )Cntr 0=, G#500=, )Cntr00= ) a, <, a, < M#ll#nG #n!#nAll P o"e((e( Pa a'ete ( No. Pa a'ete ( No. Pa a'ete ( No. 8a, 8<, 8pm, 8tm, 8a, 8<, 8pm, 8tm, 8%, 8a, 8<, 8p, 8%, 8t, 8t, 8BC#O, 8B/C), 10 8t, 8&, 8BC#O, 8B/C), 12 )Peaks, A5'10=, A5'10=, A5', Gu A5'10=, A5', Gu A5', >' A/5*0=, tp*0=, 8sk, tp 0=, tp00= 3 ) tp 0=, tp00= 2 tp 0=, tp00= 8B/C), 8pm, 8tm, 8&, 8B/C), 8pm, 8tm, 8&, 8B/C), 8pm, 8tm, 8&, 7 7 7 8BC#O, Gu, 8%m 8BC#O, Gu, 8%m 8BC#O, Gu, 8%m Pc10=, Pc20=, Pc10=, Pc20=, Pc*0=, Pc10=, Pc20=, < < < Pc*0=, Pc 0=, Pc00= Pc 0=, Pc00= Pc*0=, Pc 0=, Pc00= G#510=, )Cntr10=, G#510=, )Cntr10=, G#510=, )Cntr10=, G#520=, )Cntr20=, G#520=, )Cntr20=, G#520=, )Cntr20=, G#5*0=, )Cntr*0=, 10 G#5*0=, )Cntr*0=, 10 G#5*0=, )Cntr*0=, 10 G#5 0=, )Cntr 0=, G#5 0=, )Cntr 0=, G#5 0=, )Cntr 0=, G#500=, )Cntr00= G#500=, )Cntr00= G#500=, )Cntr00= ) ) ) a, <, a, < a, <, a, < a, <, a, <

K K&
9

5orrelation coefficients are e<ual or more than 0.E0 6PH0.007.

/ach machining operation can be deduced in an area of e!!ipse as sho%n in Figs < and , (n e!!ipse is a smooth c!osed curve %hich is symmetric about its hori3onta! and vertica! a*es $he distance bet%een antipoda! points on the e!!ipse, or pairs of points %hose midpoint is at the center of the e!!ipse, is ma*imum a!ong the ma1or a*is or transverse
!ournal of "ngineering #ciences, Assiut $ni%ersit&, 'acult& of "ngineering, (ol. 1, )o. *, +a&, 201*, "-mail address, -es.aun.edu.eg

1033 O. B. Abouelatta, Prediction of machining operations and surface roughness using artificial neural network, pp. 1021- 10

diameter, and a minimum a!ong the perpendicu!ar minor a*is or con1ugate diameter 6227 (n e!!ipse in genera! position can be e*pressed parametrica!!y as the path of a point "I6t7, J6t7#, %hereR I6t7 H Ic K a cos6t7 cos67 - b sin6t7 sin67 J6t7 H Jc K a cos6t7 cos67 K b sin6t7 sin67 "1# "2#

as the parameter t varies from 0 to 2S 5ere "Ic, Jc# is the center of the e!!ipse, and is the ang!e bet%een the I4a*is and the ma1or a*is of the e!!ipse $ab!e D !ists the e!!ipse parameters "Ic, Jc, a, b and # for different machining processes $he range of the Pc:8a and #:8a inde*es for different machining processes are given in Figs 74-, respective!y (!though the conventiona! surface roughness parameters !isted in $ab!es 24) vary considerab!y for each 'ind of surface, it is not difficu!t to observe from the figures that the inde*es identify the different machining operations c!ear!y &n addition, some genera! conc!usions for the re!ationship bet%een the inde*es and machining operations can be dra%n here %ith the he!p of Figs 74- $he inde*es may a!so vary in range, see $ab!e 24) &t can be observed that if surfaces have c!ear!y different surface roughness or height distributions, the distribution 3ones of the inde*es of the surfaces %ou!d be distinctive!y separated Moreover, the variation of the inde*es of one 'ind of surface indicates differences of surface features a!though the surfaces may have simi!ar surface roughness or height distributions
elation !et"een Ra and S 220 200 180 160 140 120 100 80 60 40 #urning $rinding Milling

M e an spa cing o fa dja ce nt loca l pe a ks (m m )

4 6 8 Average roughness height (m)

10

12

F#-. /. ( diagram of average roughness height "8a# versus mean spacing of ad1acent !oca! pea's "## of typica! machining operations
!ournal of "ngineering #ciences, Assiut $ni%ersit&, 'acult& of "ngineering, (ol. 1, )o. *, +a&, 201*, "-mail address, -es.aun.edu.eg

103) O. B. Abouelatta, Prediction of machining operations and surface roughness using artificial neural network, pp. 1021- 10
elation !et"een Ra and P c 200 180 160 )eak count (* ount) 140 120 100 80 60 40 20 #urning $rinding Milling

% 4 & 6 ' Average roughness height (m)

10

F#-. 0. ( diagram of average roughness height "8a# versus pea' count "Pc# of typica! machining operations Table 4. /!!ipse parameters "Ic, Jc, a, b and #
0arameter Ic Jc a b $urning "Facing# 8a-# 8 a- P c 0D 0D << 1<D 0 2D 0 3< 1< 37 7D 470 Mi!!ing 8a-# )D 1,2 1D ), D, 8 a- P c ,1 DD 3) 2< 477 8a-# ,1 102 33 30 0 9rinding 8 a- P c )D << 11, 0

*.*.

Anal&sis of cutting parameters

$he inf!uences of machine parameters on machined parts are not a!%ays precise!y 'no%n, and hence, it becomes difficu!t to recommend the optimum machinabi!ity data for machine process :utting parameter identification in machining operations, %hich inc!uded cutting speed, feed rate and depth of cut p!ays a very important ro!e in the efficient uti!i3ation of machine too!s and direct!y inf!uences the product qua!ity $hus, it significant!y inf!uences the overa!! manufacturing costs &n practice, the machinists se!ect cutting parameters from their specified ranges in machining handboo's, main!y based on e*perience, in order to satisfy the required accuracy of the fina! product 627 $he ana!ysis of cutting parameters %i!! be e*p!ained in detai!s in section ) 2
!ournal of "ngineering #ciences, Assiut $ni%ersit&, 'acult& of "ngineering, (ol. 1, )o. *, +a&, 201*, "-mail address, -es.aun.edu.eg

103< O. B. Abouelatta, Prediction of machining operations and surface roughness using artificial neural network, pp. 1021- 10

F#-. 1. Surface inde* "Pc:##

F#-. 4. Surface inde* "Pc:8a#


!ournal of "ngineering #ciences, Assiut $ni%ersit&, 'acult& of "ngineering, (ol. 1, )o. *, +a&, 201*, "-mail address, -es.aun.edu.eg

103, O. B. Abouelatta, Prediction of machining operations and surface roughness using artificial neural network, pp. 1021- 10

F#-. 3. Surface inde* "#:8a# ,. Neu al net5o 6 'o!el#n.eura! net%or' is a high!y f!e*ib!e mode!ing too! %ith the abi!ity to !earn the mapping bet%een input and output parameters (n artificia! neura! net%or' "(..# is capab!e of !earning from an e*perimenta! data set to describe the non!inear and interaction effects more effective!y $he net%or' consists of an input !ayer used to present data, output !ayer to produce (..Gs response, and one or more hidden !ayers in bet%een $he net%or' is characteri3ed by their topo!ogy, %eight vectors, and activation function that are used in hidden and output !ayers of the net%or' .et%or's %ith biases, a sigmoid !ayer, and a !inear output !ayer are capab!e of appro*imating any function %ith a finite number of discontinuities $he 'no%!edge is presented by the interconnection %eight, %hich is ad1usted during the !earning stage using the bac' propagation !earning a!gorithm to minimi3e the mean square bet%een the actua! output of the net%or' and the desired output pattern 6237 $he overa!! training process used in this paper is i!!ustrated in Fig 10 First!y, the roughness parameters %ere used as an input to predict the output %hich is the machining process Second!y, both of roughness parameters and the predicted machining process %ere used to predict cutting parameters %hich areR feed, speed and depth of cut

!ournal of "ngineering #ciences, Assiut $ni%ersit&, 'acult& of "ngineering, (ol. 1, )o. *, +a&, 201*, "-mail address, -es.aun.edu.eg

1037 O. B. Abouelatta, Prediction of machining operations and surface roughness using artificial neural network, pp. 1021- 10

Su 7a"e Rou-8ne(( Pa a'ete (

Neu al Net5o 6 P o"e((#n-

Cutt#nPa a'ete (

P e!#"t#on o7 Ma"8#n#n- P o"e((

A&&l.#n- Neu al Net5o 6

Average roughness height, Ra Peak count, Pc Mean spacing of adjacent local peaks, S

A&&l.#n- Neu al Net5o 6

Feed Speed Depth of cut

F#-. 12. (pp!ication of neura! net%or' to predict machining operations and cutting parameters *. . Prediction of machining process
( feed for%ard neura! net%or' %ith error bac' propagation a!gorithm %as adopted for the .. system 5ere it is used to deve!op machining process prediction mode! through surface roughness measurements From )< e*periments %ere conducted, 3, e*perimenta! datasets %ere used to train the net%or' Hefore app!ying the neura! net%or' for mode!ing the architecture of the net%or' has to be decided+ i e the number of hidden !ayers and the number of neurons in each !ayer and transfer function for each !ayer (s there are three inputs "8a, Pc and # roughness parameters# to produce one output "Machining process %hich areR $urning TFacingU, Mi!!ing or 9rinding#, the number of neurons in the input and output !ayer has to be set to 3 and 1 respective!y as sho%n in Fig 11 :onsidering one hidden !ayer, the number of neurons in the hidden !ayer is optimi3ed ( procedure %as emp!oyed to optimi3e the hidden !ayer neurons and se!ect the transfer function for %hich a program %as generated in M($>(H soft%are (ccording!y, an e*perimenta! approach %as adopted, %hich invo!ves verification of the neura! net%or' against another - sets of data from )< e*perimenta! dataset

!ournal of "ngineering #ciences, Assiut $ni%ersit&, 'acult& of "ngineering, (ol. 1, )o. *, +a&, 201*, "-mail address, -es.aun.edu.eg

103D O. B. Abouelatta, Prediction of machining operations and surface roughness using artificial neural network, pp. 1021- 10

A%erage roughness height Peak count +ean spacing of ad-acent local peaks

8a Pc # ? +achining operations 6?urning, milling and grinding7

Cnput la&er

Gidden la&er

Output la&er

F#-. 11. (pp!ication of neura! net%or' to predict machining operations


$he performance of the net%or' %as eva!uated by mean squared error bet%een the e*perimenta! and the predicted va!ues for every output nodes in respect of training the net%or' ;hen the training is comp!ete, the net%or' performance cou!d be chec'ed and determined to search for if any changes need to be made to the training process, the net%or' architecture or the data sets $he first thing to do is to chec' the training record, Fig 12 (s indicated, the iteration at %hich the va!idation performance reached a minimum %as , $he training continued for , more iterations before the training stopped $he ne*t step in va!idating the net%or' is to create a regression p!ot, %hich sho%s the re!ationship bet%een the outputs of the net%or' and the targets, Fig 13 &f the training %ere perfect, the net%or' outputs and the targets %ou!d be e*act!y equa!, but the re!ationship is rare!y perfect in practice $he four a*es represent the training, va!idation, testing and a!! data $he dashed !ine in each a*is represents the perfect resu!t bet%een outputs and targets $he so!id !ine represents the best fit !inear regression !ine bet%een outputs and targets $he 8 va!ue is an indication of the re!ationship bet%een the outputs and targets &f 8 M 1, this indicates that there is an e*act !inear re!ationship bet%een outputs and targets &f 8 is c!ose to 3ero, then there is no !inear re!ationship bet%een outputs and targets

!ournal of "ngineering #ciences, Assiut $ni%ersit&, 'acult& of "ngineering, (ol. 1, )o. *, +a&, 201*, "-mail address, -es.aun.edu.eg

103O. B. Abouelatta, Prediction of machining operations and surface roughness using artificial neural network, pp. 1021- 10
1

Best Validation Performance is 0.013347 at epoch 6


#rain ,alidation #est -est

10

10

Mean Squared Error mse!

10

+1

10

+2

10

+%

10

+4

10

+&

10

12

1" Epochs

F#-.1$. 0!ot of training, va!idation, and test performances


$ab!e - sho%s the output of neura! net%or' process $he feedbac' from that processing is ca!!ed the Taverage errorU or TperformanceU Fnce the average error is be!o% the required goa! or reaches the required goa!, the neura! net%or' stops training and becomes ready to be verified $his process %as repeated one hundred times and the !o%est MS/ run %as chosen as the best trained net (fter the training process, the mode! is tested for va!idation &n this %or', the net%or' is va!idated in terms of agreement %ith an e*tra e*perimenta! data !isted in $ab!es 24)

!ournal of "ngineering #ciences, Assiut $ni%ersit&, 'acult& of "ngineering, (ol. 1, )o. *, +a&, 201*, "-mail address, -es.aun.edu.eg

10)0 O. B. Abouelatta, Prediction of machining operations and surface roughness using artificial neural network, pp. 1021- 10

#rainin$, -'0.((3+1 %utput &' 0.(6)#ar$et * 0.0+4 %utput &' 0.(7)#ar$et * 0.1"
2.2 2 1.8 1.6 1.4 1.2 1 0.8 0.6 0.& 2 /ata 0it 12#

Validation, -'0.(+73(
/ata 0it 12#

1.&

0.& 0.& 1 1.& 2

1.&

#ar$et #est, -'0.+(0(4 %utput &' 0..4)#ar$et * 0.44


2.2 2 1.8 1.6 1.4 1.2 1 0.8 0.6 0.& 1 1.& 2

#ar$et /ll, -'0.(7+0. %utput &' 0.(4)#ar$et * 0.1


2 /ata 0it 12#

/ata 0it 12#

1.&

0.& 0.& 1 1.& 2

#ar$et

#ar$et

F#-. 1+. >inear regression of targets re!ative to outputs


!ournal of "ngineering #ciences, Assiut $ni%ersit&, 'acult& of "ngineering, (ol. 1, )o. *, +a&, 201*, "-mail address, -es.aun.edu.eg

10)1 O. B. Abouelatta, Prediction of machining operations and surface roughness using artificial neural network, pp. 1021- 10

Table 3. .eura! net%or' output


0arameter Best trail number 5oefficient of determination 687 +a3imum error A%erage error +ean absolute error 6+A"7 +ean s<uared error 6+#"7 +ean s<uared normaliBed error 6+#)"7 +ean s<uared normaliBed error with regulariBation 6+A)"8"@7 +ean s<uared error with regulariBation 6+A"8"@7 +ean s<uared error with regulariBation and economiBation 6+#"8"@"57 #um s<uared error 6##"7 )umber of Cteration 6)C7 $urning *0 0.EE*F 10.1E2E 2. *** 0.021DE 0.001FD 1.E1122 1.E2 F0 0.201*E 0.10 22 0.00EF* 1E0 Mi!!ing E* 0.E0E2 E2.*11 12.2101 0.0DFF2 0.02E0D 0.FF0FE 0.E1* 0.1ED21 0.11 FE 0.DEF * 20 1 9rinding E1 0.E*10D 1F.**** 0.2ED0 0.0 *1F 0.01111 .1D11 .0*10 0.2FD01 0.1 E1 0.2111 22E0F

*.0. Prediction of cutting parameters


(t first, an (.. mode! %as proposed to determine surface roughness in a dry environment $he capabi!ity of the (.. mode! is to genera!i3e unseen data depending on severa! factors $hese factors are appropriate se!ection of input4output parameters, the distribution of the input4output dataset, and the format of the presentation of the dataset to the neura! net%or' Se!ected input parameters are the significant variab!es that affect the surface roughness %hi!e perform machining operation under dry environment :utting speed, feed rate, and depth of cut %ere considered as input parameters $he output parameter of the mode! is the average roughness parameter "8a# @etai!s of the inputLoutput parameters of the proposed (.. mode! are i!!ustrated in Fig 1) 62)7 Second!y, possib!e interaction cutting parameters "Feed, speed and depth of cut# %ere bui!t and simu!ated using the trained net to ca!cu!ate 8a Searching for the nearest va!ues to measured 8a %ithin 0 01 error, the possib!e cutting parameters that produce measured 8a can be tabu!ated as sho%n in Prediction of 5utting Parameters 0ane!, Fig 2 $he net%or' architectureL topo!ogy or features such as number of neurons and !ayers are very important factors that determine the functiona!ity and genera!i3ation capabi!ity of the net%or' $raining of an (.. p!ays a significant ro!e in designing the direct (..4based prediction $he same procedures %ere app!ied as in machining operation prediction (fter the training, the mode! is tested for va!idation &n this %or', the net%or' is va!idated in terms of agreement %ith an e*tra 3 e*perimenta! data

!ournal of "ngineering #ciences, Assiut $ni%ersit&, 'acult& of "ngineering, (ol. 1, )o. *, +a&, 201*, "-mail address, -es.aun.edu.eg

10)2 O. B. Abouelatta, Prediction of machining operations and surface roughness using artificial neural network, pp. 1021- 10

'eed f #peed /epth of cut Ra A%erage roughness height dc

Cnput la&er

Gidden la&er

Output la&er

F#-. 1,. (pp!ication of neura! net%or' to predict cutting parameters /. Con"lu(#on(


&n this study, an artificia! neura! net%or' approaches %ere deve!oped to predict the machining processes, surface roughness, and machining parameters %hi!e cutting (&S& 10)0 stee! specimens $he obtained data %ere used to deve!op the surface roughness mode!s $he capabi!ity of the deve!oped prediction mode!s %ere eva!uated by comparing its resu!ts %ith measured va!ues $he predicted va!ues %ere found to be c!ose to the measured va!ues $he proposed mode!s can be used effective!y to predict the surface roughness in turning process $he coefficient of corre!ation "8# is more than -0 -E for the testing data in neura! net%or' mode! $he fo!!o%ing conc!usions can be dra%n from the present study 1 $%o4dimensiona! surface roughness parameters cou!d be deduced to si* groups &n other %ords, on!y si* surface roughness parameters in 2@ are enough to characteri3e machined surfaces 2 $he machining operations under investigations, %hich are turning "facing#, mi!!ing and grinding+ can be comp!ete!y separated from each other using its surface roughness signature inde*es 3 Jsing inverse artificia! neura! net%or', it %as possib!e to predict cutting parameters that produce a specified surface roughness va!ues

0. Re7e en"e(
617 Mohammed $ 5aya1neh, Montasser S $ahat, Ooachim H!uhm, ( Study of the /ffects of Machining 0arameters on the Surface Roughness in the /nd4Mi!!ing 0rocess, Oordan Oourna! of Mechanica! and &ndustria! /ngineering, vo! 1, .o 1, pp 14<, 2007

!ournal of "ngineering #ciences, Assiut $ni%ersit&, 'acult& of "ngineering, (ol. 1, )o. *, +a&, 201*, "-mail address, -es.aun.edu.eg

10)3 O. B. Abouelatta, Prediction of machining operations and surface roughness using artificial neural network, pp. 1021- 10 627 (dee! 5 Suhai!, . &smai!, S K ;ong and . ( (bdu! Oa!i!, P:utting 0arameters &dentification using Mu!ti (daptive .et%or' based Fu33y &nference SystemR (n (rtificia! &nte!!igence (pproachP, Scientific Research and /ssays, Ko! ,, .o 1, pp 1D741-<, 2011 637 Shutong8ie, Iinbiao9uo, P&nte!!igent Se!ection of Machining 0arameters in Mu!ti4pass $urnings Jsing a 9(4based (pproachP, Oourna! of :omputationa! &nformation Systems, Ko! 7, .o <, pp 171)41721, 2011 6)7 ; 9r3esi', $ ;anat, P:omparative (ssessment of Surface Roughness 0roduced by 5ard Machining %ith Mi*ed :eramic $oo!s &nc!uding 2@ and 3@ (na!ysisP, Oourna! of Materia!s 0rocessing $echno!ogy, Ko! 1,-, pp 3,)V371, 200< 6<7 Ra1endra M 0atri'ar, PMode!ing and Simu!ation of Surface RoughnessP, (pp!ied Surface Science, Ko! 22D, pp 213V220, 200) 6,7 > 5 S >uong, $ ( Spedding, P( .eura!4.et%or' System for 0redicting Machining HehaviorP, Oourna! of Materia!s 0rocessing $echno!ogy, Ko! <2, pp <D<4<-1, 1--< 677 ( ?oh!i and J S @i*it, P( .eura!4.et%or'4based Methodo!ogy for the 0rediction of Surface Roughness in a $urning 0rocessP, &nt O (dv Manuf $echno!, Ko! 2<, pp 11DV12-, 200< 6D7 W!han (si!tAr', Mehmet Bun'aC, PMode!ing and 0rediction of Surface Roughness in $urning Fperations using (rtificia! .eura! .et%or' and Mu!tip!e Regression MethodP, /*pert Systems %ith (pp!ications, Ko! 3D, pp <D2,V<D32, 2011 6-7 W!han (si!tAr', P0redicting Surface Roughness of 5ardened (&S& 10)0 based on :utting 0arameters using .eura! .et%or's and Mu!tip!e RegressionP, &nt O (dv Manuf $echno!, 0ub!ished on!ineR 0< February 2012 6107 ? Sundara Murthy, & Ra1endran, P( Study on Fptimisation of :utting 0arameters and 0rediction of Surface Roughness in /nd Mi!!ing of (!uminium under MX> MachiningP, &nt O Machining and Machinabi!ity of Materia!s, Ko! 7, .os 1L2, 2010 6117 H 5uang and O : :hen, P(n &n40rocess .eura! .et%or'4Hased Surface Roughness 0rediction "&..4SR0# System Jsing a @ynamometer in /nd Mi!!ing FperationsP, &nt O (dv Manuf $echno!, Ko! 21, pp 33-V3)7, 2003 6127 $uncay /r3urum!u, 5asan F'tem, P:omparison of Response Surface Mode! %ith .eura! .et%or' in @etermining the Surface Xua!ity of Mou!ded 0artsP, Materia!s and @esign, Ko! 2D, pp )<-V),<, 2007 6137 (ntoni ;ibo%o, Mohammad &sha' @esa, P?erne! based Regression and 9enetic (!gorithms for /stimating :utting :onditions of Surface Roughness in /nd Mi!!ing Machining 0rocessP, /*pert Systems %ith (pp!ications, Ko! 3-, pp 11,3)V11,)1, 2012 61)7 .abi! Hen Fred1, Ridha (mamou, P9round Surface Roughness 0rediction based upon /*perimenta! @esign and .eura! .et%or' Mode!sP, &nt O (dv Manuf $echno!, Ko! 31, pp 2)V3,, 200, 61<7 &ndra1it Mu'her1ee, 0radip ?umar Ray, P( Systematic So!ution Methodo!ogy for &nferentia! Mu!tivariate Mode!!ing of &ndustria! 9rinding 0rocessP, Oourna! of materia!s processing techno!ogy, Ko! 1-,, pp 37-V3-2, 200D 61,7 Oames O 9ovindhasamy, SeYn F Mc>oone, 9eorge ; &r%in, Oohn O French, Richard 0 @oy!e, P.eura! Mode!!ing, :ontro! and Fptimisation of an &ndustria! 9rinding 0rocessP, :ontro! /ngineering 0ractice, Ko! 13, pp 12)3V12<D, 200< 6177 San1eev ?umar, S ? :houdhury, P0rediction of ;ear and Surface Roughness in /!ectro4 discharge @iamond 9rindingP, Oourna! of Materia!s 0rocessing $echno!ogy, Ko! 1-1, pp 20,V20-, 2007

!ournal of "ngineering #ciences, Assiut $ni%ersit&, 'acult& of "ngineering, (ol. 1, )o. *, +a&, 201*, "-mail address, -es.aun.edu.eg

10)) O. B. Abouelatta, Prediction of machining operations and surface roughness using artificial neural network, pp. 1021- 10 61D7 / Mainsah and @ $ .dumu, P.eura! net%or' app!ications in surface topographyP, &nt O Mach $oo!s Manufact Ko! 3D, .os <4,, pp <-14<-D, 1--D 61-7 / S 9ade!ma%!a, M M ?oura, $ M ( Ma'soud, & M /!e%a, 5 5 So!iman, PSurface RoughnessP, Oourna! of Materia!s 0rocessing $echno!ogy, Ko! 123, pp 13341)<, 2002 6207 @ O ;hitehouse, P$he 0arameter Rash Z &s there a :ure[P, ;ear, Ko! D3, .o 1, pp 7<4 7D, 1-D2 6217 H > 9riffiths, PManufacturing Surface @esign and Monitoring for 0erformanceP, Surface $opography, Ko! 1, .o 1, pp ,14,-, 1-DD 6227 Hesant, ; 5 , P:hapter &&& $he /!!ipseP, :onic Sections, >ondonR 9eorge He!! and Sons, p <0, 1-07 6237 Oignesh 9 0armar, (!pesh Ma'%ana, P0rediction of Surface Roughness for /nd Mi!!ing 0rocess using (rtificia! .eura! .et%or'P, &nternationa! Oourna! of (dvanced /ngineering Research and Studies, /4&SS.22)-VD-7), &O(/RS, Ko! &, &ssue &&&, pp )74<0,(pri!4Oune, 2012 62)7 @i%a'ar Reddy K , ?rishnaiah 9 , ( 5emanth ?umar and Sushi! ? 0riya, P(.. based 0rediction of Surface Roughness in $urningP, &nternationa! :onference on $rends in Mechanica! and &ndustria! /ngineering "&:$M&/\2011# Hang'o', pp 1,<4170, 2011

9:;<=>?@A 9:BCDEA 9FBGEA HAIJKL<M N>LOA 9PQGRS T:UGKEA V<:WXDM YB=KEA


[ZJWXEA f} c}ca }dr }fhio {|}g~dh vgghi bc_defghi jk lmnio cpqc_frcs teuhi vwxyz ]^_`a v}}dgr qc}}fzi }} }}yfhi bc}}cgh c}}~hi imf}}|ho v}}dfghi }}yfhi bc}}dgr m}}m^ v}gghi {c}i j}k "ehi gro vfhio ehi vr# dghi ehi o mm^o yfhi o} m}m^o }y v}dgr }a c} v_}ck v c qohi jg hho i m }dr }ehi bcdgr ]g ^_hi ip l m^k te vwxyz dr x^dh v~ghi ehi om}}so }}dfhio }}fhio v}}ihi bc}}dgr imf}}cs (&S& -- d}}hi j}}k bx}}yk v}ugz }dr qc}fhi ]}a m}o v}chi te}uhi vwx}yz c} } m_ ix imfi }dr x}^hi }s }ho v}dfk e gro vo e bcr imfcs vr j~sqao ix}yr c}a }dr jfrx}gk }h bc}wc_hi }p }u mo vocfk te vwxyz kixr ]kmf}}i m}}o qc}}_fzi bc}}wcs v}}rxggh o qm}}fhi bc}}wcs v}}rxggh v}}r c}}k "(..# v}dfghi vrce}`i v_~hi bc_yhi jk cgw qmfh qmfhi bcwc_hi vrxgk ]kmfi cg dhi qcyfwi v_ imfcs teuhi vwxyz kixr o ehi bcdg~s _fdh v_}~hi bc_}yhi c}gh b|zm}g biqc}_fzi j}k c}~g } }fhio v}dg~hi qc}fhi bcwcs }}kixro }}ehi o}} m}}m^ h}}o v}}dfghi }}yfhi bc}}dgr j}}s o}}hi m}}m^ m}}s v}kqixz g}fh te}i vwx}yz }kixr } }dr m}gf~ } ]kmfio teuhi vwxyz }o }yfhi bc}dg~s }_fdh kc}ws }o }_e } v}dfghi yfhi bcdg~s _fdh v_ck bc}dg~s }_fhi }dr lq c} v}nfghi c}ghi a c}fhi b}a mo vdfghi tei vwxyz vdfghi tei vwxyz j~ ho ehi o mm^o yfhi

!ournal of "ngineering #ciences, Assiut $ni%ersit&, 'acult& of "ngineering, (ol. 1, )o. *, +a&, 201*, "-mail address, -es.aun.edu.eg

Vous aimerez peut-être aussi