MATHEMATICAL MODEL OF TITANIUM TURNING AND IT S APPLICATION IN MULTI-CRITERION OPTIMIZATION OF THIS PROCESS

ADVANCES IN MANUFACTURING SCIENCE AND TECHNOLOGY Vol. 4, No., 16 DOI: 1.478/amst-16-8 MATHEMATICAL MODEL OF TITANIUM TURNING AND IT S APPLICATION IN MULTI-CRITERION OPTIMIZATION OF THIS PROCESS Józef Zawora S u m m a r y The article includes information on basic properties of titanium and its alloys, features for titanium machining, the development of the experiment-based mathematical model for straight turning of the WT3-1 titanium, single- and multi-criterion optimization of the selected process features. Keywords: titanium, mathematical model, multi-criterion optimization, optimum parameters Model matematyczny toczenia tytanu i jego zastosowanie do wielokryterialnej optymalizacji procesu S t r e s z c z e n i e W pracy zaprezentowano podstawowe właściwości fizyczne i mechaniczne tytanu i jego stopów. Scharakteryzowano podstawy obróbki skrawaniem tytanu. Przedstawiono budowę doświadczalnego modelu matematycznego procesu toczenia wzdłużnego stopu tytanu WT3-1. Podano przykład optymalizacji jedno- i wielokryterialnej złożonego przebiegu procesu skrawania. Słowa kluczowe: tytan, model matematyczny, optymalizacja wielokryterialna, parametry optymalne 1. Introduction Research on the titanium turning has been conducted in order to develop a mathematical model due to the random nature of such process. Titanium, being a constructional material, features the following properties [1-4]: high level of tensile strength R m [MPa], both at room and elevated temperature, high level of relative strength R m/ρ, relatively low density ρ [g/cm 3 ], low value of thermal conductivity coefficient, high resistance to corrosion, high plasticity. Address: Józef ZAWORA, PhD Eng., Institute of Manufacturing Technology, Warsaw University of Technology. -663 Warsaw, Niepodległości, e-mail: jzawora@meil.pw.edu.pl

8 J. Zawora Titanium is typically applied in the aircraft and aerospace industry, shipbuilding, chemical engineering and in medicine. Titanium and titanium alloys are considered hard-to cut materials due to the following features [5, 6]: high degree of chemical affinity with the cutting tool edge material, particle adherence inclination, strength conservation at elevated temperatures, low thermal conductivity. The research has also been aimed at searching for new methods of constituting the properties of superficial layer for machine parts in view of optimization procedure. This process belongs to a category of special techniques because of the output quantity shaping by physical phenomena in various machining conditions. The up-to-date tool materials, machine tools and manufacturing techniques make it possible to obtain very high temperatures in the cutting zone, up to 1 degrees Celsius. The cutting fluid which has been introduced into this area evaporates instantly. This is a serious limitation on the low-pressure coolant stream introduction into the cutting tool edge proximity so that the effective cooling becomes impossible. One of the possible solutions is just abandonment of the cooling and carrying out the process at the lowered parameters of cut. Coolantfree machining process has also some advantage from the ecology viewpoint.. Experimental investigation Because of complexity of the course of physical phenomena taking place during the surface formation in the straight turning process, the experimental investigations were carried out according to the experimental design theory [7]. The five-level, composite, rotatable experimental design was developed for three independent variables (Table 1). The independent variables v c, f, a p were used within the following range: v c 1. 4. m/min surface speed (cutting speed), f.5.3 mm/rev feed rate, a p 1..5 mm depth of cut. The following designations were used in table I: ρ = 1.68 star radius value, i experiment No. (i = 1,, 3,, N), N = number of the experiments, j output quantity index (j = T, R z, F c, F f, F N, k c, P c, Q v) where: T [min] tool edge life, R z [µm] roughness parameter (the highest asperity value according to the DIN standard), F c [N] cutting force, F f [N] axial force, F n [N] radial force, k c [N/mm ] specific cutting force, P c [kw] cutting power, Q v [mm 3 /s] material removal rate (volumetric).

Mathematical model of titanium turning... 9 Table 1. Experimental design layout No Parametres vc f ap Yi,j 1-1 -1-1 Y1,j +1-1 -1 Y,J, 3-1 +1-1 Y3,j 4 +1 +1-1 Y4,j 5-1 -1 +1 Y5,j 6 +1-1 +1 Y6j 7-1 +1 +1 Y7,j 8 +1 +1 +1 Y8,j 9 -ρ Y9,j 1 +ρ Y1,j 11 -ρ Y11,j 1 +ρ Y1,j 13 -ρ Y13,j 14 +ρ Y14,j 15 Y15,j 16 Y16,j 17 Y17,j 18 Y18,j 19 Y19,j Y,j Research of the process output quantities was conducted for the samples made of titanium WT3-1, with diameter Φ = 8 mm and length L = 45 mm. The following mechanical properties have been assumed according to the authorized attestation documents: ultimate tensile strength R m = 11 MPa, elongation A 5 = 13%, Brinell hardness 35. The experiments were conducted on the standard toolmaker s lathe using standard tool-holder hr171.6-55.1 with the interchangeable triangular inserts TNMG1638 made of tungsten carbide H, with the cutting edge rounding radius r n =, mm. Flank wear VB =,3 mm was adopted as a critical measure (criterion) of the cutting tool edge dullness. The wear measurement was performed on the toolroom Zeiss microscope. The height of roughness asperities was determined by the luminous cross-section method. The cutting force components were measured by the extensometric dynamometer. Since the process output values i.e. cutting force components, the height of the surface roughness summit R z (DIN), cutting power, specific cutting force are dependent on the cutting edge wear, these values were determined for the mathematical model as average values throughout the tool life period in order to limit a number of experiments in the experimental design. It is in this case connected with the exclusion of machining time t as an independent variable which affects the tool wear. Time influence on the specified output quantities has

3 J. Zawora been taken into consideration by dimensionless relationship of the following shape: y y i i, śr = a, i + r j= 1 a i, j t T r (1) y i general designation for the output quantities, y i,śr general designation for the average value of y i, a constant factor in the formula, a j polynomial factor (regression factor), i index for physical interpretation of a given variable y, (i = T, R z, F c, F f, F N, k c, P c), t machining time corresponding to a value of y i, T tool life in a given experiment, j polynomial degree index (j = 1,,.., r). As a first step of the procedure, the measurements of flank wear VB and roughness parameter R z we taken in order to determine the tool life curves VB = f(t), determine tool live T values for critical tool wear, compute the average values of R z roughness parameter for machining conditions selected for each experiment of the experimental design. In the next step, measurements of the cutting force components for the center point of the design and then measurements of the cutting force components for machining conditions selected for the other points of the design and for the given values of tool flank wear. These measurements were aimed at determining the relationships between cutting force components and tool wear measure as well as determining their average levels within the tool life period. 3. Mathematical model for the WT3-1 titanium turning process An experimental model for the titanium turning process was built, based on typical program of polynomial regression and multiple stepwise regression. This model was used in the multi-criterion (multiple goal) optimization, allowing for the weights standing for significance of particular output quantities [6]. In this model, the following relationships are included: wear curve VB = f(t/t), dimensionless relationships in accordance to formula 1 for roughness parameter R z, for the cutting force components F c F f, F n and the regression for T, R z, Q v, P c, F c, F f, F n, k c versus cutting parameters. The wear curve for dimensionless coordinate t/t (Fig. 1) was determined with typical polynomial regression analysis program, assuming the polynomial order n = 9, which has been defined by the formula: r r t = + j j= 1 T VB a a ()

Mathematical model of titanium turning... 31 It is obvious that at the moment t =, the tool wear equals, so that the a constant factor value in the formula () should be zero. Still, the suggested value of a is above zero, according to computation, since it is nearly equal to the value of the cutting edge radius r n =,. If the elastic-plastic displacement of the tool edge area is neglected, it can be assumed that for the time close to zero (t > ), the tool edge abrasion on the tool flank takes place on the cylindrical part of the flank, height of which is equal to the tool cutting edge radius r n. With this assumption, it can be stated that right after the beginning of the cutting process, the height of abrasion VB on the flank face is equal to the radius r n. The values of the regression factors are listed in Table. Table. Regression factor values for the tool edge wear vs. dimensionless t/t ratio No. Designation Regression factor values a.1973645753 rn =. 1 a1 4.1715771536 a -56.167394875873 3 a3 38.33589913 4 a4-145.1549188373 5 a5 3173.53341436 6 a6-495.944943669 7 a7 3465.435819143 8 a8-158.8566637776 9 a9 83.469166685179 Tool Tool flank wear VB VB = f(t/t) = f(t/t), [mm] mm,35,3,5,,15,1,5 -,5 -,1 Tool flank wear VB = f(t/t) Serie1 Serie Serie3,,4,6,8 1 1, Dimensionless coordnate t/t Fig. 1. Tool edge wear VB vs. dimensionless coordinate t/t. Serie 1 measured points, Serie statistical model (approximation by a polynomial of n = 9 order), Serie 3 regression residuals graph

3 J. Zawora Statistical assessment: coefficient of multiple correlation between VB, t/t: R (VB,t/T) =.974, Fisher ratio F = 316.48, number of measurements N = 163, number of the regression factors K = 9, critical F ratio for the assumed level of significance α =.5, for the number K of the numerator degrees of freedom and the number N-K-1 of the denominator degrees of freedom (according to the F ratio definition) F cr(α,n-k-1,k) = 1.95. The adequacy of the regression was examined by the F-test, which consists in determining the ratio which has been defined by the formula 3. F( α, N K 1, K ) Fkr ( α, N K 1, K ) 1 (3) If the ratio F/F kr 1 then the regression is adequate which is the case in our calculations (F/F kr = 16.3). According to formula 1, the dimensionless relationships were determined for the roughness parameter R z, cutting force F c, axial force F f and radial force F n (formulae 4 7). The listed relationships take the following shape: Rz t t =,969 +,1, 71 R T T zśr, (4) cśr, 3 Fc t 1,13 1,3, 71 t t = + 1, 47 F T T T (5) Ff t t =,864,56 +,58 F T T f, śr (6) Fn t t =,98 +, 4 +, 3 F T T nśr, (7) The dimensionless relationship defined by the formula 4 is illustrated in Fig..

Mathematical model of titanium turning... 33 Dimensionless coordinate Rz/Rz 1,4 1, 1,8,6,4, -, -,4 Rz/Rz,śr = f (t/t) course Serie1 Serie Serie3,,4,6,8 1 1, Dimensionless coordinate t/t Fig.. Roughness parameters ratio Rz/Rz,śr vs. dimensionless coordinate t/t. Serie 1 measured points, Serie statistical model, Serie 3 regression residuals graph The mathematical model regression for describing WT3-1 titanium turning process is presented below (formulae 8 15) ln(t) = 7.634+4.486 v c 4.794 v c 1.7 f 1.384 v c f.97 v c a p 4.151 f a p (8) R z =.945 +.731a p +.3 v c + + 185.71 f.81v c a p 5.9635 f a p (9) F c = 171.96 + 16v c + 857.3f + 8.44a p.3441v c 3559.f -18.86578a p + 179.3v ca p (1) F f = 17.8 5.11v c 73.149f 73.175a p + + 9.1799 v cf +.918v ca p + 417.7516 f a p (11) F n = 69 + 599.54 f + 344.11 a p +.1196 v c 8.555 a p 3.1544 v c a p +187.81f a p (1) k c = 396.773 + 6.8496 v c 1567.671f 55.1a p 1.35 v c + 1351.75 f + 1f a p (13)

34 J. Zawora P c =.393.14 a p.1v c.133 f +.357 v c a p +.541v c f +.57348 f a p (14) Q v = a p f v c (15) The depth-of-cut quantity a p on Figures (3 1) is subject to variation within.5 do 1. mm along the velocity axis. Tool life T = f(v c, f, a p ), min Tool life T, min 1 8 6 4 1 1,9 16,6 1,3 7,4 35,3 8-1 6-8 4-6 -4 -,5,91,165,3 Fedrat f, Feed f, mm/rev mm/rev Fig. 3. Tool life relationship T = f(vc, f, ap) Surface roughnes parameter R z = f(v c,f,a p ) Roughnes parameter R z (DIN), um 15 1 5 1 15,5,9 6,4 31,8 37,3,3,189 15-1-15 5-1 -5,78 Fedrat f, Feed f, mm/rev mm/rev Fig. 4. Rz parameter relationship Rz = f(vc, f, ap)

Mathematical model of titanium turning... 35 Cutting force F c = f(v c, f, a p ) Cutting force Fc, N 14 1 1 8 6 4 1 15,5,9 6,4 31,8 37,3 1-14 1-1 8-1 6-8 4-6 -4 -,3 Posuw f, Feed,5 mm/obr f, mm/rev Fig. 5. Cutting force relationship Fc= f(vc, f, ap) Axial force F f, N 1 8 6 4 1 Axial force F f = f(v c, f, a p ) 15,5,9 6,4 31,8 37,3 8-1 6-8 4-6 -4 -,3,17,133 Fedrat f,,5 Feed f, mm/rev mm/rev Fig. 6. Axial force relationship Ff = f(vc, f, ap) Radial force Fn = f(vc,f,a p ), N Radial force F n, N 6 5 4 3 1 1 15,5,9 6,4 31,8 37,3,3,189 5-6 4-5 3-4 -3 1- -1,78 Fedrat f, Feed f, mm/rev mm/rev Fig. 7. Radial force relationship Fn = f(vc, f, ap)

36 J. Zawora Specific force of cut k c = f (v c, f, a p ), N/mm Specific force of cut k c, N/mm 35 3 5 15 1 5 1 1 3 3 3-35 5-3 -5 15-1-15 5-1 -5,5,133,17 Fedrat f,,3 Feed mm/obr f, mm/rev Fig. 8. Specific cutting force relationship graph kc = f(vcf,ap) Cutting power P c = f(v c, f, a p ) Cutting power P c, kw,45,35,5,15,5 -,5 1 15,5,9 6,4 31,8 37,3,35-,45,5-,35,15-,5,5-,15 -,5-,5,78,189 Fedrat f,,3 Feed mm/rev f, mm/rev Fig. 9. Cutting power relationship graph Pc = f(vc, f, ap) Material remowal rate Q v = f (v c, f, a p ) Material remowal rate Q v, mm 3 /s 5 15 1 5 1 1 3 3,3,189-5 15-1-15 5-1 -5,78 Fedrat f, Feed mm/rev f, mm/rev Fig. 1. Material removal rate relationship graph Qv = f(vc, f, ap)

Mathematical model of titanium turning... 37 Adequacy of the regression was tested with the F-test at the significance level α =,5 (Table 3). Significance of the individual factors in the regression was test with the t-test at the same significance level. Table 3. Assessment of the regression significance and of the regression factors significance Regression number Degrees of freedom N-K-1 K computed Value of F-ratio Critical α =.5 F F kr Coefficient of correlation R Critical tkr ratio at α =.5 8 13 6 16.677.955 5.644.945.16 9 14 5 1.457.96 78.816.994.145 1 1 7 348.51.91 119.77.998.179 11 13 6 451.435.955 15.77.999.16 1 13 6 36.395.955 1.15.97.16 13 13 6.53.955 6.947.95.16 14 13 6 13.35.955 7.93.995.16 4. Multi-criterion optimization For technological applications, there are usually expectations connected with multi-criterion optimization with limitations which is called sometimes multiplegoal optimization or poly-optimization. Problems connected with this optimization type can be found in the available references [8, 9]. Practically, the dominant problem of multi-criterion optimization is replaced by a set of single criterion optimization problems. The adopted multi-criterion optimization usually includes the following the partial goals: the possibly highest productivity and the longest tool-life as well as the possibly lowest cutting power and surface roughness measured by the parameter R z (DIN). The additional goal appears as determining the other performance features which have not been subjected to optimization, but remain technologically significant in the examined process. The limiting factors appear as some conditions precluding the use of force higher than permissible due to a number of factors like feed rate mechanism strength, interchangeable cutting insert strength, permissible deflection of the machined part, but also ensuring tool-life period T longer than 5 min. (according to the ISO standard). In order to determine the dominant optimization criterion, it is necessary to know the optimum values of the individual goal functions understood as minimum and maximum. It means that multi-criterion optimization should be preceded by single-criterion optimization for the individual functions of partial goals. For this purpose, custom optimization programs have been designed for single- and multi-criterion optimization employing a principle of systematic search. The results were presented in Table 4. In order to determine the optimum set of machining parameters for the turning process, multi-criterion optimization based on correlation and weights and

38 J. Zawora making use of dimensionless assessment of the OPT quantity as described by the following relationship: α Y Y OPT = W Y Y n i max, i i i (16) i= 1 max, i min, i max W i weight of each individual single-goal optimization criterion (i = 1,, 3, n), α coefficient; α = +1 for a maximized quantity, α = 1 for a minimized quantity. Table 4. Results of a single-goal optimization Optimum values of the output quantities and corresponding machining parameters No. Maximum values vc f ap 1. T = 834.585 min 14.848.13.7. Rz = 18.8 µm 4..3.5 3. Fc = 1391.937 N.77.93.5 4. Ff = 999.619 N 5.455.6.5 5. Fn = 59.856 N 1..3 1.818 6. kc = 351.596 N/mm 3.333.5 1. 7. Pc =.535 kw 4..3.5 8. Q = 5. mm 3 /s 3.333.5 1. No. Minimum values vc f ap 1. T = 68.334 MPa 5.455.67.44. Rz = 68.1 µm 13.939.5 1. 3. Fc = 399.568 µm 39.697.55 1. 4. Ff = 1.1 mm 1..3 1. 5. Fn =.4 g/cm.77.5.5 6. kc =.146 % 4..3 1. 7. Pc = 9.639 µm 1..5 1.3 8. Qv = 8.333 mm 3 /s 1..5 1. A value of the global assessment OPT may vary within the following limits: OPT Wj < < (17) W n j = Wi (18) i= 1 The wanted optimum set of machining parameters v c, f, a p for the adopted variant of a single-goal fulfillment can be understood as such, for which the OPT expression value will be the highest. By assuming that all performance quantities

Mathematical model of titanium turning... 39 of the process are of the same significance [1] which is equivalent to assigning them the weights W j = 1, i = 1,, 3,, n, the following set of optimum parameters was found and is presented in Table 5. Table 5. Results of multi-criterion optimization for the adopted dominant criterion Variant: Qv max, Tmax, Pc min, Rz min Optimum parameters (OPTmax): vc = 14.55 m/min. f =.53 mm/obr. ap =.91 No. Optimum values of performance quantities 1. Qvx = 6.64 mm 3 /s. Tx = 831.837 min 3. Pcx =.75 kw 4. Rzx =.78 µm 5. Fcx = 31.571 N 6. Ffx = 588.7 N 7. Fnx = 64.17 N 8. kcx = 89.43 N/mm Non optimum parameters (the worst OPTmin): vc = 3,3 m/min, f =, 3 mm/obr, ap =,485 Lp. Non optimum values of performance quantities 1. Qvn = 86.134 mm 3 /s. Tn = 5.118 min 3. Pcn =.531 kw 4. Rzn= 14.99 µm 5. Fcn= 98.136 N 6. Ffn = 588.7 N 7. Fnn = 48.487 N 8. kcn = 5.676 n/mm Y i current value of the process performance parameter subjected to optimization, quantified with the given increment, Y max, i (Y min, i) maximum (minimum) value, n number of the process parameters subjected to optimization 5. Conclusion Typical solutions presented in the paper can be used for further research of the process, for improving the elaborated program of multi-criterion optimization with limitations and for applying the other types of multi-criterion optimization prepared by other known authors in order to find the best solution to problems occurring in the examined process. The process of turning titanium and its alloys is difficult to examine, due to lack of stability. The developed mathematical model as well as optimization algorithms may become a foundation for the post-processing software in the CAM systems for automatic tool-path programming for the CNC machine tools.

4 J. Zawora References [1] P. LASKOWSKI, W. HABRAT, K. KRUPA, J. SIENIAWSKI: Toczenie wykończeniowe stopu tytanu Ti-Al.-4V z zastosowaniem HCP. Stal, Metale i Nowe Technologie, (13)11-1, 56-6. [] J. MAŁECKA: Stopy tytanu na osnowie faz międzymetalicznych TiAl(γ) i Ti 3Al(α ). Mechanik, (13)1, 888-89. [3] K. KUBIAK: The influence of forging and hit treatment conditions on the microstructure and fatigue strength of die forgings made of two-phase titanium alloys Ti-6Al-4V and Ti-Al6-Mo-Cr. Advances in Technology of Machines and Mechanical Equipment, 1(1997), 43-56. [4] P. NIESŁONY, W. GRZESIK, W. HABRAT: Experimental and simulation investigations of face milling process of Ti-6Al-4V titanium alloy. Advances in Manufacturing Science and Technology, 39(15)1, 39-5. [5] K.E. OCZOŚ: Kształtowanie ubytkowe tytanu i jego stopów w przemyśle lotniczym i technice medycznej, część I. Mechanik, 81(8)9, 888-89. [6] K.E. OCZOŚ: Kształtowanie ubytkowe tytanu i jego stopów w przemyśle lotniczym i technice medycznej, część II. Mechanik, 81(8)1, 753-756. [7] K. MAŃCZAK: Technika planowania eksperymentu. WNT, Warszawa 1976. [8] W. POGORZELSKI: Teoria systemów i metody optymalizacji. OW PW, Warszawa 1996. [9] M. KOWALCZYK: Application of Taguchi and Anova methods in selection of process parameters for surface roughness in precision turning of titanium. Advances in Manufacturing Science and Technology, 38(14), 1-35. [1] W. TARNOWSKI: Modelowanie systemów. Wydawnictwo Uczelniane Politechniki Koszalińskiej, Koszalin 4. Received in March 16