Ngày đăng: 24/11/2016, 18:08
Journalof ELSEVIER Journalof MaterialsProcessingTeelmology59 (1996) 81-94 Materials Processing Technology Flashless closed-die upset forging-load estimation for optimal cold header selection Matthew O'Connell a, Brett Painter a,*, Gary Maul b, Taylan Altan a aERC for Net Shape Manufacturing, Ohio State University, Columbus, Ohio 43210, USA blndustrial, Systems, and Welding Engineering Dept., Ohio State University, Columbus, Ohio, USA Industrial Summary In cold forming operations, the two primary considerations in determining process feasibility are load and energy required to form the part These two factors determine the press size and therefore the maximum production rate Close estimations of these values during the process planning stage allow for accurate machine sizing and lead to higher efficiency However, methods for estimating forming loads and energies of flashless closed-die upsets are not well researched The objective of this study was to develop guidelines for this purpose Experiments were conducted to measure the load during many flashless closed-die upsets These results were compared to predictions made using a mathematical model (modified slab method) and an FEM analysis (DEFORM 2D) to gauge the ability to accurately predict forging loads Additionally, FEM was used to study the sensitivity of the forging load to friction conditions, part size, and average material strain The results indicate that the modified slab method predictions are quite accurate in some cases but overestimate the load by as much as 35% in others The DEFORM 2D load predictions are more consistent but g e n e r a l l y underestimate the load by 5-20% Introduction This project was undertaken in cooperation w i t h the National Machinery Company (NMC), located in Tiffin, Ohio, which supplies cold and warm headers for the forging industry When designing cold and w a r m headers, NMC needs to estimate t h e forming loads at each station to determine t h e sequence of progressive dies and the size of t h e header necessary to produce the part To select the appropriate size header for a given application, the loads and energies of each die station are summed The die stations are used for a mix of metal forming operations, and may include forward extrusions, backward extrusions, open upsets, flashless closed-die upsets, a n d / o r trimming and piercing operations Guidelines for estimating forming loads are commonly available for most of these operations and with a variety of materials Closed-die flashless upsets, however, have not been well investigated * Corresponding author 0924-0136/96/$15.00© 1996ElsevierScience S.A.All rightsreserved PI10924-0136(96) 02289-3 It is important to correctly estimate the forming load for each die station Oversizing a header w i l l result in an unnecessarily large machine cost and lower production rate Part cost will therefore be higher than necessary On the other h a n d , undersizing a header will cause overloading of t h e press, resulting in frequent downtime for maintenance and repairs This paper describes the experiments and results obtained while investigating the flashless closeddie upset process The experiments included running closed-die upset forgings with a variety of b i l l e t materials and process conditions Two t h e o r e t i c a l models were then used to compare predicted forging loads to these experiments The theoretical models tested were the modified slab method and FEM Background The tooling investigated in this study is seen in Fig The tooling consists of a flat faced punch and a simple round die cavity The edges of the punch 82 M 'Connell et al / Journal of Materials Processing Technology59 (1996) 81-94 and die inserts are square The main forging problem in this process is the lack of complete die cavity f i 11 at the corners [1] The typical load-stroke curve for this operation has two distinct stages: an upsetting stage and a comer filling stage, illustrated in Fig At t h e beginning of the stroke, the billet is cylindrical and does not make contact with the die casing As t h e process begins, the billet undergoes a process similar to an open upset until it makes contact with the die casing (shown to the left of the dashed line in Fig 2) At this point, the forming loads increase dramatically as the comer filling occurs (shown to the right of the dashed line in Fig 2) The load required to then fill the cavity is "extremely h i g h (3 to 10 times the load necessary for forging the same part in the cavity without corner filling)" [2] [3] This corner-filling stage is the focus of interest in this project In particular, one is interested in finding the required load to produce a specified amount of comer filling The amount of comer filling is expressed as a ratio of the length of die wall in contact with the deformed billet to the length of the die wall, and is termed the percentage of die w a l l contact (%DWC) A review of upsetting technology indicates a need to further investigate flashless closed-die forging with respect to the forging loads required to a t t a i n Up o J / I Corner I filling I I I STROKE I Fig 2: Typical load-stroke curve of an enclosed upset process Filler Punch Casing Punch Die Casing Filler sufficient die cavity fill Relatively little experimental data has been published in t h e literature to aid estimating the forging loads in enclosed upsets Likewise, little theoretical work is available to estimate forging loads in enclosed upsets This project consists of three main tasks: 1) collecting experimental data, 2) estimating forging load using the modified slab method, and 3) estimating forging load using FEM The main procedures and results of each task are detailed in the following sections Experimental Program Overview Kic Fig 1: Tooling for the enclosed die upset investigated in this project For the experiments, it was necessary to collect data for a range of upset geometries and a variety of billet materials Hence, four different b i l l e t materials were selected for the study, and t h r e e different upset geometries were designed, to yield a test matrix of twelve configurations 83 M "Cormellet al /Journal of Materials Processing Technology 59 (1996) 81-94 At the start of each testing configuration, the punch-to-die distance was set such that was no material contacted the die wall after the blank was upset (an open-die upset) The BDC position of the p u n c h was then decreased until a uniform band of die wall contact was obtained At this distance, testing began b y heading three to ten blanks This BDC was then decreased by 0.005 inch to 0.010 inch Again, three to ten blanks were headed This procedure was repeated until the m a x i m u m punch tonnage (approximately 140 U.S ton / i n 2) w a s reached A data acquisition system recorded the p u n c h load and p u n c h position over time for each trial The data acquisition system consisted of a 286 PC equipped with "Labtech Notebook" software The system w a s configured to begin taking data w h e n the position sensor output of the punch reached a specified voltage Therefore, the system was triggered simply by cycling the header and data collection began at a consistent ram position each trial Data samples of all input channels were taken at the s a m e instant at a rate of 60 Hz • the ratios (Q = h0/d0) of initial upset head height (h0) to initial diameter (do) are different (QB = 0.69, Q A = 1.04, Qc = 1.50) d1 = 0.755" I I h' I ( : ) ho 0.500" = 0.500" e Fig 3: Characteristic aspects of parts produced in Tests A, B, and C Tooling Design To simplify the study, the tooling design was m a d e as simple as possible A schematic of the tooling used in the tests is shown in Fig above, divided into two halves The left half of the d r a w i n g shows the tooling with the punch not in contact with the workpiece; the right half of the d r a w i n g displays the punch at bottom dead center (BDC) In the figure, the die is stationary and only the p u n c h moves The tooling was m a d e to upset axisymmetric parts from round bar stock, illustrated in Fig In all cases, the initial blanks were formed from 0.500 inch diameter bar stock, and were headed to a diameter of 0.755 inches However, three different billet lengths were selected, namely L = 1.020, 0.845, and 1.250 inches, as indicated in Table The tolerances on the overall blank lengths were specified as +0.001 inches, in an attempt to minimize workpiece v o l u m e variations, thereby reducing load variations during forging The different initial billet lengths, from Table 1, result in the following significant changes in the final part dimensions and process: • the final head heights are different (hlB = 0.142 inch, hlA = 0.218 inch, hlc = 0.333 inch • the final-head-height to final-diameter ratios are different (R = h l / d 1, R B = 0.19, R A = 0.29, R c = 0.44) Billet Materials The headers manufactured by the National Machinery C o m p a n y produce a wide variety of parts, from different materials such as copper alloys, various steels, and aluminum alloys Therefore, the following materials were chosen for the study: • CDA 101 Copper (a 99% pure copper) • 1008-15 Steel (low carbon steel) • 1038 Steel ( m e d i u m carbon steel) • 304 Stainless Steel Tensile specimens of each material type were fabricated to determine basic material properties The results of these tests are given in Table Compression specimens were upset to determine the strength coefficients and strain hardening exponents In a flashless closed-die upset operation, lubrication between the workpiece and die wall is not generally viewed as a factor contributing significantly to the load Therefore no petroleum based lubricants were used during testing Only the standard phosphate wire coating served as a lubricant on the 1008-15 and 1038 steel No lubrication was used for CDA 101 copper A molybdenum-disulfide dry lubricant was used on the 304 stainless steel to prevent part seizure in the die cavity M 'Connell et al /Journal of Materials Processing Technology 59 (1996) 81-94 84 Table 1: Details of the upset testing configurations Part Aspect Test A Test B Test C Billet cut-off length, L 1.020 inch 0.845 inch 1.250 inch Upset head height, ho 0.520 inch 0.345 inch 0.750 inch Target head heighti h 0.218 inch 0.142 inch 0:333 inch q's upset, (ho/d0) 1.04 0.29 0.69 0.19 1.50 0.44 h i / d l ratio of h e a d Material spread, dl/d0 1.5 1.5 1.5 Average true strain, e 0.82 0.82 0.82 Table 2: Material property data from tensile and compression tests Young's Modulus, E (Mpsi) Yield Stress, Sy (ksi) Ultimate Tensile Strength, SUTS (ksi) Elongation 1008-15 Steel 30.0 1038 Steel 30.8 304 Stainless Steel 21~5 41.1 39.0 69.7 61.0 41,4 49:0 72.1 100 17 36 15 55 84 81 67 76 54 85 101 200 0!12 0:21 0:15 0.35 CDA Copper 22.5 101 (%) Reduction in Area (%) Strength Coefficient, K (ksi) Strain H a r d e n i n g Exponent, n To promote die cavity fill (bulging), material movement is typically restricted at the punchworkpiece interface Hence, the punch face was roughened with a coarse stone in a lathe and co lubricant was used at this interface during any of the tests Measurement and Characterization of Parts Fig shows a number of parts produced with varying degrees of die cavity fill The dark band around the part head is the area of die wall contact The blanks s h o w n are from Test A, 1008-15 steel The part on the far left required about 20 U.S tons of forming load (45 U.S tons/in 2) and shows little die wall contact, while the part on the far right required 60 U.S tons of forming load (135 U.S tons/in 2) and shows m u c h die wall contact After all the parts were produced, the amount of die cavity fill in each part needed to be measured It was therefore first necessary to decide h o w die cavity fill could best be characterized The characterization was selected based on what a design engineer will find valuable and by evaluating the statistical validity of particular measured dimensions For the initial evaluation of resulting part dimensions, eight parts from Test A, 1008-15 steel were selected for measurement Four of the parts chosen exhibited the minimum amount of die wall contact, and the other four parts exhibited the maximum amount of die wall contact The comer underfill dimensions were measured and named upper horizontal (UH), upper vertical (UV), die wall contact (DWC), lower vertical (LV), and lower horizontal (LH), as shown in Fig An optical comparator set at 20X magnification with digital position readouts was used to measure the dimensions Each dimension was measured in four locations around the part head (with a 90 ° spacing between measurement locations) The averages and standard deviations of each dimension were then M O'Connell et al / Journal 85 of Materials Processing Technology59 (1996) 81-94 W Fig 4: Progression of parts produced with increasing die cavity fill (Material 1008-15 steel, Test A) calculated (based on 16 measurements, parts times m e a s u r e m e n t s / p a r t ) , and are p r o v i d e d in Table At minimum fill, the UV and DWC dimensions show a variation of half or more their averages, while the UH, LH, and LV dimensions show very low variations At m a x i m u m fill, all the dimensions except DWC show variations of half their averages, while the DWC is very low Because the DWC value shows the lowest variation of the five dimensions at high loads, it suggests that the DWC value is the most reliable value for developing a relationship between punch load and comer filling It is also reasonable to assume that the value of < DWC is of most concern to the process designer Therefore, the value of DWC is used to c o m p u t e a measure of comer filling as given in Equation and Fig length o f D W C %DWC lO0 (1) total h e a d height Equation was used to find the comer filling of each upset of the study A sample set of the experimental results is presented below This UPPERHORIZONTAL(UH) UPPER ~ '/~ J = VERTICAL (UV) DIE WALL CONTACT(DWC) , ~ LOWERVERTICAL(LV) LOWERHORIZONTAL(LH) %DIE WALL CONTACT _ - Dwc * 100 UV+DWC+LV Fig 5: Characterization of enclosed (flashless) upset part dimensions and cavity fill A'I 'Connell et al /Journal o f Materials Processing Technology59 (1996) 81-94 86 Table 3: Typical measured results of comer filling for minimum D W C and maximum D W C (each value is an average of 16 measurements) Dimension Name Minimum DWC Value 3s Std D e v DWC Maximum Value 3s Std D e v U p p e r Horizontal (UH) 0.028 inch 0.007 inch 0.011 inch 0.006 inch U p p e r Vertical (UV) 0.095 0.040 0.022 0.012 Die Wall Contact (DWC) 0.037 0.026 0.175 0.023 Lower Vertical (LV) 0.104 0.014 0.018 0.011 Lower Horizontal (LH) 0:034 0.005 0:010 0.004 measure of comer filling was used to compare with predictions m a d e of the theoretical models (modified slab m e t h o d and FEM) as described in later sections of this paper 150 140 130 120 Experimental Results After all the test parts were produced and measured, the results for each material and geometry combination were compiled The upset loads and percent die wall contact (%DWC) for each BDC setting were averaged from the three to ten blanks headed at each setting Therefore, for each combination of material and test geometry, a series of data points relating punch load to percent die wall contact was found Fig shows the data taken f r o m 1038 steel for all three test geometries (tests A, B, and C) Standard deviation values for the data points were calculated, but are not s h o w n in Fig for clarity This figure shows the required punch pressure to obtain a given %DWC for each of the test geometries The punch pressures were determined by dividing the punch loads by the area of the punch face ( a p u n c h = 0.444 in2) In each series, the relationship between punch load and die wall contact is a positive one as expected An offset exists between the three tests The test C series corresponds to the part with the greatest final head height and greatest number of diameters upset ( h o / d o = 1.50), and lies beneath the other two curves Similarly, the test B corresponds to the part with the smallest final head height and least number of diameters upset ( h o / d o = 0.69), and lies above the other two curves This trend was found to be consistent for all of the materials tested, with one exception: the 1008-15 steel material exhibited little difference between the results for Tests A and B 110 ~ " 100 ~_ g 9o uJ ~ 80 •/ ~" Q 60 50 • Test B IhJd o = 0.7) • Test A (hJd o = 1.0} • Test C (hJdo = 1.5) 4O ~E W~L C ~ T ~ T ~ - 30 20 10 ENCLOSED UPSET 0 10 20 30 40 50 60 70 % Die Wall Contact 80 90 100 Fig 6: Closed-die upset results for 1038 steel (tests A, B, and C) Fig shows a summary of all the experimental results All four material types are represented in the figure The three tests for each material have been collapsed to a single curve, but a correction factor model has been supplied to account for differences in upset geometry Note that this chart is designed for upsets where the average true strain of the material is e = 0.82 (or, dl/d0 = 1.5), as described in the Tooling Design section above To use this figure to compute the required punch load to ~/~ O "Connell et al / Journal o f Materials Processing Technology 59 (1996) 81-94 obtain a specific amount of die wall contact, the following procedure should be used: Determine the m i n i m u m %DWC required in the flashless closed-die upset M o v e f r o m that %DWC on the x-axis vertically to the corresponding billet material curve Move horizontally from the curve to find the base p u n c h pressure on the y-axis Look up the difference value, D, from Table for the a p p r o p r i a t e billet material Substitute the number of diameters upset ( h o / d o ) and the difference value, D, into the correction factor equation to find CF, given in Equation Find the total p u n c h pressure by adding the base punch pressure obtained from step to the correction factor, CF Multiply the total punch pressure by the projected area of the p u n c h face to determine the total load, as in Equation 87 curve for 1038 steel gives a base punch pressure of approximately 124 U.S tons / inch 150 - 10 20 30 40 / 60 70 80 90 / "~$~Y 140 130 120 50 140 A/ / 100 150 Z i d / 130 120 110 ~"~.100 - - 100 g 90 90 8o 8O ~ 70 7O o 6O #- so • 60 5O • 4O 3O Table 4: Look-up table of difference values, D, for the billet materials studied 2O 10 Material C D A 101 Copper 1008-15 Steel 1038 Steel 304 st'ainless Steel Difference, D 10 (U.S ton / in 2) 13 14 20 27 30 40 50 60 70 % Die Wall Contact 80 90 100 Fig 7: Closed-die upset s u m m a r y CF(Correction F a c t o r ) = D [ - h~-~/ (2) L = pA (3) where: 20 L = Total p u n c h load p = Total punch pressure A Project punch area To illustrate the use of the chart, consider this example: 1038 steel is being cold formed in a flashless closed-die upsetting station The initial diameter of the bar is 0.67 inch, which is being upset to a final diameter of 1.00 inch (punch area = 0.80 inch2), to yield an average true strain of e = 0.82 ( d l / d = 1.5) for which the experimental data above was collected The process calls for h o / d o = 0.90 (diameters upset) with a requirement of 65% DWC Thus, the initial head height is 0.600 inch and the target head height is 0.267 inch From Fig 7, the The appropriate D value for 1038 steel is found in Table to be 20 U.S tons / inch Next, the correction factor, CF, is calculated by substituting the D value and the number of diameters upset ( h o / d o = 0.90) into Equation 2, resulting in CF = 15 U.S tons / inch Summing the base punch pressure of 124 U.S tons / inch a with the CF of 15 U.S tons / inch gives a total punch pressure of 139 U.S tons / inch From Equation the total required punch load of (139 U.S tons / inch2)(0.80 inch 2) = 111.2 tons is calculated Discussion of Experimental Results Fig shows that each material has a different obtainable degree of die cavity fill at typical maximum tooling pressures (140 U.S tons / inch2) This gives the process designer insight to determine whether additional operations are necessary to properly form a part For example, if a flashless closed-die upset process requires 95% DWC with 1038 steel, more than one hit will be needed to form the part, because the m a x i m u m obtainable die wall A£ O'Connell et al /Journal of Materials Processing Technology 59 (1996) 81-94 88 contact is 78% in one blow with the maximum pressure of 140 tons/inch Fig allows the required forming load to be found as functions of the billet material and the degree of upset (G/d0) However, the experiments not evaluate the forming load dependence upon the average true strain, E, of the part In all the experiments performed, the average true strain E = 0.82 (E = In ] d l / I = 0.82) In a closed-die flashless forging, this value typically is not exceeded Occasionally however, there can be requirements for E to go as high as 1.62 Additional experiments would be necessary to find this dependence empirically However, this issue was addressed during the FEM analysis of the upsetting, and is described in a later section of this paper Slab Method The slab m e t h o d analysis is used to predict the required forging load based on material properties and process geometries and conditions Slab method analysis is a common tool and is described in many texts In particular, evaluation of an axisymmetric homogeneous open-die upset can be found in [4] The details of the derivation are omitted here; however, the final formulation to compute the maximum tooling load, L, based on part geometry and material properties is given in Equations and as: mdl L = d ~ ~f (1-~ cJf = K - I n ) (4) 3a/3-hj Application Upsetting of Slab Method to Closed-Die In a case where the material is forced to fill a cavity, the operation cannot be considered simply as an open-die upset This is due primarily to the difference in the assumed and actual deformation zones Because the material flow is restricted by the cavity in a closed-die upset, less workpiece volume deforms than does in an open upset Hence, a modification to the traditional slab method is made to account for the reduced deformation zone when trying to predict the forging load The maximum tooling load occurs at the end of the upset stroke At the end of the stroke, the part is partially in contact with the die wall, and can be divided into three distinct layers (slabs) of similar deformation characteristics as shown in Fig When the part is in contact with the die wall, the middle layer (slab in Fig 8), is assumed to not deform further; rather, it is considered to undergo rigid-body motion The top and bottom layers (slabs and in Fig 8), however, continue to deform because they are not restricted by the die wall To compute the actual forging load using the modified slab method, it is important to consider only the top and bottom layers The necessary changes to the load formulation are detailed below In Equation 5, the only geometrical variables required to predict the forging load are the final diameter, dl, and the initial and final part heights, h0 and hi Therefore, these variables must be corrected to reflect only the top and bottom layers Equations and are modified and presented as Equations and below (5) •(1+ L = -~(d,)2-(~f where: L dl sf m H1 K h0 n = maximum load on tooling = final upset head diameter = material flow stress = shear friction factor = final height of upset head = Hollomon strength coefficient = initial height of billet to be upset = material strain hardening exponent However, this formulation is valid only for an open-die upset, and is therefore not usable for predicting the load in the closed-die upset discussed in this paper Equations and therefore need to be changed to make them applicable to closed-die upsets The changes result in what is called the modified slab method of analysis ~ o f = K.(ln2h~_/n m-d~ ,/ (6) 3~/3'2h,) (7) (2h;) where: = required forging load to closed-die upset part to a given DWC d = final diameter of the upset part sf = material flow stress m = shear friction factor h~= final height of layer (or 3) K = Hollomon strength coefficient h~,= initial height corresponding to the final height of layer (or 3) n = material strain hardening exponent L 89 M O'Connell et al./Journal of Materials Processing Technology 59 (1996) 81-94 A ///; B //// //// The material boundaries indicated b y A and B in Fig are assumed to be straight, thus making both layers I and truncated cones The height of layer I is assumed to be equal to the height of layer The amount of upper and lower horizontal (defined in Fig 5) is assumed to be one-half the value of hr This assumption is based on the values presented in Table 3, and will not cause significant errors if incorrect To find the volume of material in layer (or 3), the diameter of the top surface of layer (or bottom surface of layer 3), to be called d2, is found using assumption above Specifically: d = d, - hf ///, (9) Then the volume of layer (or 3) is found using the formula for a truncated cone, given as: Fig 8: Dividing the part into layers for modified slab analysis v = '"3 , where: To use the modified slab analysis, the initial head diameter and head height must be known Also, the desired final head diameter and height must be known, as well as the desired amount of die wall contact As usual, the appropriate material properties m u s t also be known Then, the final slab height, h~, is found using Equation 8: h' = (1oo% o/M'IW(~ " ~ "" ~ " h (8) where: hf = final height of layer I (or 3) h I = final target head height of whole part (same as for slab analysis) To find the corresponding initial slab height, h0, the volume constancy principle is used The head geometry at the end of the stroke is known, and therefore the volume of material in layers and can be computed The stock diameter at the beginning of the stroke is also known, and so the corresponding height of deforming material can then be found Several assumptions about the problem are m a d e at this stage to estimate the volume of material in layers and 3: B + + (10) = area of bottom surface of layer 1, B m ~d[ B ' = area of top surface of layer B'= 4622 Finally, the corresponding initial height deforming layers, h f), is found using: ho = - - ' V Tg'.do 1, of the (11) The predicted forging load can then be calculated using Equations and above To illustrate the modified slab analysis method, an example using the same values as before is presented: Compute the forging load needed to cold form 1038 steel in a flashless closed-die station to 65% DWC The initial stock diameter is 0.67 inches; the final head diameter, dl, is 1.00 inches The initial head height is 0.600 inches and the final target head height is 0.267 inches Using a friction factor of m = 0.9 (selected based on FEM results, discussed later) and with Equation 7, the forging load is predicted to be 110 tons This value is in agreement with the prediction of 111.2 tons forging load found using the experimental data of Fig 7, described earlier To obtain this result, the friction 3'I 'Connell et al / Journal o f Materials Processtng Technology59 (1996) 81-94 90 factor was assumed equal to 0.90, because, during the experiments, the friction conditions between the punch and workpiece were artificially increased to promote die filling In the DEFORM simulations detailed later, the shear friction factor was found to be m = 0.90 140 (TesiCi) 120 : ~ loo ~ 80 6o • a_ Results of Modified Slab Method 13- 40 : ; 2o : : ExperimentalTest C I (1038 Steel) I - t - M o d i f i e d Slab Method | (m = 0.9) I i Figs 9a and 9b show the results obtained using the modified slab method to predict the forging pressures for 1038 steel, using the test configurations B and C The modified slab method data points are indicated by black diamonds, connected by solid lines The results are shown along with the corresponding experimental data, with 3-sigma error bars Evaluating the results for test B (thinnest upset head height, hi = 0.142 inch) of 1038 steel, the modified slab m e t h o d overestimates the forming pressures by approximately 20% However, for test C (thickest upset head height, h = 0.333 inch) of 1038 steel, the modified slab method predictions are nearly perfect This trend was also observed for CDA 101 copper and 1008-15 steel H o w e v e r the modified slab m e t h o d greatly overpredicted the forging pressures for 304 stainless steel The modified slab method results of Figs 9a and 9b are based on an assumption of the friction factor = 0.9 Since the exact friction conditions are unknown, it is useful to examine the sensitivity of the predicted forging pressures on the friction factor Fig 10 shows the predicted modified slab method forging pressures for a range of friction factors, for 1038 steel, test C The figure shows that as friction factor is increased from 0.5 to 0.9, the predicted forging pressure is increased by as little as 20% for low amounts of DWC and as much as 40% for high amounts of DWC 140 120 "-: 100 GO v i a_ 80 • 40 20 ExPerimentalTest B I (1038 Steel) I Modified Slab Method | (m = 0.9) L 10 L 20 L 30 L 40 L 50 L 60 I L 70 L 80 L 90 100 Fig 9a: Comparison of modified slab method to experimental data (1038 steel, Test B) 10 20 30 40 50 60 70 80 90 100 Fig 9b: Comparison of modified slab method to experimental data (1038 steel, Test C) 140 : E 91 120 100 05 80 03 60 I& I 40 ,I m :0.5 J o m = 0.7 O- 20 - - e - - m = 0.9 : i : [ 10 20 30 40 50 60 70 80 90 100 Fig 10: Effect of friction factor ~ predicted forging pressure (1038 steel, Test C) using modified slab method analysis DEFORM 2D FEM Simulation Version 4.1 of DEFORM 2D was used to simulate the experiments described in this p a p e r [5] Because the process is axisymmetric, only half a cross-section of the part was modeled Four steps of an example process simulation are shown in Fig 11 The tooling and workpiece geometries used in the simulation were obtained from Fig and Table The material properties (strength coefficient, K, and strain hardening exponent, n) were taken from Table A uniform initial mesh density was assigned to the workpiece with 600 elements The punch and die were considered rigid bodies and hence did not require meshes The punch dimensions were defined to overlap the die wall to prevent nodes from penetrating between the punch and die wall during the simulation A~ 'Connell et al /Journal of Materials Processing Technology 59 (1996) 81-94 STEP (-1) 91 STEP (40) 1.2~ 1.000 t~ e c "1" 750 750 SO0 2SO i 260 I 500 Radius J 750 25O =~rJO (in.) ,750 R a d i u s (in.) STEP(80) 12GD ~0 STEP (92) 12w,~ I o(l~ 1¸ooo c -v TSO 000 R a d i u s (in.) Fig 11: D E F O R M process s i m u l a t i o n of a c l o s e d - d i e u p s e t I 750 25O R a d i u s (in.) 92 M "Connellet al / Journal of Materials Processing Technology 59 (1996) 81-94 The friction conditions of the process were not specifically known; it was therefore necessary to estimate the friction factor at the workpiece-die and workpiece-punch interfaces The friction factor at the workpiece-die interface was defined to be rn~ = 0.1, because of the lubrication conditions on the materials described earlier The friction factor at the workpiece-punch interface, however, was defined to be a high rr~, = 0.9, for two reasons First, for the experiments, the punch face was roughened with a stone purposely to increase the friction Second, FEM models generally underestimate required forging loads because the code neglects elastic deflections of the press and work losses within the press Therefore, choosing such a high value of friction factor partially compensated for these effects, and caused the predicted forging loads to be more realistic Results of FE Modeling Overall, the results obtained from DEFORM were quite good when compared to the experiments Fig 12 shows the experimental data and the DEFORM results for CDA 101 copper, test B geometry For low percentages of DWC (20% - 60%), DEFORM predicted a forging pressure less than was found by experiment For high percentages of DWC (60% - 100%), DEFORM predicted a forging pressure more than was found by experiment In both cases, the error was not more than about 10% 140 120 c 100 though, DEFORM tended to underestimate the required forging pressures The error varied among the cases from zero percent up to about 25%, and averaged around 10% to 15% Also, the slope of the DEFORM predictions at low percentages of DWC agreed very well with experiment But at high percentages of DWC, DEFORM tended to predict a steeper slope than found from experiment The experimental results given in Fig are limited in two important aspects: (1)absolute upset part size and (2)average strain during upset FEM was used in this project to investigate the effect of these two factors on upset forging load DEFORM Investigation of Absolute Part Size FEM was used to study the effect of absolute part size on required upset load The relative flow geometry would be the same, as well as the number of diameters upset (ho/do) and the average true strain, E But the impact on the required forging load to obtain a specified %DWC was unknown To study this effect, the geometry of experimental test C was magnified by 0.5 times (do= 0.250", h0= 0.375" and dl = 0.375") and by 2.0 times (do = 1.000", h0 = 1.500" and d~ = 1.510") for further FE simulations These geometries still have an average true strain E = 0.82 (marl spread = 1.5) and upset ratio h o / d o = 1.5 Fig 13 shows the results of three DEFORM simulations (0.5, 1.0, and 2.0 times the original part size) As seen the figure, the difference among the three curves is very small, illustrating that the absolute size of the geometry does not significantly affect the relation between required upset load and percent die wall contact obtained (%DWC) 80 DEFORM Investigation of Part Strain o o 60 13 40 D_ 20 Experimental Test B • • D E F O R M (m0 = 0.9) 10 20 30 40 50 60 70 80 90 100 % Die Wall Contact Fig 12: Comparison of FEM analysis to exp data (CDA 101 copper) In all cases, DEFORM was found to reasonably predict the required forging pressure In general FEM was also used to study the effect of average part strain on required forging load All the tests of this project had the same average true strain ( E -0.82, dl/d0 = 1.5) However, it is common that other head geometries will result in different average true strains ( E ¢ 0.82, dl/d0 ¢ 1.5) from the same initial stock geometries To study this effect, two additional forging geometries were modeled in DEFORM The first had a larger diameter cavity (d = 1.000", E = 1.38, marl spread = 2.0); the second had a smaller diameter cavity (d a = 0.625", E 0.45, matl spread = 1.25) Both FE models began with the same initial billet dimensions as Test C (do = 0.500" and h o = 0.750") M O'Connell et al /Journal o f Materials ProcessingTechnology$9 (1996) 81-94 Fig 14 displays the DEFORM results of these simulations Each curve represents the forging load required to obtain a specified percent of DWC for a particular head geometry The figure shows t h a t there is a distinct shift from one curve to the next The curve that ties highest on the graph corresponds to the part with the greatest part strain; the curve lowest on the graph corresponds to the part with the lowest strain This figure indicates that the g r a p h of Fig is valid only for the particular a v e r a g e part strain, ~ = 0.82 Thus, in a flashless closed-die upset, the relation between the forming load and %DWC is dependent upon two variables: (1) the upset ratio (number of diameters upset, h o / d o ) and (2) the average true strain, E, of the upset p a r t head 140 120 c • 93 DEFORM Test C d =05m / e ) / 0, • o ~ in TestC ~caleXO.§ d0"02$~ d1 • 037~ in I T e s t C S c a l e X 2,0 100 i / / ) /; /; do.lOre ~ ea ~ 6O n 2O , 10 20 30 , , , , , , , , , 40 50 60 70 % Die Wall Contact , 80 90 100 Fig 13: Effect of absolute geometry size (CDA 101 copper, ho/do = 1.5) SUMMARY The objective of this study was to study t h e flashless closed-die upset process, in order to learn the factors that influence the required forging loads Experiments were performed using four common materials and three different test geometries The results were then presented as the forging load (or pressure) required to obtain a specified percent of die wall contact These experimental results were t h e n compared with two common mathematical analyses of the process: (1) slab method calculations and (2) finite-element analysis During process development, a process designer can use the results of this project to predict required forging loads and the amount of underfilling after a hit A limitation of the project results is that t h e results are applicable only to closed-die processes where the average part strain is 0.82 A l t h o u g h this a c o m m o n value of average part strain in closeddie upsets, many processes result in higher or lower part strains Finite-element simulations were conducted to study the effect of part strain on required forging loads, and are presented in this paper, but the results were not verified against experiment Both the modified slab method and finiteelement analyses proved to predict the required forging loads reasonably for closed-die upsets The modified slab m e t h o d tended to overpredict t h e required forging loads by zero to 20% FEM analysis tended to underpredict the required forging loads a t low %DWC, and overpredict the required forging loads at high %DWC, in both cases the error being 10% or less 140 tl/a ° • 15 120 ,-082 • DEFORM TEST %/% ~s " DEFORM TEST i hoJd° , 1s dlJ % • 20 " 100 @ 80 ,ot38 o i 60 o~)? O -" 40 20 10 20 30 40 50 60 70 % Die Wall Contact 80 90 100 Fig 14: Effect of part strain on die filling (CDA 101 copper, Test C) Acknowledgment The work summarized in this paper was partially funded by National Machinery The authors also would like to acknowledge Robert Lucius and National Machinery for their generous support of this project and for the use of t h e i r facilities to conduct the experiments described in this paper 94 ~ O'Connell et al / dournal of Materials Processing Technology 59 (1996) 81-94 References [1] Wick, C (1984) Tool and Manufacturing Engineers Handbook: Volume I Forming, Society of Manufacturing Engineers, SME Drive, Michigan, USA [2] Raghupathi, P.S., Oh, S.I., & Altan, T (1982) "Topical Report No 10, Methods of Load Estimation in Flashless Forging Processes", Battelle Columbus Laboratories, 505 King Ave., Columbus, Ohio, USA 43201, p 94 [3] Lange, K (1985) Handbook of Metal Forming, New York, McGraw-Hill Book Company [4] Altan, T., Oh, S., & Gegel, H., (1983) Metal Forming: Fundamentals and ADDlications, American Society for Metals, Metals Park, Ohio, USA [5] DEFORM, Scientific Forming Technologies Corporation, (1993) Metal forming FEM code, Columbus, Ohio, USA
- Xem thêm -
Xem thêm: Flashless closed-die upset forging load estimation for optimal cold header selection, Flashless closed-die upset forging load estimation for optimal cold header selection, Flashless closed-die upset forging load estimation for optimal cold header selection