1. Trang chủ
  2. » Kỹ Thuật - Công Nghệ

Machining of High Strength Steels With Emphasis on Surface Integrity by air force machinability data center_8 doc

9 171 0

Đang tải... (xem toàn văn)

THÔNG TIN TÀI LIỆU

Thông tin cơ bản

Định dạng
Số trang 9
Dung lượng 890,74 KB

Nội dung

Figure 173. Finite Element Method (FEM), to obtain simulated, but realistic data on isother- mal temperatures within the cutting region. [Source: Tay et al., 1993] .  Chapter  Figure 174. Typical wear patterns that could be present on a cemented carbide (uncoated) cutting insert, utilised under ‘steady-state’ turning conditions . Machinability and Surface Integrity  • good quality and consistent workpiece material is to be utilised; • that the condition monitoring of machine tool en- sures that it is in an optimum state for use; • any ood coolant supply and quality – if it is to be used – is of the correct grade and dilution concen- tration; • work-holding/support is both rigid and precise/ac- curate; • expert support is available – if necessary – along with the user’s own practical experiences. ese factors oer a good ‘start-point’ in ensuring that the ‘ideal’ tool wear development takes place. Classification of Tool Wear Types Tool wear depends on several inter-related factors, some of these have been mentioned above, but are worth restating, such as: the cutting insert and work- piece material combination – plus their physical, mechanical and chemical properties; cutting insert ge - ometry; as well as cutting uid properties and pressure – if applied; together with various other operational parameters – cutting data selected, stability of the cut- ting process and work-holding application techniques. Any knowledge obtained on analytical studies of wear mechanisms, is largely based upon the results from ex- perimental trials. Simply obtaining wear data presents considerable diculties, then simply analysing these results can be somewhat onerous, due to isolating the major cause of this particular wear regime. Neverthe- less, having stated these problems, many potential so- lutions to specic wear patterns can be found, so long as the actual wear regime, or composite wear behav- iour can be singularly identied. With this in mind, the following classications for tool wear are given be- low (i.e. see Fig. 174 for of several these wear patterns), which include: • Flank wear – as its title suggests, occurs on the cut- ting edge’s anks, usually the result of an abrasive wear mechanism. Both of the clearance faces – lead- ing and trailing edges, together with the tool nose radius are subject to a parallel land wear, created by the workpiece travelling past the contact regions of the tool both during and aer chip formation. Such a wear mechanism is considered normal tribologi- cal behaviour and a progressive form of ank wear can be tolerated and subsequently dealt with, by an ecient tool-changing strategy, based upon antici- pated tool life expectancy. NB Toward the end of the steady-state and progres- sive ank wear regime, it could lead to several un- desirable factors, such as: increasing friction, which can possibly change the insert’s prole – leading to poor machined surface texture, or dimensional in- accuracies as the ‘tool dris’ 63 – creating variability in tolerances of successive parts. • Crater wear – this is present on the rake, or chip face and is normally the result of a combination of an abrasion and diusion 64 wear mechanism. 63 ‘Tool driing’ , is a term used to describe the fact that having initially set the tool to a particular dimensional size, the tool’s ank will progressively wear – under steady-state machin- ing conditions. e variability in dimensional size can be the subject of both random and systematic errors – even when the operation is behaving normally. is dimensional variabil- ity, causes for example: turned diameters to get larger, while drilled holes get smaller – as successive components are ma- chined, this is the essence of tool-driing. e term process capability* has been coined to explain the stochastic process output from a normally-operating production process – see Chapter 2, Footnote 26, for more information regarding this subject. *Process capability (C p ) can change during consecutive pro- duction output of components, being the result of the ‘vari- ables’ (i.e. as each singular part dimension is known), pro- ducing either random, or systematic errors, or both, as the production run progresses. is is why it is usual practice to utilise ‘Statistical control techniques’ to show any signicant changes in output. erefore, ‘Shewart charting techniques’ in combination with ‘Probability paper’ are employed, to esti- mate the: C p value and to determine if the process is behaving/ operating ‘normally’ – usually a ‘normal output’ is signied by establishing a ‘straight-line’ (i.e. plotted) relationship on the ‘Probability paper’. 64 ‘Diusion wear’ , was initially proposed in 1858 by the Ger- man physiologist Adolph Fick (1829–1901), where he enun- ciated laws governing the diusion of substances generally on a quantitative basis. Today, we are concerned with ‘atomic migration’ within metallic solid solutions. Fick produced two laws, with Fick’s 1 st  Law stating: ‘at the amount (J) of a ma- terial moving across a unit area of a plane in unit time is pro- portional to the concentration gradient (∂c/∂x) at the same time but of opposite sign’. It can be expressed as follows: J[atoms/m 2 .s] = −  D  [m 2 /s](∂c/∂x)[atoms/m 3 .1/m] Fick’s  1 st   Law Where: J = ux, net ow of atoms; D = diusion coe- cient; ∂c/∂x = concentration gradient. NB Assuming that X-axis is parallel to direction in which concentration gradient is operating. Fick’s  2 nd   Law was de- rived from the 1 st  Law and from the fact that matter is con- served, relating the change in concentration with time (∂c/∂t) and it can be expressed as: (∂c/∂t) = ∂/∂x (D∂c/∂x) Fick’s 2 nd  Law (General case) By dierential calculus, this 2 nd Law changes to: ∂c/∂t) = D ∂ 2 c/∂x 2 .  Chapter  e crater can be formed either via a hard-particle grinding action, which mechanically-removes rake face surface layers, or by a complex ‘atomic diusion process’ 65 interacting between the chip and the tool material (ie see Fig. 174 – top right). NB If a cutting insert has high bulk hardness, combined with ‘hot-hardness’ 66 , plus minimum af- nity between these two materials, this will dimin- ish any crater wearing tendencies. Moreover, crater wear changes the cutting insert geometry of the edge, which may impair chip formation and modify cutting forces, or lead to a weakened edge strength. Many of today’s multi-coated cutting inserts are less aected by crater wear than their uncoated coun- terparts. NB From this it can be appreciated why the nal stages of dif- fusion are somewhat slow, due to the rate of diusion decreas- ing as the concentration gradient diminishes. (Higgins, 1979) 65 ‘Atomic diusion process’ , there is strong evidence – when ferrous workpiece machining – to indicate that cratering of WC-Co cutting inserts (i.e. uncoated), occurs by diusion of the C atoms into chip at the interface (i.e see Fig. 174 – top right schematic diagram). Remembering that solid-state dif- fusion depends upon the rate at which the tool’s atoms dis- solve/diuse into the chip. For WC, the most rapid diusion is by the tool’s Co atoms – of the carbide bond and, the Fe atoms from the chip. Hence the carbide grains are undermined and swept-away for two reasons:With WC tool material, carbide grains are not isolated and constitute the bulk of the mate- rial, so support each other in a ‘rigid framework’ ,Due to Co atoms from the tool ‘diusing-out’ , so Fe atoms from the chip ‘diuse-in’ and these provide support for the carbide grains, which in turn inhibit their removal. In the chip, C atoms being small, rapidly diuse through the Fe matrix, however those in the tool are strongly-bonded to W and are not free to move by themselves. us, the rate of diusion of both W and C atoms together from the tool go into the chip and thus, will control diusion wear with respect to its temperature – as Fick’s Laws suggest. NB e distances for diusion at the tool/chip interface are between 1 nm up to 1µm. Diusion in the tertiary shear zone (i.e. ank) is normally higher than in the secondary shear zone, due to the signicantly greater workpiece surface speed in this vicinity. So, not only is attrition a mechanism for ank wear, diusion is also partly responsible – even when the rake face is hardly worn. In appearance, when the grains look to be smooth, this is a good indication of a diusion mechanism taking place. (Armarego and Brown, 1969) 66 ‘Hot hardness’ , this is the ability of a cutting insert to retain its relative bulk hardness and hence geometry at elevated tem- peratures. • Plastic deformation – occurs when high pressures (i.e. compression) are exerted on the cutting edge in combination with elevated temperatures. Con- ditions likely to create plastic deformation on the cutting insert are when high speeds and feeds are utilised on workpiece materials that are prone to work-hardening. Tool materials must have the re- quired mechanical properties to withstand plastic deformation during machining. Typically, bulging of the edge in the tool nose region, leads to: geom- etry deformation; chip ow modication; greater localised temperatures – until a critical juncture is attained. So cutting insert ‘hot-hardness’ is a vital characteristic. NB In order to combat cutting insert plastic defor- mation, a large tool nose radius, plus more robust tool geometry adds greater strength in this ‘exposed region’ of the tool. • Notch wear on insert’s leading edge – is the result of mechanical action, promoted by either machining workpiece materials that may easily work-harden, so each successive longitudinal turning pass at the same D OC leads to the previous surface condition being harder, resulting in a more abrading-action here – hence a notch will wear at this point on the insert‘s ank. is ‘notching eect‘ can be reduced, if a variable D OC is employed, to ‘even-out’ the con- tact region along the leading edge of the insert. NB ‘Black-bar stock’ having been hot-rolled from its primary processing route, tends to have a hard and abrasive oxide scale to its periphery, which may contribute to insert notching when only the surface is ‘skimmed’ by a longitudinal turning operation. • Notch wear on insert’s trailing edge – occurs by in the main, by adhesion wear, but to a lesser extent, may be the result of an oxidation wear mechanism. e notch on this ank’s trailing edge is formed where the cutting edge and the workpiece material separate. NB Notch wear here, tends to be very localised to- ward the end of the cut, enabling air to reach this cutting vicinity, which has a high temperature pres- ent, so adhesion/oxidation can be expected. • Built-up edge (BUE) formation – is usually the re- sult of tool/workpiece anity associated with tem- Machinability and Surface Integrity  perature and its respective cutting speed (i.e. see Fig. 28). Moreover, it can also transpire as a result of ‘edge agging’ , or from other wear mechanisms. is ‘cold’ pressure-welded workpiece material be- ing attached to the tool as a BUE, changes the cut- ting insert’s geometry – to its detriment. Hence, this BUE is both severely work-hardened and ‘unstable’ – it will break-away from the tool mate- rial thereby potentially ‘frittering’ the insert’s edge. NB BUE machining data conditions have been reasonably well-dened, so fortunately, these re- spective cutting speeds can be avoided, particu- larly, as most CNC machining operations happen at much higher speeds and modern insert grades and coatings, minimise this BUE eect. If BUE does oc- cur, it can create a poor surface nish on the ma- chined surface. In any BUE machining condition, if it continues without attention, then the result can be rapid edge breakdown, or even result in insert fracture. • e former conditions are in the main, conned to continuous cutting and steady-state machining conditions, albeit with single-point cutting inserts. • e latter conditions are generally restricted to in- termittent cutting multi-point machining, or inter- rupted cutting operations: • ermal cracking – is usually the result of fatigue wear, produced by thermal cycling machining con- ditions, such as when milling. ese cracks that form are normally at 90° to that of the cutting edge 67 . ese cracks are spaced out periodically along the cutting edge and when they propagate (i.e. grow) to 67 ‘ermal fatigue cracks’ , are usually termed ‘comb-cracks’ – due to their appearance is not unlike that of a hair comb. When these cracks propagate to a critical length which can be ex- plained in terms of ‘Fracture mechanics’* and in particular the ‘stress intensity factor’ (K IC ) – with the ‘C’ standing for ‘critical’. Such cracks will fracture quickly around the ‘Speed of sound’ (i.e. Mach 1, or in a steel workpiece @ 5050 ms –1 ), so little, if any warning is given of the likely failure condition as it arises – when the tool’s edge eventually catastrophically fails. *In 1957, G.R. Irwin and his co-workers, laid the foundations for ‘Fracture mechanics’ and were particularly noted for the mathematics for dening the ‘stress intensity factor’ (K), spe- cically: K = σ √ (πc) [Nm ½ ] Where: σ = fracture stress, c = half length of an internal aw. (Shaw, 1984) a critical size, bulk tool material will be pulled-out of the tool’s edge – leading to a very rapid type of cutting insert edge failure. NB Varying the chip thickness will also aect tem- peratures throughout the cut. A cautionary note here, concerning cutting uid application: if used under certain conditions, the cutting uid has a detrimental inuence in some metal cutting opera- tions, as it amplies the variations in temperature between and in- and out-of-cut. • Mechanical fatigue cracking – may be present if cutting force shock-loads are extreme. Fatigue 68 is a form of fracture which is promoted by continual variations in load, but where the load in itself, is not great enough to cause fracture. 68 ‘Fatigue’ , can be dened as a: ‘Phenomenon leading to the fail- ure of a part under repeated, or uctuating stress below the ten- sile strength of the material.’ Failure usually occurs suddenly as a result of crack propagation without plastic deformation at a stress level well below that of the elastic limit for the material. e stress can be either an: ‘alternating’; ‘repeated’; or a combi- nation of these types. At a discontinuity such as a notch, hole, or step, the stress is considerably greater and is termed a ‘stress concentration factor’ (K). Graphs can be plotted , such as: SN curves (i.e. to nd the endurance limit for steels, or for non-ferrous metals, alloys and plastics -the fatigue stress ‘σ FS ’ is specied for a nite number of stress reversals), Soderberg diagram – for steel, with alternating stress plot- ted against steady stress. Moreover, a ‘safety factor’ (FS) can be applied to the graphical result, as follows:   (Safety factor) FS = σ y σ m +(σ y �σ e )K σ r Where: σ y  = yield stress, σ m = steady stress component, σ e  = failure occurs – (i.e. above a line drawn from this value: σ e on the ‘Y-axis’ to σ u on the ‘X-axis’); Kσ r = alternating com- ponent – with ‘K’ representing the ‘stress concentration factor’ and ‘σ r ’ representing ‘alternating stress’. NB Most steels have an ‘endurance limit’ being about half its tensile strength, with an approximation oen utilised: For  steels: Endurance limit = 0.5 tensile strength (i.e. up to a tensile strength of 1400 N mm –2 ), Endurance limit = 700 N mm –2 (i.e. above a tensile strength of 1400 N mm –2 ).   For Cast steel/iron: Endurance limit = 0.45 tensile strength (i.e. up to tensile strength of 600 N mm –2 ), Endurance limit = 275 N mm –2 (i.e. above a tensile strength of 600 N mm –2 ).   Non-ferrous metals/alloys: there is no endurance limit and the fatigue stress is taken at a denitive value of stress rever- sals, e.g. 5 x 10 7 . (Carvil, 1994, et al.) – –  Chapter  NB erefore at the initiation of a cut, the varia- tions in the magnitude of the cutting force and its direction, may not be too great for both the tough- ness and strength of the cutting insert. With con- tinual usage however, these fatigue cracks grow – in the main – parallel to the cutting edge and may eventually be the cause for premature tool failure. • Cutting edge chipping – this transpires when the edge line fractures, rather than being the result of wear. It can be considered as a form of fatigue fail- ure, because of the cycles of loading and unloading during cutting, leading to particles of tool material being removed from the insert’s surface. is type of wear mechanism is generally the result of inter- mittent cutting operations. NB An investigation into whether this edge wear is either from chipping, or the result of ank wear. ‘Spalling’ (i.e. cracking, or aking of the surface) and ‘nicking’’ are also variants of this category of edge degeneration. • Fracture – is normally catastrophic conclusion to the cutting process (i.e. see Fig. 175). Here, bulk material fracture can have serious consequences obviously to the cutting insert, but also aecting the machined part. Moreover, this form of edge fracture is more oen than not, the termination of alternative wear regimes. If Fig. 175 is investigated in more detail, it may help comprehension of the nature of the serious problems associated with such a sudden failure mode. e cut- ting insert was purposely catastrophically failed in practical trials conducted by the author, using a rea- sonably robust turning and facing geometry, longitu- dinal turning P/M ferrous compacts without coolant. Here, the cutting speed was raised by 25% above the optimum, with the feedrate 40% greater than usually specied. is ‘abusive machining regime’ , created high ank wear and plastic deformation to the cutting edge, which shortly failed – catastrophically. In Fig. 175c, detail of the fracture surface indicates both duc- tile and brittle failure modes instigated from the worn leading edge’s ank. By increasing the cutting data by just the cutting speed alone and leaving the feedrate at the optimum, tool life was reduced on other simi- lar inserts, but catastrophic failure did not occur, only very high levels of ank wear. However, if the cutting speed was kept at the optimum and the feedrate was increased – as mentioned – in-line with other insert trials, then catastrophic failure eventually occurred, well before that predicted by ‘Taylor’s tool life calcu- lation’ . is conrmed the fact that the high abrasive nature to the testpieces produced from ferrous-based P/M compacts, in combination with an increased fee- drate caused premature catastrophic failure of the cut- ting inserts during these ‘harsh’ machinability trials. As previously mentioned, Appendix 11 has a con- cise ‘trouble-shooting guide’ for some of the potential wear regimes that are likely to be experienced during many machining operations. .. Tool Life Introduction It is normal practise to assess tool life according to three mutually-inuencing criteria, as any one of them could be the reason for the expensive business of sub- sequent part scrappage. ese criteria that signicantly aect machined components and can be the reason for curtailment of the cutting tool’s life are: 1. Ability to sustain workpiece tolerances – here if the tool has been in operation for too long ‘in-cut’ , then this will increase the tendency for ‘tool dri- ing’ which will amplify machined component vari- ability, while creating inconsistency in part produc- tion (Figs. 31ci and ii), 2. Maintaining machined surface texture quality – as the tool is progressively utilised, the ank and cra- ter wearing tendencies will increase, leading to de- generation of the surface texture, below that which was demanded from the designer’s direct engineer- ing requirements (i.e. see graph in Fig. 148), 3. Eciency in chip-breaking ability – if the cut- ting insert/tool has been operated for considerable time, there is every expectation that both ank and more importantly crater wear will be present. is will have an adverse eect on chip-breaking ability, leading to either poor component surface texture, or variability in component tolerances, or both (Figs. 37 and 38a and b). If a cutting insert, or tool no longer satises the above wear criteria, its useful life is ended and it should be summarily discarded. e tool life’s predictability, is a key factor in an estimation of the anticipated produc- tivity output level. Approached from a dierent direc- tion, an CNC programmer may deliberately choose Machinability and Surface Integrity  Figure 175. Catastrophic failure of a turning insert.  Chapter  the cutting insert, or tool they are most familiar with, because they know – from practical experience – that it performs and wears in a progressive manner, rather than the unpredictability associated with an insert of ‘uncertain machining capability’ that might otherwise prematurely fail. Prior to discussing criteria for determining when a cutting insert is ‘worn-out’ , it is necessary to estab - lish in practice, what this actually means. For example, does ‘worn-out’ refer to when the: dimensional accu- racy becomes unpredictable: or if the surface nish has signicantly deteriorated; or perhaps the fact that its automatic chip-breaking behaviour has become inef- cient? In many situations it is by the user’s experience that one can judge how much ank wear can be toler- ated on the cutting edge before machining is discon- tinued. As a rule, ank wear is a dependable criterion for assessing when the cutting edge is eectively worn- out. Moreover, from the previous discussion, perhaps the degree of cratering may in certain machining cir- cumstances prove to be more signicant than the ank wear, in respect to the shortening tool life. Tool wear can be established by several techniques, but the usual method is to observe and then measure the actual wear as it progressively develops. e eec- tive cutting time, or tool life ‘T’ , is specied as time- elapsed prior to a predetermined degree of wear has been reached. A typical procedure for determining ank wear can be: to observe cutting edge(s) in-situ on the machine tool; then remove from the machine and visually inspect the tooling; followed by its respective wear rate can then be optically magnied in suitable equipment allowing accurate dimensional measure- ment – against the following criterion (i.e. see Fig. 174): • Extent of ank wear from original edge – if this wear is of relatively uniform nature, it may be dis- tributed across three zones, ‘A’ ,‘B’ and ‘C’. e mean ank wear ‘V B,C–A ’ is measured over the cut- ting region of the leading edge across these zones – it is oen just referred to as simply: ‘V B ’. If excessive wear develops at one position on the cutting edge, for instance where the wear-notch ‘V N ’ occurs, this zone is usually ignored when establishing the ‘mean wear’. Here and under these conditions, it is usual to quote the maximum ank wear as ‘V Bmax ’ , • Extent of cratering – this is usually specied by the maximum crater depth from the plane of the original rake face ‘K T ’ and in some cases, by its di- mensional size: ‘K B ’- width and ‘K M ’ – length (not shown). e above wear criteria, are normally utilised for esti- mating the extent of ank and crater wear. Over many years of experimental research into tribological wear mechanisms, it has been established that progressive ank wear develops according to a xed pattern, with three distinct stages to this wear regime, they are (Fig. 176): 1. Initial, or primary wear – if a new cutting edge is used to machine a workpiece, there is a rapid breakdown of the of the cutting edge. is early ank wear on the tooling is depicted in the graph of wear against time in Fig. 176a, indicated by its preliminary high wear-rate, is wear-rate is de- pendent upon the cutting conditions and type of workpiece material, plus any cutting uid applica- tion – if utilised. Flank wear increases in relation to an higher cutting speeds, 2. Progressive, or secondary wear – occurs aer the initial ank wear has taken place. During the fol- lowing time period, there is a steady and progres- sive stage to the cutting tool’s/insert’s wear, with a much less pronounced increase than that indicated at the initial wear stage, this is when the productive machining output occurs. Toward the end of this progressive wear stage, this being the case when the ank wear ‘V B ’ reaches approximately 0.8 mm in height, here, it is normal practice to replace this old tool with a ‘sister tool’ – to continue machining the component batch, or production run. Once ank wear has reached this arbitrary dimensional value, then to all practical purposes its productive life is ended, 3. Catastrophic, or tertiary wear – will normally only become apparent if the tool is taken toward, or up to, its complete failure. Such catastrophic failure is the result of a combination of several tool wear mechanisms: high ank wear; large crater forma- tion – reaching the point where the tool has been suciently weakened for the increased tool forces now operating to cause it to fracture. Inevitably, if such an immediate breakdown occurs during the nal pass over the workpiece’s surface, it is prob- able that the component has to be scrapped. If the workpiece has a high residual raw-stock value, then aer machining, signicantly more added-value will have accrued. So, any initial savings made by using these tools into the tertiary ank wear stage, will be more than cancelled-out by scrapping this component! Machinability and Surface Integrity  Tool-life Diagrams Machinability is a subject that has yet to be fully-de- ned and analysed, in particular the interactive mech- anisms that take place at the chip/tool interface, with the user’s own experience being a good start-point for any future machining operations. As has been men- tioned above, tool wear varieties can have several dif- fering causes and eects (i.e. see Appendix 11). With any machining batch, or production run, it is custom- ary practice to establish a ‘norm’ for both the tolerable ank wear dimension and the depth/size of crater for- mation. In particular, as the ank wear pattern usu- ally takes in-cut time to progressively develop and this predictable tool/wear relationship has been well estab- lished some years ago, initially by F.W. Taylor’s pio- Figure 176. Tool wear under steady-state conditions: (a) tool wear as a function of time, (b) if cutting speed is changed, then tool life is aected, (c) amalgamation of these ‘Taylor curves’ and derivation of the ‘general taylor curve’. [Courtesy of Sandvik Coromant] .  Chapter  . ‘steady-state’ turning conditions . Machinability and Surface Integrity  • good quality and consistent workpiece material is to be utilised; • that the condition monitoring of machine tool en- sures. deformation – occurs when high pressures (i.e. compression) are exerted on the cutting edge in combination with elevated temperatures. Con- ditions likely to create plastic deformation on the. strongly-bonded to W and are not free to move by themselves. us, the rate of diusion of both W and C atoms together from the tool go into the chip and thus, will control diusion wear with

Ngày đăng: 21/06/2014, 22:20

TÀI LIỆU CÙNG NGƯỜI DÙNG

TÀI LIỆU LIÊN QUAN