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

Machining of High Strength Steels With Emphasis on Surface Integrity by air force machinability data center_5 pdf

9 316 0
Tài liệu đã được kiểm tra trùng lặp

Đang tải... (xem toàn văn)

THÔNG TIN TÀI LIỆU

Thông tin cơ bản

Định dạng
Số trang 9
Dung lượng 0,9 MB

Các công cụ chuyển đổi và chỉnh sửa cho tài liệu này

Nội dung

7.5 Machined Surface Texture Introduction to Surface Texture Parameters When a designer develops the features for a component with the requirement to be subsequently machined utilising

Trang 1

perfectly matched, allowing either a partial arc,

or circular feature to be reproduced If

non-syn-chronised motion occurs, often termed

‘servo-mis-match’  between these two axes, then an elliptical

profile – usually inclined at an 45° angle occurs,

• Squareness – when orthogonal (squareness) is not

maintained between the two interpolating axes,

then the net result will look similar to that of a

milled angular elliptical profile shape, which is

un-affected by the selected circular interpolation

rota-tional direction

Considerably more machine tool-induced factors can

affect a milled circular interpolated profile These

‘er-rors’ can be found, isolated and then reduced by

di-agnostically interrogation by using dynamic artefacts,

such as the ballbar Ballbars and their associated

in-strumentation can not only find the sources of error,

they can prioritise their respective magnitudes – to

show where the main ‘error-sources’ occur, then

in-stigate any feed corrections into the CNC controller

to nullify these ‘machine-induced errors’ As a result

of eliminating such ‘error-sources’ , this enables the

milled circular contouring and overall performance to

be appreciably enhanced

7.5 Machined Surface Texture

Introduction to Surface Texture Parameters

When a designer develops the features for a component

with the requirement to be subsequently machined

utilising a computer-aided design (CAD) system, or

by using a draughting head and its associated

draw-ing board, the designer’s neat lines delineate the

de-sired surface condition, which can be further specified

by the requirement for specific geometric tolerances

In reality, this designed workpiece surface condition

cannot actually exist, as it results from

process-in-duced surface texture modifications Regardless of the

method of manufacture, an engineering surface must

have some form of ‘topography, or texture’ associated

 ‘Servo-mismatch’ , can often be mistaken for a ‘squareness

er-ror’ , but if the contouring interpolation direction is changed,

from G02 (clockwise) to G03 (anti-clockwise) rotation, then

an elliptical profile will ‘mirror-image’ (‘flip‘) to that of the

op-posite profile – which does not occur in ‘squareness errors’.

with it, resulting from a combination of several inter-related factors, such as the:

• Influence of the workpiece material’s microstruc-ture,

• Surface generation method which includes the cut-ting insert’s action, associated actual cutcut-ting data and the effect of cutting fluid – if any,

• Instability may be present during the production machining process, causing induced chatter, result-ing from poor loop-stiffness between the machine-tooling-workpiece system and chosen cutting data,

• Inherent residual stresses within the workpiece

can occur, promoted by internal ‘stress patterns’ 

causing latent deformations in the machined com-ponent

From the restrictions resulting from a component’s manufacture, a designer must select a functional sur-face condition that will suit the operational constraints for either a ‘rough’ , or ‘smooth’ workpiece surface This then raises the question, posed well-over 25 years ago –

which is still a problem today, namely: ‘How smooth is smooth?’ This question is not as superficial as it might

at first seem, because unless we can quantify a surface accurately, we can only hope that it will function cor-rectly in-service In fact, a machined surface texture condition is a complex state, resulting from a combi-nation of three distinct superimposed topographical conditions (i.e as diagrammatically illustrated Fig 160a), these being:

 ‘Stress patterns’ , are to be expected in a machined

compo-nent, where: corners, undercuts, large changes in cross-sec-tions from one adjacent workpiece feature to another, etc., produce localised zones of high stress, having the potential outcome for subsequent component distortion ‘Modelling’

a component’s geometry using techniques such as: finite ele-ment analysis (FEA), or employing photo-elastic stress analy-sis* models or similar simulation techniques, will highlight

these potential regions of stress build-up, allowing a designer

to nullify, or at worst, minimise these potential undesirable stress regions in the component’s design.

*Photo-elastic stress analysis displays a stress-field, normally

a duplicate of the part geometry made from a thin two-dimen-sional (planar) nematic liquid crystal, or more robustly from

a three-dimensional Perspex model, which is then observed through polarised light source This polarised condition, will highlight any high-intensity stress-field concentrations in the part , which allows the ‘polarised model’ to be manipulated

by applying either an un-axial tension, or perhaps a bi-axial bending external stress to this model, showing dynamically its potential stress behaviour during its intended in-service con-dition.

Trang 2

Figure 160 Surface texture comprises of: ‘long-’, ‘medium-’ and ‘short-components’, together with the

‘direction of the dominant pattern’ – superimposed upon each other [Courtesy of Taylor Hobson]

Trang 3

1 Roughness – comprising of surface irregularities

occurring due to the mechanism of the machining

production process and its associated cutting insert

geometry,

2 Waviness – that surface texture element upon

which roughness is superimposed, created by

fac-tors such as the: machine tool, or workpiece

deflec-tions, vibrations and chatter, material strain and

other extraneous effects,

3 Profile – represents the overall shape of the

ma-chined surface – ignoring any roughness and

wavi-ness variations present, being the result of perhaps

the long-frequency machine tool slideway errors

The above surface topography distinctions tend to be

qualitative – not expressible as a number – yet have

considerable practical importance, being an

estab-lished procedure that is functionally sound The

com-bination of roughness and waviness surface texture

components, plus the surface’s associated ‘Lay’  are

shown in Fig 160a The ‘Profile’ is not depicted, as it

is a long-frequency component and at best, only its

partial affect would be present here, on this diagram

The ‘Lay’ of a surface tends to be either: anisotropic,

or isotropic in nature on a machined surface

topog-raphy When attempting to characterise the potential

functional performance of a surface, if an anisotropic

‘lay-condition’ occurs, then its presence becomes of

vital importance If the surface texture instrument’s

stylus direction of the trace’s motion over the assessed

topography is not taken into account, then totally

mis-representative readings result for an anisotropic

sur-face condition occur – as depicted in Fig 160b This

is not the case for an isotropic surface topography, as

relatively uniform set of results will be present,

regard-less of the stylus trace direction across the surface (i.e

 ‘Lay’ , can simply be defined as: The direction of the dominant

pattern’ (Dagnall, 1998).

 ‘Anisotropic, or isotropic surfaces, either condition can be

in-dividually represented on all machined surfaces Anisotropy,

refers to a surface topography having directional properties,

that is a defined ‘Lay’ , being represented by machined

feed-marks (e.g turned, shaped, planed surfaces, etc.) Conversely,

an isotropic surface is devoid of a predominant ‘Lay’

direc-tion, invariably having identical surface topography

charac-teristics in all directions (e.g shot-peening/-blasting and, to a

lesser extent a multi-directional surface-milling, or a

radially-ground surface, etc.).

see Fig 161a – for an indication of the various clas-sifications for ‘Lay’)

Returning once more to Fig 160b, as the stylus trace obliquity changes from trace ‘A’ , inclining to-ward trace ‘E’ , the surface topography when at ‘E’ has now become flat, giving a totally false impression of the true nature of the actual surface condition If this machined workpiece was to be used in a critical and highly-stressed in-service environment, then the user would have a false sense of the component’s potential

fatigue characteristics, potentially resulting in

ei-ther premature failure, or at worst, catastrophic fail-ure conditions In Fig 162, the numerical data (ISO 1302:2001), has been developed to establish and de-fine relative roughness grades for typical production processes However, some caution should be taken when utilising these values for control of the surface condition, because they can misrepresent the actual state of the surface topography, being based solely on a derived numerical value for height What is more, the

‘N-number’ has been used to ascertain the arithmetic

roughness ‘Ra’ value – with more being mentioned on

this and other parameters shortly The actual ‘N-value’

being just one number to cover a spread of potential

‘Ra’ values for that production process

Neverthe-less, this single numerical value has its merit, in that

it ‘globally-defines’ a roughness value (i.e.‘Ra’) and

its accompanying ‘N-roughness grade’ , which can be

used by a designer to specify in particular a desired surface condition, this being correlated to a specific production process The spread of the roughness for a specific production process has been established from experimental data over the years – covering the

maxi-mum expected ‘variance’  – which can be modified

 ‘Fatigue’ , can be defined as: ‘The process of repeated load, or

strain application to a specimen, or component’ (Schaffer, et

al., 1999) Hence, any engineering component subjected to repeated loading over a prescribed time-base, will normally undergo either partial, or complete fatigue.

 ‘Variance’ , is a statistical term this being based upon the

standard deviation, which is normally denoted by the Greek

symbol ‘σ’ Thus, variance can be defined as: ‘The mean of the

squares of the standard deviation’ (Bajpai, et al., 1979)

Thus, σ = √Variance, or more specifically for production

op-erations:

s=��n −  ċ �n

j=(xj− ¯x)

*s = the standard deviation of a sample from a production batch run.

Trang 4

depending upon whether a fine, medium, or coarse

surface texture is obligatory Due to the variability in

any production process being one of a ‘stochastic

out-put’ , such surface texture values do not reflect the

likely in-service performance of the part Neither the

surface topography, nor its associated integrity has

been quantified by assigning to a surface

representa-tive numerical parameters In many instances, ‘surface

engineering’ 8 is utilised to enhance specific

compo-nent in-service condition

It was mentioned above that in many in-service

engineering applications the accompanying surface

texture is closely allied to its functional performance,

predominantly when one, or more surfaces are in

mo-tion with respect to an adjacent surface This close

proximity between two mating surfaces, suggests that

the smoother the surface the better, but this is not

nec-essarily true if the surfaces in question are required to

maintain an efficient lubrication film between them

The apparent roughness of one of these surfaces with

respect to the other, enables it to retain a

‘holding-film’ in its associated topographical ‘valleys’ While

another critical factor that might limit the designer’s

choice of the smoothness of an engineering surface’s

selection, is related to its production cost (i.e see

Fig 161b) Therefore, if the designer requires a very

smooth machined surface, it should be recognised that

its manufacturing time is considerably longer – so its

respective cost will be greater to that of a rough

sur-face, this being exacerbated by a very close

dimen-sional tolerance requirement

 ‘Stochastic processes’ , are defined as: ‘A process which has a

measurable output and operating under a stable set of

condi-tions which causes the output to vary about a central value in a

predictable manner’ (Stout, 1985).

 ‘Surface engineering’ , is applying suitable discrete

technolo-gies to create surface films (e.g 10 to 100 nm thick), or by

ma-nipulating the surface atomic layers (e.g 2 to 10 atomic layers,

approximately 0.5 to 3 nm), to enhance the ‘engineered’

sur-face condition (i.e Source: Vickerman, 2000).

 ‘Surfaces’ , are recognised to have topographical features that

mimic the natural world So a regular/irregular engineering

surface can exhibit both peaks and valleys, not unlike

moun-tainous terrain.

7.5.1 Parameters for Machined

Surface Evaluation

In order that a machined workpiece’s surface texture can be determined using stylus-based (two-dimen-sional) instrumentation, three characteristic lengths are associated with this surface’s profile (i.e see Fig 163a), these are:

1 Sampling length0 – is determined from: the length

in the direction of the X-axis used for identifying

the irregularities that characterise the profile under evaluation Therefore, virtually all surface

de-scriptors (i.e parameters) necessitate evaluation over the sampling length Reliability of the data

is enhanced by taking an average of the sampling lengths as depicted by the evaluation length shown

in Fig 162a Most of today’s stylus-based surface texture instruments undertake this calculation au-tomatically,

2 Sampling length – can be established as: the to-tal length in the X-axis used for the assessment of

the profile under evaluation From Fig 163a, this

length may include several sampling lengths – typi-cally five – being the normal practice in evaluating roughness and waviness profiles The evaluation

length measurement is the sum of the individual

sampling lengths (i.e it is common practice to em-ploy a 0.8 mm sampling length for most surface texture assessments),

3 Traverse length – can be defined as: the total

length of the surface traversed by the stylus in mak-ing a measurement The traverse length will

nor-mally be longer than the evaluation length (i.e see Fig 163a), this is due to the necessity of allowing

‘run-up’ and ‘over-travel’ at each end of the

evalua-tion length These addievalua-tional distances ensure that any mechanical and electrical transients, together filter edge effects are excluded from the measure-ment

0 ‘Sampling length’ , is often termed ‘Meter cut-off’ , or simply the ‘cut-off’ length and its units are millimetres The most

common cut-offs are: 0.25, 0.8, 2.5, 8.0, 25.0 mm The 0.8 mm

sampling length will cover most machining production

pro-cesses In any surface texture evaluation, it is essential that the

cut-off is made known to the Inspector/Metrologist reviewing

this surface topographical data.

Trang 5

The number of two-dimensional surface profile

pa-rameters that have been developed over the years for

just the stylus-based instruments – discounting the

three-dimensional contact and non-contact varieties,

has created a situation where many users simply do

not understand, nor indeed comprehend the intrinsic

differences between them! A term was coined some

years ago to show exasperation by many

metrolo-gists’ with this ever-increasing development of such parameters Many researchers and companies were totally disenchanted with their confunsion and plight,

so they simply called the predicament: ‘parameter-rash’ However, here we need not concern ourselves

with this ‘vast expanse of surface descriptors’ , as only

a few of the well-established parameters and discuss just the most widely-utilised ones It is worth making

Figure 161 ‘Lay’ indicated on drawings, plus the relative cost of manufacture for different production processes

.

Trang 6

Figure 162 Anticipated process ‘roughness’ and their respective grades [Source: ISO 1302, 2001]

.

Trang 7

Figure 163 Surface texture data-capture, with techniques for the derivation of the arithmetic roughness parameter: Ra

.

Trang 8

the point, that all of these two-dimensional surface

pa-rameters can be classified into three distinct groupings

and just some of these parameters are:

1 Amplitude parameters (Ra, Rq, Wa, Wq, Pa, Pq)

– with Ra being universally recognised for the

‘international’ parameter’ for roughness It is: ‘The

arithmetic mean of the absolute departures of the

roughness profile from the mean line’ (i.e see Fig

163b and c) It can be expressed as follows:

Ra = �lrlr

NB The Ra parameter is often utilised in

appli-cations to monitor a production process, where a

gradual change in the surface finish can be

antici-pated, making it seem to be ‘ideal’ for any form of

machinability trials, but some caution is required

here, as will be seen shortly in further discussion

concerning this ‘surface descriptor’ By way of

com-parison, another previously used amplitude

param-eter is given in Appendix 10 and is the ‘R Z(JIS)’ (i.e

10-point height) parameter

Other useful parameters of the assessed profile, to be

shortly discussed in more detail, include: ‘Skewness’

(Rsk, Wsk, Psk), which is often utilised in association

with ‘Kurtosis’ (Rku, Wku, Pku), producing the

so-called: ‘Manufacturing Process Envelopes’ – as a means

of ‘mapping’ and correlating machined surface

topog-raphies

2 Spacing parameters (Rsm, Wsm, Psm) – can be

de-fined as: ‘The mean spacing between profile peaks at

the mean line, measured within the sampling length’

(i.e depicted along a machined cusp – at differing

 The designation of the letters follows the logic that the

pa-rameter symbol’s first capital letter denotes the type of profile

under evaluation For example, the: Ra* – calculated from the

roughness profile; Wa – derives its origin from the waviness

profile; with the latter in this logical sequence, namely the Pa

– being derived from the primary profile Here, in this

discus-sion and for simplicity, only the first term in the series – e.g

‘Ra’ notation – will be used.

*Ra is today shown in the International Standard (i.e ISO

4287: 1997) as being denoted in italics, while in the past, it was

usually shown as follows: ‘R a’ , but even now, many companies

still use this particular notation.

 Historically, the classification of the relative roughness of

sur-faces was initially developed in England and was then termed:

‘Centre Line Average’ (CLA), while in the USA its equivalent

term was the ‘Arithmetic Average’ (AA)

feedrates in Fig 169a and b) It can be expressed in the following manner:

Rsm = �n i=n

i=si= XS +XS+XS +XSn

n

Where:

n = number of peak spacings.

NB The Rsm parameter needs both height and

spac-ing discrimination and, if not specified the default

height bias utilised is 10% of: Rz, Wz, or Pz, – where these are the ‘Maximum height of profile’ As can be

seen from the ‘idealised’ machined surface topog-raphy given in Fig 169a and b, the spacing param-eters are particularly useful in determining the feed

marks Moreover, the Rsm parameter relates very

closely to that of the actual programmed feed rev– of either the cutter, or workpiece – depending on which production process was selected See also, Appendix

10 for a graphical representation of the previously

utilised ‘High Spot Count’ (HSC) parameter.

3 Hybrid parameters (Rmr, Wmr, Pmr, R∆q, W∆q, P∆q, Rpk, Rk, Rvk) – each of these ‘hybrids’ will now be briefly mentioned Rmr, or its alternative notation Mr is the ‘Material ratio curve’ , which can

be defined as: ‘The length of the bearing surface (ex-pressed as a percentage of the evaluation length ‘L’) at

a depth ‘p’ below the highest peak (i.e see Fig 165)

– Rmr:

It is often known as the ‘Abbott-Firestone curve’ ,

the mathematics of this Rmr-curve can be

ex-pressed in the following manner:

Rmr= b +b+b=B +bn

n �  = 

n

i =n

i=bi

NB This ‘Material ratio curve’ represents the

pro-file as a function of level More specifically, by plot-ting the bearing ratio at a range of depths in the profile trace, the manner by which the bearing ratio changes with depth, provides a method of charac-terising differing shapes present on the profile (i.e see Fig 165 and Appendix 10)

– R∆q:

The R∆q parameter, can be defined as: ‘The root

mean square (rms) slope of the profile within the sampling length’ (i.e see how its angle changes at

differing machining feedrate conditions shown

in Fig 169b and c), it can be mathematically ex-pressed as follows:

Trang 9

R � q =��lr�lr

[θ(x) − ¯θ]dx

Where:

¯θ = �lrlr

θ (x)dx

θ = slope of the profile at any given point.

• Rpk, Rk, Rvk:

These parameters (i.e see Appendix 10 for

graphi-cal representations of the parameters), were

origi-nally designed for the control of potential wear in

automotive cylinder bores in volume production

by the manufacturing industry Today, Rpk, Rk and

Rvk are employed across a much more diverse-field

by a range of industries Such hybrid parameters

are an attempt to explain – in numerical terms, the

respective form taken from the profile’s trace of the

‘material ratio curve’ (Rmr), hence:

– Rpk parameter – is the ‘reduced peak height’ ,

il-lustrating that the top portion of a bearing

sur-face will be quickly worn-away when for

exam-ple, an engine initially begins to run,

– Rk parameter – is known as the ‘kernal

rough-ness depth’ , therefore the long-term running –

‘steady-state wear’ of this surface will influence

for example, the performance and life of the

au-tomotive cylinder(s),

– Rvk parameter – is the ‘trough depth’ this

in-dicates that the surface topography has an

oil-retaining capability, specifically via the ‘trough

depths’ which have been purposely

‘cross-honed’ into the bore’s surface

Arithmetic roughness parameter (Ra)

Although the Ra ‘amplitude parameter’ has been

widely quoted ‘Internationally’ , there are a few

provi- ‘Cross-honing’ , uses either: (fine) Abrasives/CBN/Diamond –

‘stones’ , that are fitted into a ‘honing head’ which then rotates

and oscillates within a hole, or an engine’s bore The critical

parameters are the rotational speed (Vr) oscillation speed

(Vo), the length and position of the honing stroke, the

hon-ing stick pressure (Vc) The inclination angle of the

cross-hon-ing action, is a product of the up-/down-ward head motion

(Vo) and the rotational speed for the head (Vo) This complex

action of rotating and linear motion, generates the desired

cross-honed ‘Lay-pattern’ within the bore – for improved oil

retention.

sos, or conditions that must be met, if it is to be utilised satisfactorily, these are:

The Ra value over one sampling length represents

the average roughness The effect of a spurious

non-typical peak, or valley within the profile’s trace be-ing ‘averaged-out’ so will have only a small

influ-ence on the Ra value obtained;

The evaluation length contains several sampling

lengths (Fig.163a), this ensures that the Ra value is

representative of the machined surface under test;

An Ra value alone is meaningless, unless quoted with its associated metre cut-off (λc) length Repeat-ability of the Ra value will only occur at an

identi-cal length of metre cut-off;

If a dominant surface texture pattern occurs (Lay), then

the Ra readings are taken at 90° to this direction;

That Ra does not provide information as to the

shape of either the profile, or its surface irregulari-ties Different production processes generate diverse

surface finishes, for this reason its is usual to quote

both the anticipated Ra numerical value along with

the actual manufacturing process;

• Ra offers no distinction between peaks and valleys

on the surface trace

The most confusing argument concerning the use of

an Ra value alone, is that its numerical value is not

only meaningless, but it can have catastrophic conse-quences if interpreted incorrectly These opinions can

be substantiated by close observation of Fig 164a,

where an identical numerical Ra value produces

widely divergent surface topographies In addition, if a

designer’s engineering application called for a ‘bearing surface’ (Fig 164ai), rather than a ‘locking surface’ (Fig

164aiii), then the numerical value of 4.2 µm in isola-tion, becomes pointless, as it tells the designer nothing about the ‘functional’ surface topography This

prob-lem is exacerbated when the wrong surface topography

is selected for a specific engineering application For ex-ample, a ‘locking surface’ applied to a bearing industrial

application in a harsh environment, can be expected to catastrophically fail after very little in-service time

Skewness (Rsk, Wsk, Psk) and Kurtosis (Rku, Wku, Pku) Parameters

These surface descriptors of ‘skewness’ and ‘kurtosis’ are often derided as simply ‘statistical’ amplitude pa-rameters, that can introduce spurious results and as a consequence, having little use in engineering applica-tions However, when used in the correct context, they can provide a valuable insight into the overall shape of

Ngày đăng: 21/06/2014, 22:20

TỪ KHÓA LIÊN QUAN

TÀI LIỆU CÙNG NGƯỜI DÙNG

TÀI LIỆU LIÊN QUAN

🧩 Sản phẩm bạn có thể quan tâm

w