Figure 147. Machinability testing utilising an ‘accelerated testing procedure’ – a combination of the rapid facing and degraded tool tests . Machinability and Surface Integrity on a moderately short timescale. Normally in many previous testing programs, an uncoated cemented car- bide P20, or P10 grade would have been used, since these grades withstand both higher speeds and have better tool wear resistance to that of previously utilised cutting tool materials. However in this case, an P25 grade was chosen, which is a degradation from the optimum P20 grade, but it should still perform satis- factorily. Furthermore, the cutting speed was raised by >2.5 times the optimum of 200 m min –1 , with all fac- ing operations being conducted at a ‘constant surface speed’ 12 of 550 m min –1 . Typical tool-life curves produce by the AWT tech- nique are illustrated in Fig. 148, showing the expected three stages of ank wear. is ank wear being a func- tion of: the initial edge breakdown, steady-state wear – as the insert’s ank progressively degenerates and - nally, catastrophic insert edge breakdown – as the edge completely fails. Detailed metallurgical analysis can be made as to the reasons why some P/M compacts per- formed better than others, by reference to the litera- ture on the metallurgical interactions between the tool and the compact – this subject being outside the scope of the present discussion. e facing-o secondary machining operation meant that aer 10 facing passes, a pre-programmed ‘optional stop’ can then be applied, to allow both tool ank wear and compact surface tex- ture to be established. e faced-o surface texture re- sults can then be superimposed onto the same graph – for a direct comparison of ank wear and for that of the machined surface texture parameter. Without go- ing into too much detail of the specic aspects of the processing and metallurgical interactions present here on the composite graph, some compacts abraded the cutting insert more than others, while the ‘faced’ sur- face texture, generally seemed to get worse, then im- prove and nally worsen again. However, this is a complex problem which goes to the ‘heart’ of the vi- 12 ‘Constant surface speed’ , this can be achieved by employing the appropriate ‘canned-cycle’ G-code accessed from the CNC controller, which allows the testpiece’s rotational speed to in- crease as the faced diameter decreases*. * Normally there is a restriction on the rotational speed limit – created by the maximum available speed for this machine tool, which would normally be reached well before the cutting insert has coincided with that of components centre line, but because in this instance, the compacted testpiece is hollow, the rotational restriction does not present a problem. sual aspect of machined surfaces – wherein the real situation is that surface texture continuously degen- erates, and it is only the burnishing (i.e.‘ironing’) of the surface that ‘masks’ the temporary improvement in machined surface – more on this topic will be made in the surface integrity section. What is apparent from using the AWT technique is that on a very short tim- escale, considerable data can be generated and applied research assessments can be conducted both speedily and eciently. is topic of exploiting the minimum machining time and data-gathering activities to gain the maximum information, will be the strategic mes- sage for the following dialogue. Machinability Strategies: Minimising Machining Time, Maximising Data-Gathering Prior to commencing any form of machinability tri- als, parameters for cutting data need to be ascertained in order to minimise any likelihood of repetition of results, while reducing the amount of testpieces to be machined to the minimum. Data obtained from such trials must be valid and to ensure that the cutting pa- rameters selected are both realistic and signicant a disciplined experimental strategy based upon the ‘De- sign of Experiments’ (DoE) approach is necessary – see Fig. 149. Here, a ow-chart highlights the step-by- step approach for a well-proven industrial technique, to maximise the labour-intensive and costly exercise of obtaining a satisfactory conclusion to an unbiased and ranked series of machinability results. ere are a range of techniques that can be utilised to assess whether the cutting data inputs, namely: feeds, speeds, D OC ’s, etc., will result in the correct inputs to obtain an extended tool life, or an improvement in the ma- chined surface texture from the testing program. One such method is termed the ‘Latin square’ – which as- sesses the signicance of the test data and its interac- Chapter Figure 148. Graphical results obtained from the accelerated machinability test, illustrating how ank wear and surface texture degrades, with the number of facing-o passes . Machinability and Surface Integrity tions. For a practical machinability trial employing a ‘Latin square’ , it uses a two-way ANOVA 13 table, with a limited amount of ‘degrees of freedom’ , typically: fee - drate, cutting speed, D OC , plus surface nish – these parameters can be changed/modied to suit the ‘pro- gramme of machining’ in hand. By using a very lim- ited group of cutting trials, a two-way ANOVA table can be constructed and their respective ‘F-ratio’ for each interaction can be determined. is calculated ‘F- ratio’ should be greater than the 5% ‘condence limit’ of the statistical distribution to be signicant. If the F- ratio falls below –5% (i.e. for the calculated F-ratio), then the interactions are not signicant, which ne- cessitates increasing the ‘factor strength’ (e.g. increas- ing the: cutting speed, feedrate, etc.), to generate data which is >5% condence limit – as shown by the ‘feed- back loop’ in Fig. 149, or alternatively, using a dierent factor. By such means, ANOVA tests for signicance of machining data, ensures that the processing parameters utilised for the prospective machinability trial are both valid and the correct ones to use in the proposed ma- chining programme. 13 ‘Analysis of variance’ (ANOVA), or as it should be more ap- propriately termed the ‘analysis of variation about the means’ , consists of portioning the total variation present in a data set into ‘components’. Each ‘component’ is attributed to an iden- tiable cause, or source of variation; in addition, one ‘com- ponent’ represents the variation due to uncontrolled factors and random errors associated with the response measure- ments.Specically, if the data set consists of ‘n’ measurements ‘y 1 .…,y n ’ and their mean is denoted by: ‘y ’ , the total varia- tion about the mean is embodied in the ‘sum of squared de- viations’ , as following diagram depicts, for the ‘partitioning scheme’ for ANOVA: Total Sum of Squares about the mean: n � i= (y − ¯ y) ↓ ↓ ↓ ↓ ↓ Sum of squares – due to Source1 Sum of squares – due to Source2 Sum of squares – due to Source3 Sum of squares – due to Source4 Error, or residual Sum of Squares e technique of analysis of variance decomposes this total ‘sum of squares’ into the parts shown above, for a case in which four identiable sources of variation are present – in addition to the ‘error component’. e number of identiable causes of variation and the formulae for the ‘component sums of squares’ are intrinsically connected to the specic experi- mental design utilised, in the data collection and to the statis- tical model deemed appropriate for this analysis. Rather than spending considerable time, eort and indeed exorbitant expense, on a large and com- plex machining testing programme, which more oen than not, produces numerous machined components that are almost indistinguishable from each other. It might be more prudent, to conduct a ‘condensed’ series of trials, based upon a rigorous statistically-designed methodology. erefore, experiments based on the so- called ‘orthogonal arrays’ can be benecially engaged in this regard. Many applied researchers and engineers have utilised a range of factorial-designed experi- ments, typied by the ‘Taguchi-approach’. e main problem with these ‘arrays’ is that in many situations the large number of ‘interactions’ (i.e. fac- tors) have been shown to interfere with the overall re- sults – introducing ‘secondary eects’ , which will not have been anticipated for, when the original strategic programme was devised 14 . Such spurious data, could seriously aect future machining recommendations and inuence the outcome in a negative manner. e ‘interaction problem’ can have these aects consider- ably reduced by incorporating a more ‘truncated-ap- proach’ to the experimental design strategy for the machinability trials, rather than using a ‘full’ Taguchi orthogonal array (Fig. 150). For example, if all of the experiments are conducted in for example one of ‘stan- dard’ the Taguchi L 8 (2 7 ) orthogonal array, depicted in Fig. 150, then the ‘total outcomes’ (i.e. components machined), would be: 2 7 = 128 × 8 = 1,024 individual components machined. Here, in the Taguchi orthogo- nal array seven factors have been employed and with the vast amount of components produced from such a long-running and very costly machining programme, many of the pertinent details will be lost on those en- gineers/researchers attempting to de-code the vast as- sortment of machinability data collated. However, it is possible to utilise a much simpler-approach to the overall massive data-collection and analysis problem, yet still providing statistical signicance, this can be achieved by adopting a ‘Fractional factorial-designed experiment’. Here, instead of the virtually ‘mindless task’ of producing 1,024 almost identical components, 14 ‘Orthogonal array factors’ – when utilising a ‘full’ Taguchi- designed orthogonal array for a complete picture of all of the interactions, then it has been shown (Shainin, 1985 – see refer- ences), that if many factors are employed (i.e. normally >5), this results in unwanted ‘secondary eects’ which cannot be accounted for, leading to spurious results from any machin- ability trials. Chapter by using a ‘Fractional factorial-designed experiment’ with an identical matrix to that given in Fig. 150, only 8 components are produced! is testing regime is both signicantly quicker and much less costly to perform, obtaining a ‘snap-shot’ of the overall ma- chinability problem, but because considerably less tes- tpieces are produced, the ‘interaction-problem’ and its ‘secondary eects’ are not an issue, even when seven factors are utilised. Obviously, this machinability data has to be collated and investigated in a disciplined and controlled fashion. One tried-and-tested method of establishing an unbiased and ranked interpretation of these results, is to use the much misunderstood and maligned technique of ‘Value Analysis’ 15 (VA). is VA when used to show trends in competitive functions 15 ‘Value Engineering and Analysis’ (VE/VA), with VE being principally concerned with an overall improvement of design- based details on engineering components, while a more lim- ited form of this technique is termed VA – being particularly relevant for detailed interpretation of recorded data from ex- perimentation. Here, in this case, from the wide-ranging and oen seemingly unrelated output of machinability trials. Figure 149. Flow chart indicating the desigh philosophy for unbiased and ranked machinability trials. Machinability and Surface Integrity Figure 150. A fractional factorial-designed experiment, based upon a Taguchi L 8 (2 7 ) – orthogonal array. Chapter and operations, can be successfully utilised from the comparisons of cutting uids, through to complex and dicult-to-machine aerospace machinability trials. If a more sophisticated technique is required, then it is also possible to utilise ‘Quality Function Deployment’ 16 (QFD), to obtain a complete picture of the outcomes from machining trials. QFD is oen used by indus- try as a means for its ‘Continuous-improvement pro- grammes’ 17 . Here for ‘simplicity’s-sake’ , the more basic and somewhat less complex VA tabulated data-colla- tion approach, will be briey reviewed. e application of VA to a series of collated and compiled massed-data is not new. In fact, it was widely-used during the 1960’s, but fell into disfavour, partly because its function and operation were oen not well-dened – this being exacerbated by poor im- plementation of its recommendations. However, VA techniques are useful, allowing one to interpret data trends both quickly and objectively – without undue bias – at a glance of a spreadsheet. Not only can signi- cant trends be readily seen, but the spreadsheet shown in Fig. 151 – shows a typical machinability data for P/ M compacts drilled by two diering drill-point geom- etries. By using the spreadsheet, not only can overall trends be readily seen, it also can depict sub-set trends as well, giving a complete picture (i.e. globally) of the important criteria in assessing machining data. As a simple ranking system is used, considerable objectiv- ity can be gained and with little undue inuence – bias, aecting the outcome from these tabulated results. In employing the ranking of the results, it is normal prac- tice to decrement down and if two values are ranked identically, then they are given the same rankings, fol- lowed by the next lower ranking, being two numbers lower, as following example shows: 16 ‘Quality Function Deployment’ (QFD), is a general term that means the: ‘Deployment of quality through deployment of qual- ity functions’ (Akao, 1988). It is oen known as the ‘House of Quality’ , because the tabulated graphical representation looks similar to that of a house – when all the interacting factors for subsequent analysis have been included on the chart. is QFD technique, is a wide-ranging philosophy for the com- plete analysis of both simple and intricate designs and can be successfully exploited for machinability trials. 17 ‘Continuous-improvement programmes’ , can be dened as an: ‘Operational philosophy that makes the best use of resources in order to increase product, or service quality and result in more eective satisfaction of customers’ (Swanson, 1995). For example, in Fig. 151 – for the values shown in column two (i.e. le-hand side: Jobber drill, rust Force 0.254 N): Compact type: 1 2 3 4 5 6 7 8 Ranking: 6 8 5 1 5 7 3 2 NB Here, two 5’s were ranked, meaning that the next decremented value would rank as 3. Hence, in this case the Low compaction Compact type No. 2 this was best and Low compaction No. 4 worst – as jobber drilled. is ‘truncated approach’ the elementary and easily comprehended VA tabulation (Fig. 151) , enables non- specialists, together with knowlegdible experimenter, to recognize the inuence various machining param- eters have on the potential performance of the trials undertaken. By judicious use, the VA technique in conjunction with a strictly controlled and limited ma- chining strategy – based upon some form of ‘orthogo- nal array’ , in combination with the ‘strength’ (i.e >5% ‘F-ratio’) of parameters by ANOVA, this will enable a researcher to conduct a speedy, compact, realistic, yet meaningful machinability assessment. 7.2 Machined Roundness Roundness is a condition of a ‘surface of revolution’ , which can take the form of a: cylinder, cone, or sphere, where all the peripheral data points (i.e. measure- ments) intersect. In reality, the radius of say, a nomi- nally round workpiece tends to deviate – from the ‘true circle’ – around the periphery of the part, making these variations the theme to subjective interpretation of the measured results. In fact, in the past, the sim- plistic technique for the assessment of roundness was usually measuring three diameters on a workpiece, to determine the diametrical variations, then ‘averaging’ to give its overall dimensional size. Moreover, for vari- ations in a workpiece’s radius about an axis of rotation, this was oen found by positioning the part between a ‘bench-’ , or sine-centres’ – the latter equipment is em - ployed for turned tapered features, then rotating and monitoring it with dial gauges both at and along its length. In the past, this rather supercial metrologi- cal workpiece assessment was supposed to inform the inspector as to its potential in-service performance. If some radial variations occurred, this geometrical Machinability and Surface Integrity Figure 151. Value analysis – tabulation of the performance of two drilling points and a typical range of drilling data, when machining powder metal- lurgy compacts . Chapter lobing 18 , or elliptical state, may have not have proven to be detrimental to its prospective overall in-service performance. In reality, there might be a whole host of reasons for a machined part to vary in its radius – for a stated cross-sectional plane. e following list attempts to show where and why radial dierences occur: • Machine tool and its production processes – induc- ing some form of rotational imperfections from ei- ther the machine/tool/workpiece system, • Release of strain, or that induced into a workpiece – the former case may be the result of releasing the part from its clamping pressure, while the latter may result from plastic deformation promoting lo- calised surface residual ‘hoop-stresses’ 19 , • Induced radial vibration – potentially resulting from cutting forces and its eect on rigidity, in as- sociation with both tool geometry and cutting edge displacement (i.e. see Fig. 152), • Circumferential surface texture – created by the lasting eect resulting from the recent production process. It has been alluded to above that the machine tool and particularly its spindle, can create machine-induced inaccuracies of various kinds onto the machined 18 ‘Lobing’ , has a constant diameter if measured in a single plane. When attempting to measure lobing with a ubiquitous mi- crometer calliper, this is not possible, as a constant micrometer reading will result. Conversely, an ‘elliptical’ workpiece has both a major and minor diameters, allowing this diametral- dierence to be determined using a micrometer calliper. NB A ‘lobed-shape’ can be established, by either placing the workpiece in a Vee-block, then carefully rotating the part and, if any pointer motion appears on the touching dial gauge, this represents the lobed-harmonic dierence. To obtain much more detailed information on a ‘lobed’ workpiece, it is neces- sary to inspect the part on a roundness measuring machine. 19 ‘Hoop-stress’ , this can be dened as: ‘e circumferential stress in a cylinder wall under pressure, or in a rotating wheel [i.e. mass]’ (Carvill, 1997). e maximum hoop stress can be found using the following expression: σ hmax = p (r b + r a ) (r b − r a ) i.e. at the inner radius: σ L = 0) (r b 2 – r a 2 ) Where: r = radius, p = pressure. workpiece’s ‘harmonic roundness’ 20 , some of these fac- tors include: • Spindle imbalance – introducing dynamic lower- frequency harmonics on the part, • Cutting forces – can dynamically aect the machin- ing process, causing a series of high-frequency har- monics to be superimposed on the lower-frequency harmonic, resulting from imbalance (Fig. 152), • ermal growth eects – changing both the spin- dle’s growth (axially) and causing modications of an elastic nature to the relative ‘axis-orthogonali- ties’ 21 of the machine tool – which in turn, creates harmonics on the machined part, • Working clearances and motor-drive congura- tions – this is necessary to allow for relative ther- mal growth and beaing component ‘running-ts’ – within the spindle/headstock assembly, which are exacerbated by the type of motor drive system, spe- cically belt-driven systems (see Fig. 153). is latter feature of spindle inaccuracy, is present in many belt-driven CNC Lathes and turning centre headstocks being minimised by having a machine tool with a direct-drive spindle. In the case of these belt- driven headstocks, the working clearances and belt- drive, have the belts-tensioned on one side only. is arrangement, causes an irregular harmonic rotational motion to the spindle and hence, its work-holding equipment – chuck, etc., which when translated onto the resultant machined roundness and to a lesser ex- tent the surface texture, creates harmonic problems 20 ‘Harmonic roundness’ , refers to the departures from roundness of a workpiece, with harmonic eects – oen termed ‘undula- tions per revolution’ (upr) – being instigated by any number of external sources, such as those described in Table 10. NB On Roundness testing machines, the various harmon- ics are superimposed onto each other. For example, the 1 st harmonic of the workpiece, may have say, the 5 th harmonics superimposed onto it, followed in a similar fashion by 60 th harmonics. is composite harmonic behaviour can be ‘l- tered-out’ by the judicious use of double Gaussian lters, in order to see the eects of individual harmonic behaviour on the machined part. 21 Machine tool ‘axis-orthogonalities’ , relate to the fact that most of today’s 3-axes machine tools have each axis posi- tioned either on top of each other and at 90° with respect to each other (i.e. X- and Y-axes), or normal/right angles to these axes (i.e. with respect to the Z-axis) – hence the term ‘orthog- onality’. Machinability and Surface Integrity Figure 152. The harmonic departures from roundness of a component, resulting from a lack of rigidity/damping eects whilst turning . Chapter . accelerated machinability test, illustrating how ank wear and surface texture degrades, with the number of facing-o passes . Machinability and Surface Integrity tions. For a practical machinability. philosophy for unbiased and ranked machinability trials. Machinability and Surface Integrity Figure 150. A fractional factorial-designed experiment, based upon a Taguchi L 8 (2 7 ) – orthogonal. X- and Y-axes), or normal/right angles to these axes (i.e. with respect to the Z-axis) – hence the term ‘orthog- onality’. Machinability and Surface Integrity Figure 1 52. The harmonic departures