Investigation of machining tool path on surface roughness and dimensional accuracy for high-speed micro milling

This paper investigates the effects of machining tool path and cutting layer strategies on machining efficiency and accuracy in micro-milling of linear and circular micro-geometric features. Although micro-milling includes many characteristics of the conventional machining process, detrimental an effect in downscaling the process can be excessive tool wear which could, in turn, increase the machining forces and hence affect the geometrical accuracy and surface roughness. Most of the research in micro milling reported in the literature has focused on optimising machining parameters, such as feed rate and depth of cut to achieve lower cutting forces, better surface roughness, and better machining efficiency. However, is there yet little known about the effect and stability of machining tool paths and cutting layers strategies for the micro-milling process. Various tool path strategy, including lace(0°), lace(45°), lace(90°), concentric and waveform in producing linear and circular micro geometric features were compared and analysed. The effect of various cutting layer strategies in producing thin walled structure was investigated. The optimisation method with respect to surface roughness and dimensional accuracy is proposed for selection of optimum machining strategies experimentally tested. Experimental results show that the most commonly used strategy lace(0°) and concentric, reported in the literature have provided the least satisfactory machining performance, while the waveform strategy provides the best balance of machining performance for both linear and circular geometries. Adopting an optimum sequence of material removal layer in micromachining of thin walls has proven to improve the overall accuracy. This paper concludes that an optimal choice of machining strategies in process planning is as important as balancing machining parameters to achieve desired machining performance.


Introduction
The nature of manufacturing has changed to reflect the advancement on customer demand for high production rate, process efficiency, and product accuracy [1].The ongoing tendency for miniaturization of products to satisfy the modern manufacturing demand leads to new requirements that are not feasable with current manufacturing technologies [2,3].Strong desire for direct manufacturing of 3D geometric features with high aspect ratios, a wider choice of materials including the use of metallic material, and the cheaper manufacturing cost has been on demand by micromanufacturing industries such as medical devices and micro-molds.Material removal characteristic shows a significant difference at micro/meso scale, and differences are the consequence of scaling down between the principal constituents of cutting operation performed, although kinematically is similar to conventional milling [4].Current tool manufacturing restriction leads to micro tool cutting edge radius to be comparable with size of part geometry beside available equipment cannot achieve an optimum machining parameter required for micro tools within feasible cost [5].In micro milling, the cutting process is described as the transition from cutting dominated to a ploughing-cutting process where the tool edge radius has a significant effect once uncut chip thickness falls below the tool edge radius know as minimum chip thickness [6].Minimum chip thickness has stated to overrule the material removal behaviour [7] in the process of downscaling affecting the surface roughness of finished part.Hence surface roughness has been used as one of the main references to evaluate the micro-milling process and selection of appropriate machining parameter [8][9][10].Meng & Li [11] suggested there is only four machining parameters mainly affecting the surface roughness(R a )namely; spindle speed, feed rate, axial depth of cut and the length of cutting tool.Rational gray analysis was used to identify the relational degree of each parameter in refrence to surface roughness(R a ), suggesting feedrate and axial depth of cut has the most and cutter length has the least effect.Bandapalli et al. [9]studied the influence of feed rate and axial depth of cut using Taguchi method confirming the lower feedrate and a smaller depth of cut provides a better surface finish due to more stable machining enviroment.Fu et al. [12] proposed a feedrate optimisation method by analysing the cutting forces along the toolpath and utulising Newton-Raphson iteration algorithm to alter the feed rate in the original tool path file to achive a constant feedrate and steady machining.Mayor et al. [13] also developed a variable feedrate intelligent segmentation method to compensate non consistance feed rate using interpolation technique applied to segments along the tool path.Pedro and Paulo [14] investigated three commercially used tool paths; continuous overlap spiral, parallel spiral, and parallel zigzag by comparing the finished surface roughness and machining time suggesting; constant overlap spiral provide a better machining performance in compare to others.Banerjee et al. [15] suggested circular tool path can avoid the discontinuities in the tool movement providing a consistent feed rate and smooth material removal.Further to the downscaling effect on the production of micro parts using conventional process, low rigidity of parts result in a significant increase in the deformation of both micro tool and workpieace [16].Research was done on compensation methods to reduce the resultant machining deformation.Smith et al. [17] experimentally tested the effect of toolpath on machining of thin webs suggesting at process planning toolpath should be chosen with consideration that the section being machined is supported by as much unmachined workpiece as possible.Kim et al. [18] analyzed the surface error due to the deflection of the cutting tool and geometrical deformation to be compensated in the toolpath planning.Chen et al. [19] proposed an active error compensation for each layer of machining, compensation method compares the predicted deformation from the previous layer and adjust the machining depth of cut for the next layer suggesting active multilayer compensation method is more efficient than full compensation method.Gao et al. proposed a mirror machining deformation compensation using the location of the cutter and estimation of the the tool deformation to offset the toolpath.The proposed method was experimentally tested and show a 52.88% decrease in deformation in machining of thin wall structure.Based on above literature major limitation to widespread use of micro milling is the stability of machining operation.In tool path planning constant feedrate and engagement of tool are critical to stablise the machining environment.Hence this paper experimentally investigate the effect of comonly used toolpaths in micromachining of circular and linear geometries and the effect of different cutting layer strategies using constant feedrate by analysing surface roughness, geometric accuracy and machining time.An optimization method is proposed to provide an optimum tool path and cutting layer strategy selection at different machining stages with an aim to improve machining efficiency and accuracy.

2-Methodology of process planning for high-speed milling
This work consist of three phases; Modeling, experiment and optimization for optimum selection of toolpath and machining strategy by looking at 4 common manufacturing aims: machining accuracy,machining surface finish, productivity and balance of all three.The modeling phase used a well-established cutting force model to predict machining forces.Finite element enviroment was developed to assess the impact of predicted machining forces using different cutting layer strategies.Data collected on the impact of machining forces on geometric behaviour of the test sample was analalysied and used to propose optimum cutting layers strategy for low rigid parts.Second phase , the experimental phase was where the physical micro machining were conducted to first validate the cutting force model and numeric model used in modleing phase.Follow by machining of on thin wall structures using proposed cutting layer stratgies and machining of linear and circular geometrise using comonly used toolpaths to collect neccecery data on geometrical accuracy, surface roughness and machining time.Finally, optimization method was proposed for an optimum selection of machining toolpath and cutting layer strategy looking at 4 common manufacturing aim. Figure 1 presents the work flow and method used in selection of optimum Machining startegy.

1-Machine tools
Experimental work has been carried out on a standard Hurco precision CNC machining center (VM10) to ensure the industrially feasible results.A high-speed spindle(NAKANISHI -HES810) with electric drive and ceramic bearings were retrofitted to the main spindle.The high-speed spindle is capable of the continuous power output of 350W and output torque of 3cNm over the speed range of 20,000-80,000rpm allowing for higher cutting velocity with smaller diameter tools.Ultra precision collets were used to clamp the micro tool, and spindle run-out was controlled at 1µm.The machining center used to offer a single axis positioning accuracy of 5µm where the experiment has been designed to compensate for positioning error as it will be detailed in section 2.4, and will not influence the surface roughness and topography which are the main criteria of this research.Spindle error has been stated to have a significant impact on the surface roughness, in this experiment the main spindle was on mechanical lock throughout the experiment, the spindle error will be limited to the vibration and runout of the high-speed precision spindle.This experimental set up ensure that both the main spindle error such as vibration, run out and slideway error has been minimised.Therefore it will not be taken into account in the analysis.

3.2-Micro end mill
Micro flat, uncoated tungsten carbide tool (WC) end mills were used in this experiment with a nominal diameter of 1 mm.A nominal tool shank diameter of 3mm used to fit 3 mm ultra-precision spindle collet.Large tool diameter has been selected to prevent premature tool failure due to harsh machining environment.In each experiment, the new tool is chosen to ensure the endurance of the tool without excessive tool wear and chipping.Table 1 presents the geometries of selected micro end mill tool used in this experiment.Tools have also chosen from the same batch to reduce the randomization error due to the dissimilarity of the micro end mills due to different tool manufacturing techniques.Table 1: Micro tool geometries used in this experiment

3.3-Finite Element modelling
A numerical model of micromachining environment was developed in ANSYS to assess the impact of different machining layer strategies on the tool and low rigid micro-geometries.Explicit dynamic model of a micro cutting tool was drawn up in finite element environment to predict the deflection of cutting tool due to resultant machining forces.Johnson Cooks material constitutive strength model was used to account for strain hardening, thermal softening and elastic recovery of the cutting tool; described in equation 1: Where   is an effective plastic strain,   * is normalized effective plastic strain rate and   is homologous temperature.The five material constant are shown in  [20] Constant (A) is the initial yield stress, (B) and (n) represents the effect of strain hardening,(C) is the strain rate constant and (m) is the thermal softening exponent.Resultant cutting force (F) was applied as magnitude of F x and F y force in Eq.( 1) Cutting forces along the X axis, (F x ) and Y axis (F y ) in equation two have been calculated from a wellestablished mechanistic cutting force model adopted from Chang and Chen [21].
Where K te , K tc , K rc , and K re are material shearing coefficients, ℎ  (∅  ()) is uncut chip thickness and ∆ is depth of cut.The cutting force has been experimentally validated prior to FE simulation explained in section 4.1 through machining of slots with identical machining parameters used in the experiment detailed in section 3.1.The effect of machining forces on rigid structure was analysed by measuring maximum deformation of 30µm thin wall structures through applying static point load across the tool path illustrated in figure 3.

3.4-Experimental procedure
In this experiment, half immersion slot milling was used to machine circular and linear geometries; figure 2 illustrate experiment setup on Aluminum 6061-T6 today's leading non-ferrous metal in use.1mm tools were selected from a single batch to reduce the tool geometry randomization used to conduct dry milling.The surface finish and dimensional accuracy have been obtained from the experiment samples using optical 3D measurement surface profilometer; Alicona InfiniteFocusSL with a vertical resolution of 50 nm.Average measurable roughness (Ra) of the machined surface was recorded across the square and circular geometries.The circular geometry surfaces were scanned and flatten before average R a could have been obtained.Machining parameter used across both experiment were; Spindle speed of 60000 RPM, radial depth of cut of 0.5 mm and 200 mm/min feed rate.The depth of cut was fixed to 2.5mm for machining toolpath experiment while at machining layers strategies it varied from 1 to 3mm.Machining strategies one to four feature from a full sweep of material removal across the length where strategy one proposed to remove 50% of total height of the wall from each side at each stage whereas strategy three suggest a removal of all material from one side follow by remaining on the opposite side.Strategy four test the effect of a smaller layer of material removed from each side of the wall.The numerical experiment data in section 4.2 suggest a maximum geometry deformation was negligible up to 70% of material removal across the feature height.Therefore strategy two suggested removing 70% of material from each side with at a maximum depth of cut achieved by the micro tool follow by remaining of the material using only 10% of height per layers.Strategies five to eight, however, adopted a none continuous sweep of material removal along the length.Strategy five and seven remove 70% of material across the length and then remove the remaining uncut material at 10% of feature height layer by layer at strategy five and strategy seven propose removed 70% full height maximising the depth of cut follow the remaining material at 10% of full height layer by layer from each side.Strategy six and eight, on the other hand, propose the removal of material from the middle leaving uncut material on the end edge of the thin wall follow by removal of the remaining from each side inside to outside for strategy 8 and opposite approach on strategy 6.

3.4.2-Machining tool path
Figure 5 presents a schematic diagram of each tool path strategy used to remove the same volume of material using fixed machining parameters and machining layer strategy 4 presented in figure 4.

Figure 5: schematic diagram of strategies used in this experiment
Toolpath used in concentric strategy involves in a circular movement of the tool using constant diameter as the tool merge in and out of the material.Waveform, however, uses variable diameter toolpath as the toolpath diameter triple while the tool comes out of the material.Lace 0° follows a parallel toolpath to the finished geometry which removes the material from outer to inner layer by layer.lace 45° and lace 90° adopted a 45° toolpath from the tangent and perpendicular to the finished geometries respectively.For both lace 45° and lace 90°, the path begins at one end of the feature and goes round the desired geometry removing all the material from outer to inner.

4.1-Varification of cutting force model
The experiment was set up to validate the cutting forces calculated using Chang and Chen [21] mechanistic cutting force model through machining of microchannels using five different depth of cuts varies from 0.1 to 0.3 mm using matching machining parameters stated in section 2.4.
Measured average cutting force across X, Y and Z was used to plot figure 6.Average cutting forces obtained experimentally was compared with calculated forces presented in Table 3.

: Calculated and measured cutting forces
The calculated cutting forces are shown to be in good agreement with experimentally measured forces along X and Y-axis.Experimentally measured cutting force has been carried forward into numerical model FE to predict the resultant tool deflection and to predict the geometrical accuracy of the finished parts affected by tool deflection.

Figure 7: Micro Tool Deflection using numerical model
Figure 7 shows the maximum tool deflection of 0.0049mm due to experimental cutting forces, assuming there will be no deformation in the workpiece.Machining undercut of 0.0049mm from each side of the feature predicted to results in 0.0098mm in the overall geometric accuracy of the finished parts.

4.2-Numerical experiment of different machining layer strategies
Figure 8 and 9 show the effect of cutting forces on the geometrical accuracy of the thin wall using different proposed machining cutting layer strategies where maximum deflection as a result of applying static force at ten equal distance across the machining path been recorded from each side of the wall.Table 5 presents the maximum geometry deflection predicted on the finished parts assumed to be the total plastic deformation expected.Machined samples were analysed using SEM where geometrical deviation, maximum deformation and machining time from machining experiment been measured and recorded for each machining layer strategies in table 5.

Table 4: Numerical and experimental results in cutting layers strategies test
Although the machining environment in the numerical model has been optimised to provide an accurate estimation of resultant machining deformation, inconsitant difference between predicted

Strategy
No. and measured maximum deflection was observed.In respect to machining time strategies 1 and 3 advantage over lowest overall machining time, also leaving 50% uncut material to support the thin wall shown to significantly reduce maximum deflection from 0.09µm to 0.02µm.However, due to the average geometrical deviation for strategy three being below the threshold, this strategy identified to be not suitable for micromachining.Strategies four and eight have resulted in almost the same resultant machining deflection where strategy four took nearly twice as long with significanlty large geometrical deviation.Strategy two and five also have similar machining time, however, the strategy two shown to be more suitable due to lower deflection and geometrical deviation.Strategy six in compare to other strategies proposed, measured the lowest overall; deformation, average machining time and geometrical deviation.Each machining strategy has shown to be suitable for when the target of manufacturing is focused on specific task such as machining effeicent, machining accuracy or both.Due to different machining required an optimisation method to compare and select the suitable machining layer strategy for various machining scenarios.

4.3-Machining tool path strategies
Figure 8 presents tool strategies; Lace 45°, Lace 0°, Lace 90°, concentric and waveform simulated for circular and linear geometry using EdgeCAM simulator.Data on machined surface roughness and geometrical accuracy of each machine features were measured, and machining time for each strategy was recorded and summarized in Table 5.  to be significantly larger.In the linear feature, the desired finished width was 10mm according to model specification.In reference to the numeric model, prediction of the finished part was predicted to be 0.0098 undercuts across all strategies providing a 10.0098mm finish width.However, experimentally measured results have shown a significant difference summarised in Table 3. Tool strategies lace 0°, lace 45° and concentric provided a similar performance and led to the deviation of 0.4mm, 0.4mm, and 0.5mm in geometric accuracy respectively.However, lace 90° has shown a slightly better performance due to the smaller deviation of 0.1mm but still far from the predicted result.On the other hand, waveform strategy has shown a significant performance compared to the other strategies with a deviation of 0.02mm which is close enough to predict accuracy using the numeric model.In circular geometries, the desired finish diameter was also 10mm.Experimental results summarised in Table 5 show a diversity of deviations in geometric accuracy using different machining strategies.However, lace 45° and concentric strategies led to material overcut proved to be not suitable for machining of circular geometries.The deviations for geometrical accuracy lace 0° and waveform were measured to be 0.5mm and 0.3mm respectively.Nevertheless, lace 90° resulted in a significantly smaller deviation at 0.1mm shown to be more suitable for the use of machining circular geometries.The resultant surface roughness of each machining strategy was recorded in Table 5 where waveform and lace 45° provided the lowest surface roughness in machining of linear and circular geometries respectively.Machining time for each machining strategy presented in Table 5 are the over all machining time including the processing time for each tool path.In this experiment due to small chip load and use of free cutting material the tool wear assumed to be negligable.The overall machine time indicat that lace 90° is the most efficient strategy compared to the rest, scored the lowest machining time of 50 min while lace 45° showed to be the least efficient strategy, scored the longest machining time of 108 min.The machining time also include the time the CNC macine takes to process the tool path that would include cutting time.Small Due to different design requirements, an optimisation method has been adopted [14] to identify the most suitable strategy for both circular and linear geometries accordingly.

4.3.1-Optimization
Machining strategies were compared in reference to geometrical accuracy and machining time.Also, machining tool paths were compared in reference to surface roughness while machining layer strategies were compared in reference to maximum deformation.A mathematical formulation as a function of all references was developed: ( , , ) = Where x,y and z are the weight factors to add up the contribution accuracies(α), the surface roughness(β) and machining time(γ).The optimisation process approaches by evaluating the machining strategies to which variable the customer/engineer wants more, then add up the contributions and look for biggest total given.The score table 7 shows the ranking of each tool path strategy based on each design requirements, Geometric accuracy, Surface roughness and machining time individually where suitable strategies can be chosen when each of above criteria has a higher priority on design specification.The balance column presents the ranking when a balance of all criteria is requested in the design specification.Given the geometrical accuracy and surface roughness is the target of product design specification, Waveform strategies are most suitable since it scored highest at 1.22 and 1.40 respectively.Providing the productivity is the goal, Lace 90° would be the suitable strategy to be used with a score of 1.39.In the scenario where design specification requests a balance of all criteria's, waveform as scored the highest at 1.30.The impact of various tool path strategies has shown to be more significant in machining of circular geometries in compare to linear geometries due to the wide range of scores for each criterion.Considering different manufacturing goal, in both cases of high accuracy and low surface roughness criterion waveform had scored the highest However this trend changed when optimisation goal focused on productivity, Lace 90 identified to be most suitable whereas waveform was the least desired tool path to be used.The score table 8 presents the ranking of each machining layer strategies providing the goal of manufacturing is to have a balance of all the criteria's, strategy 1 scored the most suitable.However, since the geometrical deviation in Table 4 indicate the finished geometry was over cut this would be the least suitable choice.Strategy 3,4 and 5 had a very similar performance all remain non-competence with strategies 6-8.Providing the strategies feature from continues the sweep of material removal have not scored high enough to compete with others indicates conventional machining layer strategies are not suitable to be directly downscaled and used at micro scale.Strategies 7 and 8 has a noticeably high score for all criteria's.However, strategy six scored the highest for all criteria's suggesting at process planning for micro milling leaving the uncut material at the weakest point of the geometries to be machined at last will improve the overall machining performance.

5-Conclusion
This paper presents an experimental investigation of different machining strategies and usefulness of integrated toolpath and machining layers strategy optimisation methodology for a highperformance micro end milling of Aluminum 6061-T6.The proposed optimisation model provides an optimum tool path and cutting layer strategy selection based on machined part requirement.
The following conclusion can be drawn from this work: • Choosing optimal machining strategies is as equally important as choosing an optimum machining parameter to achieve the overall goal of optimum machining performance and productivity.

•
In process planning of machining tool path, the strategy has to be chosen accordingly for feature part geometry.

•
Different machining layer strategies result in variation of geometric deformation wherein machining of low rigid feature the material removal should proceed from the least supported location to most supported location • Low surface roughness and high accuracy currently are achieved in exchange of productivity, however in process planning by selection of suitable machining strategy balance of high-performance machining and productivity can be accomplished.

Figure 1 :
Figure 1: Flow chart of optimum machining strategies selection 3-Experimental set up 3.1-Machine toolsExperimental work has been carried out on a standard Hurco precision CNC machining center (VM10) to ensure the industrially feasible results.A high-speed spindle(NAKANISHI -HES810) with electric drive and ceramic bearings were retrofitted to the main spindle.The high-speed spindle is capable of the continuous power output of 350W and output torque of 3cNm over the speed range of 20,000-80,000rpm allowing for higher cutting velocity with smaller diameter tools.Ultra precision collets were used to clamp the micro tool, and spindle run-out was controlled at 1µm.The machining center used to offer a single axis positioning accuracy of 5µm where the experiment has been designed to compensate for positioning error as it will be detailed in section 2.4, and will not influence the surface roughness and topography which are the main criteria of this research.Spindle error has been stated to have a significant impact on the surface roughness, in this experiment the main spindle was on mechanical lock throughout the experiment, the spindle error will be limited to the vibration and runout of the high-speed precision spindle.This experimental set up ensure that both the main spindle error such as vibration, run out and slideway error has been minimised.Therefore it will not be taken into account in the analysis.3.2-Micro end millMicro flat, uncoated tungsten carbide tool (WC) end mills were used in this experiment with a nominal diameter of 1 mm.A nominal tool shank diameter of 3mm used to fit 3 mm ultra-precision spindle collet.Large tool diameter has been selected to prevent premature tool failure due to harsh machining environment.In each experiment, the new tool is chosen to ensure the endurance of the tool without excessive tool wear and chipping.Table1presents the geometries of selected micro end mill tool used in this experiment.Tools have also chosen from the same batch to reduce the randomization error due to the dissimilarity of the micro end mills due to different tool manufacturing techniques.

Figure 2 :
Figure 2: Point load locations across the length of thin wall

Figure 3 : 3 . 4 . 1 -
Figure 3: An illustration of experimental setup 3.4.1-Machininglayer strategies Figure 4 presents a schematic diagram of selected cutting layer strategy put forward to machine 30µm thin wall structure.

Figure 4 :
Figure 4: Schematic diagram of cutting layers strategies compared

Table 5 :Figure 10 :
Figure 10: Machining strategies experimentally tested Although the machining environment in the numerical model has been optimised to provide an accurate estimate of tool deflection, the inaccuracy due to tool deflection experimentally has shown

Table 2 :
Johnson-Cook strength model constants for cutting too material

Table 6 :
Table 7presents weight factors looking at four global design requirement aim to achieve; High geometric accuracy, Low surface roughness at tool path selection and maximum deflection at cutting layer strategy selection, high productivity and the combination of all of criteria's.Weight factors targets at each manufacturing aims Scores for different weight based on design requirement for each scenario was evaluated and recorded in Table7and 8 for both cutting tool path selection and cutting layers strategy selection; higher score presents a better performance and suitability of the strategies for a given weights.

Table 7 :
Toolpath strategies scores for different weighting of each scenario

Table 8 :
Cutting layers strategies scores for different weighting of each scenario