Comparative analysis of wobble milling, helical milling and conventional drilling of CFRPs

Due to its excellent specific mechanical properties, carbon fibre-reinforced polymer (CFRP) composite is a widely used structural material in the aerospace industry. However, this material is difficult to cut, mainly due to its inhomogeneity and anisotropic features and because of the strong wear effects of its carbon fibres. In the scope of aerospace industrial uses of this material, thousands of holes have to be machined for purposes of assembly. Nevertheless, conventional drilling technology – even if special drilling tools are used – is only moderately able to manufacture good quality holes. Wobble milling is a novel advanced hole-making technology, which has been developed to minimize machining-induced geometrical defects like delamination or uncut fibres. The main objective of the present paper is to compare wobble milling, helical milling and conventional drilling technologies concerning unidirectional CFRPs. In addition, the kinematics of wobble milling technology is discussed in detail. In the scope of this paper, numerous machining experiments were conducted in unidirectional CFRPs: herein the impact of the type of cutting tool and of process parameters on the quality of machined holes are analysed and discussed (diameter of holes, circularity error and characteristics of uncut fibres). During these investigations, experimental data were evaluated with the help of digital image processing (DIP) and with the help of analysis of variance (ANOVA) techniques. Experimental results show that the amount of uncut fibres can significantly be minimized through the application of wobble milling technology.

Specific mechanical properties of carbon fibre-reinforced polymer (CFRP) laminates are excellent [1,2]: these composite materials are therefore widely used in high-tech industries like the automobile and aerospace industries [3]. CFRP laminates can be ready-to-shape manufactured using numerous advanced and automated technologies like autoclave or closed mould vacuum bagging [4,5]. Nevertheless, after the curing processes, there are still many micro-and macro-geometrical features that have to be machined mechanically. In the aerospace industry, thousands of high-quality holes have to be machined in order to assemble CFRP parts [6,7]. However, CFRP composites are difficult-to-cut materials due to their (i) anisotropic and (ii) inhomogeneous features, and because of (iii) the abrasive wear effects of their carbon fibres on the cutting tool [8][9][10][11][12]. Furthermore, in this case, carbon-chip treatment is also problematic [13] as the chips have to be removed from the machining zone due to (i) health issues and (ii) the wear effect of the carbon fibres on the different parts of the machine tool.

Conventional drilling and helical milling of CFRPs
Conventional drilling technology does not require any difficult tool paths or machine tools; this technology can therefore be easily adapted by the industry. Nevertheless, due to the push-out effect of the drilling tool during exiting the composite, laminated layers can easily separate from each other: delamination and uncut fibres can be formed [23]. Davim et al. [53] outlined that feed rate has the highest physical and statistical impact on delamination during conventional drilling of CFRPs. Furthermore, improper geometry of the cutting tool can peel up the top layers of the composite and can also form geometrical defects [7]. Gaitonde et al. [6] claim that highspeed cutting can play a major role in reducing the damage at the entrance of the hole and that the combination of low feed rate and point angle is also important in minimizing delamination during the drilling of CFRP composites. Rubio et al. [54] also concluded that high-speed cutting in drilling is suitable for reducing damage and that it could also increase material removal rate. Given this, researchers and tool manufacturers have developed special cutting tool geometries in order to decrease peel-up and push-out effects [55]. Vigneshwaran et al. [56] observed that for achieving better drilling results, it is recommended to develop novel drilling tools. Abrão et al. [57] showed that 56% of the tools used for drilling polymer composites have a special geometry. Some of the special drilling tools developed for CFRP cutting are as follows: (i) double point angle twist drills [58][59][60][61][62][63][64], (ii) brad and spur drills [21,[65][66][67][68][69][70], (iii) dagger drills (one-shot drills) [28,31,[67][68][69][70][71][72][73], (iv) step drills [67,68,[74][75][76][77][78][79][80], (v) core drills [81,82] and (vi) special core drills [83][84][85]. These special drilling tools are reviewed and discussed in detail by Geier et al. [55].
It is often possible to produce good quality holes with the help of special drilling tools (often featuring diamond coating) using optimized process parameters. However, these technologies often require a considerable amount of optimisation time and cost. Therefore, researchers developed novel milling technologies in order to decrease tool costs and to improve hole quality. Special hole machining technologies, like helical milling, tilted helical milling or wobble milling, require longer operation times due to longer tool paths, but the amount of machining necessitated by geometrical defects as well as tool cost can significantly be decreased when such special hole machining technologies are applied.
Tool paths of helical milling (also known as orbital drilling) are more complex than tool paths of conventional drilling: operation time of helical milling is therefore longer and more time is needed for optimisation [86]. As can be seen on Fig. 1a, the axial force component of cutting force can still peel up the upper laminated layers of the composite, but the push-out effect is minimal due to the compression axial cutting force (F a ) component. Machining-induced geometrical defects can therefore be decreased at the exit of the holes, according to Wang et al. [44]. Furthermore, helical milling technology usually does not require special milling tools; thus it can be more effective than conventional drilling, as demonstrated by Geier and Szalay [7].
Conventional drilling and helical milling strategies in CFRPs were compared by Voss et al. [86]. They found that orbital drilling produces fewer geometrical defects Fig. 1 Schematic drawings of acting cutting forces in the case of (a) helical milling and (b) wobble milling of fibre-reinforced composites (delamination and uncut fibres) in CFRPs than conventional drilling, which Geier and Szalay [7] also observed. However, the operation time of orbital drilling was two to four times longer than that of conventional drilling. Furthermore, Brinksmeier et al. [87] described the kinematics of helical milling. They showed that axial cutting force can be decreased significantly if helical milling technology is used. Qin et al. [43] analysed delamination with the help of artificial neural networks during the helical milling of CFRPs. Qin et al. found that helical milling has great potentials in comparison with conventional drilling. Based on a multi-objective Pareto genetic algorithm, Wang et al. [44] optimized cutting parameters (cutting speed, tangential and axial feed rates) during helical milling in order to optimize cutting force, peel-up delamination and productivity. Chang et al. [88] studied micro helical milling using flat end mills. They concluded that the microcutting efficiency of micro helical milling is theoretically larger than that of conventional micro-cutting.
Li and Liu [32] analysed surface topography and surface roughness of hole making with respect to helical milling. They stated that material removal rate can be maximized, while the surface roughness of machined surfaces can easily be controlled. Liu et al. [45] analysed the temperature variation of helical milling with respect to CFRPs. Their proposed temperature prediction model is able to simulate the temperature distribution of CFRP workpieces. Additionally, Sakamoto and Iwasa [35] examined cutting temperature during helical milling using a ball nose end mill. They showed that compressed air could efficiently reduce heat at the cutting point. Using ultrasonic vibration and liquid nitrogen, Ishida et al. [46] analysed the thrust force, tool wear and accuracy of machined holes in the case of helical milling. They concluded that thrust force could be reduced when using ultrasonicvibration-assisted helical milling and cryogenic cooling. In addition, Haiyan and Xuda [89] developed a mechanistic force model in order to predict cutting force in the case of helical milling of CFRPs.

Wobble milling of CFRPs
Wobble milling is an advanced hole machining technology. During wobble milling, axial cutting force at the hole entry can be further decreased by tilting the cutting tool, as can be seen on Fig. 1b. By tilting the cutting tool, the acting cutting force compresses the upper and downer laminated layers of the composite; therefore peel-up and push-out effects can be minimized. Tool movements can only be implemented by a complex machine tool or an industrial robot: this technology is therefore more expensive and needs detailed design and optimisation. However, there are some industrial fields (like aerospace and space industries) where huge CFRP parts are manufactured by industrial robots. In these contexts, the implementation of wobble milling yields no additional machine tool costs.
Schulze et al. [50] observed that machining-induced damage can be reduced by advanced, complex machining strategies, which direct cutting forces inwards. Schulze and Beke [52] investigated the impacts of process and workpiece parameterssuch as feed, tool inclination, tool spiral angle and cutting-edge radiuson resultant machining force vectors. They also showed that wobble milling is insensitive to tool wear and they developed a mechanistic model to calculate chip geometry and specific forces in the scope of their study [51]. In addition, they figured out that during wobble milling the top laminated layers of CFRPs can be machined with the resulting process force vector directed towards the centre of the workpiece. Furthermore, they stated that five-axial advanced machining strategies can achieve better quality holes than conventional drilling. It can thus be concluded that the number of studies on wobble milling is fairly limited: a more extensive discussion of the kinematics and usability of wobble milling of CFRPs is therefore required.
The main objective of the present paper is to compare wobble milling, helical milling and conventional drilling technologies in the case of unidirectional CFRPs. Furthermore, the kinematics of wobble milling technology is also discussed in detail. In the scope of our investigations, numerous machining experiments were conducted in the case of UD-CFRPs using three different milling tools. In the present study, the impacts of (i) the type of cutting tool, (ii) of feed rate and (iii) of the depth of cut (i) on the machined diameter of holes, (ii) on the circularity error and (iii) on the characteristics of uncut fibres are analysed and discussed.

Setup and methods of the experiment
In the scope of this study, holes were machined in UD-CFRPs using three different drilling strategies: (i) conventional drilling, (ii) helical milling and (iii) wobble milling. The kinematics of conventional drilling and helical milling is moderately complex, but the kinematics of wobble milling merit a more thorough discussion.

The kinematics of wobble milling
Wobble milling is an advanced hole-making technology, which can effectively decrease geometrical damage (e.g. delamination, uncut fibres, fibre pull-outs) on machined surfaces. However, this technology requires a complex five-axis motion cycle. This advanced tool path can be generated by special machines, by industrial robots or by five-axis machining centres. In the case of huge CFRP parts to be machined, the parts are usually fixed and the cutting tool makes the following movements.
First, the tool centre point (TCP) assumes the starting position of the cycle (situated above the centre of the hole), as can be seen in Fig. 2a. Then the tool moves on a linear path at a feed rate (v f ) until reaching the drilling depth, as can be seen in Fig. 2b. Second, the cutting tool starts to rotate around the centre point of the hole (Ointersection point of the axis of the tool and the axis of the holewhich is the rotation point), until the axis of the tool reaches the required tilting angle (β v ). Simultaneously with performing the tilting movement, the tool starts to rotate around the hole axis (ω), as illustrated in Fig. 2c and d. After the axis of the cutting tool has reached the nominal diameter of the final tilting value (β v ), the cutting tool returns to the vertical position and stops rotating around the hole axis, as can be seen in Fig. 2e. Afterwards, the cutting tool starts to rotate around the hole axis (ω) and moves away from it (r) at a feed rate (v f ), as illustrated in Fig. 2f. Finally, the tool retracts from the surface of the hole and returns to the starting point of the wobble milling cycle, as illustrated in Fig.  2g and h.
The geometrical model of the wobble milling used in this study can be seen in Fig. 4. Based on this model, the discrete points of the tool path of the milling cycle can be generated, using trigonometrical correlations and interdependencies, as follows: (i) One of the main objectives was to describe and program the wobble milling cycle in as simple terms as possible. It was also a significant aim to develop a flexible and accurate algorithm to ensure the widespread use of this application. Last but not least, it was a further goal to automate CNC code generation. (ii) In the case of the application of conventional tool offsets (programming the tip point of the tool), five axis movements have to be defined and programmed. However, by way of the offsetting of the programmed points (TCPtool centre point) to the centre point of the hole (O), only two coordinates have to be defined and programmed. This approach is applied in the scope of this study, too. (iii) Due to the kinematics of the machine tool used in the scope of this study, rather than rotating the cutting tool, it was the workpiece that was rotated on a rotary table (ω), as shown in Fig. 2. Based on relevant simplifications, just two coordinates were calculated: the rotation of the Table ( 1) and (2), respectively.
The equations were calculated to ensure a constant cutting depth (a) during the process, as shown in Fig. 4. As a first step, the last and the first tool rotations were calculated; then, if it was necessary, further tool rotations were also calculated. When the currently calculated tool rotation reached a higher value than that of the last tool rotation, the algorithm stopped, and the tool value rotations (β i ) in each full workpiece rotation were saved for further processing.
The CNC program for wobble milling was generated based on the algorithm shown in Fig. 5. This algorithm was experimentally tested and validated. In the above figure, q is the resolution of the rotary table expressed in degrees, n is the spindle speed, F are the feed rates, f is the parameter that counts the number of rotations of the rotary table and e is the parameter that counts generated calculated controlled positions. This algorithm generates (i) the pilot hole drilling path and then (ii) the controlled points for wobble milling and then creates (iii) the helical milling path for the final machining step. (These steps were discussed in detail in relation to Fig. 2.) The algorithm is independent of any type of CNC controller; it can therefore be applied in any CNC environments with the following main limitations: (i) the axis of the hole has to be parallel to the axis of the rotary table; furthermore, (ii) the tool offset has to be specifically set, as discussed before.

Experimental setup
A unidirectional carbon fibre-reinforced polymer (UD-CFRP) composite material was applied for the machining experiments, where the matrix material was vinyl ester. The composite plate thickness was h = 5 mm. The CFRP specimens were prepared by a water jet machine: specimens with a diameter of d = 34 mm were cut.
The experiments were conducted on two different machine centres. A VHTC 130 linear machining centre (five axis) was used for the wobble milling experiments, with a maximum speed of ω = 3000°/min concerning the machine's rotary table. This feature limited the maximal feed rate to v f = 262 mm/ min (calculated for a diameter of 10 mm). A Kondia 640B machining centre (three axis) was used for the conventional drilling and helical milling experiments. Both of the cutting machine centres were equipped with a Nilfisk GB7333 vacuum cleaner in order to clear the cutting area of the carbon chips.
Two different types of compression end mills and a conventional end mill were applied in the scope of the experiments. The cutting tools can be seen on Fig. 6, and their important parameters are summarized in Table 1. Each of the cutting tools is made of uncoated solid carbide. The compression end mills differ from each other in the number and size of the cutting edges: the medium tool has more but smaller cutting edges than the coarse tool, as can be seen in Fig. 6a and b. Tools with a working diameter of d = 6 mm were applied for the helical and wobble milling experiments, and d = 10 mm tools were used in the scope of conventional drilling experiments.
A special fixture was designed and manufactured for the wobble milling experiments, as can be seen in Fig. 7. It was designed to fit in a h = 4-6 mm thick CFRP workpiece, with a diameter of d w = 30-35 mm. The α = 90°c one ensures that the tool can take up a position at the required angle (~31°). The fixture has an n = 3 socket head screw, with spring washers ensuring an appropriate clamping force. With the help of this fixture, the clamp was equipped with an evenly balanced support at a location close to the final diameter. This arrangement also minimized the vibration of the workpiece. The back-up support of the CFRP specimen was fixed to d s = 15 mm (p = 150% of the nominal hole diameter). A special backup support plate (Fig. 8) was also used for the conventional and helical milling experiments in order to minimize peel-up and push-down delamination effects.
A Zeiss UMC 850 coordinate measuring machine (CMM) was used to measure the diameter and circularity errors of the holes. N = 500 measuring points were used for each hole at a depth of the centre plane of the plates (at a depth of z = 2.5 mm). The cutting wavelength was set to 50 UPR Gaussian filter, based on the hole diameter. The speed of touch probe was set to v = 2 mm/s and the touch probe radius was r = 1 mm. Furthermore, a Dino-Lite Pro 123-012676 digital microscope was used to record images of the entry and exit sides of the holes. In order to ensure the same conditions, photos were taken in the lower half of the fixture, and for this purpose, the same settings were used.
Due to the extremely high experimental costs (cutting tool, specimens, etc.), the number of experiments of wobble milling had to be minimized. The experiments were therefore designed based on a central composite face centred (CCF) design, using Minitab software. The factors were chosen in line with previous studies: thus feed rate (v f ) and cutting width (a) were set based on previous studies. Minimum and maximum levels of the factor space of the experiments were set to vf,min = 150 mm/min, v f,max = 250 mm/min, a min = 0.05 mm and a max = 0.95 mm, based on previous studies [20,50,52] and in line with suggestions by tool producers. Cutting speed was fixed to 160 m/min. Table 2 shows an experimental matrix table for wobble milling. The centre point was repeated n = 5 times to satisfy the necessary conditions of statistical calculations. This method was implemented and repeated for each tool type (A, B and C).
The cutting speed of conventional drilling and helical milling was fixed to 160 m/min. The feed rates of (i) conventional drilling, (ii) of the pilot hole drilling before executing helical milling and (iii) of the pilot hole drilling during wobble milling were v f = 700 mm/min. The feed rates concerning the    Table 3. Each experimental setup was repeated three times due to reasons of statistical analysis.

Digital image processing
In computer science, digital image processing (DIP) consists in the use of computer algorithms for performing image processing on digital images [90]. Many researchers [23,25,26,28,29,47] characterized the quality of machined holes in CFRPs by using parameters calculated using different DIP methods. Yet, some of these parameters were not specific (diameter-independent) values. In this research, a diameterspecific parameter was used to characterize the quality of drilled holes. In the scope of the experiment, digital images were taken by a Dino-Lite microscope and then were prepared for analysis using the following method: (i) The master lightness bar was decreased by 39 points, and the saturation was increased by 70 points in order to increase the contrast, as can be seen in Fig. 9b .
(ii) Red, cyan, blue and purple bars were decreased by 200%; yellow and green bars were increased by 300% in order to segment the image (transform colours to grey pixels), as can be seen in Fig. 9c. (iii) Brightness was increased by 50 points, and contrast was increased by 100 points in order to filter unnecessary pixels (see Fig. 9d. (iv) Contrast was increased by 100 points to increase more accurate edge detection (Fig. 9e). (v) The image was further filtered by medium noise reduction using 2-pixel radius adjustments. As a final result, the images were nearly assessable (Fig. 9f). (vi) Ultimately, each image was checked and the pixel errors were fixed manually.
The above-described algorithm was programmed in Adobe Photoshop CS6 Extended graphical software in order to analyse the flexible, precise and high-level reproducibility of images.
Further image processing was executed with the help of Wolfram Mathematica software, as can be seen in Fig. 10. The algorithm used for this purpose was as follows: (i) With the help of segmentation, the formerly prepared images were converted to include only black and white pixels. (ii) The number of white pixels was counted in order to calculate the area of the hole (A hole ). (iii) Finally, the smallest covering disk was defined as follows: the number of pixels constituting the resulting disk characterizes the disk area (A disk ). The area of burr (A burr ) was calculated as the difference between A hole   (3) and is applied to characterize the quality of machined holes with respect to the amount of uncut fibres.
With this novel calculation method, each hole was examined in relation to themselves. This method was necessary so that changes in the zoom ratio would not affect evaluation.
Due to the microscope setup and image processing errors, burr-free holes showed a constant error of about 1-3%, especially during wobble milling.

Results and discussion
In this section, experimental results are analysed and discussed. First, the impact of tool geometry and drilling  Second, the impact of process parameters of wobble milling on machining-induced damage and errors is analysed. Finally, the results of the above-mentioned two analyses are discussed, and future trends in advanced hole machining technologies are described.

Impacts caused by tool geometry and drilling strategies
Holes of nominal diameter d n = 10 mm were machined using three different cutting tools and three different technologies. The deviation between the nominal diameter and the measured diameter by the CMM can be seen in Fig. 11 and is expressed in Eq. (4): where d n is the nominal diameter of holes, d m is the measured diameter of holes and n is the number of experimental runs (n = 13 in the case of wobble milling, n = 3 in the case of conventional drilling and helical milling). As expected, our results show that in almost all cases, the diameters of machined holes are bigger than the nominal diameter of holes. ANOVA results show that the type of cutting tool has a significant effect (in this study ANOVA is used at a 0.05 significance level) on D d , as can be seen in Fig. 14b. Tool C produced the biggest diameter of holes compared with tools A and B. On the other hand, no significant difference has been found as far as the effect of different drilling technologies on the diameter of holes is concerned, as can be seen in the main effects plot in Fig. 14a. The impacts of the type of cutting tool and of drilling strategies on the circularity errors (CE) of machined holes can be seen in Fig. 12. ANOVA results show that drilling strategies have a significant effect on CE (Fig. 14c): (i) in the case of wobble milling and conventional drilling technologies, circularity errors were not larger than CE = 0.07 mm; however, (ii) in the case of helical milling strategy, the value of CE = 0.1-0.17 mm was considerably high. A possible reason for the high CE value associated with helical milling can be found in the accuracy of the CNC controller's interpolation. Conventional drilling tool paths do not contain horizontal movements; circularity is possibly therefore lower than in the case of the other technologies under scrutiny. Furthermore, no significant difference was found in the effects of different drilling tools on CE, as illustrated in Fig. 14d.
With the help of the application of the burr factor (B) introduced above, the characteristics of uncut fibres were analysed. The impacts of the type of cutting tool and the effects of drilling strategies on the burr factor of machined holes can be seen in Fig. 13 as well as in Fig. 14e and f. It is clear from the results that the burr factor of holes machined by way of the application of wobble milling is minimal, compared with the other analysed technologies. The impacts of the type of cutting tool on burr factor were found to be negligible in the case of wobble and helical milling. However, in the case of conventional drilling strategy, the type of the applied cutting tool has a significant impact on burr factor. This could be explained by the "push-out" effect of the drilling tool on the last Fig. 11 Impact of drilling strategies and cutting tools on the diameter of machined holes Fig. 12 Impact of drilling strategies and cutting tools on the circularity errors of machined holes laminated layers of the composite. The burr area was expected to be larger in drilling operations than in the case of wobble or helical milling operations [50][51][52], because the axial cutting force (which bends the fibres at the exit of the hole) is higher during drilling.

Analysis of wobble milling
The impacts of analysed process parameters on the machined diameter of holes, on circularity error and on burr factor can be seen on Figs. 15, 16, and 17, respectively. Results of our experiments show that the effect of feed rate on the measured diameter of wobble-milled holes is small and the measured diameter is moderately influenced by the depth of cut. However, the interaction between feed rate and depth of cut is considerable. ANOVA results show that cutting tool type exerts the only significant impact on the measured diameter of holes (p value = 0.000). Concerning the diameter of machined holes, IT8 quality holes can be machined in CFRPs when wobble milling is used (calculated in line with the results of repeated experimental setups).
As shown in Fig. 16, feed rate does not exert a significant effect on the circularity error of machined holes, but the effects of the depth of cut are more considerable. Furthermore, the interaction is insignificant in the case of tools A and B. Nevertheless, interaction is more apparent in the case of tool C. CE of machined holes created by the compression end mill is lower than the CE associated with tool C. ANOVA results show that cutting tool type exerts the only significant impact on CE (p value = 0.000). For minimizing the circularity error, it is recommended to use either tool A or tool B with a small depth of cut and at any feed rate.
The impacts of feed rate and of cutting depth on burr factor can be seen in Fig. 17. Interaction is insignificant, and, in the analysed factor space, feed rate does not exert a significant effect on burr factor. ANOVA results show that cutting tool type exerts the only significant impact on burr factor (p value = 0.004). It is clear from the results that the amount of uncut fibres can be minimized on condition tool A or tool B is used. Tool C produces more uncut fibres than the other tools, but its associated burr factor is still small: not higher than B = 4%. Compared with other hole machining technologies, all of the analysed tools could produce better quality holes when wobble milling was used as compared to the use of conventional or orbital drilling. In order to minimize burr factor (~minimize the amount of uncut fibres), the use of tool A or tool B at small feed rates and at any depth of cut is recommended.

Discussion
In the scope of this paper, n = 57 machining experiments were carried out; thus n = 114 digital images (entry and exit side of the holes) were captured, processed and analysed. Due to the limitations of the length of this paper, a limited number of images are presented herein, and they are summarized in Fig. 19. As can be clearly seen on the figure, machining-induced macro-geometrical damage is minimal in the case of wobble milling technology. This could be achieved thanks to a proper cutting force direction setup, as shown in Fig. 1. As attested by Ahmad [3], cutting tool geometry has a significant effect on chip removal mechanisms. The cutting edges of the applied cutting tools were sharp (small cutting edge radius, positive rake and clearance angles) and exhibited good wear resistance (solid carbide tool material); the chip removal mechanisms were therefore mostly crushing-dominated. However, in the case of improper setups (e.g. first row in the figure), chip removal mechanisms become bending-dominated: fibres and matrices are bent rather than cut. Thus numerous uncut fibres and matrices are left on machined edges.
In fact, chip removal mechanism is also highly influenced by the fibre-cutting angle (angle between the direction of fibre reinforcements and the direction of cutting speed, as defined in Fig. 18) [3]. In the case of conventional drilling, fibre-cutting angle changes between 0°a nd 180°. Three-dimensional drilling can be simplified, which results in a two-dimensional cutting model [25,31,91,92], and thus the three-dimensional cutting of wobble milling could be simplified as well, as shown in Fig. 18. All of the four well-known UD-CFRP-related main chip removal mechanism types [55] appear during wobble milling, very similarly to conventional drilling. However, acting axial cutting forces are (i) smaller at the hole entry and (ii) press lower-lying laminated layers of the composite more powerfully during the wobble milling process. Accordingly, as opposed to the impact of the fibre-cutting angle, the applied technology (cutting tool path, tilting movement, etc.) exerts a more significant impact on the quality of the holes. In the future, chip removal mechanisms of wobble milling need more sophisticated experiments and numerical analyses in order to verify the statements and assumptions detailed in the present paper. Conventional drilling and helical milling technologies can easily be realized and applied in industrial environments due to their moderately difficult tool paths, which also means that their operation time is low. As for circularity error and hole diameter, these technologies are appropriate, but the amount of uncut fibres on machined edges is significant, as can be seen in Fig. 19. For minimizing the amount of uncut fibres, wobble milling technology is recommended in the industry. In this case, operation time is longer, and this technology requires more prudent programming and a five-axis machine tool or industrial robots for implementation. If operation time is a significant factor, as it usually is, further optimisation of wobble milling is required.
Future directions regarding the applicability of the present research work are as follows. (i) During machining one of the most important optimisation parameters is operation time. A future goal is to shorten cycle time during wobble milling. This can be achieved by increasing the feed rate and the depth of cut. Therefore, both of these parameters have to be analysed in a wider context and wider domains (factor space) in order to optimize wobble milling processes. (ii) In the scope of this study, maximal feed rate was limited by the rotational speed of the machining centre's rotary table. Wobble milling experiments should ideally be conducted using an industrial robot in order to eliminate such feed limitations. (iii) More advanced wobble milling algorithms should be described and examined: many tool path generation methods (spiral, cylindrical, trochoid, etc.) could be used with respect to calculating the tilt of the wobble milling cycle. (vi) The impact of wobble milling should be tested to examine the machining of hybrid composites and sandwich elements, because in that case, macro-geometrical defects could be minimized.
All in all, with the help of wobble milling, excellent quality holes can be machined: this could be a promising future technology in the aerospace industry. Furthermore, the implementation of wobble milling does not necessitate additional investments because huge CFRP parts (wings, fuselages, etc.) require industrial robots for performing associated machining tasks.

Conclusions
In the present study, hole machining technologies (conventional drilling, helical and wobble milling) were analysed and compared concerning CFRPs. According to the present study, the following conclusions can be drawn: & A wobble milling algorithm has been developed, discussed in detail and validated by numerous machining experiments. This newly developed algorithm is  holes. On the other hand, tool type has a significant impact on this value. & It has been found that the burr factor of holes machined using wobble milling is minimal in comparison with the other technologies analysed. The impact of cutting tool type on burr factor has been identified as negligible in the case of wobble and helical milling. & IT8 quality holes can be machined in CFRPs using wobble milling. & In the case of wobble milling, tool type has a significant effect on burr factor. Compared with other hole machining technologies, all of the analysed tools were able to produce better quality holes when wobble milling was used as opposed to conventional or helical milling. & In the future, additional experiments and more detailed analyses are required for optimizing wobble milling processes in order to: (i) significantly decrease operation time concurrently with an increase of feed rate, (ii) sustain and retain the good quality of machined holes and (iii) conduct numerical analyses of chip removal mechanisms during the wobble milling of UD-CFRPs.