Optimization of an Autonomous robotic drilling system for the Machining of Aluminum Aerospace Alloys

This paper aims to identify the capability of a highly exible industrial robot modied with a high-speed machine spindle for drilling of Aluminum 6061-T6. With a focus on drilling feedrate, spindle speed and pecking cycle, the hole surface roughness and exit burr heights were investigated using the Taguchi design methodology. A state of the art condition monitoring system was used to identify the vibrations experienced during drilling operation and to establish which robot pose had increased stiffness, and thus the optimum workspace for drilling. When benchmarked against a CNC machine the results show that the CNC was capable of producing the best surface nish and the lowest burr heights. However, the robot system matched and outperformed the CNC in several experiments and there is much scope for further optimization of the process. Overall the proposed drilling system is far more exible than a CNC milling machine and when considering the optimized drilling of aerospace aluminum this robotic solution has the potential to drastically improve productivity.


Introduction
The aerospace manufacturing industry is always looking for ways to reduce cost, improve quality and reduce lead time. Drilling within the aerospace sector has always been very important for the manufacture of airplanes and is the most commonly performed task in machining aerospace components. The Boeing 747 contains roughly 3 million fasteners with roughly 40,000 rivets on each wing of a Boeing [1]. With Boeing drilling roughly 1.1 million holes per day [2]. In a bid to improve e ciency, the implementation of automation within the automotive sector has increased from 20% to 80% [3]. Despite the similarities between aerospace and automotive the drive for automation has not been witnessed within the aerospace sector despite the bene ts that could be gained [4].
Airbus and Boeing released their global market forecast for 2018 -2037 in July 2018. The reports show a predicted growth of 3.5% of new airplane deliveries per annum [5]. In addition, during the same time period there is expected to be a 4.4% annual growth in air tra c [6]. Additional sources of forecasts from Airbus and Boeing highlight a continuing demand for civil aircraft between 2015 and 2034 whereby aircraft volumes will increase from 19000 to 38500 [7]. This data demonstrates the expected growth that will be witnessed within the aerospace sector moving forward. To meet the growing demand for aircraft manufacture automation has the potential to increase both productivity and component quality within the aerospace sector.

Drilling operations
Within the aerospace sector assembly work is commonly highly physical and can be monotonous. Typically, the assembly of an aircraft fuselage or wing consists of skin sheets being mounted onto supporting structures that are subsequently connected through riveting [8]. The manual riveting operation requires a pre-drilling drilling operation which can cause several potential health risks related to fatigue, dust, noise and vibration [9]. In addition, the sector traditionally has inherent high labour turnovers, training costs and recruitment di culties due to the nature of the machining and assembly procedures used [10]. Duquette et al identi ed that at least 30% of workers found that the tools used to assemble aircrafts introduced back pain after prolonged use [10]. Also, a study by Menegon et al on 522 aircraft assembly workers reported that pain, discomfort and numbness were experienced in body regions such as neck, shoulders and knees which correlated to employees seek leave of absence from work [11]. In addition, within the sector the quality of drilling operations suffers from inconsistency. Typically, quality is higher at the beginning of a shift when compared to the work produced at the end of a shift. This variation is quality can be attributed to operator fatigue and human constraints which highlights the urgent requirement to introduce more automation into labour intensive activities. One challenge for automation is the quality of the drilled holes. The precision of the connecting holes determines the riveting quality. Any introduction of imprecision will result in bending stress being introduced at the joints [12]. This can be detrimental to the fatigue life of the aircraft with 80% of structural failures being attributed due to fatigue damages from the connecting holes of the aircraft structures [13,14]. Also, tight hole tolerances are required therefore any manual drilling with inherent poor hole quality consistency will require subsequent rework [15]. These errors are a concession to the original speci cation therefore increase the manufacturing costs to rectify [16].
Vibrations during the drilling process often lead to chatter which in turn causes; poor hole quality, increased noise levels and potential damage to the component [17]. Tao et al. identi ed the importance of identifying chatter as early as possible to improve hole quality [18]. When considering different feed rates (F r ) (0.9 -7.8 mm/second) and the spindle speeds (S s ) (1800 -3600 RPM) the research showed that increasing the S s resulted in decreased stability due to rapid increase of the amplitude of vibrations, leading to higher levels of chatter. Recent evidence suggests that provided the correct machining parameters are identi ed and used, increased chatter levels can be avoided [19][20][21]. These studies go on to state that at lower S s rates the chatter levels can be reduced due to the stabilising effect of the process damping. The drilling parameters also affect the surface roughness (R a ) [22]. Zhang showed that the application of different types of drills and the setup of the different drilling parameters can have a major effect on the resulting R a of the hole [23]. Chen identi ed that F r is an important machining parameter which in uences both the accuracy and surface quality of drilled holes and that an adaptive material removal rate increase machining accuracy [24].
In an investigation into the application of industrial robotics for the assembly of commercial aircraft components, Durham found that hole quality can be controlled by adjusting the F r , S s and pecking cycle (P c ). Huang et al. provided an optimization analysis of the drilling parameters whilst dry machining Aluminium 6061 alloy. Using a Taguchi's signal / noise ratio and grey relational analysis method the investigation concluded f that the optimal drilling parameters for the desired R a is a S s of 3000 RPM and F r of 0.2 mm/rev. Whilst the optimal drilling parameters for the desired burr height, is a S s of 2500 RPM and F r of 0.15 mm/rev [25].

Hole burrs
Hole burrs are an undesired by-product produced during the drilling process. They consist of plastically deformed material at the entry and exit peripheries. Typically, a burr is an accumulation of plastically deformed work material and if left unremoved, burrs can impact the overall quality and life of a product [26,27]. In 2004, Kishawy et al demonstrated that hole burrs can have signi cant effects on the fatigue life of Aluminium aerospace alloys. The study highlighted that test pieces with burrs were found to have signi cantly reduced fatigue life by up to two orders compared to non-burred specimens [28]. Deburring is classed as a non-value-added manufacturing process that can require a signi cant amount of production time especially on large structures [29]. Studies on the in uencing factors of burr formation show that the burr height can be signi cantly in uenced by the workpiece material, tool geometry and wear and machining parameters [30,31]. L Jie observes that burrs cannot be completely prevented by changes in feed, speed or tool geometry in isolation. The research concluded that burr height can be minimized signi cantly by selecting the appropriate drilling parameters [32]. In a study by Lee the in uence of machine parameters on the resulting thrust force (F t ) in drilling was investigated [33]. The research demonstrates that the resulting thrust force determines the amount of material that undergoes plastic deformation during the nal stage of drilling [33]. Further studies on burr formation have concluded that burr height increases as a function of drill wear. Also burr height can be reduced through dry machining [34].

One Way Assembly (OWA)
A signi cant trend in aerospace manufacture is OWA, whereby a drill goes through multiple layers and sometimes hybrid stacks of materials. For the burrs produced during OWA the drilling target is to produce an interlayer burr height of 100 μm or less [35]. For quality and safety reasons it is imperative that aerospace structures are manufactured without excessive burrs or swarf within the structure. In standard aerospace body assembly, operational compliance requires the cleaning of interlayer interfaces prior to sealant applications [36]. However, when applying OWA machining strategy, disassembly and internal deburring and interface cleaning is not possible. Several studies had identi ed the challenges in drilling hybrid stacks [37][38][39]. In particular, Pardo et al. found that an increasing interlayer gap width resulted in greater abrasion of the interfacing burrs on an aluminium-aluminium stack, along with increasing scratch marks around the hole bore. This was due to the aluminium swarf chips accumulating in the interface gap around the hole bore, then subsequently being rotated by the drill [40]. Lazar et al. found that delamination in a bre-reinforced stack primarily occurs at two locations: tool entry and tool exit [41]. For a carbon ber reinforced polymer (CFRP) and aluminium stack, the work of Pardo et al and Lazar et al identify that burr abrasion during drilling could in uence the interlayer gap and facilitate a path for contamination to enter the interface between the materials within the stack [40,41].

Automation of drilling
Drilling in aerospace is predominantly done using Advanced Drilling Units, (ADUs) with some manufacturers moving to Light Electric Drilling Units (LEDUs). Durham investigated the time savings obtainable through the integration of automation within the assembly and manufacture of aerospace components and identi es the potential economic savings of automation [42]. Furthermore, Durham demonstrated that even a 1 second reduction in takt time per hole on the 8000 exterior holes required in a fuselage would save 13 hours of operation [42]. Historically, the reluctance to use automation within the aerospace sector has been attributed to the poor accuracy of early generation robots. Poor arm and joint stiffness lead to higher levels of vibrations. Traditional CNC machines have a stiffness of greater than 50 N/µm whereas a typical articulated robot usually has less than 1 N/µm. Zhang et al. state that the difference in stiffness between the two machines is a major hurdle preventing the adoption of robots within industry [17]. In recent years, there have been vast improvements in manufacturing robots and speci cally the 6-axis manipulator type robot resulting in higher accuracy and repeatability [43,44]. They have become increasingly stiffer which correlate to higher accuracy machining when compared to older generations of robots. This has permitted machine spindles to be used as end effectors and thus allowing for the possibility of highly exible automated drilling operations. However, manufacturing robots have some drawbacks, the "Fuselage Automated Upright Build" process (FUAB) was abandoned by Boeing in 2019 as the robots proved painful to set up and were error-prone resulting in fuselages having to be nished by hand [45]. Despite the challenges research has demonstrated the bene ts of robotics utilization. In particular, Yin et al concluded in their study that improving the KUKA KR500-2's stiffness lead to better accuracy of countersink drilling depth [46]. Also, by reducing vibrations during the automated drilling process tool wear and tool breakage is decreased. In addition, dimensional quality is improved and surface roughness is reduced [47,48]. This paper will investigate the utilization of a state-of-the-art Kuka KR-16 manipulator with a high-speed spindle and drill end effector for the machining of aerospace AA601 aluminum alloy. The study will utilize an aerospace compliance standard for hole accuracy and burr height drilling operations for dry machining applications [49]. The main aims are to optimize dry machining parameters through the use of the Taguchi design of experiment (DOE). The resulting machining responses of; surface roughness, burr height and concentricity of the holes produced will be empirically compared to the results obtained from a conventional 3 axis CNC milling machine. The research will then investigate the Kuka machining setup.
In particular, the resulting vibration as a function of machining position. This will be followed by a DOE to establish the machining parameters that have the largest effect on the quality of holes. Section 3 will analyse the results recorded for these experiments and benchmark the robotic machining against the CNC system. Finally, conclusions will be made on the potential of this method of exible automation for machining aerospace components.

Drilling parameters
The holes to be analyzed for this research are 8 mm diameter and 8 mm deep, drilling parameters for these dimensions have been identi es. The F r range used is between 200 -600 mm/minute, S s of 1800 -4000 RPM and the number of passes of a pecking cycle (P c ) of 2-6 mm at a time. P c is de ned as the machining method that is used in the drilling processes to prevent swarf from building up whilst drilling holes. The pecking depth improves tool life but each additional cycle increases the drilling takt time. In this research a robotic based P c that considers different pecking depths will be used to gain an insight into whether or not the surface roughness of the holes is in uenced. The robot and CNC canned pecking cycle will also be compared.

Design of experiments
Within the study an L9 Taguchi array design with 3 factors and 3 levels will be used and a total of 3 experimental repeats will be undertaken (Table 1). For the drilled material, Aluminum 6061-T6 has been selected. The alloy is used widely within the aerospace sector in the skin of aircrafts due to its excellent mechanical properties of hardness, weldability, corrosion resistance and relatively low cost. The experimental parameters used to conduct the drilling experiments are detailed in Table 2. Drill chatter and vibration can have a large effect on the quality of the hole produced. Therefore, in order to try and ensure that the holes being drilled experience the least amount of chatter, vibration tests were undertaken to measure which sections of the robot workspace had the lowest vibration amplitude. To establish the vibrations of the robot system a condition monitoring system is used. Two SignalCalc Ace Vibration sensors are connected to DAQ (V-A) and SignalCalc software is used to record data from the accelerometers. The rst accelerometer was placed on the side of the machine head to measure the vibrations in the x plane (

Metrology
Pre trials identi ed that the burrs from the experiments are to be classi ed as breakouts (negative burrs).
Due to the plastic deformation of a burr, non-destructive measurements are preferred as the specimen is not further deformed during the measurement. Once the holes were drilled, the procedure used was to photographed to allow them to be examined and ranked according to the severity of the burr. The test piece is placed under the Keyence VHX-1000 digital microscope to observe the hole exit burrs. In order to generate focused images, depth composition was used image the different burr heights. After the burr measurements had been taken, the part was cut down the middle and the specimen was then put under the Veeco Wyko NT9300 white light interferometer to measure surface roughness.

Vibration Results with the Spindle Off
The rst vibration experiments were performed without the spindle switched on, the robot was positioned in each of the 99 positions and the vibration was measured. Figure 4 shows a heat map of the robot vibrations experienced within the drilling workspace. The lowest vibration experienced (0.00097 mm) is highlighted by the light blue block in (4,2), the area of highest vibration (0.00921 mm) is shown by the black block (12,6). The all white blocks are the locations within the workspace that the robot could not reach due to the pose limitations. The vibration data indicates that although the robot is static in terms of drilling there are still vibrations and dynamic responses passing through the robot. The kinematics of the robot are based on Denavit Hartenberg parameters and there are rules that the robot must follow in order to rotate coordinate frames. Thus, the position of each of the 6 axis are based on a default kinematics unless modi ed by the operator. The variation in vibration within the work space shows that the kinematic chain of 6 axis articulated robots are stiffer in certain poses, this has implications for machining operation stability in certain poses.

Vibration results with the spindle on
The Zimmer HFL04-103-01-A-00 Machine Spindle is capable of operating up to 24,000 RPM so it was decided that the maximum, minimum and average vibrations up to this RPM range would be measured. Figure 5 shows that the vibration of the entire workspace at various RPMs doesn't uctuate dramatically between 2000 -10,000 RPM. However, there is an increase of 40% between 3000 and 5000 RPM, and an increase of 655% from 10,000 RPM and 20,000 RPM. The ndings comply with the work by Tao et al. [10]. In their paper they observed that when the spindle speed was increased, the stability decreased. The potential for speeds >3000 RPM to negatively in uence the quality of the holes means that only speeds below this will be investigated further in this study.

Vibration Results with the spindle at the selected operation range
Following on from the spindle off experiments the same was conducted but with the spindle switched on and limited up to 3000 RPM. Figure 6 shows shown by the black block (8,8). The all white blocks are the locations within the workspace that the robot could not reach due to the pose limitations. The vibration data indicates that the V a observed across the three RPMs is 40 x larger than when the spindle was off, and the V m observed across the three RPMs is Page 9/19 84 x larger than when the spindle was off. This shows that the robot experiences signi cant vibrations when the spindle is turned on.

Main distribution of the exit burr heights from the CNC and Kuka systems
To compare both drilling systems nine burr height measurements were made for each of the experiments (Figure 7). The measured burr heights for all the CNC experiments show that the mean height is 0.9 mm. The minimum observed average height (0.15 mm) is observed in experiment 8 and as this experiment also has the lowest deviation (σ of 0.021), this making it the optimum CNC setting. The settings for experiment 2 produced the highest bur heights and largest deviations with 1.85 mm and σ of 1.55 mm respectively. There is a reasonable consistency in burr height and deviation for experiments 3-7.
The mean measured burr heights for the Kuka holes for all experiments is 1.07 mm, this is 16% less than the CNC average. The best setting is experiment 1 as it produces an average minimum observed burr height of 0.29 mm with a low σ of 0.021 mm. The maximum is 3.11 mm from experiment 3 which is the highest exit height of all experiments and has a wide deviation of measured heights. Except for experiment 3 there is a reasonable distribution except for the high height observed with experiment 3 and consistency in the results between 4-9.
An important observation to be made when considering both systems is that although the average burr height for the CNC system is better throughout the 9 experiments the Kuka system can produce lower exit burrs as demonstrated by experiments 1, 2, 5, 6 and 7 and 9. It can also be seen that for nearly all experiments the Kuka drilling results have a lower variation in measured burr height. This suggests that the settings used on the Kuka system produce exit burrs that are more consistent and that the burrs may not be signi cantly in uenced by the overall stiffness of the systems. Table 4 shows the main effects of the parameters on burr heights for both drill systems. Based on the results the P c has the largest in uence on both. The ranking is showed by the difference between the lowest and highest (δ) of 0.49 mm and 1.04 mm respectively. For both systems an increase in the P c increases the burr height ( gure 8 and 9). F r is ranked second and for the CNC the burr height is reduced as the F r is increased. For the Kuka the F r increase from 200 mm/minute to 300 mm/minute reduces the burr height by 0.96 mm but there is a marginal height increase from 300 mm/minute to 400 mm/minute.  The mean measured R a for all the Kuka drilled holes show that the mean roughness is 4.00 µm. The minimum observed R a is 2.13 µm from experiment 3, this also has a low distribution (δ of 0,59 µm) making it the optimum setting. The maximum is 5.68 µm from experiment 9. Experiments 2 and 4 had wide deviations making them less predictable settings for consistent surface roughness.

Main effects of the exit burr heights from the CNC and Kuka systems
There is no consistency in the experiments and both systems have a wide deviation in results, with experiments 2 and 4 for the Kuka system producing the widest deviations making them unsuitable settings for drilling with a consistent surface roughness. An important observation to be made when considering both systems is that although the average R a for the CNC system is better throughout the 9 experiments (>19.5%), the optimized Kuka system can produce a comparable surface nish (experiments 3) to the best observed R a drilled by the CNC system (experiment 4). This result is similar to the exit burr height results and suggests that R a may not be signi cantly in uenced by the overall stiffness of the systems. Table 5 and gures 11 and 12 show the main effects of the parameters on the hole surface nish for both drill systems. Based on the results for the CNC drilled holes the S s is the most in uential factor, as the speed is increased from 2000-3000 RPM the roughness of the hole increases and a δ of 1.17 µm can be expected. The P c is the second most important and the results show that the surface roughness improves with 4 passes. For the Kuka system the P c has the largest in uence with a δ of 2.1 µm observed. However, the result is not straight forward as the surface roughness is at 3.49 µm for 2 passes but deteriorate to 5.31 µm at three passes and then improves again at 4 passes with a R a of 3.2 µm. This behavior could be due to interactions with the other factors. The second most in uential factor is S s . As the S s is increased the surface roughness improves (δ of 1.22 µm), this change is the complete opposite to that observed with the CNC and could be due to increased vibrations experienced by the Kuka resulting in a lower surface nish. Optimised Solutions For The Kuka System

Factor interactions of the exit burr heights
The main effects are instructive when optimising a process however there are often interactions between the factors that can be less easy to predict. Using ANOVA it was possible to plot interactions for the exit burr heights ( gure 13). In gure 13a there is an interaction observed with low F r and P c . It can be seen that the optimum setting for low height burrs is a low F r and low P c , however if you increase the passes at the same F r there is a signi cant increase in the burr height. There is no signi cant interaction for medium and high F r .
For the relationship between the S s and P c ( gure 13b) there is no obvious interaction except for the large increase in burr height observed at the highest S s when moving for 3 passes to 4. An optimum setting would be 2 passes and a low S s . When considering the relationship between the F r and S s ( gure 13c) the medium and high F r have no signi cant relationships with S s . However, there is a signi cant interaction with the low F r and S s , where the exit burr height increased with the change from a 2000 to 3000 RPM. The setting of a low F r and S s appear optimum for producing low height burrs.
There are three factor interactions that result in height burr heights, a F r of 200 and 4 passes, 3000 RPM and 4 passes and 3000 RPM and a F r of 200. The each of these there is evidence for increased energy at the hole exit and a change in exit velocity. To reduce the drill exit force this there is the option of a nal micro peck cycle. This optimized solution will have very little in uence on the overall cycle time and the Kuka robot has a high capability to use this function.

Factor interactions for surface roughness
In gure 12 the main effect of P c shows that three passes is detrimental to the surface nish. The same result is shown in the interaction plot for S s and P c ( gure 14a). The best surface nish can be achieved with a low F r and 4 passes followed by a medium federate and 2 passes. It is di cult to ascertain the reason for the in uence of the P c , and a possible cause could be that three passes produces step effect in the hole. For optimized machine setting it is clear that additional passes should be avoided for particularly with a high F r setting. Figure 14b shows the relationship between the S s and P c . This also shows that 3 passes is detrimental to surface nish particularly with a high spindle speed. However, the high S s setting does produce the better surface quality with 2 passes and 4 passes. Figure 14c shows the best surface nish can be achieved with a high S s and low-medium F r . There is an interaction with the high S s and F r in that it does not follow this trend and actually results in the producing the worst surface nish. In addition, a low S s and medium F r provide a poor surface nish. The low S s is shown to produce poor surface quality for all F r settings, and the high S s can result in the highest quality for low to medium F r setting. The anomaly of a high S s and F r producing the worst surface nish could be due to the F r exceeding the cutting capability and providing an increase of force at the cutting edges of the drill.

Predicted results
The best quality hole for the purpose of this research is de ned as lowest surface roughness and lowest exit burr height. An analysis was conducted on the Taguchi L9 array which included all responses from the experimental data. The focus of the research was also to observe if the Kuka drilling system could be optimized in a way that the hole quality could be comparable with that of a CNC machine.
The results for both systems show that there is no single experiment that provides a low surface roughness and low exit burr height and that for both systems post processing will be required for the holes to meet aerospace drilling speci cations. The Kuka experimental data shows that the optimum hole providing the lowest R a value is experiment 3 (2.13 µm,) and this was with a F r of 200 mm/minute, S s of 3000 RPM and a P c with 4 passes through the work piece. Experiment 1 produced the lowest exit burr height value (0.29 mm) and this was with a F r of 200 mm/minute, S s of 2000 RPM and 2 passes through the work piece. A compromise for the exit burr height and surface roughness would be experiment number 6 where the measured heights and R a are 0.77 mm and 2.35 µm respectively.
Using the main effects plots ( gure 9 and 12) it possible to take the factor settings that provide an optimum result and predict the average response for this combination of control factor levels. The objective is to determine the settings that minimize the means of the Burr height and R a . The results in table 6 show that the burr height can see an improvement of 60.68% when compared to the measured response from experiment 1. For the surface roughness no improvement is predicted and experiment 3 remains the preferred setting. Identifying the correct robot pose for drilling is important for machine optimization, in order to do this a state of the art condition monitoring system was set up on the end effector spindle of the Kuka system to establish the vibrations experienced during operation. Within the drilling workspace 99 different pose positions were measured. The initial tests were conducted without the spindle on in order to establish which robot pose had increased stiffness, this was then followed by the same positions but at different spindle speeds. The results show that due to the kinematic chain of 6 axis articulated robots certain poses within the workspace are stiffer. After condition monitoring with different spindle speeds it was possible to identify a location within the work space with the lowest amplitude of vibrations. Once this was established the machine vice was set to this position for reduced vibration machining. The Kuka drilling system includes a Zimmer Machine Spindle that can run up to 24,000 RPM. The condition monitoring system was used to establish the machine vibrations >20,000 RPM for the 99 different robot poses. It was shown that vibrations signi cantly increased when the spindle was on and this would lead to excess drilling chatter. Vibration was reduced below 3000 RPM, so the main experimental parameters were focused between 2000-3000 RPM. After monitoring all 99 poses it was possible to identify the optimum range for the drilling S s and the workspace areas with reduced vibration. All further experiments were conducted with this preprocess data.
Exit burrs due to drilling operations can have an in uence on manufacturing costs. To establish the capability of the Kuka system a design of experiments approach was used see which of the selected parameters in uenced burr height. The system was also benchmarked against a CNC drilling setup.
As expected the CNC was capable of producing the lowest burr heights. However, the Kuka matched and outperformed the CNC in 6 of the experiments and the distribution was better overall. For both systems post process deburring would be required to meet aerospace manufacture standards.
For the Kuka system the optimum settings were identi ed and the number of passes of the Pc was established as the most in uential process factor on reducing the burr height. Using ANOVA it was possible to identify interactions, and the setting of a low F r and S s can be used for producing low height burrs. Overall the results show that there is much scope for optimizing the process to control exit burr height and further work should be conducted to control the drill exit forces on the nal pecking cycle.
The second hole quality metric investigated was surface roughness. The control factors had a large effect on both the Kuka and the CNC systems. An important observation is that the CNC system produces better surface quality, but there are two settings of the Kuka that outperform the much stiffer CNC machine. For the Kuka system P c has the largest in uence followed by S s . The main effect of P c shows that additional passes is detrimental to the surface nish and the ANOVA interactions shows that this should be avoided particularly with a high F r By using a condition monitoring and a design of experiments approach it has been possible to identify a range of quality standard that can be achieved using a state of the art robotic machining system. The system is far more exible than a CNC milling machine and when benchmarked against a CNC the Kuka performs well when drilling aerospace aluminum. Optimization of selected parameter has been shown to improve the hole surface roughness and reduce exit burrs. With further research there is also potential for further improvements to be made, particularly with control of the drill exit force and control of the pecking cycles while drilling.

Declarations Acknowledgments
The authors would like to acknowledge the support of the Future Manufacturing Research Institute, College of Engineering, Swansea University and Advanced Sustainable Manufacturing Technologies (ASTUTE 2022) project, which is partly funded from the EU's European Regional Development Fund through the Welsh European Funding O ce, in enabling the research upon which this paper is based.
Further information on ASTUTE can be found at www.astutewales.com.
Funding -Not applicable as the work is not linked to direct funding.
Con icts of interest/Competing interests (include appropriate disclosures) -There are no con icts of interest.
Availability of data and material (data transparency)-Data is held and available.
Code availability (software application or custom code) -Not applicable.
Authors' contributions B Pereira -Principal researcher, conducted 60% of the tests C Gri ths -Initial proposal and conducted Minitab analysis of test results.
B Birch -Conducted experiments and drilling setup support.
A Rees -Veri cation of test results.
Additional declarations for articles in life science journals that report the results of studies involving humans and/or animals -Not applicable.
Ethics approval (include appropriate approvals or waivers) -Not applicable.
Consent to participate (include appropriate statements) -All authors consent to participate.
Consent for publication (include appropriate statements) -All authors consent to publish.