Optimization for drilling process of metal- composite aeronautical structures

Metal-composite laminates and joints are applied in aircraft manufacturing andmaintenance (repairing) using aluminum alloys (AA) and glass fiber-reinforced polymer (GFRP). In these applications, drilling has a prominent place due to its vast application in aeronautical structures’ mechanical joints. Thus, this study presents the influence of uncoated carbide drills (85C, 86C, H10N), cutting speeds (vc = 20, 40, and 60mmin), and feed rates ( f = 0.05, 0.15, and 0.25mm rev) on delamination factor, thrust force (Ft), and burr formation in dry drilling metal-composite laminates and joints (AA2024/GFRP/AA2024). Experiments were performed, analyzed, and optimized using the Box–Behnken statistical design. Microscopic digital images for delamination evaluation, piezoelectric dynamometer for thrust force acquisition, and burr analysis were considered. The major finding was that the thrust force during drilling depends significantly on the feed rate. Another significant factor was the influence of the drill type (combined or not with feed rate). In fact, it was verified that the feed rate and the drill type were the most significant parameters on the delamination factor, while the feed rate was the most relevant on thrust force. The cutting speed did not affect significantly thrust force and delamination factor at exit ( ) FdaS . However, the combination f × vc was significant in delamination factor at entrance ( ) FdaE . Based on the optimized input parameters, they presented lower values for delamination factors ( = F 1.18 daE and = F 1.33 daS ) and thrust force ( = F 67.3 N t ). These values were obtained by drilling the metal-composite laminates with 85C-tool, 0.05mm rev feed rate, and 20mmin cutting speed. However, the burrs at the hole output of AA2024 were considered unsatisfactory for this specific condition, which implies additional investigation.

Abstract: Metal-composite laminates and joints are applied in aircraft manufacturing and maintenance (repairing) using aluminum alloys (AA) and glass fiber-reinforced polymer (GFRP). In these applications, drilling has a prominent place due to its vast application in aeronautical structures' mechanical joints. Thus, this study presents the influence of uncoated carbide drills (85C, 86C, H10N), cutting speeds (v c = 20, 40, and 60 m min −1 ), and feed rates ( f = 0.05, 0.15, and 0.25 mm rev −1 ) on delamination factor, thrust force (F t ), and burr formation in dry drilling metal-composite laminates and joints (AA2024/GFRP/AA2024). Experiments were performed, analyzed, and optimized using the Box-Behnken statistical design. Microscopic digital images for delamination evaluation, piezoelectric dynamometer for thrust force acquisition, and burr analysis were considered. The major finding was that the thrust force during drilling depends significantly on the feed rate. Another significant factor was the influence of the drill type (combined or not with feed rate). In fact, it was verified that the feed rate and the drill type were the most significant parameters on the delamination factor, while the feed rate was the most relevant on thrust force. The cutting speed did not affect significantly thrust force and delamination factor at exit ( ) F daS . However, the combination f × v c was significant in delamination factor at entrance ( ) F daE . Based on the optimized input parameters, they presented lower values for delamination factors ( =

Introduction
The advent of new processing techniques and cutting tools ensures that composite and hybrid materials have been used in several engineering fields, especially in aeronautical, aerospace, and automobile industries [1,2]. Fiberreinforced polymer matrix composites (FRPs) are widely used in the aerospace industry [1][2][3][4][5][6], with carbon (CFRP) and glass (GFRP) fibers being the most used reinforcements and, consequently, the most studied [3,[6][7][8]. As it is known, in general, the mechanical properties of CFRP are superior to those of GFRP. However, the machinability of both materials is similar [9].
Manufacturers need to drill approximately one million holes in structural components to assemble one aircraft's fuselage in the aeronautical industry. In general, aeronautical structures consist of sheets made from different materials (such as sandwich panels) that are connected by mechanical joints (using fasteners) of metal/ metal, metal/composite, and composite/composite parts [10]. The aerospace industry reaches 40 million holes per year, 80% of which are obtained manually, preceding mechanical fasteners [11]. About 60% of the hybrid structural components that are rejected by the aeronautical industry have defects directly related to the drilling process [6]. Combined with the parts' high added value, this fact makes drilling composite materials and metal-composite laminate subject to constant investigations [12].
The drilling of hybrid metal/composite stack is the main challenge due to different phenomena, such as delamination [5] and damage in the polymer matrix and fibers due to thermal gradients from metal chips [6]. Furthermore, it is challenging to select optimal cutting parameters and tools to guarantee machining quality and surface integrity for all constituents (fibers and matrix) [4,5].
Zitoune et al. [6,13] verified that the most common problems in drilling metal-composite laminate CFRP/Al were (i) damage due to the interaction between composite surface and continuous chips, which comes from machining Al; (ii) separation of composite ply from the metal ply due to high thrust forces, producing an accumulation of aluminum chips and fiber fragments at the CFRP/Al interface; and (iii) adhesion of a metallic layer along the entire cutting edge.
The CFRP delamination directly affects the quality of the assembly of hybrid structures. The degree of defects represents the drilling quality characterized by the socalled spalling at the hole exit. The feed rate effect is often higher than the spindle speed; then, it is necessary to control the thrust force when the chisel edge is in contact with the exit surface of the plate [14]. The mechanical performance of the metal-composite laminate stack decreases, especially when subjected to cyclic loads (fatigue), and can be mainly identified in the region of entry and exit of the cutting tool [1]. Damage caused by delamination in the tool exit is higher than that caused in the entrance [15,16]. The delamination in the tool entrance (peel-up) can be considered a consequence of the contact between the drill tip and upper layers of laminate. Conversely, in the tool exit (push-out), damage is caused by the compressive loadings applied by the end of the drill on the upper layers of the laminate [17]. The usage of the digital analysis is suitable to estimate damages after drilling. This kind of failure can be quantified using the delamination factor (F d ) that can be calculated through the ratio between the maximum diameter of the damaged area (D max ) and the tool diameter (D o ) as shown in equation (1) [18,19], considering the contribution of cracks in the area damaged by the process drilling holes. An adjusted delamination factor (F da ) was proposed by ref. [18], as shown in equation (2), which considers the influence of the damaged area (A d ). Besides, it used A max , which is the area of the circumference calculated through D max , and A o is the area of the nominal circumference calculated using the drill diameter (D o ).
Different authors concluded that the feed rate is the most relevant input parameter on delamination factor [16,[21][22][23][24][25][26][27][28] and thrust force [20][21][22]24,[26][27][28][29] during the drilling GFRP. Others have also verified the influence of the drill. Srinivasan et al. [25] showed that the drill diameter strongly influences the delamination results. Gaitonde et al. [27] showed that a lower point angle combined with the feed rate could reduce the damage. Batista et al. [26] confirmed that the tool geometry affects delamination and thrust force. Rubio et al. [29] also observed that the drill geometry and drill wear influence delamination and thrust force. Davim and Reis [30] identified that carbide is a better choice than HSS for drilling CFRP and null wear in the flank surface, while the HSS drill presents wear of 0.012 mm. Abrão et al. [31] verified the highest thrust force values for drills with three cutting edges. Rubio et al. [32] identified high-speed machining (HSM) contribution for drilling GFRP to ensure low damage levels.
Considering the process of generating through-holes in ductile metals by the drilling process with a twist drill, burrs may occur at the holes' exit. Furthermore, drills with smaller point angles generate higher forces and plastic deformation in the holes' central part, generating crown burrs. However, larger point angles can distribute better the forces at the periphery of the holes, causing a complete or partial fracture and generating smooth burrs, which are more easily removed. If a partial fracture takes place, a hat burr occurs [33]. Moreover, Pinto [34] found that drills with smaller helix angles generate the best results when thin sheets of AA2024 are drilled, mainly at the holes' exit, decreasing the subsequent deburring steps.
The deburring step, which is often required after the manufacturing process, is not usually done in an automated way but manual. Thus, it implies the qualified workforce, specific equipment, and process time increment, causing increased production costs [35][36][37].
Considering the scenario pointed earlier, the present work's contribution consists of finding the optimal combination for values of different parameters involved in the drilling process, such as the drill type, the cutting speed, and the feed rate to minimize the delamination factor, the thrust force, and burr formation generated during drilling of a hybrid composite laminate, AA2024/GFRP/AA2024 stack, applied in aeronautical structures. Thus, the Box-Behnken design (BBD) was performed for experimental analysis and input parameters optimization. Finally, based on the results, some relevant conclusions and recommendations are highlighted for the drilling of hybrid composite laminate joined by fasteners and used in aeronautical structures.

Hybrid composite stack
The metal-composite laminate used in the tests was obtained by stacking two AA2024 (Al) plates with 1.0 mm thickness and one plate of glass fiber-reinforced polymer (GFRP) with 5.0 mm thickness, forming a sandwich Al/GFRP/Al. Table 1 presents the chemical composition of AA2024 obtained through optical emission spectrometer brand Bruker model Q2ION and standardized by ASTM B209. The GFRP plate is made of glass fibers (reinforcement) and epoxy resin (matrix). In general, fibers used for lamination are found as blankets or fabrics [38]. In this study, the reinforcements are composed of type-E fiberglass fabric. The GFRP plate used 20 unidirectional layers of glass fibers, producing a composite plate with a 50% volume fraction of fibers.

Cutting tools
Sandvik ® uncoated-carbide drills with 6.0 mm diameter and cross edging were used. These are recommended for drilling CFRP, GFRP, and stacking of plates with aluminum and titanium. Table 2 presents the drill tools specifications. The drill was mounted on a 6 mm diameter collet holder in a BT-30 morse taper (DIN 6499), both manufactured by Sandvik ® . The run-out (δ) was verified for each tool assembly change with a Digimess dial gauge (resolution of 0.01 mm).

Experimental procedure
The experiment was carried out in a three-axis vertical machining center Romi ® Discovery 308 (maximum power of 5.5 kW and maximum rotation of 4,000 rpm).
The input parameters were analyzed using three levels and combined according to the BBD, which is a method of statistical optimization that aims to change variables at three levels (−1, 0, +1) to develop a response surface. The design of experiments (DOE) consists of a combination of factor analysis and incomplete block designs [39] to reach the best levels according to the factors that influence a specific process. When threefactor, three-level BBD is used, 12 runs are required with combinations of the proposed variables, including three replicates at the center point (level 0), i.e., 15 runs. These replicates estimate the experimental variance and check the loss of linearity between the levels chosen for each variable. BBD also allows for optimization. It is based on the statistical analysis of the parameter's influence and the nature of the result [40]. Table 3 presents the parameters used and their levels.
The experimental design is presented in Table 4. From the definition of three control variables (input parameters), it is possible to identify four response variables (output parameters): (i) adjusted delamination factor at the tool entrance of the GFRP plate (F daE ), (ii) adjusted delamination factor at the tool exit of the GFRP plate (F daS ); (iii) thrust force in the cutting operation (F t ); and (iv) burr type generated at the tool exit of the AA2024 plate (Burr).
The thrust force (F t ) were measured by a Monitor System, which is composed of a Kistler ® 9272 stationary piezoelectric dynamometer four-component, a Kistler ® 5070A charge amplifier, a DAQ board, and a dedicated PC with LabVIEW™ 9.0. The acquisition rate was 1.0 kS/s. Sixteen holes were made in the metal-composite laminate for fixing ( Figure 1a) the plate as shown in Figure 1b.
Pretests (preliminary tests) were also carried out. The positions of the holes have been optimized for maximum usage of the plates, where the minimum distance between centers was equal to 2.5 D to measure the F t over the 314 cm 2 circular area of the dynamometer surface.
The hole images in detail were obtained using a digital microscope Dino-Lite AM-413ZT with 30x magnification after the experiment (Figure 2a). The Image-J  software was used for the analysis of the holes. The treatment of tones, contrast, brightness, and formatting with a binary image filter was performed to evaluate the variables used to calculate the adjusted delamination factor (F da ). The segmented image (Figure 2b) was used to determine the delamination area (A d ) and the respective diameter. F d and F da values were correspondingly calculated according to equations (1) and (2). The macroscopic analysis of the burrs was performed using a portable digital microscope Dino-Lite Pro model AM-413ZT with 50x amplification used at the hole exit. Then, the types of burrs formed during the drilling process were determined. After assessing the delamination defects associated with the drilled holes, BBD methodology, which is implemented in the Minitab ® 19 software, was used to analyze the influence of each input delamination parameter (F daE and F daS ) and thrust force (F t ), considering a 95% confidence interval (p-value ≤0.05). Besides, an evaluation of the burr type on drilling AA2024 was performed via the comparative visual analysis. Table 4 presents the control factors and response variables, and Figure 3 depicts the results for the response variables as a function of the control factors for the drilled holes.

Results and discussion
The ANOVA of the response variables is presented in Table 5. Besides, for a statistical investigation through the response surface, Figure 4 presents the 3D surface plots according to the analyzed variables.

Delamination factor
The adjusted delamination factor was calculated by using image processing techniques. In Figure 3, the error bars represent the most relevant measurement variation concerning the average value. This figure shows a comparison between F daE values generated at the tool entrance  (peel-up) with F daS values produced at the tool exit (pushout). It was noted that the second one always shows higher values than the first one [17]. The increase in F t , F daE , and F daS agrees with the increase in the feed rate ( f). However, the same behavior is not observed compared to the cutting speed (v c ), which has less influence on the response variables, as presented in Table 5. These results are similar to those obtained by Silva [16] and Malacarne et al. [21], who showed significant differences with different advances and tip angles. Gaitonde et al. [27] observed a similar behavior (reduction of delamination defects with the decrement of feed and cutting speed), but they considered the tip angle as a significant factor. Therefore, it is relevant to consider the effect of geometry on the observed results, as well.
The highest F daE values were observed when using H10N drill combined with the highest f (0.25 mm rev −1 ), as presented in Table 5 and highlighted in Figure 4, demonstrating its significance in the experiment. When the highest f value was applied, the highest F daS and F t values were also observed. Bold values indicate that the influence of controllable input factors on the output response variables is significant for 95% confidence interval ( p-value equal and lower than 0.05).
These results have shown the strong influence of the feed rate on the response variables. It should also be noted that the errors associated with the output variables can be reduced by increasing the number of runs. Table 5 presents that the feed rate ( f), the drill type (linear and quadratic effects), and combinations of the feed rate with the cutting speed ( f × v c ) and the drill type ( f × drill) have    Table 4. High friction during drilling (contact between the drill and the hole wall) promotes high temperatures that increase when increasing v c . Moreover, high temperatures reduce the composite strength, which shows polymer matrix failures, as already observed in the previous studies [21,24,29]. In some cases, as reported by Rubio et al. [32], when high cutting speeds (HSC) were evaluated, there was a highlight for improving the quality of the hole. This condition demonstrates the change in the removed material's geometry because the process conditions promote less pulling of fibers and induce the matrix's softening, facilitating its removal. Considering the adjusted delamination factor at the tool exit (F daS ) due to the push-out of GFRP (Table 5), the feed rate is also a relevant parameter (49.4%), followed by the drill type with a contribution of 18.6% (linear and quadratic effects). The R 2 was 71.1%. The surface graphs (Figure 4) for the drill 85C (0) (runs 2 and 8) show that F daS is lower than 1.40, considering v c = 20 m min −1 and f = 0.05 mm rev −1 , which is presented in Table 4. The results show the degree of relevance regarding the drill characteristics, such as the point angle, helix angle, and core thickness. Thus, it is possible to highlight that tool variables can significantly influence the composite's drilling process, as also mentioned by other researchers [16,29,41], showing the relevance of monitoring geometric characteristics. Indeed, these parameters change the cutting characteristics such as shear area, the facility of chip removal, and the tendency to increase the thrust force in the center of the tool.

Thrust force
Concerning the analysis of the thrust force (F t ) ( Table 5), the feed rate was the most significant parameter (56.5%), as mentioned by Zitoune et al. [6]. The drill type (quadratic effect) and its combination with the feed rate (f × drill) were partially significant (4.25 and 4.55%, respectively). The R 2 was 74.3%. The surface plots (Figure 4) show a decreasing trend in F t with decreasing f and increasing v c for 85C drill type. This fact can be related to the material removed rate (MRR). Moreover, the 85C drill contributes to the lowest F t values since it has modified geometry on the tool tip, resulting in less damage to the analyzed materials.
The 85C drill, compared to the others, has the smallest point angle (118°), which reduces the contact area of the tool-tip with the material. Likewise, its smallest core thickness (1.45 mm) combined with the smallest chisel edge (0.073 mm) contributes to the centralizing effect, reducing vibrations and resulting in less surface roughness, as shown by Astakhov [41]. Astakhov studied the roughness analysis to provide specifications of the surface quality level associated with the process qualification level. Borba et al. [48] showed that when cross sharpening is used, small chisel edge values (low effect of negative rake angle in the center of the drill) generate low F t values. The same effect was observed by Gaitonde et al. [27] who showed the contribution of the low feed rate combined with the smallest tip angle, resulting in lower thrust force. This effect is associated with a smaller shear area promoted by the smaller angle.
According to Kumar and Sing [3] and Wang et al. [42], thrust force (F t ) and delamination (F d ) are tightly linked since an increase in the feed rate (f) promotes an increase in F t , inducing an increase in F d . Therefore, minimizing F t reduces F d . According to Table 4, when run 2 was generated, the lowest values were observed for F t (82 N), F daE (1.07), and F daS (1.26).

Burrs
The burrs generated at the exit of the AA2024 plate were analyzed macroscopically, and two specific types were identified: crown (CRW) and uniform (UNF) ( Figure 5).
In the visual analysis, the uniform burr with drill cap (UDC) was considered a variation of the uniform burr (UNF) because they are caused by the same mechanism [43]. The hat type was not observed. As they are easier to remove, uniform burrs are preferable [33]. The UNF conditions and the absence of significant burrs were observed in 53.3% of the total holes (i.e., 1, 2, 5, 6, 8, 10, 11, 12, 13, 14, 15, 16, 20, 21, 25, and 28). In other holes, there was an undesirable CRW (46.7% of the total). The trend of better results occurred using the 86C drill with v c ≤ 40 m min −1 and f ≤ 0.15 mm rev −1 . According to Pinto [34], smaller feed rate (f) values generate lower thrust force (F t ), and a higher point angle (σ) tends to produce a distributed circular load of the holes, promoting total or partial burr fracture. This scenario can be associated with the 86C drill, which has these geometric features.
For some sets of holes (i.e., 1, 3, 5, 12, 17, 21, 22, 24, 26, and 28), differences were observed between the burr types despite having the same cutting parameters. These demonstrate the need for additional characterization of analyzes, such as checking the burr height, microstructural analysis, and hardness analysis to quantify and classify differences in their formation. According to Ko and Lee [33] and Pinto [34], in some situations, the cutting parameters can be near to promote the transition point among the type of burrcrown and uniformstaying within the margin of error of the results. For some instances, these differences may be associated with other factors, such as holes 1 and 3, which presented close thrust force values (F t ) equal to 103 and 138 N, respectively. These results may be related to the vicinity of the fracture point for this condition. Therefore, as pointed by Dornfeld et al. [44] and Manjunatha [45], even for the same type of burrs (whether UNF or CRW), burr's height, slope, shape, and mechanical characteristics need further analysis.
The holes 9, 29, and 30 resulted in similar F daE and F daS values and presented the same burr type (CRW). These characteristics can be attributed to the 85C tool geometry and, mostly, to the point angle (118°), as observed by Srinivasan et al. [25] and Durão et al. [46] who associated the burr with the geometric features of the tool. However, run 9 presented higher thrust force (369 N) in comparison to runs 29 and 30 (217 N and 240 N, respectively). Therefore, the only modification observed was the run-out (δ) of the drill 85C (run 9, δ = 0.02 mm; runs 29 and 30, δ = 0.01 mm), which may have caused an increase in F t value due to the increase of vibrations in the cutting process. According to Tsao and Hocheng [47], the vibration can even interfere in the burrs formation, and it can be related to run-out deviations. Those deviations can generate differences in roughness, but it was not investigated in the present study.
Although holes 10, 11, and 25 had been made using the same cutting conditions as holes 9, 29, and 30, and obtained similar F t values, the burr shapes were different. In other words, UNF was observed in the first one and CRW was observed in the second one, which is usually associated with the drill point angle and its sharpening [33,43]. However, the same drill type was used (85C), and any kind of failure (tool wear or tool fracture) was not detected. Thus, complementary analyses are necessary to identify and associate the burr parameters (hardness and height) because these can be set in the transition zone between the burr types.

Parameter optimization
The multivariate optimization of control variables was performed through BBD based on the response variables to minimize the delamination (tool entrance and tool exit) and thrust force simultaneously. Thus, optimized parameters were obtained such as drilling with the 85C drill applying v c = 20 m min −1 and f = 0.05 mm rev −1 , which will give the following estimated responses: F daE = 1.15, F daS = 1.36, and F t = 90.6 N.
A validation test was performed using the optimized parameters obtained via BBD. Then, three holes were drilled and evaluated to check if those parameters reach the estimated responses mentioned earlier. Table 6 presents the average of the experimental results. The experimental values of F daE and F daS were similar to those predicted values (with an error lower than 3%). The peel-up and push-out phenomena were small in the fibers and polymer matrix. Conversely, the experimental thrust force was about 25.7% lower than the estimated F t . The low parameters used (v c , f) and drill point angle 85C (σ = 118°) were relevant for good results. Similar results and observations were highlighted previously by Srinivasan et al. [25].
Although the burr type was not analyzed via BBD, the optimized condition promoted a CRW type, which is not recommended for the drilling process due to the need for rework to remove this element. Figure 6 shows the images of the burrs obtained in the validation test.
As commented in the literature [35,36], the UNF burrs, with less complicated removal, are associated with a distributed circular load of the holes, unlike CRW burrs, more irregular and laborious removal. The CRW type, predominant in optimization tests, is related to smaller point angles, as shown in the 85C drill (118°), and the consequent deformation in the hole center region. Silva et al. [37] recommend avoiding the deburring step Drilling conditions: v c = 20 m min −1 , f = 0.05 mm rev −1 , and drill 85C. because this rework leads to additional manufacturing costs, making the process unfeasible. The generation of uniform burrs with short height should be the priority goal. In most cases, this rework is carried out manually without adding value to the product.

Conclusion
It was evaluated the influence of cutting conditions over the delamination factor and thrust force of metal-composite laminate. The present work investigated a sandwich plate composite by AA2024/GFRP/AA2024 joint. According to the results for this specific case study, it is possible to have some relevant conclusions and, mainly, recommendations for drilling metal-composite laminates used in aeronautical joints: • The thrust force (F t ) during drilling depends significantly on the feed rate (f). Another partially significant factor is the drill type (combined or not with f). • f and drill are control variables that influence the adjusted delamination factor significantly at the input (F daE ) and output (F daS ) of the GFRP. • The cutting speed (v c ) does not affect significantly F t and F daS . However, the combination f × v c is significant in F daE . • The optimization of the cutting parameters via BBD can obtain values close to those estimated for the response variables (F daE , F daS , and F t ). The optimized levels were v c = 20 m min −1 , f = 0.05 mm rev −1 , and drill type 85C for the investigated metal-composite laminate. • In case of the burrs at the hole output of AA2024 using the optimized parameters, the results are unsatisfactory (crown type), which requires further investigations on the optimal hybrid stack drilling.