Influence of cooled and heated compressed air on the surface finish in face milling of CF30-reinforced PEEK

Abstract

Carbon fiber-reinforced poly (ether-ether-ketone) (CF/PEEK) is a widely used composite material in aerospace, automotive, and biomedical applications. However, achieving a high-quality surface finish during machining remains challenging due to its low thermal conductivity and high abrasiveness. This study investigates the thermal effects of using cooled compressed air (CCA at −40°C) and heated compressed air (HCA at 60°C) on surface finish during face milling of PEEK reinforced with 30% short carbon fiber (CF30), compared to dry machining (DRY). A Box-Behnken experimental design was applied with three levels of cutting speed (v_c), feed per tooth (f_z), and axial depth of cut (a_p). Surface profiles and roughness parameters (R_a and R_z) were statistically evaluated using ANOVA, Abbott–Firestone curves (AFC), and probability density functions (PDF). Scanning electron microscopy (SEM) and 3D interferometry were employed to analyze the optimized and best-performing samples. Results indicated that the lowest roughness values were achieved under CCA, while HCA yielded the highest. In both cases, f_z was the most significant factor. AFC and PDF analyses confirmed that surfaces machined with CCA exhibit a more homogeneous 2D roughness profile, with less pronounced peaks and valleys. The regression model accurately predicted roughness values under CCA, with an error of approximately 4%, aligning well with experimental results (R_a = 0.307 μm, R_z = 2.210 μm) and maintaining a suitable material removal rate (MRR = 1723 mm³/min). SEM images revealed that CCA enhances surface finish by increasing composite stiffness, facilitating cutting. 3D roughness profiles confirmed isotropic characteristics for all conditions. After optimization, CCA exhibited the lowest S_a (0.371 µm), DRY the lowest S_z (12.42 µm), and HCA the roughest surface profile, confirming the trends observed in 2D analysis.

Keywords

CF/PEEK face milling cooled compressed air heated compressed air surface finish process optimization

Introduction

Poly-ether-ether-ketone (PEEK) is a high-performance thermoplastic polymer, known for its excellent properties, making it an ideal choice for engineering applications that require high performance.¹ It is an excellent alternative to metals such as aluminum, stainless steel, and titanium due to its superior resistance to hydrolysis, corrosion, friction, and flame. To enhance the strength and performance of a polymer, reinforcements are often incorporated into its matrix.² Consequently, this improvement depends on the properties of the matrix and reinforcement, the reinforcement’s geometry and orientation, and the matrix-to-reinforcement ratio.³ Carbon fiber (CF) reinforcement is available in two configurations: short fibers and continuous fibers. Continuous fibers offer superior mechanical performance but result in higher material consumption and cost, while short fibers provide an intermediate solution, improving mechanical performance relative to the neat thermoplastic, although they do not achieve the efficiency of continuous fibers.⁴ Since CF and PEEK are highly compatible, this combination improves mechanical strength, frictional performance, and thermal properties. Therefore, PEEK composites are widely applied in the automotive, aerospace, chemical, and biomedical sectors.⁵ Polymeric composites offer excellent design versatility, allowing components to be manufactured close to their final shape. However, to achieve high geometric precision, adequate tolerances, and superior edge quality, certain machining operations remain necessary.⁶

Machining is an efficient process to produce mechanical parts, implant materials, and prostheses from PEEK composites, ensuring the dimensional accuracy and surface finish required for the desired performance and extended service life.^7,8 Cutting CF/PEEK composites is mainly required to produce components with specific dimensions and surface roughness to facilitate assembly processes.⁹ However, the abrasive nature of carbon fibers reduces their machinability.⁷ Additionally, the anisotropy and non-homogeneity of composites influence the cutting process, often causing delamination, fiber pull-out, and surface defects.^10,11 In CF/PEEK machining, feed rate is the main cutting parameter that affects surface roughness, followed by cutting speed and depth of cut.¹¹ Increasing the feed rate enhances heat generation and tool wear, damages surface roughness, and may also increase vibrations that are harmful to surface finish. Low cutting speeds can result in a large material flow of cut fibers that harms surface roughness, while high cutting speeds increase temperature, producing chips mainly composed of less deformed matrix material and cut fibers. Increased depth of cut may lead to high normal pressure and seizure.¹² Furthermore, surface roughness is a key quality characteristic that directly affects performance, aesthetic appearance, fatigue resistance, and degradation resistance.¹³ Thus, process optimization is necessary to ensure high-quality surfaces in machined parts.

The low thermal conductivity of polymers leads to heat concentration at the tool’s cutting edge, hindering the generation of a smooth surface. In CF/PEEK machining, high cutting speeds combined with abrasive reinforcement accelerate tool wear mechanisms and compromise surface quality.^11,14 Cutting temperatures can easily exceed the glass transition temperature (T_g) of the polymer matrix and even the degradation temperature around the cutting edge.¹⁵ When the cutting temperature exceeds the T_g, the resin loses its fiber-supporting capacity, causing surface defects and material degradation.^16–19 High-speed dry (HSD) milling studies of CF/PEEK indicate that cutting temperature is a primary factor in defects such as delamination and uncut fibers, highlighting the need to control thermal effects for surface integrity.^6,20 Despite cryogenic cooling being an eco-friendly machining strategy that can improve surface quality, reduce tool wear, mitigate burr formation, and enable higher cutting speeds and feed rates,¹⁵ excessive workpiece cooling may potentially induce undesirable microstructural changes in polymeric composites.⁵

Compressed air techniques are attractive for sustainability and cost-effectiveness, as they are cheap and readily available resources.²¹ Liu et al.¹⁶ demonstrated that compressed air helps limit temperature increases and preserve surface integrity, making HSD milling a promising strategy to improve quality and efficiency. Cooled compressed air (CCA) is an eco-friendly alternative that reduces energy consumption and supports high cutting performance.²² In HSD milling of CF/PEEK, a compressed-air jet removed up to 50% of the generated heat, maintaining the surface temperature below T_g and preserving the integrity of the polymer matrix, resulting in a good surface finish.¹⁶ Despite these benefits, Ensinger²³ recommends preheating short CF/PEEK (30/70) workpieces from 80 to 120°C before semi-finish machining, followed by finish machining after cooling at room temperature. This recommendation motivated the investigation of heated compressed air (HCA). John and Kumaran²⁴ reported that HCA is economically viable, sustainable, and beneficial in drilling unidirectional CFRP laminates, reducing thrust force and delamination at low feed rates. However, a scientific gap remains in the process optimization of CF/PEEK milling with CCA and HCA to enhance the surface quality.

To optimize the machining process and achieve the best workpiece surface finish, the Box-Behnken Design (BBD) emerges as a multi-objective method for identifying the optimal combination of input parameters for a specific process.²⁵ BBD is a second-order rotational or nearly-rotational design based on incomplete three-level factorial designs for each controllable factor, generating a response surface.²⁶ Complementarily, the Abbott-Firestone curve (AFC) is a helpful statistical tool for evaluating surface texture, providing a cumulative profile function that describes the distribution of peaks and valleys. The flatter the curve, the lower these peaks and valleys, and the more uniform the roughness profile.²⁷ Additionally, the probability density function (PDF) assesses skewness (peak-valley relationship) and kurtosis (degree of flatness) of the roughness profile.²⁸ The PDF describes the probability of occurrence of a given vertical amplitude in the roughness profile relative to a reference position. The closer the skewness is to zero, the more symmetric the number of peaks and valleys on the surface. Furthermore, the closer kurtosis is to zero, the better the balance between few sharp peaks and shallow valleys in the surface profile.^29,30

Table 1 presents some studies on CF/PEEK milling. Considering that existing studies on CF/PEEK milling were predominantly conducted under dry conditions, a gap in the literature exists regarding the effects of heated compressed air (HCA, +60°C) and cooled compressed air (CCA, −40°C) on the face milling of CF30-reinforced PEEK. These thermal conditions in compressed air have not been previously reported for this material, representing an original contribution to the machining of fiber-reinforced thermoplastics. Additionally, the study employs the Box–Behnken Design (BBD), a robust statistical method for evaluating surface finish and identifying optimal cutting conditions, thereby reinforcing the scientific and industrial relevance of the investigation.

Table 1.

Summary of various studies addressing the roughness issues in milling CF/PEEK.

Ref.	Tool details	Compositetype	Thermal condition	Cutting parameters (typical ranges)	Surface roughness results
6	PCD tool	Unidirectional CF/PEEK laminate (0°/90°)	Dry (high-speed milling)	v_c: 100–2000 m/min	R_a reduced by optimization (0.4–0.8 µm)
				f_z: 0.02–0.12 mm/tooth
				a_p: 0.02–0.12 mm
11	PCD end mill	PEEK reinforced with 30% short CF	Dry	v_c: 50–150 m/min	R_a reduced by optimization (0.5–0.9 µm)
				f_z: 0.003–0.18 mm/tooth
				a_p: 0.2–1.8 mm
16	PCD tool	Unidirectional CF/PEEK laminate (0°/45°/90°/135°)	Dry and compressed-air jet	v_c: 250–1500 m/min	Significant R_a reduction (temperature suppression)
				f_z: 0.07–0.12 mm/tooth
				a_p: 3 mm
19	Carbide end mill	Unidirectional CF/PEEK laminate (−45°/0°/45°/90°)	Dry, MQL and nanofluid-based MQL	v_c: 50–150 m/min	Significant R_a reduction under N-MQL compared to dry cutting
				f_z: 0.001–0.01 mm/tooth
				a_p: 3 mm
20	PCD tool	Unidirectional CF/PEEK laminate (0°/90°)	Dry (high-speed milling)	v_c: 100–1950 m/min	Lower surface defects
				f_z:0.02-0.12 m/tooth	S_a reduction
				a_p: 2 mm	S_a reduction
Present study	PCD face mill	PEEK reinforced with 30% short CF	Dry, CCA and HCA	v_c: 100–250 m/min	Lowest R_a = 0.307 µm
				f_z: 0.02–0.10 mm/tooth	R_z = 2.21 µm
				a_p: 0.1–0.9 mm	S_a = 0.371 µm under CCA

Material and methods

The composite material used is a natural PEEK reinforced with 30% short carbon fibers (TECAPEEK CF30 black), manufactured by Ensinger Plastics. This material has a density of 1.41 g/cm³, a tensile strength of 112 MPa, a thermal conductivity of 0.66 W/(m⋅K), a glass transition temperature of 147°C, a melting point of 341°C, and a service temperature range from 260°C (long term) to 300°C (short term).³¹ Face milling was performed on a Romi Discovery 308 machining center (maximum power 5.5 kW, maximum spindle speed 4000 rpm) using diverse cutting parameter combinations. A Sandvik R390-11T304E-P4-NL CD10 positive rhombic insert with a polycrystalline diamond (PCD) tip, tool nose radius (r_ε) of 0.4 mm, cutting edge radius (r_β) of 4.5 µm, sharp wedge, and wiper geometry (2.2 mm length) was used. The inserts were mounted on a Sandvik CoroMill^® 390 end mill (20 mm nominal diameter and two-insert capacity). To ensure process accuracy, the end mill was clamped in a Sandvik Corochuck^® 930 hydraulic chuck, minimizing runout error (less than 2.0 μm). To optimize material usage and prevent damage, the workpiece (46.5 × 38.0 × 31.5 mm) was machined in steps with a radial depth of cut of 11.5 mm. Over 45 runs, the total cutting length was approximately 1418 mm, resulting in a cutting time of 6.75 min without replacing the tool edges. No measurable tool-edge wear or fracture was observed throughout the experiments. Figure 1 illustrates the experimental setup under the CCA application. A plastic push-to-connect Y-splitter for an 8 mm tube was used to direct the compressed air. It was positioned at a 45° angle and 20 mm from the cutting zone.

Figure 1.

Experimental setup with CCA application.

The Box-Behnken Design (BBD) was employed in this study to evaluate the effects of three controllable factors – cutting speed (v_c), feed per tooth (f_z), and axial depth of cut (a_p) – each at three levels, on surface roughness. Factor levels were selected based on a literature review, considering the limitations imposed by the spindle speed of the machining center, workpiece clamping device, and workpiece size. Experiments were conducted in blocks to control variability across different testing days. Besides, each parameter combination was used under three thermal conditions: DRY, cooled compressed air (CCA), and heated compressed air (HCA). The experiment was designed using Minitab^® 22 (free trial). Table 2 details the cutting combinations.

Table 2.

Cutting parameter combinations for 15 machining runs.

Runs	v_c(m/min)	f_z(mm/tooth)	a_p(mm)
1	100	0.06	0.9
2^a	175	0.06	0.5
3	100	0.06	0.1
4	250	0.06	0.9
5	100	0.02	0.5
6	250	0.06	0.1
7	175	0.02	0.9
8^a	175	0.06	0.5
9	175	0.02	0.1
10	175	0.10	0.9
11	250	0.02	0.5
12	175	0.10	0.1
13	100	0.10	0.5
14^a	175	0.06	0.5
15	250	0.10	0.5

^aCentral points defined by BBD.

The cooling device used in this study (Figure 2a) consists of an aluminum helical coil tube with 30 m length and 8 mm diameter placed inside a thermally insulated container filled with dry ice (solid CO₂) at −78 °C. Compressed air at 400 kPa circulates through the coil, exchanging heat with the dry ice. A type-K thermocouple (with a measurement range from −100 to 1000 °C) was employed to monitor the air temperature at the coil outlet. Temperature data were collected using an Agilent 34970A data acquisition system. To initiate the process, 7.0 kg of dry ice were required, with a continuous refill rate of approximately 1.0 kg/h. The runs began 1 minute after the air temperature stabilized at −40 ± 1 °C, as indicated by the thermocouple. Due to water droplets in the compressed air and the low SCO₂ temperature, air-line regulators were installed to prevent coil obstruction. In addition, the cutting strategy was optimized to execute cutting runs more quickly, and an anemometer was used to verify the air flow.

The heating device (Figure 2(b)) has a configuration similar to the cooling device. It consists of a thermally insulated container filled with water, heated to the boiling point (100°C) by two electric resistors. An aluminum coil with dimensions similar to those of the cooling device, fully immersed in boiling water, transfers heat to the compressed air circulating through it. The air temperature at the coil outlet defines the heated compressed air (HCA) condition. The runs under HCA started 1 min after the temperature stabilized at 60 ± 3°C.

Figure 2.

Scheme for compressed air devices: (a) cooling; (b) heating.

To evaluate the 2D roughness profile and the amplitude parameters R_a and R_z of the machined surfaces, a Mitutoyo SJ-201P portable profilometer (0.01 µm resolution) was used to collect three roughness profiles and parameters collected on each sample, employing a sampling length (l_e) of 0.8 mm and an evaluation length of 4.0 mm. This setup is recommended for roughness values between 0.1 < R_a < 2.0 µm and/or 0.5 µm < R_z < 10 µm for non-periodic profiles as specified in ISO 21920-3.³² Data were collected with the dedicated software SJ Tools, which allows graphical analysis of the distribution of peaks and valleys generated during face milling.

The response variables (R_a and R_z) were statistically analyzed using Minitab^® 22 (free trial) to perform analysis of variance (ANOVA) and to optimize the process via the Box-Behnken Design (BBD). Matlab R2014a (MathWorks^®) was used to generate the Abbott-Firestone curves (AFC) and probability density functions (PDF) from the 2D roughness profiles. Surface textures of the machined samples were analyzed using a Zeiss EVO MA10 scanning electron microscope (SEM). For complementary analysis of optimized surfaces (3D profile and parameters S_a and S_z), a Taylor Hobson Precision Talysurf CCI Lite M112-3993-03 white-light interferometer was employed. Data processing and analysis were conducted using MountainsMap^® software, which is integrated with the equipment.

Results and discussion

Ra and Rz Roughness

The R_a and R_z roughness values measured after face milling of CF30-reinforced PEEK under dry, CCA, and HCA conditions are shown in Figure 3. In most runs, the R_a and R_z values obtained with cooled compressed air (CCA) were lower than those produced under the other conditions. Across all runs, the application of CCA resulted, on average, in reductions of 5.12% in R_a and 8.77% in R_z compared with those values generated by DRY machining. The HCA condition also resulted in reductions of 5.12% in R_a and 9.72% in R_z when compared with the DRY condition.

Figure 3.

Surface roughness values after CCA, DRY, and HCA milling: (a) R_a; (b) R_z.

Grouping the data by cutting speed (v_c) and analyzing the mean R_a and R_z values in both CCA and HCA conditions, the best results were obtained at v_c = 100 m/min. However, under DRY conditions, the lowest roughness values were obtained at v_c = 175 m/min. It is noteworthy that the combination of CCA with v_c = 100 m/min produced the lowest roughness values among the nine evaluated cases, with a mean R_a value of 0.34 μm (representing a reduction of 8% compared with HCA and 10.5% with DRY at the same v_c) and an average R_z of 2.46 μm (14% and 16.6% lower than HCA and DRY, respectively). A lower cutting speed tends to generate less heat in the cutting zone, which is beneficial for polymers sensitive to high temperatures.⁶ The feed per tooth (f_z) is theoretically the cutting parameter with the highest influence on surface roughness.^11,13 This trend is confirmed for both DRY and HCA conditions, where the lowest R_a and R_z values were achieved with the smallest f_z (0.02 mm/tooth). However, under the CCA condition, the best performance was observed with the middle f_z (0.06 mm/tooth), indicating a distinguishing behavior. According to Cao et al.,⁶ this f_z level may represent an equilibrium between stable chip formation and tool overload, i.e., smaller values do not remove material sufficiently, while larger values increase cutting forces, vibration, deformation, or surface defects.

Table 3 summarizes the lowest roughness values measured under the three thermal conditions. The average roughness (R_a) is classified into 12 grades according to ISO 21920-3.³² In this study, N5 (0.2 < R_a ≤ 0.4 µm) and N6 (0.4 < R_a ≤ 0.8 µm) grades prevailed, typically associated with components such as piston rods, shafts, gears, and screws. Specifically, 75.6% of the 45 samples fell into class N5, while 24.4% into class N6. As previously mentioned, CF30-reinforced PEEK has been increasingly used as a base material for such components and for biomedical applications such as prostheses.⁵ Therefore, achieving a high-quality surface finish is essential to ensure the efficiency and safety of components in service.

Table 3.

Lowest R_a and R_z roughness values for each thermal condition.

Condition	Sample	v_c (m/min)	f_z (mm/tooth)	a_p (mm)	R_a (μm)	R_z (μm)
CCA	01	100	0.06	0.9	0.317	2.270
DRY	07	175	0.02	0.9	0.337	2.400
HCA	05	100	0.02	0.5	0.343	2.517

The feed per tooth in the best CCA condition (0.06 mm/tooth) disagrees with the expectation that a lower feed per tooth generates better surface quality.^11,19 As the uncut chip thickness is proportional to f_z, the f_z/r_β ratio (13.3) facilitates the chip breakage, promotes a more effective cutting regime, and reduces surface deformation at low temperatures. Besides, a higher a_p (0.9 mm) improves the coolant penetration to the cutting interface. At lower f_z (0.02 mm/tooth), the f_z/r_β ratio (4.4) can make chip cutting difficult and impair surface finish due to the ploughing effect.³³

Surface texture assessment

Figure 4 shows the 2D roughness profiles of samples CCA₀₁, DRY₀₇, and HCA₀₅. When comparing the profiles, it was observed that the distributions of peak and valley amplitudes were similar across the three thermal conditions. Although some peaks exceed ±1.5 µm, most height variations remain within ±1.0 µm, regardless of the cutting condition.

Figure 4.

Surface roughness profiles of samples: (a) CCA₀₁; (b) DRY₀₇; (c) HCA₀₅.

A distinct feature is the presence of more pronounced undulations in DRY₀₇, which are not observed in the profiles of CCA₀₁ and HCA₀₅, indicating greater surface regularity when compressed air is applied, thereby facilitating more effective chip evacuation.¹⁶

Figure 5 shows the AFC and PDF plots for the samples with the lowest R_a values in each of the thermal conditions analyzed.

Figure 5.

AFC and PDF plots of samples: CCA₀₁, DRY₀₇, and HCA₀₅.

The AFC curves for these cases exhibit similar profiles, with only slight variations in slope. Analysis of the PDF functions indicates that the CCA condition has a higher concentration of data near the mean line (high kurtosis) and no skewness, reflecting a more uniform distribution of peaks and valleys and a more controlled roughness profile.^29,30 R_sk and R_ku values obtained for sample CCA₀₁ are closer to zero, indicating that the distribution of peaks and valleys in the roughness profile is approximately normal and symmetric. Comparatively, sample DRY₀₇ exhibits a slightly asymmetric distribution, with more peaks than valleys (R_sk > 0) and higher peaks and deeper valleys (R_ku < 0). Sample HCA₀₅ also shows a slightly asymmetric distribution, with more valleys than peaks (R_sk < 0). However, in these cases, the R_sk and R_ku values have been within the interval [−1, +1], indicating excellent surface quality.³⁰

Figure 6 presents the SEM images corresponding to the best surface roughness results in each thermal condition, captured at 500x and 2000x magnifications. The machined surface of CCA₀₁ at 500x shows no significant defects, whereas at 2000x, small voids are observed. This fact may be related to the strong temperature dependence of the mechanical behavior of polymers. Below the T_g, polymers tend to exhibit brittle behavior.⁵ In this context, natural PEEK has been reported to undergo a ductile–brittle transition at low temperatures, resulting in an abrupt change in impact behavior and a significant reduction in energy absorption capability.³⁴ In the DRY₀₇ sample, the 500x image reveals a surface quality poorer than that of the CCA₀₁ sample; at 2000x, smearing, voids, and exposed carbon fiber reinforcement are evident. The voids were possibly formed due to plastic deformation and material flow during cutting, which is directly associated with smearing. The heat generated during face milling softens the PEEK matrix, causing it to spread over the machined surface (smearing) that directly affects surface texture.^35,36 For the HCA₀₅ sample, the SEM image at 500x magnification shows that plastically deformed material was dragged by the cutting edge, resulting in feed marks with smearing. Under HCA conditions, the increase in temperature induced by the cutting process, further intensified by the application of compressed air at 60°C, may have raised the workpiece temperature above its T_g (147°C), thereby driving the composite into a rubbery regime. This behavior promotes material adhesion and flow, thereby favoring the formation of pronounced feed marks and surface smearing as indicated in the SEM image at 2000x magnification.

Figure 6.

SEM images of the samples with better surface finish.

Statistical analysis and optimization

To quantify the influence of the controllable factors on the response variables, an analysis of variance (ANOVA) was performed, considering a 95% confidence interval (α = 0.05). Table 4 presents the significant effects of the machining conditions on the two-dimensional amplitude parameters (R_a and R_z). For all conditions, the coefficient of determination (R²) and the lack-of-fit (LOF) test for the roughness parameters are within the expected range, except for R_z under the DRY condition, where R² was below 70%, and the LOF reached 28.29%. The complete ANOVA table is available in the supplementary material.

Table 4.

ANOVA results for R_a and R_z (α = 0,05).

Condition	Factor	R _a			R _z
Condition	Factor	p-value	Contribution	R ²	p-value	Contribution	R ²
CCA	f _z	<0.001	40.34%	96.58%	0.002	43.45%	94.32%
	v _c	0.002	23.48%		0.012	17.17%
	a _p	0.006	13.93%		0.014	15.45%
	v_c x f_z	0.01	11.11%		0.036	9.20%
	a_p x f_z	0.05	4.54%		-	-
	LOF	0.118	3.15%		0.704	2.53%
DRY	v_c x v_c	0.047	19.19%	85.09%	-	-	68.78%
DRY	LOF	14.31	0.06%	85.09%	0.138	28.29%	68.78%
HCA	f _z	0.007	39.02%	90.08%	0.025	29.69%	85.32%
	v_c x a_p	0.035	16.24%		-	-
	LOF	0.353	7.42%		0.571	8.35%

Figure 7 presents the multiple comparisons of the paired means of the significant first-order variables obtained using the Tukey test (α = 0.05). As reported by Aita et al.,³⁷ this test was used as a complement to ANOVA to assess whether changes in the levels of the input parameters lead to statistically significant differences in the surface roughness of CF/PEEK. The graphical interpretation is based on identifying confidence intervals that do not intersect the zero value, indicating statistically significant differences among the evaluated means. All of Tukey's graphs are available in the supplementary materials.

Figure 7.

Tukey pairwise comparisons.

The f_z interval of 0.02–0.10 mm/tooth was the only range that exhibited statistically significant differences in both R_a and R_z under CCA. In addition, the cutting speed interval of 250–100 m/min showed a statistically significant effect on R_z. Zhang et al.¹¹ reported that the most significant parameters affecting surface roughness generated by CF/PEEK milling are feed per tooth (f_z) and cutting speed (v_c). This trend was confirmed under the CCA condition (Table 4 and Figure 7), where f_z had the strongest influence on both R_a and R_z, followed by v_c.

Under DRY conditions, none of the parameters exhibited a significant linear effect (Table 4). The only relevant factor was the quadratic effect of cutting speed (v_c x v_c) on R_a, which may also be attributed to the temperature increase in the cutting zone induced by this parameter.¹² This behavior was also confirmed by the Tukey test (Figure 7), which showed that the v_c interval of 250–175 m/min was the only range that exhibited statistically significant differences. This effect may be further intensified by the absence of cooling, which hampers temperature control and efficient chip evacuation.^16,21

Liu et al.²⁰ stated that v_c is the parameter that most influences the heat generated in the cutting zone during dry face milling of unidirectional CF/PEEK laminates, although fiber orientation also affects this factor. Liu et al.³⁸ concluded that v_c is the primary factor influencing the temperature rise in the cutting zone during the milling of carbon fiber-reinforced polymer (CFRP). Klein et al.²¹ noted that increasing the axial depth of cut (a_p) increases the cutting-zone temperature during CFRP milling, leading to higher tool wear, reduced matrix support for the fibers, and poorer surface finish. These studies indicate that the cutting-zone temperature should be limited through appropriate machining conditions to ensure a high-quality texture, i.e., surfaces with minimal defects. Therefore, the significance of v_c and a_p (and/or a_e) is understood to depend on variations in the cutting-zone temperature.

Finally, under the HCA condition, f_z again proved to be the most influential parameter. However, in addition to f_z, the interaction between cutting speed and axial depth of cut (v_c x a_p) was significant (Table 4). This effect can again be related to changes in the cutting-zone temperature. Moreover, the removal of the layer from this heated surface depends directly on the a_p. The f_z exhibited two statistically significant intervals for R_a and R_z in the Tukey test, which is most likely associated with the pronounced feed marks observed. These marks result from increased plastic deformation during cutting caused by high temperatures, as shown in Figure 6.

Table 5 presents the results of desirability-based optimization of cutting parameters for the CCA, DRY, and HCA thermal conditions. The desirability function (D) is a statistical technique proposed by Derringer and Suich³⁹ that converts multiple experimental responses into a single numerical value ranging from 0 (worst condition) to 1 (best condition), reflecting how closely the responses approach their ideal targets. A desirability value of D = 1 was achieved under CCA and HCA conditions, whereas for DRY, the desirability function reached D = 0.917.

Table 5.

Desirability function results.

Condition	Sample	v_c (m/min)	f_z (mm/tooth)	a_p (mm)	R_a (estim) (μm)	R_z (estim) (μm)	D
CCA	16	100.0	0.0523	0.90	0.318	2.305	1.000
DRY	17	186.36	0.020	0.698	0.346	2.459	0.917
HCA	18	250.0	0.020	0.90	0.343	2.452	1.000

As mentioned before, a higher f_z (0.0523 mm/tooth) facilitates the chip breakage at low temperatures more than a lower f_z (0.02 mm/tooth) due to the ploughing effect.³³ It allows for improving the surface finish.

Table 6 compares the R_a and R_z values measured with the profilometer during the multi-objective optimization model validation tests and those estimated by the model for the three thermal conditions. Comparing the measured values, it is observed that sample CCA₁₆ was produced with cutting-parameter levels very similar to those of sample CCA₀₁, differing only by a reduction in f_z from 0.060 to 0.052 mm/tooth. This reduction led to a slight decrease in R_a and R_z for the optimized sample. On the other hand, sample DRY₁₇ exhibited slightly higher roughness values than sample DRY₀₇, probably due to the increased heat generation resulting by the higher v_c (from 175.0 to 186.4 m/min)⁶; the reduction in a_p (from 0.90 to 0.70 mm) likely did not affect this variation, as it does not significantly influence R_a and R_z (see Table 4). Additionally, the absence of compressed air may have further hindered efficient chip removal.²¹

Table 6.

Measured and predicted roughness values.

Condition	Sample	R_a (meas) (μm)	R_a (estim) (μm)	Error (%)	R_z (meas) (μm)	R_z (estim) (μm)	Error (%)
CCA	16	0.307	0.318	3.8	2.210	2.305	4.3
DRY	17	0.370	0.346	6.4	2.563	2.459	4.1
HCA	18	0.390	0.343	11.9	2.653	2.452	7.6

The substantial increases in v_c (from 100 to 250 m/min) and a_p (from 0.50 to 0.90 mm) may have caused higher roughness values (particularly R_a) in the HCA₁₈ sample compared to the HCA₀₅ sample, also due to the higher temperature in the cutting zone. The analysis of the material removal rate (MRR) (Figure 8) shows that, in addition to providing the best surface finish, the CCA condition also exhibits superior MRR. The MRR results for all runs are in the supplementary materials.

Figure 8.

Comparison between material removal rates for better surfaces.

Figure 9 shows the SEM images of the surface with the optimized roughness values results for each thermal condition, captured at 500x and 2000x magnifications. As previously mentioned, the texture of the CCA₁₆ sample appears more uniform than that of the CCA₀₁ sample, although small voids are still observed at 2000x. The cooling effect promotes stiffening of the composite, facilitating efficient cutting while reducing the spreading and plastic deformation that are predominant under dry or less effective cooling conditions.⁴⁰ In the DRY₁₇ sample, the surface texture retained characteristics like those of the DRY₀₇ sample, but with reduced smearing and no tearing (exposing the CF). Conversely, in the HCA₁₈ sample, no feed marks were observed, unlike in the HCA₀₅ sample, where smearing of higher magnitude and adhered material were detected.

Figure 9.

SEM images of the samples generated under optimized parameters.

The R_a parameter, widely used for evaluating metallic surfaces, is limited in composite machining due to the material’s heterogeneity, making surface characterization less reliable (it does not differentiate between peaks and valleys nor define the shapes of profile irregularities).¹⁷ Therefore, the S_a parameter, which evaluates an area, provides a value closer to the real surface. In this study, S_a values remained consistent with the R_a measurements. While R_z denotes the arithmetic mean of the five highest peak-to-valley distances in each sampling length of a 2D profile, the S_z indicates the maximum height of the limited area surface. Thus, S_z is more consistent with R_t (maximum height of the 2D profile) than with R_z.

Figure 10 presents the 3D and 2D roughness profiles obtained by interferometry for the previously best-finish samples (CCA₀₁, DRY₀₇, HCA₀₅) and for the optimized samples (CCA₁₆, DRY₁₇, HCA₁₈), considering a predefined square with an area of 1.96 mm². It is observed that the surfaces of the optimized samples CCA₁₆ (S_a = 0.371 μm; S_z = 18.23 μm) and DRY₁₇ (S_a = 0.400 μm; S_z = 12.42 μm) exhibit similar characteristics, differing mainly in the amplitude of the peaks and valleys (smaller under the DRY condition). A similar behavior is observed in samples CCA₀₁ (S_a = 0.385 μm; S_z = 23.09 μm) and DRY₀₇ (S_a = 0.402 μm; S_z = 15.48 μm), which display comparable morphology. However, for the HCA₁₈ sample (S_a = 0.394 μm; S_z = 19.19 μm), in addition to a higher roughness amplitude, the feed marks highlighted in Figure 6 are observable. High amplitudes were also observed in the HCA₀₅ sample (S_a = 0.415 μm; S_z = 20.39 μm), although, in this case, no feed marks were evident.

Figure 10.

3D and 2D roughness profiles of the samples generated under optimized parameters and those with the best surface finishes from the initial runs.

The S_tr parameter denotes surface isotropy when its value exceeds 0.50. As cited previously, anisotropy in composites can induce surface flaws.^10,11 It is known that unreinforced PEEK is isotropic, but adding aligned CF would make it anisotropic (the strength and stiffness of the composites depend on the fiber directions). The high dispersion of short fibers and the low incidence of structural defects may have contributed to the surface isotropy of this CF30-reinforced PEEK workpiece. Moreover, the values obtained for the optimized and the best-performing combinations were similar. The 2D profiles complement the analysis of the 3D surface. The optimized CCA₁₆ sample exhibits a lower number of peaks and valleys than CCA₀₁, leading to an improvement in the S_tr parameter. The profiles of the DRY₀₇ and DRY₁₇ samples show a high degree of similarity, which is consistent with their identical S_tr values. The HCA₁₈ sample shows fewer peaks and valleys and a more uniform profile than HCA₀₅. Overall, the 2D profiles are in good agreement with previously obtained results.

Thus, based on the surface quality characterization employing the S_a parameter from the 3D roughness profile, it is evident that the optimization process yielded slightly better results than those observed in the best runs for each condition.

Conclusions

The study analyzed the influence of cooled compressed air (CCA) and heated compressed air (HCA) on the face milling of CF30-reinforced PEEK using a PCD tool, comparing these conditions with dry machining (DRY) and evaluating their effects on surface finish under different cutting conditions. The main conclusions are summarized as follows:

(1) Sample CCA₀₁ exhibited the lowest roughness values (R_a = 0.317 μm and R_z = 2.27 μm) among the 15 initial samples, indicating that cooled compressed air at a low v_c reduced the cutting-zone temperature, limited thermal softening of the matrix, and preserved the surface integrity.

(2) AFC and PDF confirmed a more homogeneous and stable roughness profile of sample CCA₀₁, with fewer sharp peaks and shallower valleys than samples DRY₀₇ and HCA₀₅.

(3) Multi-objective optimization identified CCA as the optimal thermal condition for minimizing R_a and R_z values with v_c = 100 m/min, f_z = 0.052 mm/tooth, and a_p = 0.9 mm. Validation tests demonstrated accuracy for predicting roughness values (error ≈4%), providing good agreement with the experimental results.

(4) SEM analyses confirmed that CCA₀₁ achieved the best surface quality, with no smearing and only small voids. After optimization, the texture of sample CCA₁₆ became more uniform than that of CCA₀₁, while DRY and HCA conditions exhibited increasing feed marks and smearing.

(5) 3D roughness analyses confirmed isotropic behavior of this composite under all better conditions. After optimization, CCA₁₆ presented the lowest S_a value (0.371 μm), DRY₁₇ the lowest S_z value (12.42 μm), and HCA₁₈ the most irregular texture (highest roughness values).

Although cooled compressed air (CCA) improves surface quality, its industrial implementation should also consider the cost-benefit relationship, as the additional complexity of the cooling system may not always justify the cost compared to dry machining (DRY). In addition, the present study is limited to a short CF30-reinforced PEEK composite, a single cutting-tool material/geometry, and the absence of in situ temperature measurements. Future investigations should therefore include temperature and tool-condition monitoring, cooling strategies, and the assessment of different tool materials/geometries to establish performance-productivity trade-offs better and support broader industrial applications.

Supplemental material

Supplemental Material - Influence of cooled and heated compressed air on the surface finish in face milling of CF30-reinforced PEEK

Supplemental Material for Influence of cooled and heated compressed air on the surface finish in face milling of CF30-reinforced PEEK by Fernando Pasquali, Carlos A. G. Aita, André J. Souza, and Milton L. Polli in Journal of Thermoplastic Composite Materials

Footnotes

Author’s note

This manuscript is original and has not been submitted for publication elsewhere. The manuscript has no associated data from any data repository.

Acknowledgements

The authors thank Ensinger^® Co. for the donation of PEEK/CF30 workpiece, Sandvik Coromant Co. for the donation of PCD milling inserts, Minitab^® 22 for the trail test license for statistical analysis, Multiuser Center for Materials Characterization (CMCM/UTFPR) for the 3D roughness analysis and the Federal Institute of Santa Catarina (Process No. 23292.030882/2022-26) for granting leaves for study.

Author contributions

Each author contributed to the research presented in this manuscript, approved the contents, and agreed to comply with the ethical standards.

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

ORCID iDs

Fernando Pasquali

Carlos A. G. Aita

André J. Souza

Milton L. Polli

Supplemental material

Supplemental material for this article is available online.

References

Choi

B-K

Kang

C-S

Yoo

M-H

, et al. Effect of processing parameters on bonding performance of a carbon fiber/polyetheretherketone thermoplastic composite prepared by induction welding. Materials 2023; 16: 3954.

Qin

, et al. Numerical and experimental investigation of orthogonal cutting of carbon fiber-reinforced polyetheretherketone (CF/PEEK). Int J Adv Manuf Technol 2022; 119: 1003–1017.

Alarifi

. A review on factors affecting machinability and properties of fiber-reinforced polymer composites. J Nat Fibers 2023; 20: 2154304.

Liampas

Kladovasilakis

Tsongas

, et al. Recent advances in additive manufacturing of fibre-reinforced materials: a comprehensive review. Appl Sci 2024; 14: 10100.

Khoran

Amirabadi

Azarhoushang

. The effects of cryogenic cooling on the grinding process of polyether ether ketone (PEEK). J Manuf Process 2020; 56: 1075–1087.

Cao

Liu

, et al. Process optimization of high-speed dry milling UD-CF/PEEK laminates using GA-BP neural network. Compos B Eng 2021; 221: 109034.

Geier

Poór

, et al. A review on advanced cutting tools and technologies for edge trimming of carbon fibre reinforced polymer (CFRP) composites. Compos B Eng 2023; 226: 111037.

Nisar

Banday

Amatullah

, et al. An investigation on effect of process parameters on surface roughness and dimensional inaccuracy using grey based taguchi method. Mater Today Proc 2021; 46: 6564–6569.

Liu

Tao

, et al. Influence of cenosphere on tribological properties of short carbon fiber reinforced PEEK composites. J Appl Polym Sci 2019; 136: 47245.

10.

Caggiano

. Machining of fibre reinforced plastic composite materials. Materials 2018; 11: 442.

11.

Zhang

Wang

, et al. Multi-objective optimization of machining parameters during milling of carbon-fiber-reinforced polyetheretherketone composites using grey relational analysis. Adv Mech Eng 2020; 12: 1687814020966232.

12.

Bhushan

. Effect of tool wear on surface roughness in machining of AA7075/10 wt.% SiC composite. Composites Part C: Open Access 2022; 8: 100254.

13.

Abellán-Nebot

Vila Pastor

Siller

. A review of the factors influencing surface roughness in machining and their impact on sustainability. Sustainability 1917, 2024; 16: 1917.

14.

Vieira

Pereira

RBD

Lauro

, et al. Multi-objective evolutionary optimization of extreme gradient boosting regression models of the internal turning of PEEK tubes. Expert Syst Appl 2024; 238: 122372.

15.

Ferreira Batista

Basso

de Assis Toti

, et al. Cryogenic drilling of carbon fibre reinforced thermoplastic and thermoset polymers. Compos Struct 2020; 251: 112625.

16.

Liu

Zhang

, et al. Modeling of workpiece temperature suppression in high-speed dry milling of UD-CF/PEEK considering heat partition and jet heat transfer. Front Mech Eng 2024; 19(30): 30.

17.

Wang

, et al. Investigation of machining characteristics and parameter optimization for laser-assisted milling of CF/PEEK composites. Micromachines 2025; 16: 151.

18.

Zhou

Pang

Zhang

, et al. Experimental investigation of the mechanical properties of carbon fiber-reinforced polymer (CFRP) tendons during and after exposure to elevated temperatures. Mater Struct 2022; 55(2): 82.

19.

Özcan

Vahapoğlu

Öztürk

, et al. Evaluation of slot milling processes on carbon fiber reinforced polyetheretherketone thermoplastic composite utilizing nanofluid-enhanced minimum quantity lubrication technique. J Thermoplast Compos Mater 2025; 38(11): 4329–4360.

20.

Liu

Cao

, et al. Process optimization of high machining efficiency and low surface defects for HSD milling UD-CF/PEEK with limited thermal effect. J Manuf Process 2022; 76: 532–547.

21.

Klein

Hoffmann

Souza

, et al. Optimization of cutting parameters for finish end milling CFRP under vortex-cooled compressed air. Mater Res 2022; 24(Suppl 2): e20220231.

22.

Domingo

de Agustina

Marín

. Study of drilling process by cooling compressed air in reinforced polyether-ether-ketone. Materials 1965, 2020; 13: 1965.

23.

Ensinger . Machining recommendations for semi-finished engineering plastics. Cham, Germany: Ensinger GmbH, 2017.

24.

John

Kumaran

. A feasible strategy to produce quality holes using temperature-assisted drilling on CFRP. Int J Adv Manuf Technol 2020; 110: 3113–3127.

25.

Velavan

Palanikumar

Senthilkumar

. Experimental investigation of sliding wear behaviour of boron carbide and mica reinforced aluminium alloy hybrid metal matrix composites using box-behnken design. Mater Today Proc 2021; 44: 3803–3810.

26.

Gulyás

Al-Tayawi

Horváth

, et al. Methods for experimental design, central composite design and the box-behnken design, to optimise operational parameters: a review. Acta Aliment 2023; 52: 521–537.

27.

Pawlus

Reizer

Wieczorowski

, et al. Material ratio curve as information on the state of surface topography – a review. Precis Eng 2020; 65: 240–258.

28.

Alsoufi

Bawazeer

. Probabilistic analysis of surface integrity in CNC turning: influence of thermal conductivity and hardness on roughness and waviness distributions. Machines 2025; 13(5): 385.

29.

Policena

Devitte

Fronza

, et al. Surface roughness analysis in finishing end-milling of duplex stainless steel UNS S32205. Int J Adv Manuf Technol 2018; 98: 1617–1625.

30.

ECT

Dumont

Martins

, et al. Investigation of the effects of skewness (R_sk) and kurtosis (R_ku) on tribological behavior in a pin-on-disc test of surfaces machined by conventional milling and turning processes. Mater Res 2021; 24(1): e20200435.

31.

Ensinger . Technical data sheet: TECAPEEK CF30 black. Cham, Germany: Ensinger GmbH, 2023.

32.

ISO 21920-3 . Geometrical product specifications (GPS) – surface texture: profile, part 3: specification operators. Geneva, Switzerland: International Organization for Standardization, 2021.

33.

Jin

Gao

Wang

, et al. Micro-milling mechanism and surface roughness of high volume fraction SiCp/Al composites. Int J Adv Manuf Technol 2021; 115: 91–104.

34.

Garcia-Gonzalez

Rodriguez-Millan

Rusinek

, et al. Low temperature effect on impact energy absorption capability of PEEK composites. Compos Struct 2015; 134: 440–449.

35.

Liu

Sun

, et al. Investigating the material removal mechanism and cutting performance in ultrasonic vibration-assisted milling of carbon fibre reinforced thermoplastic. Mater Res Express 2023a; 10: 095603.

36.

Liu

. Comparative study of tool wear and machining quality in drilling of thermoplastic CF/PEEK and thermoset CF/epoxy composites. J Thermoplast Compos Mater 2025, In Press (First Online).

37.

Aita

CAG

Souza

Passari

, et al. Effect of cooled compressed air on the machining force and surface roughness generated by face milling of CW50/PEEK. Int J Mach Mach Mater 2025; 27: 345–368.

38.

Liu

Zhang

Zhao

, et al. A comparative study on milling-induced damages and residual tensile strength during milling of thermoplastic and thermoset carbon fibre reinforced polymers. Polym Test 2023b; 125: 108132.

39.

Derringer

Suich

. Simultaneous optimization of several response variables. J Qual Technol 1980; 12: 214–219.

40.

Cococcetta

Jahan

Schoop

, et al. Post-processing of 3D printed thermoplastic CFRP composites using cryogenic machining. J Manuf Process 2021; 68: 332–346.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.91 MB