Strategic cost and sustainability analyses of injection molding and material extrusion additive manufacturing

Abstract

Economic and environmental costs are assessed for four different plastics manufacturing processes, including cold and hot runner molding as well as stock and upgraded material extrusion three dimensional (3D) printers. A larger stock 3D printer was found to provide a melting capacity of 14.4 ml/h, while a smaller printer with an upgraded extruder had a melting capacity of 36 ml/h. 3D printing at these maximum melting capacities resulted in specific energy consumption (SEC) of 16.5 and 5.28 kWh/kg, respectively, with the latter value being less than 50% of the lowest values reported in the literature. Even so, analysis of these respective processes found them to be only 2.9% and 3.8% efficient relative to their theoretical minimum energy requirements. By comparison, cold and hot runner molding with an all-electric machine had SEC of 1.28 and 0.929 kWh/kg, respectively, with efficiencies of 9.9% and 13.6% relative to the theoretical minima. Breakeven analysis considering the cost and carbon footprint of mold tooling found injection molding was preferable at a production quantity of around 70,000 units. Parametric analysis of model inputs indicates that the breakeven quantities are robust with respect to carbon tax incentives but highly dependent on mold costs, labor costs, and part size. Dimensional and mechanical properties of the molded and 3D printed specimens are also characterized and discussed.

1 INTRODUCTION

Society is increasingly aware of the need for sustainable solutions across all aspects of modern life. To the extent that this increased awareness does not drive organic growth in consumer demand for sustainable products, manufacturers will likely be forced to account for the sustainability impacts of their products through environmental, social, and governance (ESG) standards and related tax incentives or penalties.^{[1, 2]} Accordingly, firms should perform a formal analysis of sustainability at the product design phase to ensure strategic process and material selection.^[3]

This article investigates the cost and sustainability of injection molding and additive manufacturing (AM) by material extrusion of polymers. Injection molding is the most common method for mass production of complex products, while material extrusion (also referred to as fused filament fabrication [FFF], and Fused Deposition Modeling™) is the most common form of polymer three dimensional (3D) printing. There is a growing body of literature, summarized in Table 1, upon which the presented work is based. The article advances this state of the art by (1) using modern machines operated with a defined protocol to minimize energy consumption, (2) characterizing observed and minimum theoretical energy consumption required for heating and forming in these processes, and (3) evaluating cost and sustainability break-even points from a strategic viewpoint based on application characteristics available early in the design stage. Interested researchers are encouraged to first read the literature reviews and meta-analyses by Suárez^[12] and Khalid^[13] to understand the scope of prior work before delving into the more detailed references.

TABLE 1. Summary of related literature by category

Literature reviews Gebler et al.^[^4^] evaluated the sustainability of 3D printing on a global scale, estimating only a 5% potential impact due to a lack of technological capability. Yoon et al.^[^5^] found the specific energy consumption (SEC) for FFF was 23–346 kWh/kg, about 100-fold higher than that of conventional forming processes. Liu et al.^[^6^] provides a literature review and design methodology for sustainable AM optimization. Cerdas et al.^[^7^] provides a meta-analysis of filament-based AM literature indicating that printing energy accounts for 70% of environmental impact. Kellens et al.^[^8^] compares SEC of different AM processes, indicating 23–344 kWh/kg for FFF. Khosravani et al.^[^9^] provides a broad review of sustainability concerns across AM, including air pollution, recycling, and energy reduction approaches. Ribeiro et al.^[^10^] provided a broad review of life-cycle research on AM with environmental, economic, and societal metrics. Mastura et al.^[^11^] suggested a life-cycle analysis framework for FFF with four phases: (1) goal and scope definition, (2) inventory analysis, (3) impact assessment, and (4) interpretation. Suárez et al.^[^12^] reviews the sustainability of AM, emphasizing the need to optimize the design of the part to be manufactured as well as reduce energy consumption in processing. Khalid et al.^[^13^] provides lifecycle assessment, including a predictive framework for energy modeling and sustainability improvement. Jandyal et al.^[^14^] reviewed sustainability of 3D printing in view of industry 4.0. Cost, life cycle, and sustainability assessment Luo et al.^[^15^] evaluates environmental impact of rapid tooling methods with a focus on the embodied energy of Keltool versus electroplated, selectively laser sintered plastic molds. Hopkinson et al.^[^16^] compares the cost of injection molding with layer based AM processes, with breakeven quantities of ~700 units for a larger cover and ~8000 for a smaller lever. Peng et al.^[^17^] considers the life cycle of FFF printing, including initial filament extrusion, 3D printing, post-processing, and transportation. Dev et al.^[^18^] investigates the effect of infill parameters on material usage and properties in FFF. Chu et al.^[^19^] pursued more sustainable FFF using 100% recycled blends of PS and PP. Energy characterization and optimization Baumers et al.^[^20^] provides a review of energy consumption in AM processes, reporting a SEC of 144 kWh/kg for FFF deposition of PC. Weissmann et al.^[^21^] describes general energy modeling considerations across different families of processes with a focus on injection molded parts. Peng et al.^[^22^] investigates primary energy (melting, forming) and secondary energy (controllers, fans) usage in FFF. Ajay et al.^[^23^] found that electricity is about 30% of total cost for the FFF process with the primary consumers including 51% motors, 42% heating, and 7% overhead. Fok et al.^[^24^] describes a tool path optimizer focused on surface/detail quality and print time. Feng et al.^[^25^] performed a 27 run (3^3) DOE to characterize the bed temperature, layer height, and print speed on energy consumption in FFF. Peng et al.^[^26^] investigate the power profiles of three different FFF printers, finding their behaviors to vary with respect to relative heating and motor power consumption. Yang et al.^[^27^] developed a power consumption model for FFF, finding a standby power of 20 W, combined stepper motor power of 42 W, 262 W bed heating power, and 80 W hot end heating power. Xu et al.^[^28^] developed an energy efficient path optimization method for 3D printing, evaluating layer thickness, infill pattern, and infill trajectory are evaluated to achieve the optimization. Hernandez et al.^[^29^] investigates energy and the CO2 equivalent emissions in FFF for varying materials, specimen orientations, and number of parts per build. Nguyen et al.^[^30^] modeled the energy consumption of FFF, finding that energy consumption was 47% for the printbed, 25% for the hot end, 19% for the motors, and 9% for the fans and controller. Hopkins et al.^[^31^] found the volumetric specific energy in FFF was 24.8–85.7 kJ/cm^3 proportioned as 56% bed, 25% hot end, and 19% motors, fans, and control boards. Ma et al.^[^32^] developed energy models for FFF, SLA, and SLM, with FFF energy consumption driven by bed heating at 62.6%, auxiliary systems at 24.7%, nozzle at 9.6%, XY motors at 1.6% each, and Z motor at 0.1%. Mele et al.^[^33^] performed a sustainability analysis of the Arburg Freeformer^[^34^] indicating that the machine build and process energy consumption were the primary issues. Enemuoh et al.^[^35^] evaluates effects of FFF factors, including infill density, infill pattern, layer thickness, print speed, and shell thickness, on energy consumption, production time, and priorities. Enemuoh et al.^[^36^] compares the energy consumption of FFF and a desktop injection molding of PLA tensile specimens with a SEC of 50.4 kWh/kg for FFF and 36.5 kWh/kg for injection molding. Rojek et al.^[^37^] performs neural network optimization of extrusion-based AM. Espach et al.^[^38^] optimizes FFF process to increase tensile from 45 to 55 MPa and reduce material consumption from 34 to 29 g. Yan et al.^[^39^] provides a methodology for estimating SEC in AM with the parsing of g-code to provide real-time estimates of energy consumption.

• Abbreviations: AM, additive manufacturing; DOE: design of experiments; FFF, fused filament fabrication; PLA, polylactic acid; PP: polypropylene; PS, polystyrene;  SLM: selective laser melting. 

While there are many performance metrics for sustainable design,^[40] two of the most common are “embodied energy” and “carbon footprint.” Ashby^[41] defines embodied energy as the energy consumed to produce 1 kg of material. By comparison, the carbon footprint represents the total direct and indirect CO₂ given to a product, process, or activity over its entire lifetime.^[42] While energy consumption during product use is a significant concern in the transportation industries, Rydh et al.^[43] found that the emission of carbon dioxide in product design and manufacturing is practically linear with the embodied energy of the consumed materials; the same correlation between embodied energy and carbon footprint was also evident in many of the case studies described by Morini et al.^[40] For this reason, the article focuses on economic cost, with carbon footprint (kg CO₂) as the primary measure of sustainability.

2 MATERIALS AND METHODS

2.1 Materials and processing temperatures

All materials were processed from commercially available, extruded filament comprised of polylactic acid (PLA, Matter-Hackers part no. M-UAA-MXWX). Ten spools of 1 kg of filament were received and confirmed to be of the same lot. Four of the spools were pelletized for injection molding, with pellet lengths ranging from 5 to 8 mm. The processing temperatures for all processes were set according to the material supplier recommendation, with the hot end temperature for 3D printing processes set to 210°C while the bed temperature was set to 40°C. Similarly, the molding machine was set to provide a melt temperature of 210°C with the nozzle, forward, middle, and rear barrel zones respectively set to 210, 210, 190, and 170°C while the mold coolant (i.e., water) temperature was set to 40°C resulting in a cycle time of 48 s. The use of a hot runner mold with a larger feed system allowed the PLA to be molded at lower temperatures of 180°C and shorter cycle times of 26 s. All materials were processed consistently and without difficulty.

2.2 Injection molding

Injection molding was performed on a 2021 Milacron Fanuc Roboshot α-S130iB all-electric molding machine shown in Figure 1. This machine, including installed options, has a list price of USD246,610. Power consumption metering is built into the machine's user interface, with reporting of the servo motors, heaters, and auxiliary systems as shown by the inset images at the top of Figure 1. While the machine assumes a default equivalence of 0.555 kg CO₂ per kWh of energy as shown in Figure 1, later analysis converts the recorded energy values using a conversion of 0.433 kg CO₂ per kWh.^[44] A script for MATLAB (Mathworks, Cambridge, MA) was written to perform the analysis and is provided in Section S.2.

FIGURE 1

(bottom) Milacron Fanuc Roboshot a-S130iB all-electric molding machine used for this study with a (top) power consumption panel including (inset) CO₂/kWh conversion.

An ASTM mold, shown at left in Figure 2, was used to mold sets of test parts. The injection molded ASTM Type I tensile specimen had a thickness of 3.2 mm and a degated part mass of 10.5 g. The cold runner molding process was optimized to minimize cycle time and energy input. Accordingly, the minimum feasible cycle time was found to be 48 s with the lower bound constrained by the rate of material cooling. At lower attempted cycle times, the parts and runner system were too hot and flexible to be ejected from the mold.

FIGURE 2

(left) Cold runner ASTM mold, and (right) hot runner two-cavity tensile bar mold. Tensile specimens are highlighted (red) at right and comparable to those produced from the large tensile cavity shown in the mold at left.

In plastic mold design, the choice of cold or hot runners is a strategic decision. Hot runner systems add significant cost but lower material usage and cycle time through the avoidance of molding in the cold runner system.^{[45, 46]} To support the cost and sustainability analysis, a valve gated hot runner design to make tensile bars was also considered. The hot runner mold (design shown at right in Figure 2) was designed to produce two 9.6 g tensile specimens in every molding cycle. Since the hot runner did not need to cool the feed system, a cycle time of 26 s was achieved before the parts became too flexible to demold.

The listed price of the received hot runner mold was USD50,000 while the cold runner mold cost was estimated at USD20,000. Later analysis will explicitly investigate the role of machine and mold mass and cost on process selection according to specific application requirements. To ensure consistency across processes with slightly different part geometries, the results for all processes were normalized to a de-gated part mass of 10.5 g. Hence, the cold runner molding process was modeled as a two cavity mold with a characterized 48 s cycle time.

2.3 Material extrusion AM

Tensile specimens were produced from the PLA using two common but different 3D printer systems. The larger CR10S-Pro (Creality, ShenZhen, China) costs USD520 and has build dimensions of 300, 300, and 400 mm in the X, Y, and Z directions. This printer was unmodified and uses its stock hot end with a Bowden style filament drive, as shown at left in Figure 3. The Ender 5 (also Creality, shown at right in Figure 3) costs USD259 and provides slightly smaller build dimensions of 220, 220, and 300 mm in the X, Y, and Z directions. The Ender 5 printer was upgraded with an E3D Volcano hot end (Matter Hackers part no. M-CD8-H4NE, costing USD65) to increase melting capacity as well as a Micro-Swiss direct drive extruder (Matter Hackers part no. M-29Z-9VUL costing USD55) for improved control.

FIGURE 3

(left) Creality CR10S-Pro with stock Bowden filament drive, and (right) Creality Ender 5 upgraded with Micro-Swiss direct drive extruder and E3D Volcano hot end.

Prior to printing, the maximum print speed was determined based on the melting capacity of the 3D printer's hot end. Specifically, g-code was programmatically generated to print a serpentine pattern while increasing the volumetric flow rate in increments of 0.1 mm³/s. At higher flow rates, the material flowing through the hot end does not have sufficient residence time or heat transfer, and so the temperature of the material flowing through the nozzle decreases. As a result of the lower temperature and associated increases in viscosity and drive pressure/torque, the extruder drive gear slips and the flow rate of the extrudate temporarily decreases. The temporary cessation of flow causes a necking defect but allows the material additional time to reheat and continue printing. These defects are cyclic and have increasing frequency with increased flow rates (e.g., 10.1, 10.2, and 10.3 mm³/s), such as shown at the bottom of Figure 4. While this defect can be formally detected by monitoring the melt pressure in the hot end,^[47] it is an important best practice to characterize the melting capacity in FFF for specific materials and processing conditions to ensure rapid but robust deposition rates.

FIGURE 4

Material extrusion melting capacity characterization for upgraded Ender 5 in which print speeds are incrementally increased in a serpentine pattern from 9.3 to 10.3 mm³/s. Defects indicate extruder skipping due to insufficiently heated polymer above 10 mm³/s (36 ml/h).

The characterized maximum flow rates were found to be 4.0 mm³/s (14.4 ml/h) for the CR10S-Pro and 10.0 mm³/s (36 ml/h) for the upgraded Ender 5. Ultimaker Cura 4.13.1 (Utrecht, Netherlands) was used to generate g-code for printing five fully dense tensile bars in parallel at the maximum melting capacity using a road width of 0.5 mm and a road height of 0.2 mm. The effect of the in-fill pattern was also investigated with the CR10S-Pro using a hatch pattern with rastering at ±45° relative to the tensile specimen's load direction, while the Ender 5 was programmed to use a concentric fill pattern to minimize the print time. The CAD geometry was created to provide a batch of five 3D printed specimens with the same thickness as the molded specimens. The resulting print time was 225 min for the CR10S-Pro printing at 4 mm³/s with the ±45° raster pattern, compared to 85 min for the upgraded Ender 5 at 10 mm³/s. Given the deposition rates and observed print times, the serpentine pattern results in a print reduction of 5.6% relative to the ±45° raster pattern.

2.4 Characterization of energy consumption

A power meter (Poniie PN2000, Amazon part no. B0777H8MS8) was used to record the power usage in 3D printing the tensile specimens. The power consumption was found to be 16.5 kWh/kg for the CR10S-Pro and 5.28 kWh/kg for the upgraded Ender 5. Given this specific energy consumption (SEC), additional experimentation was performed to characterize the power usage. A set of seven blocked designs of experiments was performed according to the settings of Table 2 using an Acuvim II-M-333-P1 power meter (Toronto, ON) connected to a National Instruments NIUSB6212 (Dallas, TX). A g-code script was programmatically generated to perform all 28 runs with a starting and ending idle period to investigate changes in the X print speed and acceleration, Y print speed and acceleration, Z print speed and acceleration, extruder (E) print speed and acceleration, hot end temperature, hot end fan speed, extrusion flow rate, bed temperature, and enclosure condition. The factors, level settings, and observed power statistics, including the mean and standard deviation (SD), are provided in Table 2.

TABLE 2. Seven blocks of four run (2² full factorial design) investigating 13 factors

Run #	DOE block	DOE run	Factor 1		Factor 2		Power, W mean ± SD
			Descriptor	Setting	Descriptor	Setting
1	0	1	Printer Idle	N/A	none	N/A	32.64 ± 17.90
2	1	1	X Speed	20 mm/s	X Acceleration	500 mm/s^2	30.27 ± 0.56
3	1	2	X Speed	20 mm/s	X Acceleration	3000 mm/s^2	30.35 ± 2.39
4	1	3	X Speed	100 mm/s	X Acceleration	500 mm/s^2	30.27 ± 0.57
5	1	4	X Speed	100 mm/s	X Acceleration	3000 mm/s^2	30.32 ± 0.52
6	2	1	Y Speed	20 mm/s	Y Acceleration	500 mm/s^2	30.30 ± 0.74
7	2	2	Y Speed	20 mm/s	Y Acceleration	3000 mm/s^2	30.14 ± 0.32
8	2	3	Y Speed	100 mm/s	Y Acceleration	500 mm/s^2	30.04 ± 1.06
9	2	4	Y Speed	100 mm/s	Y Acceleration	3000 mm/s^2	30.15 ± 0.47
10	3	1	Z Speed	2 mm/s	Z Acceleration	200 mm/s^2	30.26 ± 0.93
11	3	2	Z Speed	2 mm/s	Z Acceleration	1200 mm/s^2	30.20 ± 0.78
12	3	3	Z Speed	10 mm/s	Z Acceleration	200 mm/s^2	30.21 ± 0.57
13	3	4	Z Speed	10 mm/s	Z Acceleration	1200 mm/s^2	30.25 ± 0.46
14	4	1	E Speed	1 mm/s	E Acceleration	25 mm/s^2	30.20 ± 0.50
15	4	2	E Speed	1 mm/s	E Acceleration	150 mm/s^2	30.28 ± 0.71
16	4	3	E Speed	5 mm/s	E Acceleration	25 mm/s^2	30.29 ± 0.42
17	4	4	E Speed	5 mm/s	E Acceleration	150 mm/s^2	30.26 ± 0.74
18	5	1	Hot End Temp	190°C	Fan Speed	5%	84.67 ± 62.97
19	5	2	Hot End Temp	190°C	Fan Speed	100%	87.63 ± 63.43
20	5	3	Hot End Temp	230°C	Fan Speed	5%	82.20 ± 62.99
21	5	4	Hot End Temp	230°C	Fan Speed	100%	86.46 ± 63.90
22	6	1	Hot End Temp	190°C	Flow Rate	2.5 mm^3/s	80.18 ± 62.49
23	6	2	Hot End Temp	190°C	Flow Rate	10 mm^3/s	93.52 ± 65.37
24	6	3	Hot End Temp	230°C	Flow Rate	2.5 mm^3/s	89.77 ± 64.55
25	6	4	Hot End Temp	230°C	Flow Rate	10 mm^3/s	93.87 ± 64.86
26	7	1	Enclosure	Open	Bed Temp	60°C	81.23 ± 62.27
27	7	2	Enclosure	Open	Bed Temp	100°C	102.05 ± 61.86
28	7	3	Enclosure	Closed	Bed Temp	60°C	68.40 ± 59.37
29	7	4	Enclosure	Closed	Bed Temp	100°C	96.47 ± 63.05
30	0	2	Printer Idle	N/A	none	N/A	31.44 ± 17.82

Data were recorded at a sample rate of 10 kHz and filtered with a first order lowpass digital Butterworth filter with a normalized cutoff frequency of 0.1. Section S.1 provides a plot of the acquired power and cumulative energy across the seven blocked DOE. A multiple linear regression model was fit to the power consumption data summarized in Table 2 with a coefficient of determination, R², equal to 0.992 and a slope of 1.006 between the predicted and observed power consumption. Section S.1 also provides the main effects plots for the CR10S-Pro and upgraded Ender 5 printers, which were used to confirm that the implemented printer settings were selected to yield the minimum SEC.

2.5 Characterization of product performance

Five molded and printed specimens were tested with an Instron Universal Testing Machine (Norwood, MA, USA) at standard conditions with an extension rate of 0.1%/s, such that strain rate effects are unlikely to be significant. The measured part dimensions, yield stress, and elastic modulus are later reported with process capability measures in Section 4.5.

3 THEORY AND CALCULATIONS

3.1 Cost and carbon footprints

Section S.2 provides a Matlab script for calculating the economic cost and carbon footprint for the four different processing strategies: (1) cold runner injection molding, (2) hot runner injection molding, (3) stock CR10S-Pro 3D printing, and (4) upgraded Ender 5 3D printing. This script starts with a parametric analysis using function calls to Calc_ESG_USD_Per_Part() to generate the presented results. In that function, matrix calculations are performed to evaluate the results for vectors of production quantities, Q, between one hundred and one million units.

Generic production equipment, such as molding machines and printers are typically not purchased exclusively for a single application, and so cost accounting is performed on an amortized hourly basis given the lifetime of the equipment. For each processing strategy, the economic cost contribution of the machine is evaluated on a total cost basis as ${\mathrm{USD}}_{\text{machine}}=\mathit{Qt}{C}_{\text{machine}/\text{hour}}$ given the production quantity, $Q$, average part processing time, t, and the hourly machine cost, $C_{\text{machine}/\text{hour}}$. This hourly machine cost is calculated using straight line depreciation given each machines' list prices. Ten-year and 5-year asset lives for the molding machines and 3D printers were assumed.^[48] For all machinery, 4000 operating hours per year were considered.

For molding processes, the cost of the mold, ${\mathrm{USD}}_{\text{mold}}$, is initially fixed as previously described (USD20,000 and USD50,000 for cold and hot runner molds for the tensile bars). The amortized mold cost (on a per part basis) is simply calculated by dividing the total production cost by the target production quantity. It is noted that mold maintenance costs were neglected in the analysis but can be significant.^[49] The implemented models can be readily extended by including the mold maintenance cost in the upfront purchase cost of the mold or otherwise modeling the maintenance cost as a function of the production quantity.

The cost of the material, energy, and labor consumption are also modeled. The total material costs are ${\mathrm{USD}}_{\text{material}}=\mathrm{Qm}{\mathrm{C}}_{\text{material}/\mathrm{kg}}$ given the part mass, $m$, and cost of material per kg, $C_{\text{material}/\mathrm{kg}}$. The total energy costs are ${\mathrm{USD}}_{\text{energy}}=\mathrm{QE}{\mathrm{C}}_{\text{energy}/\mathrm{kWh}}$ given the energy consumption per part, $E$, and cost of energy per kWh, $C_{\text{energy}/\mathrm{kWh}}$. The total labor costs are modeled as ${\mathrm{USD}}_{\text{labor}}=\mathit{Qt}{C}_{\text{labor}/\text{hour}}/n_{\text{machines}/\text{operator}}$, where $C_{\text{labor}/\text{hour}}$ is the average total burdened hourly rate per operator is equal to USD24.97/h.^[50] Since each operator can supervise multiple machines, the number of machines per operator is respectively set to 10 for injection molding and 40 for 3D printing. It is noted that the post-processing costs for 3D printing are not explicitly modeled but can be readily modeled through the assignment of the number of machines per operator, wherein the operator finishes parts when not otherwise actively supporting their assigned printers.

The carbon footprint analysis generally proceeds in parallel with the cost analysis, albeit with a different set of coefficients per unit of consumption as set forth in Table 3. The carbon footprint of equipment can be estimated from the components used in the machine assembly, such as implemented by Yi et al.^[59] However, a rough estimate of the carbon footprint for equipment such as machinery and molds can be estimated from the equipment cost as:

$${{\mathrm{CO}}_2}_{\text{equipment}}={{\mathrm{CO}}_2}_1{C}_{\text{equipment}}+\sum _{i=4}^{5}m{p}_i{{\mathrm{CO}}_2}_i-\sum _{i=4}^5{\mathit{mp}}_i{{\mathrm{CO}}_2}_1{C}_i$$ (1)

wherein ${{\mathrm{CO}}_2}_1$ is the U.S. national average of 0.263 kg CO₂^[51] per USD for the U.S. gross domestic product (GDP), $C_{\text{equipment}}$ is the cost of the assessed equipment, $i$ is an index of the type of resource indicated in the left column of Table 3, $m$ is the mass of the assessed equipment, $p_i$ is the consumed resource's estimated mass fraction of the total equipment mass, ${{\mathrm{CO}}_2}_i$ is the amount of CO₂ produced per unit of resource, and $C_i$ is the USD cost per unit of resource. The latter two terms in equation (1) are implemented to avoid double counting of the carbon footprint of the materials that are inherently factored into the equipment purchase price. Model data for machines and molds are provided in Table 4 as implemented in the analysis.

TABLE 3. Cost and carbon footprint per unit of consumption with supporting references

Resource #, $i$	Description	USD [reference]	kg CO2 [reference]
1	1 USD	1.0	0.263^[^51^]
2	1 kWh	0.151^[^52^]	0.433^[^44^]
3	1 kg of PLA	2.00^[^53^]	1.47^[^54^]
4	1 kg of Steel	1.65^[^55^]	1.40^[^56^]
5	1 kg of Aluminum	3.85^[^57^]	12.9^[^58^]
6	1 h labor	24.97^[^50^]	6.57^[^50, 51^]

TABLE 4. Model data for machine and mold equipment

Item	Molding machine	Stock CR10	Upgraded ender 5	Cold runner mold	Hot runner mold
Purchase cost, USD	246,610	459	379	20,000	50,000
Mass, kg	4900	13.3	6.6	157	393
$p_{4=\text{Steel}}$	90%	20%	20%	100%	100%
$p_{5=\text{Aluminum}}$	10%	80%	80%	0	0
Estimated kg CO2	69,750	140.4	80.1	5420	13,540

3.2 Theoretical minimum energy consumption

Supplemental Materials Section S.3 provides a Matlab script for evaluating the theoretical minimum energy consumption driven by polymer melting and forming requirements. Therein, the minimum melting energy is modeled as:

$$E_{\text{melting}}=c_P\left(T_{\text{melt}}-T_{\text{room}}\right)+h_F,$$ (2)

wherein $c_P$ is the material's specific heat capacity, $h_F$ is the heat of fusion, and $T_{\text{melt}}$ and $T_{\text{room}}$ are the respective melt processing and room temperatures. The minimum theoretical heating requirements are then modeled as:

$$E_{\text{heating}}=f_{\text{heating}}{E}_{\text{melting}},$$ (3)

where $f_{\text{heating}}$ is the fraction of the melting energy that actively comes from heaters and is assumed to be 100% for 3D printing processes. In screw-based plastication processes, such as single screw extrusion, power monitoring indicated that half of the melting energy is generated by viscous dissipation (shear heating) during plastication with the rotating screw.^[60] In injection molding, $f_{\text{heating}}$ is estimated as 0.6 given greater heat transfer by conduction to the material in the screw channels since the screw is not rotating for the majority of the cycle time.

For both injection molding and 3D printing, the minimum energy required to form the material is:

$$E_{\text{flowing}}=V_{\text{part}}{p}_{\text{process}},$$ (4)

where $V_{\text{part}}$ is the volume of the part (m³) and $p_{\text{process}}$ is the average processing pressure (Pa). This relation is derived from the estimation of pumping power as the product of flow rate and pressure, wherein the pressure is driven by the viscosity of the material^[61] and feed system resistance. Typical pressures are 5 MPa for polymer 3D printing by material extrusion and 100 MPa for injection molding. Melt pressures are low for polymer 3D printing, given the short nozzle length and the use of external X, Y, and Z motors to attach the extrudate to the growing substrate. Conversely, melt pressures are high for injection molding as the viscous melt is forced through the relatively thin and long mold cavity that constrains and forms the material. Even so, the resulting energy input to form the material is very small compared to the heating requirements. For the 3D printers' motion axes, the minimum energy input is calculated based on the power required to maintain the stepper motor torque:

$$E_{3\mathrm{DP}\text{motors}}=t_{\mathrm{kg}}{v}_{\mathrm{ref}}{i}_{\max },$$ (5)

where $t_{\mathrm{kg}}$ is the time to print 1 kg of material, $v_{\mathrm{ref}}$ is the reference drive voltage for the stepper motor, and $i_{\max }$ is the maximum current for the stepper motor. This energy is added to $E_{\text{flowing}}$ to estimate the total servo motor energy for 3D printing.

For injection molding, the total minimum servo motor energy is:

$$E_{\text{servomotor}}=\left(1-f_{\text{heating}}\right)E_{\text{melting}}+E_{\text{flowing}}+E_{\text{moving mold}}.$$ (6)

Here, the first turn represents the energy required to turn the screw for viscous heating of the melt. The last term is the minimum motor energy to open and close the mold, estimated as:

$$E_{\text{moving mold}}=2L_{\text{travel}}{m}_{\text{mold}}{a}_{\text{mold}},$$ (7)

where $L_{\text{travel}}$ is the mold opening distance, $m_{\text{mold}}$ is the mass of the mold, and $a_{\text{mold}}$ is the uniform acceleration required to open and close the mold within 1 s for each direction.

4 RESULTS AND DISCUSSION

Results are presented for economic cost (USD) and environmental cost (kg CO₂) as a function of the production quantity of tensile specimens. While the focus on tensile specimens may seem academic, the described processes, costs, and related modeling techniques are typical of new product development programs for plastics applications. The presented techniques and results are extensible to specific applications by substituting mold/machine pricing, cycle times, and other model inputs as appropriate and parametrically investigated in Section 4.5.

4.1 SEC relative to theoretical minima

The SEC values of the investigated processes are indicated in the top subplot of Figure 5. The cold and hot runner molding, respectively, consume 1.28 and 0.929 kWh/kg. These values are on the same order of magnitude as processing surveys by Yoon et al.^[5] and Kanungo and Swan.^[62] Still, relative to the estimated minimum theoretical energy consumption described in Section 3.2, the cold runner and hot runner molding are respectively only 9.9% and 13.6% efficient. The cold runner molding is less efficient than the hot runner molding due to the additional heating and motor energy consumption associated with the longer molding cycle times (48 s vs. 26 s) and lower machine productivity.

FIGURE 5

Energy consumption of injection molding and 3D printing processes relative to theoretical minima including (top) absolute kWh per kg of part production, and (bottom) relative energy expended. Process efficiencies relative to theoretical minima are indicated by text above the bottom sub-plot.

By comparison with injection molding, 3D printing with the stock CR10S-Pro and upgraded Ender 5 respectively consume 16.5 and 5.28 kWh/kg. These results are significantly lower than the range of 23–346 kWh/kg reported by Yoon et al.,^[5] the 144 kWh/kg for PC printing reported by Baumers et al.,^[20] and the 50.4 kWh/kg for PLA printing reported by Enemuoh et al.^[36] The primary reason is that the described methodology produced the tensile specimens at the maximum melting rate of the 3D printers to thereby reduce the printing time and energy as recommended by Ajay et al.,^[23] Fok et al.,^[24] Peng,^[26] Hernandez et al.,^[29] and Espach et al.^[38] Even with this improved 3D printing performance, energy use is substantially greater than injection molding. The bottom subplot of Figure 5 indicates that the CR10S-Pro and upgraded Ender 5 respectively operate at 2.86% and 3.81% efficiency, with the majority of the variance associated with heated bed and motor inefficiencies. These results suggest that performance may best improved by using smaller 3D printers that have smaller heated beds and motors. The results also support the recommendation by Hopkins et al.^[31] to minimize processing temperature as well as the recommendation by Ajay et al.^[23] for 3D printer manufacturers to implement improved motion controls (e.g., brushless DC motors with worm gears) to reduce energy consumption when idle.

4.2 Breakeven quantities to recoup mold investment

New product development (NPD) requires a product to be designed for a target manufacturing process to ensure the product is fit for use.^[63] As such, identification of the most appropriate process is of strategic importance with long-term ramifications for evolving core competencies and supply chains. Injection molding has historically been a preferred process for its ability to economically produce complex part geometries with excellent surface replication and good tolerances. However, the upfront investment in a mold and its ongoing maintenance costs are significant disincentives that companies increasingly wish to avoid if possible.

Figure 6 provides a plot of the USD cost per part as a function of the production quantity given the described models and parameters in Tables 3 and 4. The stock CR10S-Pro and upgraded Ender 5 3D printers provide a constant part cost of 0.849 and 0.344 USD, respectively. These results imply the importance of printer selection and adaptation. The Ender 5 is a smaller and lower cost printer that, when upgraded, provides much faster printing and thus lower costs. Accordingly, the upgraded Ender 5 would always be preferable to the CR10S-Pro for all applications not requiring the CR10S-Pro's larger build size.

FIGURE 6

USD Cost per part as a function of production quantity for cold runner injection molding, hot runner injection molding, stock CR10 3D printing, and upgraded Ender 5 3D printing. Cross-over from 3D printing to injection molding occurs near 79,000 parts, with the underlying cost factors shown in Figure 7.

Compared to part production with the upgraded Ender 5, cold runner molding becomes preferable around 79,000 units. The cold runner and hot runner molding have much higher part costs at lower production quantities given higher amortized mold costs. The hot runner provides improved molding productivity with lower wastage but these benefits are not economically realized until production of 900,000 units given the higher upfront cost of the hot runner mold. where both cold and hot runner molding result in an average part cost of 0.109 USD. As such, the cold runner mold would generally be preferable at lower production quantities.

The cost drivers for injection molded and 3D printed products are extremely different, as shown in Figure 7. At a production quantity of 79,000 units, the amortized cold runner mold costs are 0.252 USD per part and much greater than the labor, machine, and material contributions for the molding processes. By comparison, the labor costs in 3D printing with the upgraded Ender 5 are 0.306 USD, dominating the part cost of 0.344 USD even though the modeling assumes only one operator per 40 3D printers. This 0.306 USD of labor cost represents a touch time of 44 s per part given the burdened hourly labor rate of 24.97 USD^[50] and suggests that labor costs are likely to be a continuing barrier to 3D printing adoption.

FIGURE 7

Breakout of USD cost per part at a production quantity of 56,234 parts for cold runner injection molding, hot runner injection molding, stock Creality CR10S-Pro 3D printing, and upgraded Creality Ender 5 3D printing.

4.3 Carbon footprints

Energy costs are surprisingly low for injection molding, representing 0.002 USD per tensile specimen, and still relatively small for the 3D printed parts (0.0249 and 0.0080 USD for the CR10S-Pro and upgraded Ender 5, respectively). However, energy consumption has a dominating effect on carbon footprint. Figure 8 plots the average kg CO₂ per part as a function of the production quantity. The general behaviors are similar to those for the economic costs, but the breakeven quantity is reduced to 63,000 from 79,000 parts given the higher SEC of the 3D printing processes since energy carries a relatively high carbon footprint of 0.433 kg CO₂ per kWh.

FIGURE 8

kg CO₂ cost per part as a function of production quantity for cold runner injection molding, hot runner injection molding, stock Creality CR10S-Pro 3D printing, and upgraded Creality Ender 5 3D printing. Cross-over from 3D printing to injection molding occurs near 14,000 parts, with a breakout shown in Figure 9.

In Figure 8, parts produced with the CR10S-Pro and upgraded Ender 5, respectively, have a carbon footprint of 0.305 and 0.122 kg CO₂ per part. The injection molded parts have a much higher carbon footprint at lower production quantities given that the cold runner and hot runner molds have estimated carbon footprints of 5420 and 13,540 kg CO₂ that must be amortized. With respect to minimizing carbon footprint, the upgraded Ender 5 is preferable up to about 63,000 units, then cold runner injection molding is preferable to 450,000 units, after which hot runner molding is preferred. At high production volumes, the hot runner mold's large carbon footprint is overcome given its high efficiency with respect to molding productivity, low resulting energy consumption, and efficient material use.

Figure 9 provides a breakout of the carbon footprint at a production quantity of 63,000 units, including material, energy, mold, machine, and labor contributions. Labor and energy consumption are the dominating drivers of the carbon footprint for the 3D printers with the upgraded Ender 5, which far outperforms the stock CR10S-Pro due to its smaller size, higher melting capacity, and thus improved productivity with respect to SEC. At the 63,000 breakeven quantity, the carbon footprint of the upgraded Ender 5 is essentially equal to the carbon footprint of the amortized cold runner mold and the energy consumption of the cold runner molding process.

FIGURE 9

Breakout of kg CO₂ per part at a production quantity of 56,234 parts for cold runner injection molding, hot runner injection molding, stock CR10S-Pro 3D printing, and upgraded Ender 5 3D printing.

The carbon footprints of the processed materials are relatively uniform though the cold runner molding is 37% greater due to wastage associated with processing of the cold runners. The carbon footprint of the machinery is relatively consistent and relatively small, though this result assumes full utilization of 4000 hours per year across the lifetime of the equipment. Decision makers are encouraged to perform sensitivity analysis across utilization levels as well as production quantities in assessing preferred processing strategies with respect to minimization of economic cost and carbon footprint.

4.4 Dependence on carbon tax incentives

Carbon taxes have been proposed to account for carbon emission externalities associated with climate change.^[64] In this accounting paradigm, firms are charged for each unit of carbon emissions so as to guide economic decisions that indirectly have a global impact.^[65] While research has shown that consumers prefer low-carbon products if identifiable,^[66] explicit accounting of carbon emissions in production costs can provide improved decision support. In particular, new product development decisions will be better supported on a global basis with consistent carbon cost accounting that improves supply chain visibility.^[67]

Given that, on average, the U.S. economy outputs 0.263 kg CO₂ per USD,^[51] the carbon footprint can be assessed as having a literal economic equivalent of 3.80 USD per kg CO₂. As such, the ESG cost, including the direct cost and indirect cost of CO₂ is:

$${\mathrm{USD}}_{\mathrm{ESG}}=\mathrm{USD}+{\mathrm{CO}}_2/\left(0.263\mathrm{kg}{\mathrm{CO}}_2\mathrm{per}\mathrm{USD}\right)$$ (8)

where $\mathrm{USD}$ is the unburdened cost per part and ${\mathrm{CO}}_2$ is the kg of CO₂ produced per part. Figure 10 provides the burdened cost per part, including this carbon tax, as a function of the production quantity. The general behavior of the burdened cost follows that of the unburdened cost (Figure 6), with a slight reduction in the breakeven quantity to 71,000 units for injection molding given the carbon footprint behavior (Figure 8) that penalizes 3D printing's high SEC as previously described.

FIGURE 10

Total cost per part including cost of CO₂ production assuming the US average of 0.263 kg of CO₂ per USD output. Relative to Figure 6, costs are increased with a reduction in breakeven quantity to 74,000 units given the CO₂ production of Figure 8.

The impact of a full carbon tax is a dramatic shift of the cost upward by 135%. Specifically, the tensile specimen that is 3D printed with the upgraded Ender 5 had an economic cost of 0.343 USD with 0.122 kg CO₂, which results in a burdened cost of 0.807 USD including the modeled carbon tax. At a production quantity of 600,000 units, hot runner molding becomes most economical, with a total cost of 0.329 USD driven by direct costs of 0.138 USD and 0.0504 kg CO₂. While the foregoing analysis assumes the literal equivalence of 3.80 USD per kg CO₂, the question remains as to what levy governments will apply and how consumers or producers will react. As such, the following section evaluates break-even quantities with a parametric evaluation of the carbon tax and other application-specific parameters.

4.5 Dependence on application-specific parameters

While the described analysis is specific to the production of tensile specimens, the structure of the analysis is extensible with changes to model coefficients and inputs. To investigate the behavior of the model with respect to the most significant model parameters, a series of parametric analyses were conducted in which the 3D printer cost, molding machine cost, mold cost, part size, labor cost, and carbon tax were varied from 10% to 1000% of the reference model for the tensile specimen. Figure 11 plots the breakeven quantities at which point the ${\mathrm{USD}}_{\mathrm{ESG}}$ per equation (8) for cold runner molding becomes preferable to 3D printing with the upgraded Ender 5 as a function of the relative changes in the model parameters. When all model factors are set to one, the breakeven quantity from 3D printing with the Ender 5 to cold runner molding is 74,000 units, as described in the prior section.

FIGURE 11

Breakeven quantities at which cold runner molding is preferred to 3D printing with the upgraded Ender 5 as a function of parametric variation of six model inputs including 3D printer cost, molding machine cost, mold cost, part size, labor cost, and carbon tax.

The behavior of each of the model inputs is demonstrated by the various curves. The breakeven quantity is relatively constant as a function of the 3D printer cost factor since the 3D printer cost is a small contributor to the total cost. By comparison, the cost of the molding machine and mold are significant determinants of the total cost. As such, higher breakeven quantities result for molding with increases in the parametric machine and mold cost factors. The breakeven quantity is a linear function of the mold cost, which motivates the strategy to reduce mold costs using modular mold bases and lower cost mold inserts.^[68] The molding machine cost factor is less linear since the hourly rate of the machine is computed based on the total life of the machine and not on an application-specific basis.

Part size and labor costs are inversely proportional to the breakeven cost. At the same print resolution, larger parts will require proportionally longer time to 3D print, whereas injection molding cycle times are driven by wall thickness.^[45] As such, the breakeven quantity will tend to decrease from 1,040,000 units for a part weighing 2.5 g–6200 units for a part weighing 105 g. Thus, larger parts are less likely to be 3D printed by polymer extrusion until greater melting capacity and/or parallel printing within a single printer are available to reduce print time and improve energy efficiency. Labor costs also have a significant effect on breakeven quantity. At a labor cost of 2.5 USD/h, cold runner molding for the reference tensile specimen only becomes competitive at production quantities above 1,560,000 units. Increased labor costs quickly then decrease the breakeven quantity to justify the mold investment given the much faster cycle times and lower labor content of injection molding.

Lastly, the carbon tax factor has a significant impact on the total burdened cost per quation (8) but a relatively little impact on the breakeven quantities. The reason is that the direct cost behavior plotted in Figure 6 is quite similar to the CO₂ cost plotted in Figure 8. Still, as the carbon tax increases, injection molding becomes preferable at lower production quantities given 3D printing's higher energy consumption and the associated higher carbon tax burden.

4.6 Dependence on quality levels

The results thus far have assumed that all processes provide the same quality levels. However, we recognize that this is an invalid assumption and that 3D printing by material extrusion will tend to provide lower surface finishes, looser tolerances, lower stiffness, and lower strength than molded parts. The reason is that the 3D printing process by material extrusion relies on the deposition of roads having rounded sides that result in a limited contact area between roads as well as potential void inclusion.^[69]

The quality of the processes was assessed with respect to the consistency of the part width and thickness, the part yield stress, and the apparent elastic modulus along the length of the tensile specimen; lengths of the specimen were not measured as part of the mechanical testing protocol. For each of the quality attributes, a lower specification limit (LSL) was assessed. For the part dimensions, the LSL was set to 99.6% of the nominal dimension, as consistent with a standard dimensional tolerance of ±0.4%.^[70] The LSL for the yield stress and elastic modulus was assessed as 50% of the average value for the injection molded samples, which corresponds to a factor of safety of two.

The part quality measurements and statistics are provided in Table 5. In the table, the process capability index, $\mathrm{CP}$, is evaluated as $\mathrm{CP}=\left(\mu -\mathrm{LSL}\right)/\left(3\mathrm{SD}\right)$ given the mean, $\mu$, and the $\mathrm{SD}$. A process capability index of 1 corresponds to a centered manufacturing process with three SD to the closest specification limit and an expected yield of 99.87%.^[71] The results indicate that injection molding is nominally capable with respect to the part dimensions but outstanding with respect to the consistency of the structural properties. An increased sample size and improved part metrology would be necessary to precisely characterize the process capability, which was not the primary focus of this study. Regardless, the injection molding results provide a useful reference for the evaluation of the 3D printed samples.

TABLE 5. Part quality measurements and analysis results

Process	Specimen number	Part length, mm	Part width, mm	Part thickness, mm	Yield stress, MPa	Elastic modulus, MPa
Injection molding	3	91.3	13.00	3.24	50.30	2445
	5	91.3	13.06	3.29*	47.62	2111
	7	91.3	13.04	3.25	50.30	2420
	9	91.3	13.03	3.25	51.08	2429
	11	91.3	13.00	3.25	48.45	2379
	Mean ± SD	91.3 ± n/a	13.03 ± 0.03	3.256 ± 0.02	49.55 ± 1.45	2357 ± 139.6
	LSL	90.93	12.97	3.24	25.15	1222.50
	CP	n/a	0.842	0.868*	5.612	2.709
Stock CR10, vertical specimen	1	61.3	13.18	3.73	11.8	1515
	2	61.3	13.18	3.69	16.30	1406
	3	61.3	13.19	3.77	13.37	1450
	4	61.3	13.2	3.64	13.18	1564
	5	61.3	13.19	3.68	16.37	1508
	Mean ± SD	61.3 ± n/a	13.19 ± 0.01	3.70 ± 0.05	14.20 ± 2.04	1487 ± 61.38
	LSL	61.05	13.14	3.69	25.15	1222.50
	CP	n/a	2.102	0.099	−1.791	1.445
Stock CR10, horizontal specimen	1	63.04	13.08	3.53	35.89	1931
	2	63.04	13.10	3.58	35.93	1827
	3	63.04	13.10	3.57	35.4	1877
	4	63.04	13.05	3.59	34.95	1833
	5	63.04	13.07	3.60	34.24	1796
	Mean ± SD	63.04 ± n/a	13.08 ± 0.02	3.57 ± 0.03	35.28 ± 0.71	1853 ± 52.40
	LSL	62.79	13.03	3.56	25.15	1222.50
	CP	n/a	0.822	0.176	4.777	4.009
upgraded Ender 5, horizontal Specimen	1	63.17	13.03	3.57	46.97	2161
	2	63.17	13.38	3.65	47.27	2006
	3	63.17	13.22	2.36	38.21	2048
	4	63.17	13.1	3.57	48.49	2138
	5	63.17	13.43	3.71	44.07	2112
	Mean ± SD	63.17 ± n/a	13.23 ± 0.17	3.37 ± 0.57	45.00 ± 4.12	2093 ± 64.43
	LSL	62.92	13.18	3.36	25.15	1222.50
	CP	n/a	0.102	0.008	1.603	4.504

• The asterisk “*” indicates an outlier at the 95% confidence level.

Samples were produced with the tensile specimen oriented in the print direction for the stock CR10S-Pro and upgraded Ender 5, as well as in the vertical direction for the stock CR10S-Pro. The results generally indicate that the structural properties of the printed parts are acceptable except for the part strength in the Z direction, which is a commonly known issue. However, the printed parts also exhibited greater dimensional variation, especially in the thickness direction, that could preclude acceptance of the 3D printed parts without finishing machining^[72] and/or polishing.^[73] Such post-processing is possible, but it further increases the cost of the 3D printing processes. Accordingly, the new product development efforts must contemplate the suitability of 3D printing processes in production applications from both cost and quality perspectives.

5 CONCLUSIONS

The selection of production processes is of strategic importance in new product development, with a significant impact on cost structures as well as evolving core competencies and supply chains. Three primary conclusions can be drawn from this work. First, regarding SEC relative to theoretical minima, 3D printing is only 3% efficient but can be improved by reducing the size of the printer, increasing melting capacity to print faster, printing at lower temperatures and in insulated enclosures, and otherwise implementing more efficient motion controls. Second, regarding breakeven quantities at which cold and hot runner molds are justifiable, the results support the long-standing consensus that injection molding is a preferred choice for complex parts having significant production quantities, for example, above 50,000 units. Still, the costs of 3D printing equipment and materials have dropped significantly during the past 10 years, with further technological advances continuing to increase the speed and quality of 3D printing to support higher production quantities across more applications. Third, the carbon footprint was driven by the consumed materials and energy per part. Because of molding's better efficiency than 3D printing, injection molding becomes more preferable with increasing carbon taxes. Mold cost and part size were the most significant determinants of the breakeven quantity, with 3D printing preferable for smaller parts and more expensive molds. Consistent with prior research,^{[23, 24, 26, 29, 38]} the competitiveness of 3D printing is largely limited by the long print times that would require unacceptably long lead times or large print farms in order to meet production requirements.^{[4, 10, 65, 67]}

In conclusion, the authors estimate that 20% of molded plastic products are potentially addressable by AM with constraints related to part size, feature/surface detail, and strength requirements. While research in volumetric printing is proceeding,^[74] the results of this analysis suggest that injection molding will remain more economical and sustainable for high volume applications until AM's high energy consumption is reduced. The suitability of the candidate processes relative to application-specific quality requirements must also be considered.

ACKNOWLEDGMENTS

The authors acknowledge financial support from the National Science Foundation (NSF) under Grant Number 1914651. The authors would also like to thank NSF/GOALI partner Stratasys as well as Milacron Inc. for the consignment of the Roboshot molding machine and Mold-Masters Ltd. for the consignment of the hot runner system. The NSF, Milacron, and Mold-Masters did not contribute to the performance of this research or the authoring of this article.

CONFLICT OF INTEREST

The authors declare no competing interests.

Open Research

DATA AVAILABILITY STATEMENT

Data and implemented analyses are provided in the supplementary information.