Comparative analysis of ventilation efficiency on ultrafine particle removal in university MakerSpaces

Abstract

The proliferation of 3D printing MakerSpaces in university settings has led to an increased risk of student and technician exposure to ultrafine particles. New MakerSpaces do not have standardized specifications to aid in the design of the space; therefore, a need exists to characterize the impacts of different engineering controls on MakerSpace air quality. This study compares three university MakerSpaces: a library MakerSpace operating ≤4 devices under typical office space ventilation with no engineering controls, a laboratory MakerSpace operating 29 printers inside grated cabinets, with laboratory-grade ventilation, and a center MakerSpace operating ≤4 devices with neither engineering controls nor internal ventilation. All MakerSpaces were studied under both controlled (using a standard print design) and uncontrolled (real-time user operation) conditions measuring emitted particle concentrations in the near-field. Additionally, volatile organic emissions and the difference between near-field and far-field particle concentrations were investigated in multiple MakerSpaces. The center MakerSpace had the greatest net increase in mean particle number concentration (+1378.9% relative to background during a print campaign using polylactic acid (PLA) filament in a MakerBot (MakerBot-PLA)). The number-weighted mean diameter had the greatest change relative to background during the library campaign, +37.1% for the Lulzbot-PLA and −56.1% for the Ultimaker-PLA studies. For the standard NIST design with MakerBot-PLA, the laboratory’s particle removal ratio was 30 times greater than in the library with open cabinets and 54 times greater when the cabinet doors were closed. The average particle removal rate from the center MakerSpace was up to 2.5 times less efficient than that of the library for the same MakerBot-PLA combination. These results suggest ventilation as a key priority in the design of a new university MakerSpace.

1. Introduction

The proliferation of university MakerSpaces parallels the increasing role of additive manufacturing in industry, research, and entertainment (Farritor, 2017; Herron and Kaneshiro, 2017; Krummeck and Rouse, 2017; Lotts, 2017). These MakerSpaces are a place to incorporate classroom theory within a hands-on, creative environment for the design and building of prototyped items, particularly through 3D printing (Forest et al., 2014). Since university MakerSpaces are not standardized, facilities vary widely in ventilation capability, user traffic, and background (ambient) airborne particle concentrations. Human exposure to 3D printer emissions has been a growing concern (Pelley, 2018), but the variability of university MakerSpaces makes it challenging to characterize student/technician exposure risks. To date, studies of 3D printer emissions have explored the effects of filament type and color (Azimi et al., 2016; Floyd et al., 2017; Vance et al., 2017; Yi et al., 2016), printing temperature (Stabile et al., 2017), printer type (Azimi et al., 2016), vicinity from printer (Zontek et al., 2017), and print environment (Vance et al., 2017). Most of these emission factors were studied within highly-controlled print environments (Azimi et al., 2016; Floyd et al., 2017; Yi et al., 2016) with fewer than ten printers in simultaneous operation (Bharti and Singh, 2017; Patel, 2016).

Recent studies investigated typical MakerSpace operation, for which not all experimental variables can be controlled. Patel compared ultrafine particle concentrations between a university library MakerSpace and a commercial print shop during uninterrupted facility operation (Patel, 2016). However, different filaments were used at each location and the MakerSpaces had only at most six printers operating simultaneously (Patel, 2016). McDonnell et al. compared seven university 3D printing environments, one of which contained 18 MakerBot Replicator devices. This latter study was also conducted during typical MakerSpace operation, but did not attempt to quantify the relationship between facility ventilation and particle concentration (Mcdonnell et al., 2016).

To begin filling this gap, Azimi et al. simulated ultrafine particle (UFP) and volatile organic compound (VOC) emissions in a model office for various printers, filaments, and engineering controls. The simulated control strategies included (1) MERV 16 filters with activated carbon inside the central air handling units; (2) portable air cleaners with (2a) 100 m³/h and (2b) 300 m³/h clean air delivery rates; (3) spot ventilation located 1.5 m above the printers with (3a) 90 m³/h, (3b) 360 m³/h, and (3c) 1800 m³/h flow rates; and (4) a sealed enclosure with internal air recirculation/filtration. For a constant printer/filament combination, the 1800 m³/h spot ventilation and the sealed enclosure were the most effective controls (Azimi et al., 2017). Such simulations, if supported with experimental data, would help to develop guidelines for effective particle emission control.

Here we present a comparative study quantifying the efficacy of room ventilation in three university MakerSpaces, in order to understand experimental discrepancies that arise when comparing data collected from different MakerSpaces. To the authors’ knowledge, this is the first assessment that attempts to simultaneously include:

1. Time-dependent concentrations and sizes of airborne particles.

2. Normalized quantitative comparison of particle removal efficiency in three MakerSpaces.

3. Assessment of a large-scale MakerSpace facility (up to 29 printers operating simultaneously).

4. Both uncontrolled and controlled print campaigns within the same facility, to characterize the effects of real-time operation with minimal interference.

2. Methods

2.1. Library MakerSpace

The layout of the library MakerSpace is shown in Fig. 1. Particle count data were collected for three 3D printers (MakerBot Replicator 5th Generation, Ultimaker 2, and Lulzbot TAZ 5) and one laser cutter (Epilog Zing with Purex Xbase 400 fume extraction system). Only white acrylonitrile butadiene styrene (ABS), high impact polystyrene (HIPS), or PLA filaments were used with the library MakerSpace 3D printers. All library print campaigns utilized the NIST Additive Manufacturing Test Artifact print object (Moylan et al., 2012). The NIST design was printed at the same scale (10 cm × 10 cm × 1 cm) for all printers and same print speed (80 mm/s) for all MakerBots. The print speed for the Ultimaker 2 was 70 mm/s and Lulzbot TAZ 5 was 50 mm/s. The 3D printers were first tested individually, then simultaneously with the laser cutter in operation. Data was collected at both near-field (collection at 1 m from devices) and far-field (collection greater than 1 m from devices, placed in center of room) for each print campaign. For two tests, air sampling using gas canisters were used and sent to an external lab to screen for a panel of volatile organic compounds using EPA Method TO-15, (US Environmental Protection Agency, 1999). Emitted particles were captured on formvar/carbon 200 Cu mesh placed on top of a glass fiber filter set within a 37 mm cassette operated at a flowrate of 10 L/min. The grids were then analysed for particle size, size distribution and shape by TEM (Zeiss Libra 120).

Fig. 1.

Library MakerSpace layout. A. Schematic of MakerSpace. (MB = MakerBot; UM = Ultimaker; LB = Lulzbot; LC = Laser Cutter; 1 = Near-Field Measurement Location; 2 = Far-field Measurement Location; R = Air Return Location; S = Air Supply Location) B. MakerSpace with particle sizers in the foreground.

2.2. Laboratory MakerSpace

The laboratory MakerSpace layout is illustrated in Fig. 2. Laboratory studies were conducted with 27 MakerBot and 2 Ultimaker 3 Extended Printer printers in simultaneous operation, using the filament colors pre- loaded in each printer (Table S2) and same print specifications as the library study. Two “uncontrolled studies” were performed by measuring particle concentrations at near-field during normal classroom operation, when cabinets were closed as per established classroom procedures. These uncontrolled studies were based on print objects created by students, as opposed to the standard NIST Test Artifact. Two additional “controlled” studies were performed by measuring particle concentrations with the cabinets fully opened while printing the NIST Test Artifact. VOC data was not successfully obtained due to instrumentation failure. Particles were collected on formvar/carbon 200 Cu mesh for qualitative sample analysis by TEM using the same procedure described in section 2.1. The cassette housing the TEM grid was placed at the “Near Field Measurement Location” (1; Fig. 2a) adjacent to cabinet C5. Particle collection on TEM grids continued for the full duration of a single near-field MakerBot particle emission study. TEM samples were compared against the control TEM grid collected in an atrium, which had no known particle emission sources and was therefore representative of typical, ambient conditions absent of maker nanoparticle contaminants.

Fig. 2.

Laboratory MakerSpace layout. A. Schematic of MakerSpace. Experiments performed in rightmost room where C=Cabinet; 1 = Near-Field Measurement Location; 2 = Far-field Measurement Location; R = Air Return Location; S = Air Supply Location. B. Laboratory MakerSpace with cabinet-enclosed printers.

2.3. Center MakerSpace

The center MakerSpace layout is illustrated in Fig. 3. Studies were conducted for two 3D printer types (MakerBot Replicator 5th Generation and Up Box +), using the standard filament and colors pre-loaded in each printer. All center print campaigns utilized the NIST Additive Manufacturing Test Artifact print object, printed to the same specifications as described in 2.1 with an UpBox + print speed at 60 mm/s. The 3D printers were first tested individually, then all four simultaneously with and without a portable HEPA filter/activated carbon filter system (Sentry Air Systems, Inc.) directed toward the Up Box + printer exhaust. Particle count data was collected at near-field (approximately 27 cm from devices) for each print campaign. Single print runs were collected for each scenario listed due to limited availability of the MakerSpaces.

Fig. 3.

Center MakerSpace layout. A. Schematic of MakerSpace. (MB = MakerBot; UB = UpBox+; 1 = Near-field Measurement Location) b. Center MakerSpace with particle sizers on left side.

2.4. Ventilation and emission rate estimations

The experimental parameters for each print campaign are outlined in Table 1. The air changes per hour (ACH) for the library and laboratory MakerSpaces were calculated from room dimensions and HVAC system specifications (Table S1), provided by the Virginia Commonwealth University Engineering and Utilities Department. ACH values were calculated from equation (1):

$$ACH\left(h^{-1}\right)=\frac{Exhaustairflow\left({\text{m}}^3/min\right)}{RoomVolume\left({\text{m}}^3\right)}\times 60$$ (1)

Table 1.

Experimental parameters for library and laboratory print campaigns. “MakerBot” refers to MakerBot 5th Generation Replicator device.

Parameter	Library MakerSpace		Laboratory MakerSpace*		Center MakerSpace
	Single Device	Simultaneous	Uncontrolled Studies	Controlled Study	Single Device	Simultaneous*
ACH	3.125 h−1		8.7 h−1		0.18 h−1
Devices	MakerBot	MakerBot	MakerBot (26)		MakerBot	MakerBot (3)
	Ultimaker 2	Ultimaker 2	MakerBot Replicator+ (1)		Up Box +	Up Box +
	Lulzbot TAZ 5	Lulzbot TAZ 5 Epilog Zing Laser Cutter	Ultimaker 3 Extended (2)
Filament Type	white PLA white ABS white HIPS (Lulzbot)	white PLA white ABS white HIPS clear acrylic (Laser Cutter)	PLA *		PLA ABS (Up Box)	PLA ABS (Up Box)
Print Object	NIST Test Artifact		Class Projects	NIST Test Artifact	NIST Test Artifact
Print Duration (Approximate)	4 h		2–22 h	4 h	4 h
Measurement Vicinity	Near-Field (NF) Far-Field (FF)		Near-Field (NF)	Far-Field (FF)	Near-Field (NF)
VOC Detectors	MultiRAE Gas Monitor		N/A	N/A	N/A
	MIRAN SapphIRE
	Time Integrated Canister Air Sampling
Particle Sizers	TSI OPS 3330 (NF)		TSI OPS 3330	TSI OPS 3330	TSI OPS 3330
	TSI NanoScan 3910 (NF) Graywolf PC-3016A (FF)		TSI NanoScan 3910 (Both NF and FF)	TSI NanoScan 3910 (Both NF and FF)	TSI NanoScan 3910 (Both NF)

^*See Table S2 for complete color listing.

^ǂWith or without filter.

For all studies, the size distribution of emitted particles was measured using a scanning mobility particle sizer (TSI NanoScan 3910) and optical particle sizer (TSI OPS 3330). In the library study, the emitted mass concentration was monitored using a Graywolf PC-3016A instrument. These size distributions were used to evaluate ventilation and emission rate estimations. For each print campaign, at least 5 min of background data were collected prior to print start. These background measurements allowed for determination of baseline air quality. Data for particle concentrations during printing were considered valid for the time period during which devices were in active operation. Table S3 lists the timeframes for background and experimental data collection for each print campaign.

A lumped loss parameter was determined following the work of Stephens et al., (2013) and a well-mixed room (Stephens et al., 2013; Wallace et al., 2004). Briefly, after prints were completed and particle concentrations began to decay toward background concentrations the loss rate was determined using a log-linear plot of time-varying data (Stephens et al., 2013).

$$\ln \left(\frac{C(t)-C_{meanbarkground}}{C(t=0)-C_{meanbackground}}\right)=-\beta t$$ (2)

Equation (2) relates the particle concentrations in #/cm³ during the decay period, beginning at t = 0, to the average background concentration measured prior to printing. This loss parameter, β, accounts for the combined removal of air filtration, ACH, and deposition on indoor surfaces and is assumed to be constant during the print durations. Then this parameter is used to determine the time-varying emission rate, following Floyd et al. (2017), equation (3). Key assumptions for the calculation of emission rate using this equation include: 1) signficant losses occur within the chamber, 2) significant losses occur due to outflow of air through the system, 3) influx of particles into the chamber not from the emission source are negligible (Byrley et al., 2019). These assumptions account for key parameters of particle losses within the system while maintaining simplicity of calculations.

$$E_n=\frac{V\left(C_n-C_{n-1}\right)}{t_n-t_{n-1}}+\beta V{C}_{n-1}$$ (3)

A comparison of emitted particles was performed using equation (4), determining the difference between expected emitted particles and the number of emitted particles compared to the mean background concentration. This equation takes into account removal effects through the lumped loss parameter; however, it does not account for spatial variation in concentration or additional particle dynamics. Additionally, mean values were used therefore peaks values due to printer malfunction are not accounted for.

$$ParticlesRemoved=E_{mean}t-V\left(C_{meanprint}-C_{meanbackground}\right)$$ (4)

3. Results

3.1. Library MakerSpace

Individual 3D printers were first characterized using white PLA and the NIST standard printing artifact, in order to determine particle concentration per printer type (Fig. 4a). The highest-emitting printer was the Ultimaker, producing peak concentrations over three times greater than those of the Lulzbot and MakerBot. There was an observable difference in print quality between instruments, and it is possible that the increased emissions resulted from printer malfunction. When operated simultaneously, the particle concentration increased beyond a peak concentration of 50,000 particles/cc. In the near-field, only the Lulzbot had a decrease in mean concentration during printing relative to background. Additionally, the Lulzbot increased the mean particle diameter during printing nearly 40%.

Fig. 4.

Library MakerSpace Study Measurements. A. Time-dependent concentration data for PLA studies in library MakerSpace. B. Time-dependent concentration data for varying filaments in the Lulzbot printer. C. Time-dependent concentration and size data for “Simultaneous” near-field library MakerSpace study.

The Lulzbot 3D printer allowed for filament variation as part of this study. Three filaments were printed: HIPS, ABS, and PLA (Fig. 4b). Initial inspection suggests that the HIPS filament results in the highest particle number concentration. However, the HIPS filament has a high initial background concentration (29,802 particles/cc), nearly 5 times greater than the PLA and ABS background concentrations. Increases in mean concentration were observed for both the ABS and HIPS prints, 5.4% and 31.6%, respectively. The HIPS print decreased particle size slightly from 36 nm to 33 nm, suggesting the increase in concentration is due to smaller particles.

To compare near-field versus far-field data for the library MakerSpace, analysis was restricted to particles with diameters >300 nm due to the rating of the Graywolf PC-3016A used for far-field measurements. As shown in Table 2, near-field data obtained with the TSI Nanoscan SMPS 3910 and TSI OPS 3330 were based on particle number concentrations, whereas far-field data obtained with the Graywolf PC-3016A were based on particle mass concentrations. The printer and filament type impacted the particle sizes introduced. At all near-field measurements, the mean diameter decreased an average of 24% for particles larger than 300 nm. Far-field measurements indicated an overall increase in particle diameter for particles larger than 300 nm, with the exception of the Lulzbot PLA print. For simultaneous printing of PLA, the average particle diameter is higher for the background measurements than during the operation (Fig. 4c). There is a simultaneous drop in particle diameter and spike in particle concentration at ~150 min, suggesting that the particles emitted by the 3D printers are substantially smaller than particles during background measurements prior to print start. This is confirmed through the analysis of particle concentration based on size, the background weighted mean diameter was 50 nm and decreased, over 40%, to 29 nm during printing.

Table 2.

Near-field and far-field concentration and diameter data for library MakerSpace for particles >0.3 μm

Campaign	Background				During Print				% Δ During Print
	Concentration (#/cc)			Weighted Mean Diameter (μm)	Concentration (#/cc)			Weighted Mean Diameter (μm)	Mean Concentration	Mean Diameter
	Mean	SD	Max		Mean	SD	Max
Library
Near-Field1
MakerBot (PLA)	107	191.66	838.36	0.43	21.02	6.30	86.43	0.33	−80.4	−23.1
Ultimaker (PLA)	15	35.6	164.88	0.526	4.27	0.37	5.25	0.36	−71.3	−29.8
Lulzbot (PLA)	20	23.85	102.03	0.47	16.25	12.60	167.32	0.38	−18.9	−19.7
Lulzbot (ABS)	18	0.57	19.64	0.35	17.45	22.18	228.56	0.34	−4.5	−3.0
Lulzbot (HIPS)	198	200.55	515.93	0.77	25.11	57.87	503.35	0.35	−87.3	−55.2
Simultaneous	11	24.53	159.49	0.43	7.91	18.85	168.03	0.37	−25.9	−13.5
Far-Field2
MakerBot (PLA)	46.90	15.82	76.54	5.99	7.16	8.16	58.46	6.56	−84.7	+9.5
Ultimaker (PLA)	32.37	17.56	81.54	5.85	4.07	5.23	36.76	7.31	−87.4	+25.0
Lulzbot (PLA)	53.14	27.39	161.23	6.18	18.25	13.25	86.36	5.69	−65.7	−8.1
Lulzbot (ABS)	18.31	8.27	42.61	6.54	4.92	3.70	24.69	6.82	−73.1	+4.3
Lulzbot (HIPS)	20.41	9.25	42.22	6.68	6.24	5.96	53.07	7.19	−69.4	+7.7
Simultaneous	58.64	42.21	201.16	6.94	8.57	8.15	49.55	7.67	−85.4	+10.5

¹Near-field concentrations are number concentrations (#/cc), and corresponding mean diameters are number-weighted.

²Far-field concentrations are mass concentrations (μg/m³), and corresponding mean diameters are mass- weighted.

VOC concentrations were determined for the laser cutter, Lulzbot printer, and the MakerSpace (Table 3). Measurable VOCs included 1,3- butadiene, ethanol, 2-propanol, acetone, acetonitrile, and ethyl acetate. Overall, highest values detected were for 2-propanol, at a concentration of 860 ppb during Lulzbot operation. However, the highest emitter was ethanol, which increased by 50% during Lulzbot operation and 20% during laser cutter operation. Mean values for all detected VOCs are less than 1% of available OSHA and ACGIH limits (Administration, O.S. and H., 2006; Hygienists, 2017).

Table 3.

Volatile organic compound (VOC) emission data for library MakerSpace.

VOC	Volume Concentration (ppbv)						Mass Concentration (μg/m3)
	Laser Cutter	Room	Lulzbot	Mean	OSHA	ACGIH	Laser Cutter	Room	Lulzbot	Mean	OSHA
1,3-Butadiene	8.8	7.1	5.3	7.1	1000	2000	20	16	12	16.0	-
Ethanol	71	58	90	73.0	1,000,000	-	130	110	170	136.7	1,900,000
2-Propanol	640	690	860	730.0	250,000	-	1600	1700	2100	1800	950,000
Acetone	41	36	29	35.3	1,000,000	250,000	100	85	68	84.3	2,400,000
Acetonitrile	22	38	35	31.7	40,000	20,000	37	63	58	52.7	70,000
Ethyl acetate	6.5	14	20	13.5	400,000	400,000	23	52	73	49.3	1,400,000
Total	790	840	1000	876.7	2,691,000	-	1900	2000	2500	2133	-

3.2. Laboratory MakerSpace

Two uncontrolled studies were performed, measuring near-field particle concentrations during normal use of the MakerSpace. The first study provided the greatest elevated particle concentrations, increasing by 131.2% during printer operation. However, all mean particle sizes were approximately 50–60 nm in diameter, decreasing slightly during printing for the first study (from 60 nm to 56 nm) and increasing slightly during the second study (from 48 nm to 58 nm). Opening the cabinets while performing far-field measurements provided similar particle concentrations to the closed cabinet, near-field measurements, Fig. 5a. As observed in Fig. 5b, the particle size during the open cabinet study was consistent after the initial increase in particle concentration, around 60 min. This decrease in size is confirmed as a shift from 47 nm to 36 nm in mean particle size.

Fig. 5.

Laboratory MakerSpace Study Measurements. A. Time-dependent concentration data for all laboratory MakerSpace print campaigns. B. Time- dependent concentration and size data for open-cabinet, far-field MakerBot- PLA study in laboratory MakerSpace.

3.3. Center MakerSpace

Individual testing of the MakerBot and UpBox + printers was performed prior to simultaneous operation (Fig. 6a). A single MakerBot printer provided greater peak concentrations (453,602 particles/cc) and therefore a greater increase in mean concentration during printing, approximately 13 times greater than background. However, the UpBox printer provided a more consistent particle concentration profile during printing. During simultaneous operation, the profile of the UpBox printer was observed within the overall particle concentration measurements (Fig. 6a). Through using a portable HEPA filter, the particle concentration was decreased during simultaneous operation compared to without (Fig. 6b) from a nearly 5000% increase in particle concentration to less than 2000%. Although the mean particle diameter decreased for each study, with an average of 41.1%, there appear to be times of aggregation, observed from 100 to 300 min in Fig. 6c where there is an increase in particle diameter.

Fig. 6.

Center Study Measurements. A. Time-dependent concentration data for all near-field center MakerSpace studies. B. Time-dependent concentration data for “Simultaneous” near-field center MakerSpace study with and without the portable HEPA filter. C. Time-dependent concentration and size data for near-field MakerBot-PLA study in center MakerSpace.

3.4. MakerSpace comparison

A comparison of particle number concentrations and number- weighted mean diameters across all print campaigns involving the entire particle size range (10 nm–10 μm) is presented in Table 4. Most (8 of 9) campaigns exhibited an increase in particle number concentration. Although the Lulzbot-PLA campaign exhibited a drastic decrease in particle number concentration during printing (59.1% decrease), the respective background concentration (7345 particles/cc) was impacted by nearby construction. The construction created the scent of wood smoke in the room, and the particles resulting from construction caused the background to be higher until the doors closed, and the air return/ supply cycles removed the contamination.

Table 4.

Particle concentration and diameter data for 10 nm to 10 μm particles in all MakerSpaces.

Campaign	Background				During Print				% Δ During Print
	Concentration (#/cc)			Weighted Mean Diameter (nm)	Concentration (#/cc)			Weighted Mean Diameter (nm)	Mean Concentration	Mean Diameter
	Mean	SD	Max		Mean	SD	Max
Library
Near-Field
MakerBot (PLA)*	1987	274	2846	100	2008	607	5960	68	+1.1	−31.4
Ultimaker (PLA)	1707	242	2706	70	12,426	5710	23,330	31	+627.9	−56.1
Lulzbot (PLA)	7345	442	8319	51	3007	996	6673	70	−59.1	+37.1
Lulzbot (ABS)	5560	771	7119	41	5860	2763	11,835	42	+5.4	+3.7
Lulzbot (HIPS)	29,802	2488	34,190	36	39,227	6609	52,372	33	+31.6	−9.3
Simultaneous	3148	388	4313	50	13,409	14,483	52,089	29	+316	−42.2
Laboratory
Open Cabinets
Near-Field*	3276	163	3718	56	5402	1123	9049	51	+64.9	−8.9
Far-Field *	1815	142	2169	47	3007	711	5416	36	+65.7	−23.9
Closed Cabinets
Near-Field 1	3149	74	3248	60	3459	791	7106	56	+9.8	−5.7
Near-Field 2	2194	92	2530	48	3733	723	6198	58	+70.1	+21
Center
Near-Field
MakerBot (PLA)*	1130	17	1156	58	16,704	61,487	453,602	37	+1378.9	−36.0
Up Box (ABS)*	2982	79	3153	58	15,237	7042	36,471	47	+411.0	−18.2
Simultaneous with Filter	1759	66	1835	76	32,585	18,448	95,423	29	+1752.5	−61.8
Simultaneous without filter	1452	19	1485	58	71,527	28,294	136,031	30	+4824.9	−48.4

^*Campaigns using the standard NIST Test Artifact print object.

Each MakerSpace housed a MakerBot 3D printer for PLA filaments, and utilizing the NIST printing artifact, the MakerSpaces can be compared (Fig. 7). The highest peak concentrations were observed within the center, whereas the library and laboratory MakerSpaces produced similar changes in mean concentrations (+1.1% and +65.7%, respectively). The peaks observed during the center study are due to misprinting or printer malfunction (Stefaniak et al., 2019). However, the library data reflects the concentration resulting from a single MakerBot, and the laboratory concentrations follow from up to 29 MakerBots in simultaneous operation. Therefore, the mean particle concentration per printer in the laboratory is far lower than the concentration per printer in the library. For example, at far-field with open cabinets, the laboratory’s concentration per printer is 94.8% lower than the library’s concentration (compare 104 particles/cc per printer for 29 printers, versus 2008 particles/cc for one printer). Particle sizes were generally larger in the library MakerSpace, yet a similar percent decrease in size was observed, 31.4% from 100 nm to 68 nm. Similarly-sized particles were measured within the laboratory and center MakerSpaces, with a mean background diameter of approximately 50 nm that decreased to 36 nm during printing.

Fig. 7.

Time-dependent concentration data for MakerBot-PLA studies using NIST Test Artifact print object.

TEM images of particle samples captured on formvar/carbon 200 Cu meshes were collected within the library (Fig. 8a) and laboratory (Fig. 8b) MakerSpaces. The separate control TEM grid collected in a large atrium space appeared clean. The collection within the library shows several, sharp, large, fiber-like structures. Since nanoparticles emitted during printing are <1 μm, these electron-dense, fiber-like particles will require further characterization to identify. Collections within the laboratory showed many small particles, ranging from 54 to 109 nm in diameter (8b2 and 8b3). In image 8b4, decoration of the particle with smaller particulates is observed.

Fig. 8.

TEM images from MakerSpace campaigns. A. Library MakerSpace. Print Campaigns: (1) Simultaneous; (2) Lulzbot-ABS; (3) Lulzbot-PLA; (4) Lulzbot- PLA. B. Laboratory MakerSpace. All from open-cabinet print campaigns, collected inside printer cabinets during operation.

Focusing on the MakerBot PLA print campaigns using the NIST Test Artifact, the particle removal ratios can also be compared. Both gave a net increase in particle number concentration and a net decrease in number-weighted mean diameter. Table 5 compares the effectiveness of particle removal between the different MakerSpaces, based on particle removal ratios calculated from equation (4). Using the library particle removal amount of 6.2 × 10¹³ particles as a basis, the laboratory MakerSpace was found to be nearly 30 times more effective at removing particles. The center MakerSpace, however, had up to 2.5 times worse particle removal effectiveness ratio.

Table 5.

Comparison of particle removal rates in all MakerSpaces.

Specifications	Library						Laboratory			Center
	MakerBot (PLA)	Ultimaker (PLA)	Lulzbot (PLA)	Lulzbot (ABS)	Lulzbot (HIPS)	Simultaneous	Open Cabinets (NF)	Open Cabinets (FF)	Closed Cabinet (NF 1)	MakerBot (PLA)	UpBox (ABS)	Simultaneous with Filter	Simultaneous without Filter
Room Volume (m3)	130	130	130	130	130	130	198	198	198	37	37	37	37
Print Duration (h)	3.9	4.0	3.2	3.6	3.9	3.4	5.4	3.1	7.4	2.7	1.8	4.0	3.5
ACH (h−1)	3.125	3.125	3.125	3.125	3.125	3.125	8.7	8.7	8.7	0.18	0.18	0.18	0.18
Mean Background Concentration (#/cc)	1987	1707	7345	5560	29,802	3148	3276	1815	3149	1130	2982	1759	1452
Mean Print Concentration (#/cc)	2008	12,426	3007	5860	39,227	13,409	5402	3007	3459	16,704	15,237	32,585	71,527
Ratio of Concentration Increase	1.0	7.3	0.4	1.1	1.3	4.3	1.6	1.7	1.1	14.8	5.1	18.5	49.3
Mean Emission Rate (#x 1011/min)	2.7	16.4	3.6	7.5	51.8	17.7	56.7	31.2	75.9	7.3	5.0	14.3	31.2
Particles Removed (#x 1013)	6.2	38.8	9.7	17.3	124.2	36.1	185.4	57.0	338.5	11.7	2.0	32.8	62.5
Removal Ratio Compared to Library MakerBot	1.0	6.2	1.5	2.8	19.9	5.8	29.7	9.1	54.3	1.9	0.3	5.3	10.0

4. Discussion

Three university MakerSpaces of varying ventilation rates were assessed within this study, using a variety of 3D printers and filaments. The UpBox+ and Ultimaker 3D printers have not yet been evaluated within literature.

Systems printing ABS have been shown to emit a high number of nanoparticles throughout operation (Floyd et al., 2017; Kwon et al., 2017; Steinle, 2016; Vance et al., 2017; Yi et al., 2016), reported between 2.4 × 10⁸ particles/min (Steinle, 2016) to 1.9 × 10¹¹ particles/min (Kwon et al., 2017). Using a MakerBot Replicator, the emission rate of nanoparticles is between 7.4 × 10⁹ particles/min and 1.1 × 10¹¹ particles/min with an average particle size of 51–78 nm (Vance et al., 2017; Yi et al., 2016). Emission rates in this study, observed in Table 5 and calculated using the average emission rate equation (He et al., 2004; Stabile et al., 2017) found in Table S4, were 1.89 × 10⁸ particles/min to 2.17 × 10⁸ particles/min for ABS filaments. Our results align with those reported by Steinle, (2016); however, they are lower than other reported emission rates. Differences in emission rate calculations, experimental room volume, and ACH can each affect the reported results, leading to these discrepancies. Emissions from PLA filaments are typically lower, 1.0 × 10⁷ particles/min (Floyd et al., 2017) to 6.7 × 10¹⁰ particles/min (Kwon et al., 2017). A smaller emitted particle size, approximately 30 nm, is observed using a MakerBot Replicator printer (Yi et al., 2016). In our study, emission rates range from 7.46 10⁷ particles/min to 2.06 × 10⁹ particles/min for PLA filaments, in agreement with rates from Flyod et al. (2017) and Kwon et al. (2017).

Emitted VOCs vary depending on the filament used; emissions for the thermal degradation of common thermoplastics were reported by Unwin et al. (2013). Elevated concentrations of ethanol, 2-propanol, and acetone were measured in our study which aligns with Gu et al. (2019) when using Gray ABS with a Zortrax M200 3D printer. Ethanol was similarly reported as a low emitter, less than 40 μg/min, during a Lulzbot-PLA study (Azimi et al., 2016). Although not tested in our study, styrene is commonly emitted during printing with ABS (Azimi et al., 2016; Floyd et al., 2017; Gu et al., 2019; Mcdonnell et al., 2016; Steinle, 2016) and lactide during printing with PLA (Azimi et al., 2016; Mcdonnell et al., 2016). It is important to note that this study looked at room concentrations during the specific test parameters in the library only and not individual employee exposure risks. According to industrial hygiene monitoring practices, true exposure of VOCs must take into account duration, concentration, and work parameters (ex: how close worker is standing to emitted VOCs, etc.) of employees in the work areas.

Emitted particles are generally below 500 nm in diameter, either as agglomerates (Kwon et al., 2017; Steinle, 2016; Zontek et al., 2017) or small particles (Vance et al., 2017; Yi et al., 2016). Our study supports this finding, as 90% of emitted particles, across all prints, were less than 120 nm based on number concentration. Floyd et al. (2017) found decorated, rod-shaped particles from 3D printer emissions, with a length in the micron range (Floyd et al., 2017), similar to the observations from the Lulzbot printer within the library MakerSpace. Particles collected in the laboratory MakerSpace were similar in shape to the agglomerates above 1 μm in size, observed by Zontek et al., (2017) and Vance et al., (2017).

Of the thirteen campaigns, ten datasets exhibited a decrease in the number-weighted mean diameter and only one a decrease in mean concentration during operation. This increase in concentration with a decrease in number-weighted mean diameter at near-field suggests that particles emitted during printing were on the lower end of the size range (<300 nm). Of the campaigns for which an increase in mean diameter was observed, Lulzbot-PLA, Lulzbot-ABS, and Near-Field 2, small changes in the mean concentration were observed. Of the individual printer campaigns, the MakerBot-PLA campaign in the center MakerSpace exhibited the largest net increase in mean particle number concentration (14 times greater). This was similar to the campaign in the same MakerSpace for which all devices (three MakerBots and one UpBox) were used with the portable filter (18 times greater increase in particle number concentration relative to background). However, the maximum concentration for this latter campaign (95,423 particles/cc) was nearly five times less the maximum number concentration for the MakerBot-PLA alone (453,602 particles/cc). In contrast, although the MakerBot print campaign exhibited a larger peak deviation from background, the increase in particle concentration was delayed. This delayed concentration elevation seems counterintuitive as the multi-device campaign included the same MakerBot-PLA pairing; therefore, the mean concentration was expected to far exceed that of the lone MakerBot-PLA campaign. However, this hypothesis assumed a constant background concentration; that is, the initial concentration data collected prior to print initiation could be subtracted from the during- print concentration to calculate a simple net increase in particle count. The presented results suggest this assumption may not hold for all print environments, likely due to loss parameters.

As ventilation flow rates change temporally, the calculated ACH values are representative of a long time period within the room relative to the print duration. The assumption of a well-mixed environment provides a uniform particle concentration that would require a relatively moderate flow rate for the duration of the print. Realistically, particles are actually more dilute at the exhaust grill entrance and more concentrated in the immediate vicinity of the printers. Alternative to ACH, ventilation effectiveness describes the quality of distribution of the supply vents within the room (Rim and Novoselac, 2010). The particle removal rate ratios give a reasonable relative comparison between the print campaigns.

The particle removal concentration ratios presented strongly support differences in ventilation to substantially impact particle exposures in university MakerSpaces. The increased removal rate of the laboratory MakerSpace is consistent with the larger ACH, whereas the lower ACH observed in the center MakerSpace provided a far less efficient removal rate. Accumulation of particles within a MakerSpace can be determined using the AER. In evaluating studies using ABS filament that report both the AER and particle emission rate, an exponential decay of particle emission rate is observed with increasing AER (Floyd et al., 2017; Gu et al., 2019; Mendes et al., 2017; Stefaniak et al., 2018; Steinle, 2016). Emission rates allow for the calculation of loss rates, using models described in Table S4. The first term for all the tabulated emission rate models is a simple averaged concentration change rate, and subsequent loss terms account for effects of particle surface deposition, particle coagulation, and room ventilation. Loss parameters vary by model, but all were calculated from concentration data collected after 3D printers had been turned off, as room/chamber particle concentrations settled back to ambient levels. Comparing to Stephens et al. (2013), the lumped loss rates were similar or increased across all studies, averaging over the nanoparticle size range to 1.0 in the library study, 5.4 in the open cabinet laboratory study, 10.7 in the closed cabinet laboratory study, and 1.2 in the center study. Calculated this way, the lumped parameter accounts for ventilation, filtration by HVAC systems, and deposition within the MakerSpace (Stephens et al., 2013). The increased loss rate values in the laboratory suggest that particle removal is more effective than in the library or center MakerSpaces. This, however, does not take into consideration losses to instrument lines, double-counting by measurement instruments, and additional particle dynamics.

Although much analysis is performed on the effects of ventilation alone as an engineering control, the laboratory MakerSpace used grated metal cabinetry as an additional control. Separate print campaigns were performed to elucidate the effects of partial enclosures; however, the comparison is difficult due to a lack of far-field data. According to Table 5, within the near-field, the closed-cabinet campaigns had +10% and +70% concentration deviations from background and the open- cabinet campaign had +65% deviation from background, suggesting that the cabinets may decrease or have no effect on the emitted concentration. However, the closed-cabinet print 2 had an uncharacteristic increase in the number-weighted mean particle diameter (+21%) compared to all other laboratory campaigns, implying additional external factors impacted the results. Although explicit conclusions cannot be drawn from this study on the effects of cabinetry as a control, the use of a sealed enclosure during operation has been shown as an effective control to mitigate the release of aerosols within the MakerSpace (Afshar-Mohajer et al., 2015; Yi et al., 2016).

A major limitation in this study is the assumption of a constant background particle concentration. Since the final post-print particle concentration for the Lulzbot-HIPS campaign dips below its initial background concentration, the assumption of a constant background cannot hold for this case. It is apparent that the background concentration varies, especially in environments like the library MakerSpace, for which there exists no isolating HVAC system to separate the room air from external influences on ambient particle concentrations. This inference is further supported by the Lulzbot-PLA campaign, for which it appears that the particle number concentration actually decreases during printing (−59% deviation from background). In the case of the Lulzbot-PLA campaign, external influence on room particle concentration were clearly observed by the presence of wood particles and wood smoke scent from nearby construction. The decreased background post-print indicates that closing the door reduced interference from construction particles over time. This is also observed in the comparison ratios of particle removal in the center MakerSpace as the particle concentration increases dramatically above the background but is not reflected in the comparison. The caveat of a non-constant background is noted in the literature suggesting additional background concentration measurements are necessary (Methner et al., 2010). Future studies should further explore the nature of background fluctuations in uncontrolled environments.

Additionally, there is a lack of understanding regarding the composition of emitted aerosols and potential inhalation hazards. While several studies have investigated what VOCs are emitted, few studies have investigated the compounds found in the particulate component or biological response to exposure. Investigation into the hazards of inhalation during 3D printing has been performed using cell lines, mice, and human subjects (Farcas et al., 2019; Gümperlein et al., 2018; Zhang et al., 2019). Collections of emissions from 3D prints using either polycarbonate, ABS, or PLA have been exposed to human small airway epithelial cells (Farcas et al., 2019) or rat alveolar macrophages (NR8383) and carcinomic human alveolar epithelial cells (A549) (Zhang et al., 2019). Each study observed increases in cytotoxic and inflammatory responses and upon investigation of the emitted particles observed additive substances including transition metals that are associated with these responses. In vivo intratracheal aspiration of collected ABS and PLA emissions were given to mice and cytotoxic responses were observed (Zhang et al., 2019). Human subjects were exposed to emission from either a PLA or ABS print for a duration of 1 h while sitting and no clinically relevant acute inflammatory effects were observed due to this exposure (Gümperlein et al., 2018). These emission factors are likely also influenced by filament blend, filament source, or age of printer on emission amounts (Farcas et al., 2019; Zhang et al., 2019; Zontek et al., 2019). Therefore, it is difficult to predict potential health hazards from working in these spaces.

Ventilation is key to minimizing the aerosol concentrations for MakerSpace users. Based on this study, the authors recommend the use of a minimum of 6 ACH within the MakerSpace and/or the use of a portable HEPA filter to lower ultrafine particle concentrations during printer operation. For future studies exploring the impact of cabinetry, it is hypothesized that cabinets without local ventilation will affect airborne particle concentrations from two standpoints: (1) the extent of cabinetry closure (fully-sealed or partially-sealed), and (2) material of construction. For the latter factor, metal cabinets may produce greater electrostatic effects that cause particles to cling to the inner cabinetry instead of releasing into the surrounding air.

5. Conclusion

An analysis of 3D printer emissions in three university MakerSpaces was presented. Within the same MakerSpace, there was an influence on particle concentration and emitted size based on the printer type and filament used. When attributing only ventilation to the particle concentration reduction in the laboratory MakerSpace, the laboratory was 30 times more effective at reducing airborne particles compared to the library. The MakerSpace with the lowest ventilation provided up to 2.5 times less effective reduction in airborne particles compared to the library. The analysis presents a method of deducing the potential benefits of ventilation when setting up a MakerSpace. Toxicity studies would help put this data into context by specifying what particle concentration levels should be the limit for human exposure. Such toxicity data would help guide ventilation calculations by setting a target airborne particle concentration limit.

HIGHLIGHTS.

• ACH of at least 6 per hour are a key priority in the design of a MakerSpace.

• Net particles emitted were highest for HIPS filament, followed by PLA then ABS.

• There is a linear decay in net particles emitted versus HVAC flow rate.