Performance evaluation of a 3D-printed sharp-cut cyclone

A sharp-cut cyclone with an aerodynamic cut-o ﬀ diameter of 1 m m, when operated at a ﬂ ow rate of 1 L min (cid:1) 1 , was built by 3D-printing and tested against a metallic (aluminum) counterpart having the same design and dimensions. The penetration e ﬃ ciency of both cyclones was experimentally determined using quasi-monodisperse aerosol particles having aerodynamic diameters from ca. 100 nm to 2 m m. The aerodynamic cut-o ﬀ diameter for both cyclones was very similar and in accordance with the expected design value. The penetration e ﬃ ciency curve of the 3D-printed cyclone was less steep compared to that of its metallic counterpart. This di ﬀ erence is most likely attributed to the higher surface roughness of the inner parts of the 3D-printed cyclone – as also indicated by the greater pressure drop it exhibits compared to the aluminum cyclone when operated at the same ﬂ ow rate – and not by higher deviations from its design dimensions resulting from the tolerances of the 3D printer. Despite that, the substantially low cost, speed, and ease of manufacturing, make the 3D-printed cyclone a highly promising solution for applications in aerosol metrology.


Introduction
Cyclones are frequently employed as particle separators in aerosol metrology, 1 and in a number of industrial applications. 2 great advantage of cyclones is that (i) they introduce only a small pressure drop in the aerosol transportation lines they are employed, and (ii) they are very easy to operate and maintain, making them highly favorable for use in a wide number of systems. 3,4Oentimes cyclones are employed in regulatory measurements of atmospheric particulate matter (PM) for removing the fraction of the larger/heavier particles (i.e., in PM 10 , PM 2.5 , or PM 1.0 monitoring for removing respectively particles larger than 10, 2.5 or 1.0 mm) before their mass concentration is determined. 5,6In addition, cyclones are used to collect aerosol particles, including bioaerosols, having diameters larger than a specic size for further analysis, avoiding the use of lters and any artefacts from contamination and/or chemical reactions that can affect the follow-up analysis. 7,8To suit the task, cyclones need to have a well-dened cut-off diameter and steep penetration curve. 9,10yclones consist of two main parts: a cylindrical piece (namely the barrel) that serves as a pre-separator, and a conical part attached to the barrel where the separation of the large/ heavier particles takes place.For some applications were very large/heavy particles need to be collected (e.g., in bioaerosol samplers) cyclones include a cup below the conical separator in order to collect the fraction of particles that can slide down the cone. 11Depending on the way that the gas stream enters and exits the conical part, cyclones can be classied as tangential (also referred to as ow-reverse), where the direction of the ow changes by 180 , or axial where the ow enters and exits axially. 12The ow-reverse operational mode of cyclones is as follows.The incoming aerosol enters tangentially through the inlet of the cyclone and the particles experience a centrifugal force that pushes them towards its walls; the larger/heavier particles deposit on the walls, while the smaller/lighter particles exit the cyclone through an outlet on its upper part. 13he ability of a cyclone to separate particles based on their size/weight depends on its design and the operating ow rate.Kenny & Gussman showed that it is possible to dimensionally scale cyclones of a specic design based on empirical models in order to have a specic aerodynamic cut-off diameter at a certain ow rate. 14As a results, it is possible to design a "family" of cyclones (i.e., a group of cyclones whose dimensions derive by following specic proportions of their body diameter), 15 having a specic cut-off diameter when operated at the desired ow rate. 16or regulatory monitoring of atmospheric PM 2.5 , the US Environmental Protection Agency (EPA) recommends that the employed aerosol separators (i.e., cyclones or impactors) should have a steep drop in their penetration efficiency for particles having an aerodynamic diameter of 2.5 mm and above, when operated at a standard ow rate for ambient aerosol sampling (16.67 L min À1 ). 17 Based on the design proposed originally by Smith et al., Kenny et al. developed a novel Sharp-Cut Cyclone (SCC), having a penetration curve that is closer to that of the Well Impactor Ninety Six (WINS), 17,18 which is employed as a standard in regulatory monitoring of the ambient aerosol in the US. 9 Due to its unique characteristics, the SCC family of cyclones is currently employed in many applications, (e.g., regulatory sampling of particles emitted by diesel engines), as it fullls the requirements of many regulatory authorities. 19yclones are in general inexpensive compared to other components used in aerosol measurements.However, for applications where the cut-off diameter is required to be accurate, their price can become signicant, ranging from hundreds to a couple of thousands USD due to the required manufacturing precision, adding to the overall cost of the measuring system.Reducing the cost of cyclones to the order of a few tens of USD, will allow us to increase the number of sampling points and thus expand the capabilities of air quality monitoring networks.For this reason, it is important to develop cost-effective SCCs without compromising their precision signicantly.
An elegant way of fabricating cost-effective system components, including parts employed in particle processing and analysis systems, is by using 3D printing.For example, Yi et al. has recently designed and tested a 3D-printed minihydrocyclone separator (i.e., a cyclones for separating solids or different phase uids from the bulk uid). 20In this spirit, Loizidis et al., has employed 3D printing to build a ow laminarizer for use in high-precision instruments for aerosol size classication. 21In both these cases, the performance of the 3D printed components was comparable to that of counterparts produced by classical methods, while their cost of manufacturing was substantially lower.
In this study, we designed and manufactured a cost-effective SCC by 3D-printing, which offers low manufacturing cost and short production time.We designed a cyclone based on the semi-empirical model proposed by Gussman et al. having a cutoff diameter of 1 mm when operated at a ow rate of 1 L min À1 . 22he performance of the 3D-printed cyclone was determined using particles having aerodynamic diameters from ca. 100 nm to 2 mm, and compared against a metallic (aluminum) counterpartbuilt at a machine shophaving the same design and dimensions.

Cyclone design
The SCC 3D-printed cyclone was built out of conductive Acrylonitrile Butadiene Styrene (ABS) laments using the Fused Deposition Modelling technique. 23The metallic cyclone was built out of aluminum using a lathe and a drill with a custommade drilling tool having the same shape of the required conical part.The body diameter of the cyclone, D c , was determined as: where D ae50 is the aerodynamic cut-off diameter in mm, and Q the ow rate in L min À1 .To achieve an aerodynamic cut-off diameter of 1 mm at a ow rate of 1 L min À1 , the body diameter has to be 4.85 mm according to eqn (1).Fig. 1 shows an illustration of the cyclone, whereas Table 1 provides the dimensions of its parts expressed in proportion to the cyclone body diameter, including the designed and the measured (using a Vernier caliper) dimensions of both cyclones.The cyclone consists of three main parts; drawings ready for 3D printing all these parts are provided in the supplement.All dimensions of the 3D-printed cyclone have a tolerance of AE0.2 mm, while the respective tolerance for the metallic cyclone is AE0.1 mm.The tolerance for the 3D-printed cyclones was corroborated by printing and measuring 10 copies of all cyclone components.The outside surface of the 3Dprinted cyclone was covered with dissolved ABS in order to close any voids and thus prevent potential leaks through the material. 24g. 1 Schematic illustration of the cyclones (3D printed and metallic) developed and tested in this work, including their characteristic dimensions (see Table 1 for the specific values).View Article Online downstream the DMA in order to achieve an aerosol-to-sheath-ow ratio of 10 : 1.A Condensation Particle Counter (CPC; Model 3786; TSI Inc., Shoreview, MN, USA) was used to measure the particle number concentration of the monodisperse aerosol particles upstream and downstream the tested cyclones in order to determine their penetration curve.Because the tested cyclones required a ow rate of 1 L min À1 for achieving the desired cut-off diameter, while the CPC was operated at a total ow rate of 0.6 L min À1 , an additional vacuum pump was added downstream the cyclone, pulling a constant ow rate of 0.4 L min À1 .The ow through the cyclones was tested at the beginning and at the end of each experiment using a primary standard volumetric ow meter (Model Gilibrator 2; Sensidyne, St. Petersburg, FL, US).
In order to produce particles having diameters up to a few microns we used atomized solutions of Polystyrene Latex (PSL) spheres (Magsphere Inc; cf.Table 2 for details).In those measurements, instead of measuring the particle number concentrations with the CPC, we used an Optical Particle Sizer (OPS; Model 3300; TSI Inc., Shoreview, MN, USA) that measures the size distribution of aerosol particles having optical diameters in the range of 0.3 to 10 mm (cf.Fig. 2b).We should note here that the OPS is calibrated with PSL spheres, thus no correction was needed between to convert the measured optical equivalent aerosol diameters to nominal sizes.In those measurements (i.e., when the OPS was employed to measure the concentration of the PSL spheres upstream and downstream the cyclones), the additional pump downstream the tested cyclone was not operational, since the OPS provided the required ow rate of 1 L min À1 .

Cyclone penetration efficiency and pressure drop
Three sets of independent measurements were carried out in order to determine the penetration efficiency of the cyclones, which was determined by: Here N d and N u are the particle number concentrations of the tested aerosol measured respectively downstream and upstream of the cyclone.First we measured the number concentration upstream the cyclone by opening the valve for path A and closing path B as illustrated in Fig. 2, adjusting the ow rate through the cyclone at 1 L min À1 .Measurements were recorded over periods of 20 min, with 1 s intervals when using the CPC and with 6 s intervals when employing the OPS.Subsequently we measured the concentration downstream the cyclone by opening path B and closing path A, while readjusting the ow rate through the cyclone at 1 L min À1 .To conrm that the concentration upstream the cyclone remained constant throughout the experiment, we switched back to path A and repeated the measurements at the end of each experiment.To check the reproducibility of the results we repeated the measurement over three different days aer cleaning the cyclones.
The following logistic equation was used to t the penetration efficiency measurements: 25 Here D ae is the aerodynamic diameter, D ae50 the aerodynamic cut-off diameter, and b the slope of the curve.A non-linear leastsquare tting algorithm based on the interior-reective Newton method 26,27 was employed for tting eqn (3) to the experimental observations and for determining D ae50 .Eqn (3) was also used to determine the aerodynamic diameter of the particles exhibiting 16 and 84% penetration efficiency (i.e., D ae16 and D ae84 , respectively), which in turn were used to determine the sharpness of the cyclone as: 16 Because cyclones are inertial separators, their penetration efficiency depends on the aerodynamic diameter, and not on the mobility or physical diameter, of the sampled aerosol particles.For this reason, both the electrical mobility diameters of the AS particles (i.e., classied with the DMA) and the optical equivalent/physical diameters of PSL spheres, were converted to aerodynamic diameters as: 18 where D p stands for the physical diameter (i.e., equal to the electrical mobility diameter of the spherical AS particles and the optical equivalent/nominal diameter of the spherical PSL), l is the air mean free path (i.e., equal to 66 nm at 1 atm pressure), r p is the particle density, and r 0 corresponds to the unit density (i.e., 1 g cm À3 ).The physical and calculated aerodynamic diameters of both the AS and PSL aerosol particles used in our experiments are reported in Table 2.
To determine the dependence between the cut-off diameter and the ow rate, D ae50 was determined experimentally at four different ow rates (namely at 0.6, 1.0, 1.5 and, 2.0 L min À1 ), following the experimental procedures described in Section 2.2.In these experiments, the ow through the additional vacuum pump, located downstream the tested cyclones (cf., Fig. 2), was adjusted in order to achieve the required ow rates through the cyclones.The experimentally determined D ae50 values were compared with those predicted by eqn (1).
In addition to the penetration, we determined the pressure drop in both cyclones using a differential pressure manometer (Model GDH 200-07; GHM Group-Greisinger, Regenstauf, Germany).For these measurements we connected the high and low pressure node of the manometer respectively upstream and downstream the cyclone.We should note here that the pressure drop of the cyclone can be affected by the roughness of its inner surface similarly to the way that the inner surface of pipes can affect the pressure drop of the ow passing through them, 28 and consequently affect its overall performance.

Results and discussion
Fig. 3 provides the penetration efficiency curves of the 3Dprinted and metallic cyclones, when sampling AS and PSL aerosol particles at a ow rate of 1 L min À1 .The experimentally determined aerodynamic cut-off diameters of the 3D-printed and metallic cyclones are respectively 0.96 AE 0.05 mm and 1.06 AE 0.07 mm.Evidently, these values are similar, within experimental uncertainty, with the cut-off aerodynamic diameter of 1 mm set by the design model (i.e., determined by eqn ( 1)).Small deviations (up to 6%) from the designed cut-off diameter can be attributed to variations of the ow through the cyclone, uncertainties from the calibration of the particle counter, or the day-by-day variation of the conditions during the experiments. 22he achieved and/or recommended tolerances in the cut-off diameters of cyclones are typically within less than 10%.Cauda et al., for instance, designed an SCC to have a D ae50 of 0.8 mm while the actual/measured value turned out to be 0.74 mm, corresponding to a difference of 7.5% from the design value. 14ccording to the Federal Reference Methods (FRM) of the US Environmental Protection Agency (EPA), cyclones employed in PM 2.5 regulatory measurements should have a tolerance in their cut-off diameter of up to AE8% (i.e., D ae50 of 2.50 AE 0.25 mm) Fig. 3 Measured and fitted (using eqn (3)) penetration as a function of particle aerodynamic diameter for (a) the metallic, and (b) the 3D-printed cyclones.

View Article Online
when operated at a ow rate of 16.67 L min À1 . 17Evidently, the tolerance of 6% exhibited by the 3D-printed cyclone built and tested in this work is within the typical recommended values.Fig. 4 shows the measured cut-off diameters for both the 3Dprinted and the metallic cyclones when operated at 0.6, 1.0, 1.5, and 2.0 L min À1 , together with predictions using eqn (1) for a cyclone having a body diameter of 4.85 mm.The agreement between measurements and predictions is within less than 6% for both cyclones, except for the case of the 3D-printed cyclone when operated at 2.0 L min À1 that deviated by 15% from the design value.
The penetration efficiency curve of the metallic cyclone has a sharpness of 1.35, as determined by eqn (4), which is similar to the one reported by Kenny et al. and by Gussman et al. for SCCs having cut-off diameters of 1.0 and 2.5 mm. 17,22In contrast, the 3D-printed cyclone has a curve that is less steep, having a sharpness of 1.64, resembling the ones of the University Research Glassware (URG) cyclone as reported by Kenny et al. 17 As reported by Kenny et al., the discrepancy between the sharpness of the metallic and the 3D-printed cyclone could in principle be explained by differences in their dimensions caused by manufacturing tolerances. 29Liu et al. showed that the penetration curve and especially the cut-off point of a cyclone are affected by the cone contraction angles, with the effect being more dominant as the ow rate through the cyclone increases. 30his is because the swirl of the ow becomes stronger, thereby decreasing the cut-off aerodynamic diameter, as the cone contraction angle is increased.In fact, it has been argued that the sharpness in short or small-coned cyclones is dominated by specic inlet and outlet dimensions (i.e., parameters D in , D e , and h shown in Fig. 1), whereas in long or wide-coned cyclones it is dominated by the cone and base dimensions (i.e., parameter B and Z in Fig. 1). 15This explanation, however, seems unlikely for our results as all respective dimensions of the two cyclones are almost identical (cf.Table 1), yielding cone contraction angles that are equal within less than 1.5%.In addition, the measured cut-off diameters are very similar, as discussed above, supporting that differences in the dimensions (which are within less than 0.1 mm) are unlikely to affect the overall performance of the cyclones.
The discrepancy in the sharpness of the two cyclones can also be attributed to differences in the ow eld caused by the roughness of their inner walls.Simulation studies have shown that increasing the inner roughness of cyclones can change the ow eld within the cyclone due to the increased ow resistance, and thus the deposition efficiency of the particles. 31,32To investigate whether this could be the main reason for the observed discrepancy in the sharpness of the two cyclones, we measured the pressure drop (which is a proxy of ow resistance through conned ows; cf.Dzarma et al., for the case of pipes 28 ) caused by both systems as a function of operating ow rate (cf.Table 3).For the metallic cyclone the pressure drop ranged from 0.049 to 0.973 kPa, as the ow rate increased from 0.6 to 0.3 L min À1 .For the same range of ow rates the pressure drop through the 3D-printed cyclone increased from 0.067 to 1.345 kPa, whereas the difference in the pressure drop between the two cyclones when operated at 1.0 L min À1 was 0.06 kPa.Such a difference is not surprising as 3D-printed materials produced by Fused Deposition Modelling have surfaces with higher roughness compared to metallic surfaces, 33 and can in principle justify variabilities in the ow elds within each cyclone, supporting the hypothesis that those can be the reason for the difference in the sharpness of the penetration efficiency curves.Testing this hypothesis, however, requires elaborate simulations for determining the performance of the cyclones, and explaining why the roughness of their inner surfaces affects only the sharpness and not the cutoff diameter.Without such simulations, which are outside the scope of the current work, attributing the deviation in the sharpness between the 3Dprinted and the metallic cyclones to differences in their roughness can only be speculative.
We should note here that the sharpness of the penetration efficiency curves of cyclones (and other similar size separators) can be important for a number of applications including air quality monitoring where high accuracy and precision is required.For other applications (e.g., personal sampling), where sizing accuracy is not so important, the sharpness in the penetration efficiency curves can be traded for other features such as low cost and ease/speed of manufacturing, which become especially important when such systems have to be employed in large numbers.This is particularly true for the 3Dprinted cyclone developed and built here, which exhibits a cutoff size similar to that of its metallic counterpart, but with a considerable lower cost and easiness in manufacturing.We should highlight here that the time required to construct the 3D-printed cyclone was a fraction (around 1 4 ) of that required to build the metallic version, and most importantly it did not require the constant supervision of a person, thereby reducing substantially the cost of production.Furthermore, the cost of the materials and the investment for the tools (i.e., a 3D printer that costs a few hundred USD) required to build a 3D-printed cyclone, are at least one order of magnitude lower compared to those required for manufacturing a metallic cyclone, making it a much more attractive and feasible solution.

Conclusions
We have developed and tested a compact SCC built entirely by 3D printing, and compared its performance with a metallic counterpart having the same design and dimensions.Both cyclones were designed to have a cut-off diameter of 1 mm when operated at a ow rate of 1 L min À1 .The penetration efficiency curves of both the 3D-printed and metallic cyclones were experimentally determined using quasi-monodisperse AS and PSL aerosol particles having aerodynamic diameters from ca. 100 nm to 2 mm.The aerodynamic cut-off diameter of both cyclones were very similar (0.96 AE 0.05 mm for the 3D-printed and 1.06 AE 0.07 mm for the metallic cyclone) and in accordance to the expected designed value of 1 mm.In contrast to the cut-off diameter, the sharpness of the penetration efficiency curve of the cyclones exhibited a substantial difference, having values of 1.64 for the 3D-printed and 1.35 for the metallic cyclone.This discrepancy cannot be attributed to differences in the dimensions of the cyclones due to manufacturing tolerances, but most likely to differences in the roughness of their inter parts (i.e., the 3D-printed cyclone has a higher roughness), that can cause variations in the ow eld as indicated by the difference in the pressure drop that they exhibit.Despite that, the 3D-printed cyclone provides a good alternative for applications where low-cost and fast manufacturing are needed without sacricing the accuracy on the required cut-off diameter, making it a promising candidate in aerosol metrology.

Paper
Environmental Science: Processes & Impacts

Fig. 2
Fig.2shows the experimental setup used to test the performance of the two cyclones.In brief, an atomizer (Model AGK 2000; Palas GmbH, Karlsruhe, Germany) was employed for producing polydisperse ammonium sulfate (AS) particles.The resulting aerosol was subsequently dried using a silica diffusion dryer (Model 3062; TSI Inc., Shoreview, MN, USA), and then charge-neutralized by passing it through a 85 Kr source aerosol neutralizer (Model 3077 A; TSI Inc., Shoreview, MN, USA).To obtain a monodisperse aerosol with particles in the mobility diameter range of ca.60 to 750 nm, the dried charge-neutralized aerosol was passed through a Differential Mobility Analyzer (DMA; Model 3081; TSI Inc., Shoreview, MN, USA) operated at a constant sheath ow of 3.2 L min À1 .The sample ow passing through the DMA was regulated at 0.32 L min À1 by adding a simple diluter (comprised of a valve and a HEPA lter)

Fig. 2
Fig. 2 Schematic layout of the experimental setup employed to determine the penetration of the cyclones.Both the 3D-printed and the metallic cyclones were tested with quasi monodisperse (a) AS particles, and (b) PSL spheres.Key: AT: atomizer; HF: HEPA filter; SD: silicon drier; NT: neutralizer; DMA: differential mobility analyzer; 3-WV: three-way-valve; SCC: sharp-cut cyclone; CPC: condensation particle counter; OPS: optical particle sizer.

Fig. 4
Fig. 4 Cyclone cut-off diameter as a function of operating flow rate.Values predicted by eqn (1)correspond to a cyclone having a body diameter of 0.485 cm.

Table 1
Proportions in respect to the cyclone diameter, including the design and real dimensions of the SCCs (3D-printed and metallic) designed and built in this work for a flow rate of 1 L min À1

Table 2
Electrical mobility, optical equivalent and aerodynamic diameters of the particles used to determine the penetration efficiency curves of the SCCs developed and tested in this work

Table 3
Pressure drop through the 3D-printed and the metallic cyclones for different flow rates Flow rate [L min À1 ]