Identification of Emission Source Using a Micro Sampler Carried by a Drone

Abstract

A micro needle trap sampler (NTS) was carried by a mini quadrotor drone (Mavic Pro, DJI) to collect volatile organic compounds (VOCs) from industries. The NTS was fabricated using a 7 cm long, 22-gauge stainless steel needle by packing powdered divinylbenzene (DVB) adsorbents (60–80 mesh diameters). The telescoping sampling shaft was installed on the drone to extend the NTS beyond the downward air turbulence that was caused by the rotation of its propellers. The total mass of the sampling device, including an NTS, a telescoping shaft, a mini-air pump, and an ABS (acrylonitrile butadiene styrene) rack, was not more than 200 g. The emitted VOCs, those from a steel processing plant, including aromatic hydrocarbons (toluene of 15 ppb, ethylbenzene of 9 ppb and p-xylene 12 ppb), and those from a semiconductor processing factory, including trace amounts of methanol (1.96–2.00 ppm), acetone (0.05–0.10 ppm), and toluene (1.04–2.00 ppm), were extracted by the NTS on the drone and identified using a gas chromatography-mass spectroscopy (GC-MS) system in the laboratory. According to the results of VOC detection during the sampling flight of a drone, the stationary pollution sources were successfully identified.

1. Introduction

Before 2000, tethered balloons were used to collect and identify the concentrations of atmospheric pollutants. A series of balloon samplings were performed by Greenberg et al. [1] from 1985 to 1996 to collect biological volatile organic compounds (BVOCs), and to estimate BVOC emission rates from terrestrial vegetation. To provide mobility, a UAV with fixed wings was used to carry electrochemical and optical detectors of particulate matter (PM) [2]. Owing to the rapid development of multi-rotor drones and the need to identify individual organic pollutants, various samplers have been installed on drones, including adsorptive tubes [3], canisters [4,5,6], and sampling bags [7]. The take-off load of a large multi-rotor drone typically exceeds 5 kg, so various types of large sampling devices (canisters, air pumps and pipelines) can be carried in flight to take atmospheric samples. Due to its small size and easy operation, the mini drone is very suitable for sampling and identifying sources of pollutants in industrial areas. The micro sampler, NTS, meets the requirement of a mini drone in that its take-off mass is lower than 750 g.

Volatile organic compounds (VOCs) are major air pollutants in industrial areas of Taiwan [7,8], and their negative effects on human health have been studied for the last 30 years [9,10,11,12]. The Taiwan Environmental Protection Agency (Taiwan EPA) stipulates that the emissions of VOCs from industrial processes must meet the control efficiencies and exhaustion limits. Kaohsiung City is the largest industrial harbor city in Taiwan, and most of the large VOC industrial emission sources are located in the industrial zones around Kaohsiung Metropolitan. In order to effectively improve the air quality in the urban area of Kaohsiung, the Environmental Protection Bureau of Kaohsiung City (KEPB) has continued to monitor VOC emissions to ensure that various pollution sources have been improved to achieve the reduction of VOCs. KEPB had installed 1250 micro sensor boxes in the industrial zones that surround Kaohsiung City by 2019 [13]. Micro sensors, which monitor temperature, humidity, PM₁₀, PM_2.5, and total volatile organic compounds (TVOCs), have been installed in areas with a high density of factories at a height of 3 m, and in communities adjacent to industrial areas at a height of 1.5–2 m. The obtained concentration data were delivered from a micro sensor box every 3 min to the management center via a wireless network. The main function of the micro sensors is to monitor emission from sources of air pollution in an air pollution hot zone, assisting the KEPB enforce limits on industrial pollution emissions [14].

To increase the three-dimensional mobility of sensing devices for air pollution, the Environmental Protection Bureau of Taoyuan City, northern Taiwan, uses an unmanned drone to carry sensors of TVOCs, and the technology has been developed by Chung Yuan Christian University to monitor pollutants in the atmosphere [15]. However, identifying an industrial emission source by measuring only concentrations of TVOCs, whose individual organic components cannot be exactly examined, is difficult.

Owing to the inadequacy of currently used air monitors, several air sampling devices have been developed for transport on a drone to take air samples for the subsequent qualitative and quantitative analysis of individual organic components in the laboratory. The NTS, carried on a drone with a telescopic sampling device, that extends the NTS beyond the strong downdraft stream that is generated by the propellers of a drone, has been proved to be feasible means of taking representative air samples for VOC analysis [16,17]. A mini drone (Mavic Pro, DJI), carrying an NTS (Figure 1), is a small, unmanned air vehicle (UAV) for sampling, and it is especially suitable for collecting VOCs that are emitted around industrial areas. Two real cases of the use of a drone with an NTS to identify industrial VOC emission sources are presented herein.

2. Method

2.1. Sites for Sampling

Samplings were performed at Sites A and B, illustrated in Figure 2. Site A was sampled in March 2021, and Site B was sampled from December 2021 to January 2022. Site A is located close to the edge of a hi-tech park in Kaohsiung City, Taiwan. The residents of the dormitory in the hi-tech park have frequently complained about the odors of the factory exhausts and reported them to the KEPB. The KEPB has made an inventory of the factories that surround the dormitory community and selected three manufacturers that were suspected to be responsible for the pollution for flight sampling. Plant C is a chemical manufacturing factory; Plant S is a steel processing plant, and Plant P is a panel factory. Figure 2 shows the locations of Plants C, S, and P, and the dormitory community (denoted as D). Figure 3a presents the drone as it hovers to take a downwind sampling at Plant S, Site A.

Site B is located close to the boundary of an electronic processing park in Kaohsiung City. The exhaust gas from the equipment, regenerative thermal oxidation (RTO), which is equipped to decompose thermally the organic vapors from a semiconductor processing factory (denoted as SC in Figure 2), was suspected by the residents in the adjacent communities (denoted as R in Figure 2) to be an emission source of organic solvents. The manager of the semiconductor processing factory asked our team to take gas samples using the drone with an NTS to determine its pollution emission. Figure 3b presents drone sampling above the resident community close to the semiconductor processing factory. Figure 4 shows locations of VOC samplings at Site B. Air samples were taken inside the exhaust chimney of the RTO using a Teflon bag with an air pump, following the standard procedures, NIEA A722.76B, as specified by the Environmental Protection Administration, Taiwan [18]. The air samples, which were taken by the drone, were collected in active sampling mode. Air samples were collected on the ground in passive sampling mode over sampling periods of approximately 12 h.

2.2. Drone and Sampling Device

The DJI Mavic Pro quadrotor drone is small and very lightweight. It cruises and hovers at a fixed point at a desirable altitude above the industrial zone. The mini drone has a 335 mm diagonal wheelbase (excluding propellers), and a maximum horizontal voyage speed of 65 km/h. It uses a lithium battery (LiPo 3S) of 11.4 V [19]. The launching mass of the drone was 734 g [19], and the overall mass of the sampling device, including the mini air pump, Teflon tube, ABS (acrylonitrile butadiene styrene) rack, and an NTS, was not more than 200 g. The maximum flight period of the drone was limited but not exceeding 15 min. The detailed specifications of the sampling device were as presented elsewhere in an earlier work [16,17]. The operator uses a remote controller to initiate all of the VOC sampling steps one by one, as shown in Figure 5. The control program of the sampling action is written by Arduino nano via touching the light switch. The optimal location close to the drone for extracting VOCs while avoiding the downward air flows that are generated by the rotating propellers was simulated in SolidWorks [20], and the results of the simulation were evaluated by Cheng et al. [16] using a pilot experiment. The location for adsorbing VOCs by an NTS was exactly below the head of the drone, at a vertical distance of 75 mm and a horizontal distance of 173 mm from the center of the propeller. Figure 1 shows the telescopic sampling device, which extends beyond the zone of air turbulence.

2.3. Preparation of Micro Samplers

The NTS collected VOCs through a needle by molecular diffusions. A linear concentration profile of a VOC was obtained along the diffusion path (C(Z) in Figure 1), and the amount of the extracted VOC was determined from the area (A) of the needle opening and the length of its diffusion path (Z). The total mass (n) of the VOC that was adsorbed during a time interval (t) is given by Equation (1) [21].

$$n=D_m\frac{A}{Z}\int C\left(t\right)dt$$

(1)

where D_m is the diffusion coefficient of a VOC that was extracted by an NTS. Accordingly, the quantity (n) of extracted VOCs is assumed to be proportional to the time-weighted average (TWA) concentration of the VOCs in a time interval, C(t).

The NTS is composed of a 22-gauge stainless steel needle that is packed with DVB adsorbent (60–80 mesh). DVB particles were packed by aspiration to a desired length of 7 mm and the length of diffusion path was 3 mm (Figure 1). A very small amount of epoxy glue was applied to the inlet part of the adsorbent layer to immobilize DVB particles. Finally, the DVB in the NTS was thermally conditioned by heating at 280 °C at the injection port of a GC for 30 min.

The uniformity of the packing adsorption phase in an NTS was evaluated by following procedures that have been established elsewhere [22,23]. The desired air flow rate was obtained by drawing air through the packing phase in an NTS via an aspiration pump. When the relative standard deviations (RSD) of the flow rates (mL/min) in triplicate tests did not exceed 5%, the packed phase inside the NTS was assumed uniform and immobile. VOC matrixes of ethanol, acetone, benzene, toluene, ethylbenzene, and p-xylene with the individual concentrations of approximately 10 ppm were prepared in a 550 mL glass bulb, in which the NTS was inserted for 1 h and 2 h to extract VOCs. When the RSD of extracted VOC mass that were analyzed by a GC in triplicate tests was less than 5%, the VOC adsorption performances for the tested NTS were qualified [21,22]. NTS has been successfully used for sampling organic vapors from indoor sources (essential oil, electrical cigarettes) [24,25], and for assessing workers’ exposure (painting and printing) [26,27].

2.4. Materials and Chemicals

22 G needles (L = 7 cm and ID = 0.41 mm) for use in preparing the NTS were provided from Herling Co., Ltd. (Pingtung, Taiwan). Adsorbent DVB particles were purchased from Supelco (Bellefonte, PA, USA). The epoxy glue that was used for fabricating an NTS was produced by Nao-Pao Applied Material Co., Ltd. (Taoyuan, Taiwan). The aspiration air pump, which was used for testing the flow rates through an NTS, was manufactured by Kitagawa (AP-20, Kawasaki, Japan). Tedlar air bags (SKC, Blandford, UK) were used for sampling exhaust air from the RTO in the semiconductor processing factory. All gases (Jing-De Gas Co., Ltd., Kaohsiung, Taiwan) that were used for the GC-MS analysis were of ultra-high purity, and the standard chemicals (Merck, Darmstadt, Germany) for the quantification analysis were of analytic grade.

2.5. Instrumentation and Calibration Procedures

The air compounds collected in the fields were examined by a GC-MS system (6890N and 5973, Agilent, Wilmington, DE, USA). The GC capillary column was HP19091Z-413 HP-1 PDMS (30 m × 320 μm × 0.25 μm) (Agilent, Wilmington, DE, USA).

After VOC sampling, the NTS and Tedlar air bags were analyzed using the GC-MS. The desorption time and temperature at the injection port of GC were 30 s and 250 °C, respectively. The temperature of the GC oven was increased from 50 °C in increments of 15 °C/min to 180 °C, and was held for 2 min. The carrier gas was helium, and the flow rate was 1.8 mL/min. Notably, based on the earlier works by Cheng et al. [21,22], no carryovers of analytes were available in the packing phase when an NTS was thermally desorbed, as following the above VOC analysis procedures.

The calibration analysis for VOCs that were collected by NTS carried by a drone was performed using a Tedlar bag with a volume of 1 L. A micro syringe was used to inject 0.05 µL of liquid target chemical into the Tedlar bag, which had been filled with zero air, and the bag was then placed in an oven at 40 °C. After 1 h, the compound had completely evaporated in the bag, and its concentration was C₁. The standard samples for calibration had a concentration range of 0.5–10 ppm. A sampling pump with the same specifications equipped on the drone, was used to connect the NTS to the bag in active sampling mode. As the Tedlar bag was compressible, and its volume gradually decreased with the increasing of exhaustion time of pump, the concentration of the VOCs in the Tedlar bag remained constant. When the exhaustion time of the pump was 1 min, and the NTS analysis area by the GC-MS was A₁. The concentration C₂ of the target VOC that had been sampled by the NTS at the site is given by

$${\mathit{C}}_{\mathbf{2}}=\left[\left(\frac{{\mathit{A}}_{\mathbf{2}}}{{\mathit{A}}_{\mathbf{1}}}\right)\times \frac{\mathbf{1}}{\mathit{t}}\right]\times {\mathit{C}}_{\mathbf{1}}$$

(2)

where A₂ was the VOC analysis area via the NTS sampling at the site, and t was the sampling time at the site. An example of the relevant calculation is provided in the Supplementary Materials.

To conduct a calibration analysis for the air sample that was taken using an NTS in passive mode, the assistant prepared chemicals at specified concentrations in a glass bulb, and used a 1 mL gastight syringe to withdraw a VOC sample from the bulb for analysis by the GC-MS. The integral areas for the specified concentrations of target VOCs were obtained. The VOCs, that were extracted by an NTS, were also determined by a proportional relationship; that is, substituting the data from the calibration analysis into Equation (2). Based on the minimum detected area which were examined by the GC-MS, the method detection limits (MDL) for the VOCs that were sampled by the NTS were for active sampling, benzene and toluene = 0.005 ppm, methanol = 0.18 ppm, and acetone = 0.01 ppm, and for passive sampling, benzene and toluene = 0.008 ppm, methanol = 0.15 ppm, and acetone = 0.022 ppm.

3. Result

3.1. Identifying VOC Emission Sources around a Dormitory Community by Drone Sampling

After the residents of the dormitory at Site A had contacted the KEPB regarding specific chemical emission events, the auditors from the KEPB rushed to the dormitory community and took air samples on the reporters’ balcony. The air samples were brought to the laboratory, where their constituents and their concentrations were determined.

Table 1. Concentrations of VOCs at the specific sampling points around the target dormitory at Site A.

Sampling Upwind or Downwind	Concentrations (ppb) of VOCs at Different Sampling Points a						Dormitory
	Plant C		Plant S		Plant P		Urea d	1.0
(a) Upwind	Acetaldehyde	1.0	Methane	1.4	See note b		Toluene d	14.0
							Ethylbenzene d	5.5
(b) Downwind							p-Xylene d	12.5
			Methane	See note c
	Acetaldehyde	2.0	Toluene	15.0	PGMEA	2.0
	Ethylene oxide	2.2	Ethylbenzene	9.0
			p-Xylene	12.0

Notes: ^a One air sample was collected for each sampling point and on the reporters’ balcony. Chemical compounds of concentrations larger than 1.0 ppb are presented. ^b All concentrations of detected chemical compounds are lower than 1.0 ppb. ^c Methane at downwind of Site B was lower than 1.0 ppb. ^d Air samples were collected on the balcony of the dormitory on the day when the residents reported to the office of Kaohsiung Environmental Protection Bureau.

presents the results of the analysis. The main emitted VOCs were aromatic hydrocarbons, including toluene, ethylbenzene, and p-xylene (5.5–14.0 ppb). KEPB selected three suspected factories that used organic solvents as raw materials, and their space distance (d_s) were less than 1 km from the dormitory. Plant C (d_s = 388 m) is a chemical manufacturing factory, whose main products are epoxy resin and polymers. Plant S (d_s = 598 m) is a steel plant, whose main products are steel drums and galvanized drums. Plant P (d_s = 677 m) is a panel factory, which fabricates high-end panels using the low-temperature poly-silicon (LTPS) technology.

An assistant operated the drone, which carried an NTS, to collect air samples upwind and downwind of each of the above plants. Wind speeds at each sampling were determined to be lower than 5 m/s. Those wind speeds are based on the data of the Taiwan Central Weather Bureau. Table 1 summarizes the concentrations of the VOCs from the three potential stationary sources as determined using the NTS sampling device. Downwind of Plant C, the concentration of ethylene oxide, which was emitted in the manufacture of epoxy resin, was as low as 2.2 ppb. More important, ethylene oxide was not the compound that had been detected at the dormitory. Downwind of Plant P, propylene glycol methyl ether (PGMEA), which is a solvent that is used in fabricating panels, was identified at a low concentration of 2.0 ppb. PGMEA also did not match the emission that was detected at the dormitory. Finally, downwind of Plant S, toluene, ethylbenzene, and p-xylene (TEX, 9.0–15.0 ppb) were identified in the air samples, and these VOCs were characterized as similar to the aromatic hydrocarbons that were detected at the dormitory.

3.2. Monitoring the Vertical Profile of VOC Concentrations along an Emission Source

Table 2. Concentrations of VOCs at the specific sampling points at Site B.

VOCs	Concentrations (ppm) at Different Sampling Points a
	RTO Exhaust Chimney Point a (@ 65 m) b	Community (Near Boundary Wall)			Community (in a Park) Point c (Ground) d
		Point b1 (@ 65 m) c	Point b2 (@ 20 m) c	Point b3 (Ground) d
Methanol	1.96–2.00	<0.18	<0.18	<0.15	<0.15
Acetone	0.05–0.10	<0.01	<0.01	<0.022	<0.022
Benzene	<0.005	<0.005	<0.005	0.13–0.24	<0.008
Toluene	1.04–2.00	<0.005	<0.005	0.20–0.35	<0.008

Notes: ^a Two air samples were collected for each sampling point. Values of concentrations are displayed as the analyzed ranges, and no detections of specific compounds are presented in lower than the method detection limits (MLD). ^b Air samples were collected by active sampling using Teflon bags, which were connected to an air pump. ^c Air samples were collected by active sampling using NTS with an air pump, which were carried on a drone. ^d Air samples were collected by passive sampling using an NTS, which were fixed on the pillar of a road sign.

shows that three VOCs, methanol, acetone, and toluene, were detected at concentrations from 0.05 to 2 ppm upwind in the exhaust pipe of RTO in the semiconductor manufacturing factory; downwind, trace VOCs were examined, except benzene and toluene were detected on the ground near the boundary wall of the community. The sampling point, where benzene (0.13–0.24 ppm, average 0.185 ppm) and toluene (0.20–0.35 ppm, average 0.275 ppm) were detected, was around 3 m from the side of the road, on which residents frequently drive motorcycles and cars.

4. Discussions

TEX emissions may be related to the degreasing of steel plates in Plant S [28], and this plant should be responsible for the emission event based on the concept of specific emission fingerprints from industries [8]. The other components in Table 1 were detected in concentrations that did not exceed 2 ppb. Acetaldehyde was found in the ambient air around Plant C, methane was detected around at Plant S, and urea was found at the dormitory. Acetaldehyde is one of main products of burning biomass [29], and agricultural lands and scattered livestock farms with large areas are located around Site A. Methane and urea are also produced by farming and livestock activities [30]. These organic trace components are not similar to the air pollutants that were detected at the dormitory, and they are not emitted by the industry. Hence, KEPB officials determined that Plant S was the source of emission of the air pollution in the community, and so KEPB arranged a pollution control audit of Plant S.

Typically, an RTO thermally decomposes VOCs in air streams, and the continuous exhaust emits at a flow rate of 1400 Nm³/min with organic solvent vapors at an average concentration of 330 ppm. The thermal destruction rate exceeded 98.9%, based on the detected VOC concentrations in the exhaust chimney, which are provided in Table 2. The detected concentrations of methanol, acetone, and toluene in the community, 300–430 m from the RTO exhaust chimney, were all lower than the method detection limits (<0.005–0.15 ppm). Most importantly, the VOC emissions that were detected in the community do not match those of exhaust from the semiconductor manufacturing process. The ratios of concentrations benzene to toluene (B/T) are between 3:4 and 3:5, which are characteristics of aromatic hydrocarbons in vehicular exhausts in Hong Kong, as reported by Lee et al. [31], and in Kaohsiung, as reported by Yuan et al. [8]. The values of (B/T) in Table 2 are very close to the concentration ratios from vehicular exhausts. Therefore, the chemical odors about which the community residents complained were not generated by the suspected semiconductor factory.

5. Conclusions

Detectors and samplers must be light and low-cost for their successful use on mini drones to monitor pollution in industrial areas [32]. In this work, a telescoping shaft extended an NTS beyond the air turbulence zone, and the sampling device was carried by a DJI Mavic Pro quadrotor drone to identify emission sources of VOCs. A mini drone with an NTS was successfully used to sample air pollutants and identify sources of industrial emissions. The lightness (20 g) and smallness (7 cm long) of an NTS enable its installation as a micro passive sampler on a mini drone that cruises between factories in a highly industrialized zone. Figure 6 shows the DJI Mavic Pro quadcopter as it hovers and samples inside an industrial plant with densely arranged equipment and pipelines, which present obstacles for the flight of large multi-rotors. The low take-off weight of a mini drone limits the lithium battery to supporting a sampling time of no more than 12 min. Owing to the limited flight time of a mini drone, the novel micro sampler can be used to extract representative pollutants for analysis to identify sources of industrial emissions. Notably, the NTS is an environmentally friendly sampler because it is reusable, and thermal desorption in the injection port of a GC-MS system for analysis requires no solvent. Therefore, a drone that carries an NTS is promoted to collect samples of air that is emitted from industrial sources. Even though an NTS has been regarded as an ultra-trace organic pollutant sampler, a sampler which is packed with adsorbents of very high capture efficiencies will be developed in the future.