Investigation of Long-Term Performance of a Proposed Cost-Effective HCl Non-Dispersive Infrared Analyzer at Real Stationary Sources

Abstract

The zero drift, interference, and sensitivity of an HCl analyzer based on a non-dispersive infrared (NDIR) technique can be improved to develop a cost-effective solution for continuous emission monitoring systems (CEMSs). To achieve these improvements, the same bandpass filter technique, negligible interference bandpass filter, and optimal path length are applied to the analyzer. Laboratory inspections and long-term field trials are conducted to evaluate the performance of the analyzer. A metalworking factory and a cement factory are selected for field trials. In laboratory inspections, the relative error of the analyzer is less than 1%, aligning closely with the results obtained from standard ion chromatography methods. Moreover, the basic specifications of the proposed analyzer are comparable to those of commercial HCl analyzers. In field trials, the NDIR analyzer shows a significant bias compared to the standard method. However, when considering the difference between HCl emission levels and HCl emission standards, the relative errors are less than 10%. These results suggest the proposed NDIR analyzer is a practical option for the CEMS of metalworking and cement factories. However, seasonal variations should be considered when the temperatures of gas emissions are low.

1. Introduction

Hydrogen chloride (HCl, CAS number 7647-01-0) is a colorless, corrosive gas with an acrid smell at room temperature [1]. HCl negatively impacts human health. Exposure to HCl typically results in irritation of the respiratory tract, eyes, and skin [2]. Inhalation of high concentrations of HCl (1000–2000 ppm) can cause death [2]. Particulate Cl⁻ and gaseous hydrogen chloride (HCl) are important in atmospheric chemistry [3]. Particulate Cl⁻ can form through the process of equilibrium redistribution from gaseous HCl [3]. Gaseous HCl is a precursor to ozone and particulate matter (PM), which are closely related to human health [3,4]. Moreover, similar to SO₂ and NO_X, HCl is a contributor to not only boiler corrosion but also acid rain [5,6]. Therefore, gaseous HCl must be regulated.

Gaseous HCl is emitted from industrial sources, including the combustion of chloride-containing fuels (such as waste, biomass, and coal) and chemical production [4,7]. The primary emissions of HCl are from waste incineration (38%), biomass burning (19%), energy production (19%), and residential (13%) sectors [8]. Current regulations generally do not recommend the continuous monitoring of HCl in most plants [9]. The level of HCl in plants with well-controlled stack emissions is expected to be much less than 5 ppm [9,10]. Therefore, monitoring the emissions of gaseous HCl is essential in managing and controlling its emissions from stationary sources [4,7,11]. Air quality depends on monitoring anthropogenic sources, and accurate measurements are essential for monitoring emissions [7,12,13]. The critical component of an effective monitoring system is an instrument capable of maintaining accuracy under diverse operating conditions [13].

For measuring gaseous HCl, methods such as tunable diode laser absorption spectroscopy (TDLAS), Fourier-transform infrared (FTIR) spectroscopy, and the non-dispersive infrared analyzer integrated with gas filter correlation (GFC/NDIR) have been widely used to develop an HCl analyzer [4,10,13]. On the other hand, HCl sensor-types have been recently developed based on electrochemical and optical technologies [14,15,16,17,18,19,20,21,22,23,24]. Electrochemical sensors detect the variations of electrical signals generated by chemical reactions of gas compounds [25], whereas optical sensors measure the variations of light intensity due to the interaction between light and gas compounds [26].

Tunable diode laser absorption spectroscopy (TDLAS) consists of a tunable diode laser, transmission optics, an optical gas chamber, and a detector. This technique, based on a tunable diode laser, adjusts the diode laser’s injection current or temperature to align the emission wavelength with the absorption lines of a specific gas [9,27,28]. TDLAS is relatively simple in its experimental setup and capable of performing online in situ measurements using cross-stack configurations [10,27,29]. It exhibits high sensitivity and can achieve a specific detection limit to meet measurement requirements by selecting appropriate gas absorption lines [27,30]. The limitations of TDLAS include its high cost, the requirement for line-of-sight—which prevents spatially resolved measurements—limited detection bandwidth, and limited adaptability of detectors in severe environments [29,31].

Fourier-transform infrared (FTIR) spectroscopy consists of an infrared source, a beam splitter, a detector, and an interferometer [13,32]. By using a broad range of IR frequencies, the FTIR light source can simultaneously acquire comprehensive infrared spectrum information rather than being limited to a few frequencies at a time [33,34,35]. The Fourier transformation significantly enhanced IR spectrometry’s ability to solve a wide range of analytical issues, converting raw data into an actual spectrum [13,36]. FTIR is a susceptible sensitive analytical technique that can identify trace substances in a sample within a few minutes. It can also be applied to a quantitative analysis to determine the concentration of a compound in a sample [37]. FTIR’s main drawbacks include its high cost, primarily due to computer requirements, potential spectrum interference from complex sample mixtures or other molecules, and water sensitivity, necessitating drying materials before analysis [38,39]. Due to FTIR spectroscopy’s water sensitivity, materials must be thoroughly dried before analysis to prevent contamination [40].

A non-dispersive infrared (NDIR) technology, which is a type of optical technology based on the infrared absorption of an analyte, is another potential approach for HCl measurement because of its precise accuracy, long life, fast response, and high selectivity [41,42,43]. Recently, NDIR sensors have been widely developed to measure gaseous compounds because of their compact and low cost, such as NO₂ emitted from diesel engines [42], CO₂, CO, and C₃H₈ emitted from automobiles [44], CO₂ in air [45,46,47,48], CH₄ in air [49], SF₆ emitted from the power industry [50], and CO₂ and CH₄ emitted from the thermal runaway process of lithium-ion batteries [51]. However, NDIR-based HCl sensors have not been extensively documented. Furthermore, while an NDIR sensor represents a potential approach for developing an HCl sensor suitable for a CEMS, its durability and the interference effects of complex emission gases in flue gas from industrial stacks must be carefully considered. Notably, to date, no sensor of this type has been approved or regulated for use in CEMS applications [52,53,54,55,56].

An NDIR analyzer, integrated with a gas filter correlation (GFC), consists of a gas chamber, an infrared source, a detector equipped with a bandpass filter (BPF), and a filter wheel with GFC [57,58]. A single infrared source and detector allow monitoring numerous gases simultaneously by using different BPFs [32,33]. Three components make up the GFC wheel: (1) a high concentration of the gas to be detected, (2) a neutral gas (usually N₂), and (3) a window for background signal determination [58,59]. The GFC technique improves the NDIR analyzer’s sensitivity and selectivity by introducing a significant volume of sample gas. In addition, the GFC offers spectral resolution and a high degree of multiplexing, with performance requirements for zero drift and resolution [58,59]. Because IR sources with a wide range of wavelengths can run at a lower temperature than other sources, NDIR benefits energy usage [32]. Water vapor interference is one of its disadvantages, which is important since stack gases contain a considerable quantity of water vapor and can lead to the formation of artifacts for target gases [60]. Although the NDIR used a GFC to mitigate interference, the GFC indicated that elevated concentrations of interfering gases at specific levels decreased the analyzer’s accuracy, ultimately causing bias in the analytical results [59]. Although NDIR is simpler and cheaper than TDLAS and FTIR, it is used less commonly due to this issue. Therefore, the NDIR analyzer for HCl needs improving.

In terms of other HCl sensor technologies, the recent sensor development is shown in

Table 1. Summary of recent developments in HCl sensors.

No.	Technique	Operating Range	Limit of Detection	Application	Ref.
1	Electrochemical sensor: lyocell-based activated carbon fibers with a high surface area for electrochemical detector	0~20 ppm	-	Environmental and industrial gas management	[14]
2	Optical sensor: colorimetric sheath/core-type polyamide 6 (PA6)-RhYK/polypropylene bicomponent fiber	1~100 ppm	1 ppm	Industrial sites and daily life	[16]
3	Optical sensor: samarium-doped Smx: Mn0.8Zn0.2Fe2-xO4 (SMZFO) (x = 0.0–0.1) nanocrystals	10~100 ppm	10 ppm	Workplace safety	[15]
4	Optical sensor: sol–gel spin-coated V2O3 thin films	0~32 ppm	-	Workplace safety	[17]
5	Optical sensor: thin, homogeneous polypyrrole layers on flexible textile polyamide fabrics	20~100 ppm	20 ppm	Personal protection	[19]
6	Optical sensor: Mn-doped graphene	-	-	Industry and safety	[20]
7	Optical sensor: tunable perovskite nanowire laser based on a CsPbBr3 nanowire integrated with a nanostructured Al2O3 substrate	5~500 ppm	1.12 ppm	Workplace in chemical industry	[18]
8	Optical sensor: UiO-66 three-dimensional photonic crystals	0~2.5 ppm	10.9 ppb	Atmosphere	[22]
9	Optical sensor: fluorescent nanofilms based on imine linkage using 4,4′-[4,4′-Biphenyldiylbis (oxy)] dianiline (DAPODP) and 4,4′,4″,4‴-(ethene-1,1,2,2-tetrayl) tetrabenzaldehyde (TPE-CHO)	0~96 ppm	150 ppb	Industrial workplace	[24]
10	Optical sensor: 3D Ln-Metal–organic frameworks based on the H4TBAPy ligand [H4TBAPy = (1,3,6,8-tetrakis (p-benzoic acid) pyrene)]	0~80 ppm	0.18 ppm	Workplace in chemical industry	[21]
11	Optical sensor: thermoelastic spectroscopy based on a custom low-frequency quartz tuning fork	0~500 ppm	0.148 ppm	Chemical processing	[23]
12	Optical sensor: multi-gas photoacoustic sensor	0~25 ppm	24 ppb	-	[61]

.

As shown in Table 1, HCl sensors have been significantly developed in recent years. However, many of these are colorimetric-type sensors, which are primarily suitable for safety applications. Some proposed sensors can measure HCl in real time within practical operational ranges and detection limits. However, these sensors were designed for ambient air monitoring rather than monitoring emission gases from industrial stacks. The photoacoustic sensor based on near-infrared technology is also a potential method, but its sensitivity to vibration [62] is a disadvantage for industrial applications. In addition, ensuring consistent performance over a long period leaves a critical concern for these sensors because the CEMS regulations require the durable and consistent performance of analyzers to pass a yearly quality inspection [52,53,54,55,56]. Notably, as mentioned earlier, sensor-type devices have not yet been approved for CEMS applications. Therefore, an HCl analyzer type should be considered to be practically and immediately applicable for CEMS. However, there is limited documentation on the development of HCl analyzers specifically for CEMS. Instead, several HCl analyzers have been developed for ambient air monitoring using techniques such as tunable infrared laser differential absorption spectroscopy (TILDAS) [63] and cavity ring-down spectroscopy [64]. While these technologies are also suitable for HCl analyzers in CEMS applications, their high manufacturing cost is a significant concern.

Accordingly, a highly accurate cost-effective NDIR with low zero drift was developed in this study, offering a promising solution for the continuous HCl monitoring of industrial emissions. An optimal path length, a correction factor of humidity interference, and a negligible zero-drift method [57] were applied. Long-term field trials at a metalworking factory and a cement factory were conducted to investigate the performance of the analyzer.

2. Materials and Methods

2.1. Apparatus

The NDIR analyzer comprised a 14 W infrared source (PL-314K, Hawkeye Technologies Inc., Milford, CT, USA), a pyroelectric detector (LME-335, Infratec GmbH, Dresden, Germany), an aluminum anodized gas chamber with gold-coated mirrors, and a waveguide connected to a motor (PKP566FMN24A, Oriental Motor Co., Ltd., Tokyo, Japan). The general schematic diagram of the analyzer is shown in Figure 1. The reference BPF center wavelength was 3.95 μm with a 90 nm half-bandwidth (Seoul Precision Optics Co., Ltd., Seoul, Republic of Korea) applied as a reference channel. The initial cost of the analyzer was approximately 10,000 USD, which can be cost-effective for a CEMS analyzer compared to high-cost commercial analyzers.

2.2. Development of HCl Analyzer

2.2.1. Negligible-Interference Bandpass Filter

The HITRAN database [65] was used to analyze the spectral characteristics of various chemicals, a method adapted from previous research [33,34]. As illustrated in Figure 2, the infrared absorption spectra of HCl gas extend from 3.2 µm to 3.9 µm. Within this spectral range, SO₂ and H₂O present significant spectral overlap (refer to dashed circle in Figure 2), thereby interfering with the accurate detection of HCl. In particular, H₂O introduces substantial interference due to its high concentration in the stack emission gas. A new HCl-specific bandpass filter (BPF) was developed to mitigate this challenge by targeting spectral regions exhibiting minimal cross-sensitivity to other compounds. The BPF features a center wavelength (CWL) of 3.60 μm and an 80 nm half-bandwidth (HBW).

To test the performance of the new BPFs, a comparison test was conducted with a commercial product, a BPF with 3.55 μm of CWL, and 70 nm of HBW (Seoul Precision Optics Co., Ltd., Seoul, Republic of Korea). The experimental setup is presented in Figure 3.

The IR intensity at a specific wavelength decreased in the gas chamber due to the gas’ absorption of IR light at that wavelength. The detector reported the infrared intensity and produced an electrical (detector) signal. The signal-processing algorithm transformed logarithmically the ratio of the signal detected by the analyzer. The logarithm of the ratio (R:M) of the analyzer signal (R is the signal from the reference BPF channel, and M is the signal from the target gas BPF channel) was calculated. N₂ (99.999%, DongA Co., Ltd., Gwangju, Republic of Korea) was used to calibrate the zero point. Then, HCl and other standard gases were used to calibrate the span points of the NDIR. The standard gas concentration is shown in

Table 2. Concentrations of standard gases used in this study.

Compound	Concentration (ppm, %-CO2, %RH–H2O)
HCl	0, 5, 15, 30, 40, 50
NO	0, 1, 10, 50, 100, 180, 300, 400, 500, 600, 700, 800, 900, 1000
NO2	0, 1, 10, 50, 100, 180, 300, 400, 500, 600, 700, 800, 900, 1000
SO2	0, 1, 10, 50, 100, 180, 300, 400, 500, 600, 700, 800, 900, 1000
CO	0, 1, 10, 50, 100, 180, 300, 400, 500, 600, 700, 800, 900, 1000
CO2	0, 1, 2, 3, 4, 6, 10, 12, 23, 50
H2O	0, 5, 10, 20, 30, 40, 50, 60, 70, 80

Note: Standard gas was manufactured by Rigas Co., Ltd., Daejeon, Republic of Korea.

. H₂O (vapor) was generated using the bubble method [66,67]. Humidity was measured by a humidity sensor (645, Testo SE & Co. KgaA, Lenzkirch, Germany).

Each gas was injected into the analyzer for 15 min. The initial 5 min data, prior to reaching equilibrium, were excluded, and the average signal levels from the last 10 min were used. Relative sensitivity is a measure of the rate of change of the calibration curve y = f(x), which relates analyzer signals to concentrations of the target gas [68].

2.2.2. Improvement of the Zero-Drift Issue

Zero drift refers to a variation in the analyzer’s signal that displaces the entire calibration curve of the analyzer [69]. An HCl GFC with 99.99% HCl gas was used to reduce the drift of the analyzer. Using different wavelengths of reference BPF and measurement BPF causes zero drift [28,36]. To reduce the zero drift, the negligible zero-drift method for the NDIR analyzer was adapted from a previous study [57]. HCl GFC coupled with HCl BPF was used as a reference channel. This method was compared with a typical reference channel using a 3.95 µm BPF. HCl BPF was used as a measurement channel. To evaluate the stability of the NDIR analyzer, it was operated continuously for 14 days, during which continuous analytical signals were recorded for each of the three channels. Zero drift was calculated using the logarithm of the ratio (R:M) of the analyzer signal.

2.2.3. Investigation of Optimal Path Length

The analyzer needs a long optical path length to achieve sensitivity and improve the limit of detection [32,70]. Therefore, the feasible path length that can provide linear calibration curves must be determined. The analyzer signal was defined as the logarithm of the ratio between the HCl BPF measurement signal and the HCl GFC reference signal. The correlation between the analyzer signal and the target gas concentration was evaluated with respect to various optical path lengths, including 6, 10, and 12 m. An experimental setup is presented in Figure 3.

2.3. Performance of Proposed HCl NDIR Analyzer at Laboratory-Scale Inspection

To investigate the performance of the proposed HCl NDIR analyzer, its operation specifications, including detection limit, zero drift, span drift, reproductivity, linearity, and response time, were first demonstrated following the national standard (REF). N₂ gas (99.99%, Rigas Co., Ltd., Daejeon, Republic of Korea) was used for the detection limit and zero test. HCl standard gas of 39.5 ppm (Rigas Co., Ltd., Daejeon, Republic of Korea) was used as a span gas because the analyzer measurement range was developed as 0~50 ppm. HCl standard gases of 15.6, 29.5, and 39.5 ppm (Rigas Co., Ltd., Daejeon, Republic of Korea) were applied to investigate the analyzer’s linearity.

Second, the analyzer’s performance was determined using nine gas mixes chosen based on the emission pattern of a combustion process. The concentrations of mixed gases are shown in

Table 3. Mixture gas information on HCl test for a combustion process.

Mixture	HCl (ppm)	NH3 (ppm)	CO (ppm)	NO (ppm)	NO2 (ppm)	CO2 (%)	SO2 (ppm)	H2O (%RH)
1	1	1	1	1	1	10	1	10
2	1	5	25	20	2.5	15	25	20
3	1	10	50	45	5	20	50	40
4	5	1	1	1	1	10	1	10
5	5	5	25	20	2.5	15	25	20
6	5	10	50	45	5	20	50	40
7	10	1	1	1	1	10	1	10
8	10	5	25	20	2.5	15	25	20
9	10	10	50	45	5	20	50	40

. We selected concentrations of interference gases using data from CleanSYS [71], a continuous emission monitoring system for combustion in the Republic of Korea. HCl concentrations ranged from 0 to 13.5 ppm [71]. Consequently, the HCl concentrations in the mixes were 1, 5, and 10 ppm. The other compounds were also mixed at concentrations based on CleanSYS data [71]. The relative humidity (RH) of the flue gas was assumed at 10, 20, and 50%, since water pretreatment equipment was typically used to remove water vapor in the flue gas before entering the analyzer [72,73]. A 25 L Tedlar bag (232-25, SKC Inc., Eighty Four, PA, USA) was used to prepare each mixture. The experimental setup is shown in Figure 3. The ion chromatography method was used to measure the amounts of HCl [74].

2.4. Long-Term Field Trials of the Proposed HCl NDIR Analyzer

The first field trial was conducted in a metalworking factory in Incheon, Republic of Korea. Since the factory used HCl solvent for the coating process of metals, HCl gas was found in the emission gas. The factory is appropriate for NDIR performance evaluation because its HCl concentration was between 0 and 4 ppm. The field experiment was conducted by installing a shelter next to the stack according to the standard for installing continuous emission monitoring systems (CEMSs) [75], and was operated for four seasons from June 2022 to May 2023. The emission gas from the factory was collected and treated with a wet scrubber before being released into ambient air through a stack. The average flue gas temperature in the stack was approximately 15.5 ± 7.5 °C. The emission flow rate was approximately 1200 ± 100 m³/min. The HCl emission standard of the factory was 4 ppm. The second field trial was conducted at a cement factory in Gwangyang City, Republic of Korea. The emission gas from the factory’s kiln process was considered. The average flue gas temperature in the stack was approximately 56 ± 6 °C. The emission flow rate was approximately 870 ± 50 m³/min. Its HCl emission standard was 5 ppm. In accordance with the factory’s permission, the field trials at the cement factory were conducted for 6 months from November 2022 to April 2023.

The experimental setup for the field trial is shown in Figure 4.

The measurement hole is approximately 30 m high, and a probe is connected to the chimney to collect the sample. A conduit heated to 140 °C [74] is used to transfer the flue gas to the analyzer so that the emission gas can be transported without condensation. To remove the high moisture levels in the flue gas, a humidity pretreatment device based on the Nafion^TM membrane [74] was used before its introduction into the analyzer. The humidity was sustained below 40% relative humidity at 25 °C. A data logger was connected to the NDIR to collect data at 30 min intervals for analysis based on the national standard [74]. The test bed and the NDIR analyzer were checked and calibrated every two weeks.

Since there were no standard CEMS data for HCl at these factories, to evaluate the NDIR analysis data from the site, random samples were collected [76] and analyzed using the ion chromatography (IC) method [74]. It was repeated 12 times randomly during the field experiment at the metalworking factory and six times at the cement factory. Each time, three samples were collected and compared. The relative mean squared error (RMSE), mean normalized error (MNE), and mean normalized bias (MNB) were evaluated using Equation (1), Equation (2), and Equation (3), respectively, to assess the bias of the proposed analyzer compared relatively to results from the IC. If RMSE and MNE are lower, the HCl concentration observed by the proposed NDIR analyzer is closer to that obtained by the ion chromatography method. An MNB of zero indicates no systematic bias between the HCl concentration obtained from the proposed analyzer and the IC method [77,78]. A positive MNB indicates that the HCl concentration measured by the NDIR analyzer is higher than that measured by the IC method and vice versa [77,78].

$$RMSE=\sqrt{{\displaystyle \frac{1}{n}}{\sum }_{i=1}^n{(C_{NDIR}-C_{IC})}^2}$$

(1)

$$MNE={\displaystyle \frac{1}{n}}{\sum }_{i=1}^n{\displaystyle \frac{\left|C_{NDIR}-C_{IC}\right|}{C_{IC}}}\times 100$$

(2)

$$MNB={\displaystyle \frac{1}{n}}{\sum }_{i=1}^n{\displaystyle \frac{C_{NDIR}-C_{IC}}{C_{IC}}}\times 100$$

(3)

where CNDIR is the HCl concentration measured by the proposed HCl NDIR analyzer, CIC is the HCl concentration measured by the IC method, and n is the number of values.

In addition, the relative percentage difference between the proposed NDIR analyzer and the IC standard method was also calculated based on Equation (4) to relatively compare the performance of the analyzer at a field performance [79].

$$RPD={\displaystyle \frac{\left|C_{NDIR}-C_{IC}\right|}{Emission standard}}\times 100$$

(4)

Microsoft Office Excel (version 2307, Microsoft Cooperation, Redmond, WA, USA) was used to create the charts.

3. Results and Discussion

3.1. Development of HCl Analyzer

3.1.1. Negligible-Interference Bandpass Filter

The interference effects in the new HCl BPF and the commercial BPF were measured using varying concentrations of target gases. The effects of NO, NO₂, SO₂, CO, CO₂, and H₂O interference on the HCl BPFs are shown in Figure 5 and Figure 6.

As shown in Figure 5, commercial BPF was not substantially affected by NO, NO₂, and CO₂. On the other hand, SO₂, CO, and H₂O demonstrated significant effects on the commercial BPF (r² > 0.9). In flue gas, humidity is usually high. Therefore, H₂O has a significant impact on low HCl concentrations.

As reflected in Figure 6, SO₂ showed significant effects on the new HCl BPF (r² > 0.9), while other gases showed no interference effects on the new BPF. Both the new and commercial HCl BPF were significantly affected by SO₂. Flue gas desulfurizer (FGD) is widely applied to treat SO₂ emitted from the stack, and its removal efficiency is over 95% [80]. Consequently, SO₂ is discharged from the stack at a low concentration (

Table 4. Various SO₂ concentrations from different stacks in the Republic of Korea [71].

No.	City	Min (ppm)	Max (ppm)	Emission Standard
1	Sangju	0.0	0.2	16
2	Seoul	0.3	4.2	16
3	Jangheung	2.1	5.6	16
4	Muan	0.1	4.1	24
5	Yangsan	0.06	0.058	20
6	Gyeongju	0.0	0	16
7	Pocheon	0.0	11	24
8	Seosan	0.0	1.6	30
9	Bucheon	0.0	2.3	16
10	Gochang	0.0	1.1	24

) [71]. As shown in Table 4, most SO₂ emissions were less than 10 ppm. Thus, SO₂ cannot cause a significant effect on the NDIR analyzer. A cross-interference factor could compensate for interference to improve the accuracy of the analyzer [81]. These results indicate that the new HCl BPF has less interference with other gases than the commercial BPF.

3.1.2. Improvement of the Zero-Drift Issue

Commercial reference BPF and HCl GFC coupled with HCl BPF were compared for the reference channel. The zero drift of the HCl NDIR analyzer was measured continuously for 14 days, as shown in Figure 7. It was found that zero drift with the reference BPF over 14 days was approximately 0.35%. This result is a small percentage, but it is about a 1.6 ppm drift when converted to concentration (Figure 8a). On the other hand, 0.03% of zero drift with HCl GFC occurred, which is approximately 0.3 ppm at a concentration (Figure 8b). Since the HCl emission from the stack is mainly discharged at low concentrations, even the smallest error can cause a significant effect. Therefore, the drift of the HCl NDIR analyzer must be reduced, which can be resolved by applying the negligible zero-drift method [57].

3.1.3. Investigation of Optimal Path Length

The HCl calibration curves, which present relationships between the optical path length and the analyzer signals, are shown in Figure 9. The logged ratio values between the signals from the measurement channel (i.e., BPF of HCl gas) and the reference channel (i.e., HCl GFC coupled with HCl BPF) were used to calculate the relative absorbances. The linearity of the calibration curve indicates that it is effective in measuring high concentrations. All path lengths showed a good coefficient of determination in terms of calibration (r² > 0.97). However, for the 12 m path length, most of its infrared radiation was absorbed. Consequently, at long path lengths, there was a nonlinear relationship between the analyzer signals and HCl concentrations (Figure 9c), while it was linear for the 6 and 10 m path lengths (Figure 9a,b). Therefore, the 6 and 10 m path lengths could be used to measure a high concentration, and a 12 m path length could be used to measure a relatively low concentration. Longer path lengths may not always result in better outcomes when it comes to absorption over the entire concentration range.

As the optical path lengths increased, the analyzer’s sensitivity also increased. The increase in path length led to an increase in absorbance, based on the Beer–Lambert law. However, the longer path length can result in a slower response time or recovery speed of the analyzer. Thus, the optimal path length is an important issue. As a result, the sensitivity of the analyzer has been improved. However, the sensitivity at the 10 m path length did not significantly differ from that at 12 m, as the slope values are shown in Figure 9. Therefore, the 10 m path length was selected, considering the measurement range, cost, response, and noise.

3.2. Performance of Proposed HCl NDIR Analyzer at Laboratory-Scale Inspection

Table 5. Specifications of various commercial HCl analyzers and the proposed analyzer in this study.

No.	Type	Manufacturer	Model	Price (USD)	Limit of Detection	24 h Zero Drift	24 h Span Drift	Reproductivity	Linearity	Response Time	Range	Ref.
					(ppm)	(%)	(%)	(%)	(%)	(s)	(ppm)
1	FTIR	S.Fac Inc. (Daejeon, Republic of Korea)	TMS300-FTIR	>56,000	0.31	2	-		2	-	~100	[82]
2	NDIR-GFC	Envea (Poissy, France)	MIR-9000	>40,000	-	2	1	2	1	-	~15/5000	[83]
3	FTIR	Gasmet (Vantaa, Finland)	CX4000	>70,000	-	2	-	-	2	120	-	[84]
4		Thermo Scientific (Waltham, MA, USA)	15i (HCP-PKI)	>30,000	0.2	0.2 ppm	2	-	2	120	~5000	[85]
5	FTIR	Protea Ltd. (Middlewich, Cheshire, UK)	AtmosFIR CEM	-	0.2	-	-	1	2	120	~100	[86]
6		EcoChem (League, TX, USA)	MC3	>47,000	1% of scale	-	-	-	-	10	~100	[87]
7		SICK (Waldkirch, Breisgau, Germany)	MCS200HW	>100,000	2% of scale	3	3	-	-	200	~1840	[88]
8	TDLAS	HORIBA (Kyoto, Japan)	TX-100	>59,000	-	2	2	1	1	2	~500	[89]
9	TDLAS	Neo Monitors AS (Skedsmokorset, Norway)	LaserGas-II SP	-	0.05	-	-	1	-	2	-	[90]
10	TDLAS	Envea (Poissy, France)	LAS 5000XD	-	1% of scale	-	-	-	1	1	~10, 5000	[91]
11	TDLAS	METTLER TOLEDO (Columbus, OH, USA)	Gpro 500	-	0.6	2	2	0.25	1	4	~1%	[92]
12	NDIR	This study	-	10,000	0.07	0.02	0.6	0.26	0.96	75	0~50	-

shows the basic specifications of the proposed HCl NDIR analyzer. The 2 h zero drift and 2 h span drift were 0.02% and 0.3%, respectively. The representative performance data are presented in Figure 10.

Table 5 and Figure 10 demonstrate that the performance of the proposed HCl NDIR analyzer was comparable to other commercial HCl analyzers employing diverse technologies. Although the limit of detection and measurement range of some commercial products, such as FTIR analyzers, surpass those of the proposed analyzer, the lower cost of the proposed device offers a significant advantage in terms of cost-effectiveness. Furthermore, the proposed analyzer’s specifications meet the requirements for CEMS applications despite its relatively low price. It is worth noting that 10,000 USD is the production cost of a single device used in this study. Thus, massive production has the potential to further reduce this cost.

Since the negligible zero-drift method was employed, the 2 h zero drift and 24 h zero drift were similar. This suggests that the zero-drift issue of an NDIR analyzer was improved. Although its measurement range was 0~50 ppm, its detection limit was approximately 0.1 ppm. Therefore, the selected path length was acceptable for an NDIR analyzer.

To investigate the effect of interference gases, mixture gases that were simulated by emission compositions from a combustion process were used to examine the analyzer’s operation (

Table 6. Various concentrations of HCl (ppm) with respect to the NDIR analyzer coupled with the combined method and ion chromatography (combustion process simulation).

Mix	STD	NDIR Analyzer					Ion Chromatography
		Min	Max	Mean	RSD	RPD	Min	Max	Mean	RSD	RPD
1	1.00	0.98	1.02	1.01	1.50	0.57	0.98	1.01	0.99	1.54	1.00
2	1.00	0.98	1.01	1.00	1.22	0.14	1.00	1.01	1.00	0.58	0.00
3	1.00	0.98	1.03	1.01	1.89	0.57	0.99	1.02	1.00	1.73	0.00
4	5.02	4.98	5.10	5.02	0.82	0.00	5.00	5.05	5.02	0.50	0.00
5	5.02	4.98	5.10	5.04	0.89	0.40	4.98	5.07	5.03	0.90	0.20
6	5.02	4.96	5.04	5.01	0.63	0.26	5.00	5.04	5.02	0.41	0.00
7	10.0	9.97	10.1	10.0	0.56	0.04	10.0	10.13	10.1	0.55	0.50
8	10.0	9.99	10.1	10.0	0.43	0.21	10.0	10.12	10.1	0.40	0.60
9	10.0	9.95	10.0	9.99	0.37	0.34	10.0	10.1	10.0	0.20	0.20

Note: STD is the concentration of HCl standard gas in the mixture. RSD (%) is the relative standard deviation of the repeated experiments. RPD (%) is a relative percentage difference between standard gas and analytical concentrations.

). Table 6 indicates that the NDIR analyzer’s accuracy was comparable to the ion chromatography method (RPD < 1.0%). Furthermore, the NDIR analyzer demonstrated very good precision (RSD < 2%). The sensitivity of the analyzer was improved, because it could measure HCl at 1 ppm even though its full scale was 50 ppm. The HCl NDIR analyzer can be used not only in metalworking factories but also in combustion processes in which various gas components are emitted.

3.3. Long-Term Field Trials of the Proposed HCl NDIR Analyzer

3.3.1. Long-Term Field Trials of the Proposed Analyzer at a Metalworking Factory

For a year, the performance of the HCl NDIR analyzer was confirmed in the metalworking factory. Variations of HCl emitted from the factory stack are shown in Figure S1 (Supplementary Information). Since the metalworking factory was not a 24 h operation, it was demonstrated that the gas emission was not constant and changed depending on operating hours.

Random samples were taken each month and analyzed using the IC method to verify the NDIR performance in the field.

Table 7. Average HCl concentrations obtained from the proposed NDIR analyzer and IC method.

No.	NDIR	Standard Method	RMSE	MNE (%)	MNB (%)	RPD (%)
1	0.21 ± 0.16	0.24 ± 0.03	0.15	51.0	−5.75	0.66
2	0.21 ± 0.16	0.22 ± 0.03	0.14	58.2	−5.56	0.30
3	0.46 ± 0.08	0.39 ± 0.06	0.14	31.9	31.9	1.64
4	0.33 ± 0.01	0.49 ± 0.16	0.12	19.8	−13.5	3.92
5	0.11 ± 0.14	0.24 ± 0.01	0.14	45.3	−31.2	3.32
6	0.17 ± 0.15	0.22 ± 0.05	0.09	36.3	17.6	1.19
7	0.06 ± 0.02	0.29 ± 0.14	0.29	82.2	−82.2	5.91
8	0.17 ± 0.10	0.23 ± 0.05	0.10	38.9	−38.9	1.51
9	0.30 ± 0.09	0.28 ± 0.05	0.04	13.0	−13.0	0.50
10	0.59 ± 0.20	0.57 ± 0.26	0.26	47.2	39.3	0.52
11	0.53 ± 0.20	0.40 ± 0.06	0.03	6.80	1.33	3.16
12	0.27 ± 0.26	0.35 ± 0.01	0.24	62.9	−62.9	2.03

shows a comparison of the results of NDIR and ion chromatography methods.

As shown in Table 7, although RMSE values were lower than 1, MNE values were high in most cases. Moreover, most MNB values were negative. This suggests that the HCl concentrations measured by the HCl NDIR analyzer were relatively smaller than those using the IC method. This occurred because the HCl concentrations were less than one ppm and close to the detection limit of the NDIR analyzer of concern (i.e., 0.07 ppm). Therefore, a slight difference in concentration could also result in considerable bias based on MNE and MNB. However, the RPD based on the actual field performance considering the emission standard denoted the acceptable relative errors with all values < 6%. Note that the acceptable RPD is ≤20% [79]. Consequently, although the proposed HCl NDIR analyzer could not measure low HCl concentration perfectly at this factory based on uncertainty analysis, it was still acceptable as an analyzer for a CEMS based on the national standard in terms of cost-effectiveness.

An additional comparison was conducted using third-party inspection data. Since the factory does not have a CEMS for HCl, a standard measurement institute randomly conducted a monthly HCl inspection. Their measurement at that time was compared with the NDIR data, as depicted in Figure 11.

As shown in Figure 11, most HCl concentrations obtained from the NDIR analyzer were lower than those from the IC standard method. The RMSE values for spring, summer, fall, and winter were 0.14, 0.07, 0.32, and 0.37, respectively. The MNE values for spring, summer, fall, and winter were 15.4, 7.31, 36.1, and 47.6, respectively. Likewise, the MNB values were 4.11, 7.31, −36.1, and 47.6, respectively. This suggests that the performance of the NDIR analyzer in spring and summer was better than in fall and winter. Therefore, the performance of the analyzer might be affected by seasonal variations. The wide range of stack gas temperatures led to the variation of moisture content in the flue gas, which might explain this difference. This small humidity effect is negligible if the HCl concentration is relatively high compared to full-scale measurement. However, because the concentration of HCl was lower than 1 ppm, this slight bias could cause a significant difference between the NDIR analyzer and the IC standard method.

3.3.2. Long-Term Field Trials of the Proposed Analyzer at a Cement Factory

The variations of HCl concentrations emitted from a cement factory are presented in Figure S2 (Supplementary Information). Because the trial permission was for only 6 months, the four-season performance of the NDIR analyzer could not be conducted. Performance was assessed during a part of fall and spring and the whole winter. As shown in Figure S2, the HCl concentrations emitted from the cement factory exhibited a lower trend compared to the metalworking factory.

In contrast to the case of the metalworking factory, at the cement factory, the HCl concentrations obtained from the NDIR analyzer were higher than those obtained from the IC method based on MNB values (

Table 8. Average HCl concentrations obtained from the proposed NDIR analyzer and IC method.

No.	NDIR	Standard Method	RMSE	MNE (%)	MNB (%)	RPD (%)
1	0.41 ± 0.19	0.39 ± 0.07	0.12	23.9	3.68	0.45
2	0.84 ± 0.25	0.36 ± 0.02	0.59	156	156	9.55
3	0.73 ± 0.18	0.34 ± 0.04	0.31	97.8	97.8	7.84
4	0.47 ± 0.15	0.32 ± 0.01	0.24	69.2	69.2	3.06
5	0.59 ± 0.65	0.51 ± 0.37	0.28	28.7	13.8	1.65
6	0.54 ± 0.66	0.89 ± 0.27	0.52	49.2	−41.9	7.07

). Although the MNE values were high, indicating the high bias between the NDIR analyzer and the IC standard method, the RPD values, considering a significantly small concentration compared to the measurement range, were less than 10% for 6 months of field trials. Hence, the proposed HCl NDIR analyzer could also be applied in this kind of industry.

Figure 12 presents the HCl obtained from the NDIR analyzer and the standard method in comparison with third-party data using the IC standard method.

As shown in Figure 12, the HCl concentrations obtained from the NDIR analyzer were lower than those obtained using the IC method. The RMSE values in fall, winter, and spring were 0.02, 0.06, and 0.02, respectively. This suggests that the values measured by the two methods were close. In fall, winter, and spring, the MNE values were 15, 37.5, and 6.7%, respectively. The MNB values were −15, −37.5, and 3.3%, respectively. These results suggest that the effect of seasonal variations needed to be clarified in this case. This might be because the field trials were conducted for only 6 months. The stable temperature of the stack emission might also be a reason for the lower influence of the seasonal temperature.

Advancements in HCl analyzers for CEMS, aside from a few for ambient air measurement, have not been documented recently. Halfacre et al. (2023) developed a HCl-TILDAS using a 3 µm interband cascade laser for ambient air [63]. The analyzer path length was 204 m, and its detection limit was 0.5 ppt. The measurement range was 0~7 ppb. The response time of the analyzer was 4.4 s. The bias caused by humidity was −5.8% with a humidity range of 60~93% [63]. Since the HCl-TILDAS had a very long path length, its detection range was much better than the current HCl NDIR analyzer in this study. A cavity ring-down spectrometer (CRDS) was developed to measure total gas chlorine [64]. The furnace temperature of the CRDS was 825 °C to convert chlorinated compounds to HCl gas. The sampling flow rate was 2 L/min. Although this CRDS measures HCl, it was developed to measure chlorinated compounds. The measurement range of generated HCl from chlorinated compounds was 2 to approximately 500 ppb [64]. Hence, this analyzer also had a higher sensitivity than the proposed NDIR analyzer in the current study. However, in terms of the cost effectiveness of applying the analyzer for a CEMS to reduce investment costs, the proposed NDIR analyzer was used for measuring HCl emissions practically.

4. Conclusions

A cost-effective NDIR analyzer for measuring HCl emitted from stationary sources was developed in this study. To enhance its performance, a zero-drift mitigation method, an optimal path length, and a negligible-interference bandpass filter were applied. Long-term field trials at a metalworking factory and a cement factory were conducted to evaluate the performance of the analyzer in real-world conditions. Furthermore, uncertainty and bias analyses were performed to compare the measurement results of the NDIR analyzer with those obtained using the IC standard method. It was found that a negligible-interference bandpass filter at 3.6 µm could mitigate the interference effects of humidity and other gases. The optimal path length for a measurement range of 0~50 ppm was determined to be 10 m. Regarding laboratory inspection, the zero and span drift were both less than 1%. The limit of detection, reproductivity, and linearity of the analyzer were 0.07 ppm, 0.26%, and 0.96%, respectively. These specifications are comparable to those of other commercial HCl analyzers. In field trials, although the concentrations of HCl emitted from metalworking and cement factories were low, the HCl NDIR analyzer achieved a relative error of less than 6% for the metalworking factory and 10% for the cement factory. However, MNE and MNB were high due to the HCl concentration being below 1 ppm. Seasonal effects were observed when the temperature of the emission stack varied significantly across the seasons. Although MNE and MNB suggested a high bias in the NDIR analyzer’s measurements, this deviation was deemed acceptable, considering emission standards and cost-effectiveness, especially since the measured HCl concentrations during testing were below 1 ppm. Consequently, the proposed HCl NDIR analyzer shows potential for application in CEMS at metalworking and cement factories. The current study’s limitation was the HCl’s low emission range during field trials, despite the analyzer’s measurement range being 0~50 ppm. Future studies should focus on testing the analyzer at higher HCl emission levels and developing a lower measurement range for the analyzer to enhance its versatility.