Noise analysis of respirable dust by photoacoustic spectroscopy

Respirable dust is present in the atmosphere with an aerodynamic diameter of 7.07 μm or less, and the collection efficiency of dust with an aerodynamic diameter of 5 μm is 50%. Since the industrial revolution, with the rapid advancement of industrialization and urbanization, the world economy has been developing rapidly, and at the same time, it has brought a series of environmental pollution problems. Among them, atmospheric aerosol pollution has become a major source of environmental pollution,¹ causing photochemical smog,² respiratory diseases,³ and other problems that have a great impact on human health and normal life. Most cities have started to use smoke and dust emissions as one of the important indicators for testing and managing the environment. In particular, in the “13th Five-Year Plan” period, the country pays more and more attention to environmental pollution control, energy conservation, and emission reduction and continuously increases its efforts on environmental control, so the national air pollution control has achieved remarkable results.

At present, atmospheric environmental protection has become the consensus of all human beings, so air pollution control is gradually being paid attention to. Furthermore, research on dust detection sensors is beginning to emerge worldwide. One of the dust detection systems based on Mie scattering theory⁴ was designed to reduce the cost of dust detection using this system. After this, Zhou et al.⁵ used MODIS/VIIRS algorithm to inverse perform aerosol dust detection technique, and Brindley et al.⁶ used an infrared dust detection method based on SEVIRI algorithm, and the application of these techniques can make a significant improvement in dust detection efficiency. The key mainstream respiratory dust detection methods are electrostatic method,⁷ weighing method,⁸ and dust detection methods based on machine learning.⁹ However, these dust detection methods cannot meet the demand for sustainable real-time monitoring with high sensitivity, high stability, and low cost. A respiratory dust detection system based on photoacoustic spectroscopy technology is designed for this requirement. The photoacoustic spectroscopy technique uses a modifiable laser signal shot into a photoacoustic cell, and the gas in the photoacoustic cell undergoes absorption of light energy from the ground state to the excited state. Therefore, in using the photoacoustic spectroscopy technique, Zhang et al.¹⁰ proposed an online calibration technique for atmospheric oxygen based on the photoacoustic spectroscopy system, which was applied to atmospheric carbon dioxide concentration measurement and reduced the measurement error of the detection system. Yu et al.¹¹ used the properties of wavelets to process the photoacoustic signal of water in oil by the Mallat algorithm in wavelets and achieved a good noise reduction effect. Lewis¹² designed a photoacoustic cell system that can measure atmospheric aerosols and has good applications for atmospheric aerosol measurements. Based on these previous studies, a differential frequency generator¹³ was designed, and this design enabled weak gas sensing and improved the detection of trace gases. The subsequent emergence of cavity decay spectroscopy¹⁴ enhanced the detection accuracy by removing possible interference from other trace gases. Jin et al.¹⁵ developed a highly reflective mirror sensing system with a low-power blue diode laser to further improve the sensitivity of the detection system. Jean et al.¹⁶ used a new idea to improve the detection sensitivity of photoacoustic spectroscopy techniques by changing the internal structure of the photoacoustic cell and analyzing the effect of buffer gases on the photoacoustic signal.

These photoacoustic spectroscopy techniques are problematic in the detection of respiratory dust in the ambient atmosphere. In the process of respiratory dust detection using photoacoustic spectroscopy, subtle changes in the airflow velocity and changes in the surrounding external environment can cause fluctuations in the photoacoustic signal, and such fluctuations are unavoidable. The study of optimizing the detection system was conducted to filter out the noise signal generated during the experiment and make the data detected by the whole experimental system closer to reality. The influence factors of gas flow rate and pumping pump on the experiment were analyzed and combined with specific experiments. The function of prediction and correction of the Kalman filtering method is used to process the noise of the experimental system, and then, the combination of Kalman filtering and physical noise reduction is used to measure the ambient atmospheric respiratory dust.

The schematic diagram of the photoacoustic spectroscopy experimental system is shown in Fig. 2, including mass flow meter (Seven Star Huachuang), blue diode laser (DL-405, Shanghai Xilong Photoelectric), photoacoustic cell (self-researched by the group), signal generator (self-researched), pump, filter membrane (0.2 and 5 μm), preamplifier (MA221), lock-in amplifier (OE1022, CUHK), temperature and humidity sensor, reflector, and microphone. The signal generator emits two square wave signals with the same frequency of 1601 HZ, one way is delivered to the 405 nm diode laser with a power of 120 mW and excitation amplitude of 5 V, and the other way enters the lock-in amplifier to become the reference signal. The diode laser emits the laser beam, which is reflected by the reflector layer by layer and then enters the photoacoustic cell cavity perpendicular to the quartz window piece, and then interacts with the aerosol particles to produce the photoacoustic effect. The experimental system enhances the power by increasing the number of mirrors to improve the detection sensitivity, and the specific experimental bench of the photoacoustic system can be seen in Fig. 3.

Due to processing and assembly, there is a gap between the actual, theoretical, and software simulation values of the photoacoustic cell resonance frequency. Furthermore, the calibration of the photoacoustic cell resonance frequency is related to the resonance amplification effect of the photoacoustic signal in the photoacoustic cell. The resonance amplification of the photoacoustic cell can be realized to the maximum only when the modulation frequency of the signal generator is tuned to the same value as the resonance frequency of the photoacoustic cell. Since the aerosol has no obvious absorption peak, NO₂ gas with a known absorption cross-section was used to calibrate the system for the study.

During the experiment, a NO₂ branch mass flow meter was used to continuously introduce NO₂ gas into the photoacoustic cell for 10 min at a flow rate of 200 SCCM to remove the interfering gas. After the photoacoustic signal stabilized, the signal generator was adjusted to increase the modulation frequency in the range of 1400–1800 Hz in sequence. The measured photoacoustic signal was fitted by Lorentz, and the curve is shown in Fig. 4. As can be seen from Fig. 4, the photoacoustic signal amplitude increases and then decreases with the increase in the resonant frequency and takes the maximum value at 1601 HZ, so the photoacoustic cell resonant frequency is 1601 HZ. According to the fitting curve, the full linewidth of the frequency at $Smax2$ of the photoacoustic signal is about 56.86, so the quality factor is about 28.16.

From the function relation (6), y is the photoacoustic signal value, and x is the NO₂ concentration. 35.80 is the intercept of the function, which indicates that the background noise of the whole detection system is 35.80 μV. The slope of the function is 8.817, which indicates that the relationship between the photoacoustic signal value and NO₂ concentration is proportional to 8.817 μV/ppm.

The slope after linear fitting is 0.024, indicating that the ratio between the photoacoustic signal and the absorption coefficient of NO₂ is 0.024 µV/Mm⁻¹, that is, the product of the established photoacoustic spectrum detection system P_lightS_mC_cell is 0.024. Despite the small error, the correlation R² after linear fitting is 0.985, indicating that the established system for measuring the absorption coefficient of photoacoustic spectrum aerosol has good linearity and accuracy. For this reason, the aerosol absorption coefficients described in the following paper are obtained by the inversion of Eq. (6) calibration function.

For photoacoustic spectroscopy detection systems, increasing the inlet gas flow rate can increase the diffusion speed of aerosol particles in the photoacoustic cell and shorten the time for aerosol particles to reach concentration equilibrium, thereby improving the dynamic response speed of the photoacoustic system. However, an excessive inlet flow rate may bring additional flow noise and large data fluctuations, thus affecting the signal-to-noise ratio of the photoacoustic system. For this reason, experiments on the effect of inlet flow rate on background noise were conducted. The mass flow meter size is adjusted to 100, 200, 300, 400, and 500 SCCM, and the data were recorded and saved in real time, as shown in Fig. 9.

As can be seen from Fig. 9, with the increase in the inlet flow rate, the background noise fluctuates even more, and when the inlet flow rate reaches 400 SCCM, the variance increases by an order of magnitude. It can be seen that when the dynamic response speed of the photoacoustic spectroscopy system is considered, the stability of the photoacoustic signal should also be considered, so the inlet gas flow rate of the photoacoustic spectroscopy detection system should not be greater than 200 SCCM.

In the actual experiment, the small pumping pump is prone to blocking the suction phenomenon. To visually illustrate the noise and vibration generated by the small pumping pump and the noise generated by the gas flow on the photoacoustic signal, two sets of comparison experiments are designed as follows: in the first set of experiments, the diode laser and the small pump are turned off; and in the second set of experiments, the diode laser is turned off and the small pump is turned on. Because there is no light source, there is no photoacoustic signal due to the photoacoustic effect, and the photoacoustic signal displayed by the lock-in amplifier is all the background signal generated by various noises. The data results are shown in Fig. 10.

As can be seen from Fig. 10, because the pumping pump is turned on, the mean value of the background noise signal increases from 0.087 to 62.984 μV, and the variance increases from 0.047 to 10.350. It can be seen that when the small pumping pump is used as the system gas circuit power source, the background noise signal caused by the small pumping pump noise and vibration and gas flow noise are all superimposed in the photoacoustic signal, increasing the system measurement error.

From the comparative analysis of Figs. 11(a) and 11(c), it can be seen that the magnitude of the background noise signal due to the noise and vibration of the small pumping pump is about 62.954 μV, and the interference of the small pumping pump increases the fluctuation of the amplitude of the photoacoustic signal. From Figs. 11(a) and 11(b), it can be seen that the noise reduction method reduces the average value of the background noise signal of the small pumping pump by 11.221 μV, which is 17.82%, and the variance is reduced from 10.431 to 7.054. The amplitude of the photoacoustic signal is more stable.

During the detection of respiratory dust in the atmosphere using photoacoustic spectroscopy, it can be seen from Fig. 9 that the inlet flow rate and the pumping pump will have a great impact on the detection results, and both will generate noise superimposed on the photoacoustic signal during the operation. In the actual outside world for environmental testing, as the outside world may suddenly produce strong noise, this will have an impact on the detection results. To further improve the detection accuracy and maintain the stability of the detection results, the Kalman filtering noise reduction method is used. The strong noise generated by the entire detection system is reduced so that the noise superimposed on the photoacoustic signal is further reduced. Kalman filtering does not lie in how much smaller the deviation it estimates but in its clever integration of the observed data with the estimated data. Calibration of estimated data with observed data, so that the estimated data for closed-loop feedback management, so the Kalman filter the advantage of the time is that it can still maintain stability. This experimental detection system needs to carry out environmental detection for a long time, the components in the ambient atmosphere have complexity, and the traditional lock-in amplifier output cannot meet the needs of this experimental system. Therefore, this experimental system uses a combination of lock-in amplifier and Kalman filtering, as shown in Fig. 13.

When state quantity X_k and observation quantity Y_k meet the conditions of Eqs. (8) and (9), and system process noises Q and R are the Gaussian noises, then the estimated value X_k at time k can be solved according to the following steps.

Kalman filtering simulation is performed in LabVIEW, and a triangular waveform with the amplitude of the photoacoustic signal of 1 is set in LabVIEW. Meanwhile, to simulate complex working conditions, a frequency interference signal with an amplitude of twice the photoacoustic signal is superimposed.

As shown in Fig. 14, the signal value after adding noise increases three times compared with the original signal, and if the real signal cannot be extracted efficiently, it will lead to serious distortion of the measurement results and cannot get the required photoacoustic signal value. Figure 15 shows the signal value after Kalman filtering and the traditional lock-in amplifier output signal for comparison after filtering out 88.7% of the noise. It is proved that Kalman filtering has a good filtering effect and can be applied to this experimental system.

The standard NO₂ was diluted to 4.29, 3.85, 3.33, 2.00, 1.11, and 0.566 ppm by precisely controlling mass flow meter 1 and mass flow meter 2, and the NO₂ absorption coefficients were recorded at this time at a flow rate of 200 SCCM, as shown in Table I.

The aerosol generator (Met One) was used to study the relationship between aerosol absorption coefficient and particle size and number concentration. The target microsphere densities were diluted to 3.3%, 8.3%, 13.3%, and 18.3%, forming different numbers of particles with different concentrations at the same particle size. The diluted particles with different numbers of concentrations below 5 µm were passed into the photoacoustic cell at 200 ml/min, and the absorption coefficients at different numbers of concentrations were recorded, as shown in Table II.

The ambient atmospheric test for respiratory dust was conducted for 12 hours using 5 and 0.2 μm filter membranes. As shown in Figs. 16(a) and 16(b), the absorption coefficients of atmospheric NO₂ and its related chemicals were measured under the 0.2 μm filter membrane, while the data under the 5 μm filter membrane were measured for atmospheric aerosols and NO₂ and its related chemicals. Figure 16(c) shows the difference between the data measured under the 5 μm filter membrane and the data measured under the 0.2 μm filter membrane. The data in Fig. 16(a) minus the data in Fig. 16(b) is the atmospheric aerosol respiration coefficient for the removal of NO₂ and its chemicals. As shown in Fig. 16(a), the absorption coefficients of ambient atmospheric respiratory dust for a 5 μm filter membrane ranged from ∼90 to 124 Mm⁻¹ during the measurement period, with an average value of 108 Mm⁻¹ and a deviation of 9.7 Mm⁻¹ when analyzing the data. Figure 16(c) shows the trend of atmospheric aerosol absorption coefficients based on the difference between Figs. 16(a) and 16(b), and from Fig. 16(c), it can be seen that the highest atmospheric aerosol absorption coefficient is 45 Mm⁻¹ and the lowest absorption coefficient is 22 Mm⁻¹.

A system for measuring respiratory dust by photoacoustic spectroscopy based on a low-power blue diode laser with a highly reflective mirror sensor was developed. By conducting experiments to analyze the effects of the inlet flow rate and pumping on the detection system, it was concluded that the inlet flow rate and pumping would generate more noise in the experimental system. After the experiments, it was found that the inlet air flow rate was set to 200 SCCM, and a combination of physical noise reduction and Kalman filtering was used to reduce the influence of the pump on the experimental system. After a series of experiments and simulations, it can be seen that the noise generated by the test system can be reduced to the ideal state by using a combination of physical noise reduction and Kalman filtering. Ambient atmospheric testing was conducted, and the variation curve of respiratory dust absorption coefficient was obtained by using the difference method, and the variation curve of absorption was compared with the data published by the local environmental department, showing that it could meet the ambient atmospheric respiratory dust detection.

This work was supported by the Major Project of Natural Science Research in Universities of Anhui Province (Grant No. KJ2021ZD0052) and the National Foundation of China (Grant No. 51904009).

The authors have no conflicts to disclose.

Hua-wei Jin: Funding acquisition (lead); Conceptualization (lead). Hao-wei Wang: Writing-review & editing (equal).

Hua-wei Jin: Writing – original draft (equal). Hao-wei Wang: Writing – original draft (equal).

The data that support the findings of this study are available from the corresponding author upon reasonable request.