A biological aerosol particle classification device

Abstract

A biological aerosol particle classification device, the device comprises a humidity module, an airflow particle capture module, a dynamic refractive index measurement module and a data processing and identification module; the method induces the hygroscopic refractive index response of biological particles by precisely controlling the humidity environment (0-95% RH), synchronously collects multi-angle scattering light and polarization signals, inverses the particle refractive index under different humidity by using a hybrid optimization algorithm, constructs a dynamic response curve and extracts four characteristic parameters of initial refractive index, maximum change, response rate and inflection point humidity, and finally realizes particle classification through a random forest model. The utility model makes use of the dynamic refractive index response characteristics of biological particles, solves the misjudgment problem caused by refractive index overlap in static optical measurement, and the classification accuracy of 10 common biological aerosol particles reaches 96.3%, the single particle detection time is less than 5 minutes, and the utility model has the advantages of high precision, high specificity and automation.

Description

Technical Field

[0001] This invention belongs to the field of bioaerosol detection technology, and particularly relates to a bioaerosol particle classification device and method, which is suitable for high-precision and high-specificity identification of bioaerosol particles by controlling the ambient humidity and monitoring the changes in particle refractive index in real time. Background Technology

[0002] Bioaerosol particles are ubiquitous in the atmosphere, and their rapid and accurate identification is crucial for environmental monitoring, public health and safety, and allergy early warning. Particle identification technologies based on optical scattering characteristics largely rely on static refractive index measurements, i.e., obtaining a single refractive index value of the particle under specific environmental conditions as a classification criterion. However, biological particles (such as pollen, bacteria, and fungal spores) typically contain hydrophilic components (such as polysaccharides and proteins), and their refractive index dynamically changes with environmental humidity, exhibiting species-specific humidity response behavior. For example, pollen swells after absorbing moisture, leading to a decrease in refractive index, and the magnitude of this decrease and the corresponding response curves differ among different species.

[0003] In the existing technology, the identification equipment has the following defects: it does not build a humidity-controlled measurement environment, and cannot capture the dynamic change law of refractive index with humidity, so it is difficult to use this specific response feature to distinguish particles; traditional methods only use a single refractive index value as an identification feature, which makes it difficult to distinguish different particles with overlapping refractive indices; and there is a lack of feature extraction and classification methods for the "refractive index-humidity response curve".

[0004] Therefore, it is necessary to develop a bioaerosol particle identification technology and device that can regulate humidity in real time, monitor dynamic refractive index changes, and use response curve characteristics for high-precision classification, in order to make up for the shortcomings of existing technologies and improve identification capabilities and applicability. Summary of the Invention

[0005] To address the problems in existing bioaerosol particle identification technologies, such as insufficient consideration of the influence of environmental humidity, difficulty in distinguishing particles with similar refractive indices by static refractive index measurement, and lack of dynamic response feature extraction and classification mechanisms, this invention provides a bioaerosol particle classification device and method. By constructing a humidity-controlled measurement environment, the device acquires the refractive index change curves of particles under different humidity levels in real time, extracts multi-dimensional feature parameters, and combines machine learning algorithms to achieve accurate and automated identification of particle types.

[0006] The technical solution of this utility model is as follows:

[0007] A bioaerosol particle sorting device, characterized in that it includes:

[0008] Humidity module, used to generate and maintain a stable measurement environment with a relative humidity that varies continuously or in steps within the range of 0% to 95%;

[0009] An airflow particle capture module is connected to the airflow output end of the humidity control module. It is used to receive the controlled airflow and aerodynamically focus and stably confine aerosol particles to the center of the optical detection area.

[0010] The dynamic refractive index measurement module has its optical detection area spatially overlapping with the detection area of ​​the particle capture and positioning module, and is used to collect the scattered light signal of a single particle under different humidity conditions.

[0011] The data processing and recognition module is electrically connected to the signal output terminal of the dynamic refractive index measurement module. It is used to invert the particle refractive index based on the scattered light signal, construct the refractive index-humidity dynamic response curve, extract feature parameters, and perform classification and recognition.

[0012] Furthermore, the humidity module includes a humidity controller, a gas mixing chamber, and a temperature and humidity sensor disposed in the gas mixing chamber. The humidity controller is connected to the temperature and humidity sensor through a closed-loop feedback mechanism to adjust the mixing ratio of dry gas and humidified gas in real time.

[0013] Furthermore, the airflow particle capture module includes:

[0014] The gas sheath flow device has an inner flow channel connected to the aerosol sample inlet and an outer flow channel connected to the airflow output end of the humidity control module.

[0015] Furthermore, the dynamic refractive index measurement module includes:

[0016] A monochromatic laser source is positioned on one side of the optical detection area, with its optical path parallel to the particle flow direction;

[0017] A multi-angle scattering light detector array is arranged around the optical detection area at multiple scattering angle positions ranging from 0° to 180°.

[0018] A light trap is positioned opposite the monochromatic laser source to collect the laser light.

[0019] A polarization detector (which is not required for this device, but should not be used to limit the protection range of the device) is placed in the optical path at a 90° scattering angle to measure the polarization characteristics of the scattered light.

[0020] Furthermore, the multi-angle scattering light detector group includes at least five detectors at different angles, respectively positioned at scattering angles of 10°, 30°, 90°, 150° and 170°.

[0021] Furthermore, the data processing and recognition module includes:

[0022] The high-speed signal acquisition card is electrically connected to the signal output terminals of each detector in the dynamic refractive index measurement module;

[0023] An industrial computer, connected to the high-speed signal acquisition card, has a built-in refractive index inversion algorithm and a classification model trained based on machine learning.

[0024] Secondly, this utility model also provides a method for classifying bioaerosol particles using the above-mentioned device, characterized by comprising the following steps:

[0025] Aerosol particles are transported and fixed in the optical detection area by sheath flow aerodynamic focusing;

[0026] The humidity control module adjusts the humidity of the measurement environment from low to high according to a preset mode, and keeps it stable at each humidity gradient point until the particles and the ambient humidity are in balance.

[0027] At each stable humidity gradient point, the intensity and polarization state of multi-angle scattered light from the particles are simultaneously acquired using a dynamic refractive index measurement module.

[0028] Based on the collected optical signals, the complex refractive index of the particles under the current humidity is calculated using refractive index inversion.

[0029] By integrating refractive index data under different humidity levels, the refractive index-humidity dynamic response curve of the particle was plotted.

[0030] A set of characteristic parameters are extracted from the dynamic response curve, including the initial refractive index value and the maximum refractive index change Δn. max The refractive index decreases to 50% Δn max The required response time, and the humidity value corresponding to the inflection point of the curve;

[0031] The feature parameters are input into a pre-trained classification model, which outputs the classification results of particle types.

[0032] Furthermore, the refractive index inversion employs an optimization method based on Mie scattering theory or the T-matrix algorithm, wherein the optimization algorithm includes one or more combinations of gradient descent, genetic algorithm, or Levenberg-Marquardt algorithm.

[0033] Furthermore, the inversion algorithm is as follows:

[0034] 1) The core formula is as follows: For a spherical particle with radius r (r = d / 2) and complex refractive index n, under monochromatic light illumination with wavelength λ, the ratio of the scattered light intensity I(θ) to the incident light intensity I0 (relative scattered light intensity) can be expressed as:

[0035]

[0036] Where i1(θ) and i2(θ) are the angular distribution terms of the Mie scattering function, and are related to the particle's "scale parameter" x = 2πr / λ and relative refractive index (m = n / n). medium The ratio of the particle's refractive index to the medium's refractive index is directly related; the calculation of i1 and i2 requires an infinite series expansion (Mi series), which includes special functions such as Bessel functions and Hankel functions, and needs to be implemented through computer programming (such as using the Mi scattering module in the SciPy library of MATLAB or Python).

[0037] 2) Define a "theoretical vs. measured difference index" to measure the degree of match between the theoretical scattered light intensity corresponding to the currently assumed refractive index and the measured data. A commonly used objective function is the mean square error (MSE).

[0038]

[0039] Where: N is the number of angles measured (e.g., 10 angles); θ k It is the kth measured angle; I 理论 (θ k ,n) is θ calculated using the Mie forward model when the refractive index is assumed to be n. k The intensity of scattered light at an angle; I 实测 (θ k ) is the actual θ measured by the detector. k The intensity of scattered light at the angle. The smaller the objective function value, the better the match between theory and measurement, and the closer the corresponding n is to the true value.

[0040] 3) Since the objective function is usually nonlinear and multi-peaked (different n may correspond to similar scattering modes), an optimization algorithm is needed to search for the optimal solution within a reasonable refractive index range (e.g., n for biological particles is usually between 1.3 and 1.6). Commonly used algorithms include:

[0041] Gradient descent: Starting from the initial guessed value n0, n is iteratively updated along the direction of decreasing gradient of the objective function (e.g., ...). (α is the step size) until the MSE converges to its minimum. The advantage is that it is fast and suitable for scenarios where the initial value is close to the true value.

[0042] Genetic algorithms simulate biological evolution by iteratively optimizing the population through initialization (a set of possible n values), calculation of fitness (higher fitness means lower MSE), selection, crossover, and mutation, ultimately retaining the n value with the highest fitness. Their advantage is the ability to escape local optima, making them suitable for complex, multi-modal problems.

[0043] The Levenberg-Marquardt algorithm combines the advantages of gradient descent and Gauss-Newton methods. It is suitable for nonlinear least squares problems, widely used in Mie scattering inversion, and has a fast convergence speed and high stability.

[0044] 4) When the objective function MSE is less than a preset threshold (e.g., 10) -6 When the number of iterations reaches the upper limit, the optimization stops, and the corresponding value of n is the inversion result.

[0045] Compared with the prior art, the present invention has the following technical effects:

[0046] 1) Actively apply a controlled humidity change (stimulus) and monitor its refractive index change trajectory (response) in real time. Transform environmental interference factors (humidity) into favorable classification features, turn disadvantages into advantages, explore the specific response features of bioaerosol particles, and improve recognition accuracy.

[0047] 2) For the first time, the "refractive index-humidity response curve" is used as a specific identification feature for bioaerosol particles, solving the identification problem caused by static refractive index overlap; through joint classification of 4 types of feature parameters, the identification accuracy of 10 common biological particles reaches 96.3% (compared to 78.5% for traditional static methods); it is applicable to the identification of various bioaerosol particles such as pollen, bacteria, and fungal spores, with a wide identification range and universality; it realizes full automation from humidity control and signal acquisition to classification and identification, with a single particle detection time of ≤5 minutes.

[0048] 3) It can generate a humidity environment with continuous gradient or step change (0-95% RH, ±1%) and maintain stability through closed-loop feedback;

[0049] 4) Inversion methods based on optimization theory (such as genetic algorithm and LM algorithm) are used to solve the inverse problem of Mie scattering. Attached Figure Description

[0050] Figure 1 This is a YOZ plane schematic diagram of the bioaerosol particle identification device of this utility model.

[0051] Figure 2 This is a schematic diagram of the XOY plane of the bioaerosol particle identification device of this utility model.

[0052] Figure 3 This is a schematic diagram of the XOZ plane of the bioaerosol particle identification device of this utility model.

[0053] Figure 4 This is a three-dimensional schematic diagram of the bioaerosol particle identification device of this utility model. Detailed Implementation

[0054] The present invention will be further described below with reference to the accompanying drawings and embodiments, but this should not be construed as limiting the scope of protection of the present invention.

[0055] A bioaerosol particle identification device based on dynamic refractive index and humidity response characteristics, comprising:

[0056] Humidity Module 1: Composed of a humidity controller, a gas mixing chamber, and a temperature and humidity sensor, it can generate a continuous humidity gradient of 0-95%RH (relative humidity) (adjustment accuracy ±1%RH) and maintain stable humidity inside the chamber through closed-loop feedback;

[0057] Airflow particle capture module: includes a gas sheath flow device 2, which focuses and captures particles in the detection area and mixes them with sheath gas generated in the humidity module, thereby changing the humidity of the particles;

[0058] Dynamic refractive index measurement module: including laser 4, multi-angle scattered light detector 3 (covering 0°-180°), light trap 5, polarization detector (if any), to collect the intensity distribution and polarization state of scattered light from particles under different humidity levels in real time;

[0059] Data processing and recognition module: includes a high-speed signal acquisition card and an industrial computer, with built-in refractive index inversion algorithm and "response curve" feature classification model.

[0060] The identification method steps of the bioaerosol particle identification device based on dynamic refractive index and humidity response characteristics are as follows:

[0061] 1) Allow the bioaerosol particles to be tested to pass through the aerosol sampler and be focused onto the center of the detection area by the sheath flow device 2;

[0062] 2) The humidity module 1 causes the sheath gas humidity to change from low to high in a gradient. Because the sheath gas encapsulates the sample gas, the sample gas humidity will also change accordingly. The humidity stabilizes for a period of time at each gradient, ensuring that the sample gas humidity keeps up with the sheath gas humidity.

[0063] 3) Under each humidity gradient, the refractive index of the measured particles is inverted by the scattered light signal of aerosol particles in the sample gas (based on Mie scattering theory and T matrix algorithm, morphology correction is performed for non-spherical particles).

[0064] 4) Plot the "Refractive Index-Humidity Response Curve" and extract the characteristic parameters: initial refractive index (value at 30% RH), maximum refractive index change (Δn). max = Initial value - value at 90% RH), response rate (refractive index decreases to 50% Δn) max Required time), humidity at the curve inflection point (humidity value corresponding to the sudden change in refractive index);

[0065] 5) Input the extracted feature parameters into the trained random forest classification model and output the particle type (such as birch pollen, Escherichia coli, etc.).

[0066] Example:

[0067] Please see Figures 1 to 2 As shown, the bioaerosol particle sorting device in this embodiment includes a humidity module 1, an airflow particle capture module, a dynamic refractive index measurement module, and a data processing and identification module. The humidity module 1, which forms the environmental basis for dynamic measurement, consists of a humidity controller, a gas mixing chamber, and temperature and humidity sensors. The humidity controller (e.g., a PID controller) uses two mass flow controllers (MFCs) to control the flow ratio of dry gas (e.g., nitrogen) and humid gas (treated with a water bubbler), ensuring thorough mixing of the two gases in the gas mixing chamber to form an airflow with the target humidity. The gas mixing chamber is equipped with temperature and humidity sensors to monitor humidity values ​​in real time (accuracy ±1% RH) and feed them back to the humidity controller, forming a closed-loop control circuit to ensure the humidity of the output airflow remains stable at the set value. Adjusting the humidity controller can generate a "continuously gradual" or "step-like" humidity airflow, stabilizing at each step point for at least 30 seconds to ensure sufficient balance between the particles flowing through the detection area and the ambient humidity. The humidity airflow is connected to the particle capture module via a pipeline.

[0068] The airflow particle capture module includes a gas sheath flowmeter 2. The bioaerosol sample to be tested (such as particles of bacteria, viruses, pollen, etc.) enters the inner channel of the gas sheath flowmeter 2 through the sampler. A humidified airflow (sheath flow) with a specific humidity, generated by the humidity module 1, enters the outer channel of the gas sheath flowmeter 1. By adjustment, the sheath gas flow rate is made much greater than the sample gas flow rate (e.g., >10:1). Utilizing the hydrodynamic focusing effect, the particle sample flow is compressed into an extremely fine "core flow," arranged along the central axis to form a fine particle flow, allowing each particle to pass through the optical detection window individually, avoiding overlapping interference.

[0069] Dynamic refractive index measurement module, such as Figure 3As shown, the detection laser source is a helium-neon laser 4 with a wavelength of 632.8 nm and an output power of ~5 mW. Its emitted light, after beam expansion and collimation, passes through the aforementioned optical detection window and horizontally illuminates a single particle. The emitted light is collected by an optical trap 5. The scattered light from the particle is synchronously collected by multiple photomultiplier tube (PMT) detector groups 3 arranged at fixed angles. In this embodiment, detector group 3 includes five detectors located at 10° (small forward angle), 30° (forward), 90° (lateral), 150° (backward), and 170° (nearly 180° backward). These angles cover the region most sensitive to changes in scattered light intensity. A narrow-band filter (bandwidth ±2 nm) with a center wavelength of 633 nm is installed in front of each PMT to suppress stray light. Simultaneously, a polarization analyzer (consisting of a Wollaston prism and two photodiodes) is inserted in the optical path of the 90° detector to measure the polarization characteristics of the scattered light. This parameter is sensitive to the asymmetry of biological particles (such as capsid viruses and flagellated bacteria) and can serve as a supplementary feature for refractive index inversion, enhancing the feature dimension. As the humidity control module changes the gradient airflow (e.g., from low humidity to high humidity), the biological particles will expand due to hygroscopic absorption (e.g., protein shells absorbing water, pollen walls rupturing), causing the refractive index to change in real time. The above optical measurements are repeated at different humidity points, and the "dynamic response curve" (e.g., refractive index-humidity relationship graph) of refractive index changing with humidity is recorded.

[0070] The data processing and recognition module includes a high-speed signal acquisition card and an industrial computer, with a built-in refractive index inversion algorithm and a "response curve" feature classification model. The high-speed signal acquisition card (sampling rate typically ≥1MHz) converts optical signals (analog quantities) such as scattered light intensity and polarization state into digital signals, which are then transmitted to the industrial computer. The computer runs an inversion algorithm based on Mietheory—inputting the laser wavelength, particle size (estimated from the light intensity signal of the positioning system), and multi-angle scattered light intensity distribution—to calculate the refractive index of the particle at the current humidity. Simplified example of the inversion process: The scattered light intensity at 5 angles (10°, 30°, 90°, 150°, 170°) is measured: [I 10 ,I 30 ,I 90 ,I 150 ,I 170 Given λ = 632.8 nm, d = 2 μm (r = 1 μm), and the medium is air (n medium =1.0003); Assuming the search range of n is 1.3-1.6, initialize a set of candidate values ​​(e.g., 1.3, 1.35, ..., 1.6); For each candidate n, calculate x = 2πr / λ ≈ 9.948, m = n / 1.0003, and obtain the theoretical scattered light intensity [I'] through the Mie forward model. 10 ,I' 30 ,...,I' 170Calculate the MSE corresponding to each n, and find the n with the smallest MSE (e.g., 1.45), which is the inversion result. For the same particle, integrate the refractive index data under different humidity levels to generate a "dynamic refractive index-humidity response curve". For example, bacterial spores have a higher refractive index at low humidity (n≈1.58), while the refractive index decreases at high humidity due to water absorption and swelling (n≈1.52), while pollen grains may show a more significant abrupt change in refractive index (due to cell wall rupture and release of contents).

[0071] The model training and validation method for the bioaerosol particle identification device based on dynamic refractive index-humidity response characteristics is as follows:

[0072] 1) The training set collects dynamic refractive index data of 10 biological particles (5 pollen, 3 bacteria, and 2 fungi) at 30%-90% RH, with ≥50 samples of each particle type;

[0073] 2) The weights of the four types of characteristic parameters were determined by principal component analysis (PCA) (the maximum refractive index change has the highest weight, accounting for 35%).

[0074] 3) Random forest algorithm (100 decision trees), cross-validation accuracy of 96.3%, confusion matrix shows that only birch and hazel pollen have a 3.2% misclassification rate (12.7% misclassification rate of traditional method).

[0075] Experimental results show that the device of this invention achieves an average recognition accuracy of 96.3% for ten common bioaerosol particles, and the analysis time for a single particle is controlled within 5 minutes. Compared with the traditional method based on static refractive index, the performance is significantly improved.

[0076] The above description is merely a specific embodiment of this utility model, but the protection scope of this utility model is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this utility model should be included within the protection scope of this utility model.

Claims

1. A bioaerosol particle sorting device, characterized in that, include: Humidity module, used to generate and maintain a stable measurement environment with a relative humidity that varies continuously or in a stepwise manner within the range of 0% to 95%; An airflow particle capture module is connected to the airflow output end of the humidity module and is used to receive the regulated airflow and aerodynamically focus and stably confine aerosol particles to the center of the optical detection area. The dynamic refractive index measurement module has its optical detection area spatially overlapping with the detection area of ​​the particle capture and positioning module, and is used to collect the scattered light signal of a single particle under different humidity conditions. The data processing and recognition module is electrically connected to the signal output terminal of the dynamic refractive index measurement module. It is used to invert the particle refractive index based on the scattered light signal, construct the refractive index-humidity dynamic response curve, extract feature parameters, and perform classification and recognition.

2. The bioaerosol particle sorting device according to claim 1, characterized in that, The humidity module includes a humidity controller, a gas mixing chamber, and a temperature and humidity sensor disposed in the gas mixing chamber. The humidity controller is connected to the temperature and humidity sensor through a closed-loop feedback mechanism to adjust the mixing ratio of dry gas and humidifying gas in real time.

3. The bioaerosol particle sorting device according to claim 1, characterized in that, The particle capture and localization module includes: The gas sheath flow device has an inner flow channel connected to the aerosol sample inlet and an outer flow channel connected to the airflow output end of the humidity module.

4. The bioaerosol particle sorting device according to claim 1, characterized in that, The dynamic refractive index measurement module includes: A monochromatic laser source is positioned on one side of the optical detection area, with its optical path parallel to the particle flow direction; A multi-angle scattering light detector array is arranged around the optical detection area at multiple scattering angle positions ranging from 0° to 180°. A light trap is placed on the opposite side of the monochromatic laser source to collect the laser light.

5. The bioaerosol particle sorting device according to claim 4, characterized in that, The multi-angle scattering light detector group includes at least five detectors at different angles, respectively positioned at scattering angles of 10°, 30°, 90°, 150° and 170°.

6. The bioaerosol particle sorting device according to claim 1, characterized in that, The data processing and recognition module includes: A high-speed signal acquisition card is electrically connected to the signal output terminals of each detector in the dynamic refractive index measurement module; An industrial computer, connected to the high-speed signal acquisition card, has a built-in refractive index inversion algorithm and a classification model trained based on machine learning.

Images