Method and system for multi-modal monitoring of dust accumulation in dust removal pipeline

Abstract

The application discloses a dust-removal pipeline dust-accumulation multi-mode monitoring method and system, and relates to the technical field of dust-accumulation monitoring. The monitoring method comprises the following steps: using a microwave detection module to detect microwave characteristic values; using an optical auxiliary module to obtain optical characteristic values; using a density estimation algorithm based on the microwave characteristic values and the optical characteristic values to estimate equivalent density rho; and calculating deposition equivalent mass M based on the equivalent density rho. The deposition equivalent mass M is a quantitative index that can more accurately reflect dust-accumulation working conditions and can provide judgment of special working conditions such as caking, loose dust accumulation and local blockage. Therefore, the application can adapt to complex dust-conveying working conditions to accurately quantify dust-accumulation information, and is beneficial to accurately realizing risk assessment and providing maintenance suggestions. Moreover, the multi-mode monitoring method is convenient for self-diagnosis and self-calibration, and is beneficial to avoiding the situation that single means influences dust-accumulation information feedback due to material characteristics, environmental changes and the like.

Description

Technical Field

[0001] This invention relates to the field of dust accumulation monitoring technology, and in particular to a multimodal monitoring method and system for dust accumulation in dust removal pipelines. Background Technology

[0002] Dust collection ducts are key facilities in dust collection systems for transporting dust between equipment. Dust accumulation within them is unavoidable, increasing energy consumption, reducing transport efficiency, and potentially causing safety accidents. Therefore, precise monitoring of dust accumulation is essential for timely cleaning and ensuring the safe and efficient operation of the dust collection system. Currently, the main dust accumulation monitoring technologies for dust collection ducts include the following.

[0003] The first method is ultrasonic ranging, which involves installing an ultrasonic sensor through an opening in the top of the pipe and measuring the distance from the sensor to the dust surface to indirectly calculate the dust thickness. However, this method is easily affected by the dust material, temperature, and humidity, resulting in unstable accuracy and the inability to determine the density of the deposits.

[0004] The second method is the pressure difference monitoring method. This method indirectly infers the changes in the flow cross section caused by deposition by monitoring the pressure difference between the two ports of the pipeline. However, it cannot reflect the deposition location and deposition morphology, making it difficult to achieve early warning and location treatment based on it. Moreover, the inference results are easily affected by the total air volume of the system.

[0005] The third method is the optical imaging method. The latest technology of this method involves inserting an optical prism and an image acquisition unit into the pipe to obtain a cross-sectional image of the dust accumulation, and then obtaining dust information through image analysis. Although this method can obtain information such as dust thickness, area, and texture, it cannot obtain dust density, and therefore cannot accurately conduct risk assessments or provide cleaning recommendations. In addition, because the optical prism is still in the harsh environment of the pipe, it is prone to image inaccuracy due to dust contamination, and image inaccuracy is difficult to detect in a timely manner, which will continuously affect the accuracy of the detection data.

[0006] The fourth method is acoustic emission, which determines the deposition by analyzing the acoustic signals of particles hitting the pipe wall, but it requires complex signal processing and is difficult to accurately quantify the thickness.

[0007] In summary, existing mainstream monitoring technologies have their own limitations. These limitations prevent a single technology from accurately quantifying dust accumulation information, especially dust density information, in complex operating conditions such as dust removal pipelines. Consequently, they cannot accurately achieve risk assessment and provide maintenance recommendations. Summary of the Invention

[0008] The technical problem to be solved by the present invention is to provide a multimodal monitoring method and system for dust accumulation in dust removal pipelines that can accurately quantify dust accumulation information in complex dust conveying conditions, facilitate accurate risk assessment and maintenance recommendations, and facilitate self-diagnosis and self-calibration.

[0009] To solve the above-mentioned technical problems, the technical solution of the present invention is: a multi-modal monitoring method for dust accumulation in dust removal pipelines, including using a microwave detection module to measure microwave characteristic values, using an optical auxiliary module to obtain optical characteristic values, using a density estimation method based on the microwave characteristic values ​​and the optical characteristic values ​​to estimate the equivalent density ρ, and calculating the deposition equivalent mass M based on the equivalent density ρ, wherein the deposition equivalent mass M is used to reflect the deposition conditions;

[0010] The density estimation method includes fusing the microwave feature value and the optical feature value to construct a feature vector, inputting the feature vector into a trained Gaussian process regression model to obtain the density prediction mean; applying physical constraints to the density prediction mean to limit it to a range that is not lower than the air density and not higher than the maximum possible density determined by the spectral characteristics of the material, and finally outputting the equivalent density ρ.

[0011] The deposition equivalent mass M is calculated using the following formula:

[0012] M = k•ρ•S_cross;

[0013] in:

[0014] k is the pipe cross-sectional shape correction factor;

[0015] ρ is the equivalent density;

[0016] S_cross is the deposition cross-sectional area, obtained using the optical auxiliary module.

[0017] As a preferred technical solution, the method further includes a density estimation verification step, which includes density value comparison and confidence level calculation.

[0018] As a preferred technical solution, the method also includes a model calibration step, which includes an initial calibration stage and an online calibration stage.

[0019] As a preferred technical solution, the online calibration stage includes: measuring the ultrasonic dust accumulation thickness H using the ultrasonic ranging module, obtaining the optical deposition thickness H_opt using the optical auxiliary module, and periodically or when the deviation between the optical deposition thickness H_opt and the ultrasonic dust accumulation thickness H exceeds a threshold, fitting a new calibration curve with the historically accumulated optical deposition thickness H_opt and the ultrasonic flight time T, and updating the calibration coefficient of the ultrasonic ranging module.

[0020] As a preferred technical solution, the online calibration stage further includes: using the optical auxiliary module to obtain a reference density ρ_opt_ref, and using the reference density ρ_opt_ref to fine-tune the parameters of the Gaussian process regression model so that the density estimation method can adapt to changes in material properties.

[0021] As a preferred technical solution, the initial calibration stage includes the following steps:

[0022] BS1: Under clean pipeline conditions, the optical auxiliary module acquires a zero-deposition reference surface;

[0023] BS2: Inject a standard material of known thickness, and collect data using the ultrasonic ranging module, the microwave detection module, and the optical auxiliary module respectively;

[0024] BS3: Establish an initial mapping relationship for data collected after at least two injections of standard material.

[0025] As a preferred technical solution, the microwave characteristic values ​​include microwave attenuation A and phase shift value Φ, and the optical characteristic values ​​include optical deposition thickness H_opt, surface roughness Ra, and spectral characteristics.

[0026] A multi-modal monitoring system for dust accumulation in dust removal pipelines, using the aforementioned multi-modal monitoring method for dust accumulation in dust removal pipelines, and comprising:

[0027] Ultrasonic ranging module, used to measure ultrasonic deposition thickness H;

[0028] Microwave detection module, used to generate microwave characteristic values;

[0029] An optical auxiliary module is used to acquire cross-sectional images of dust accumulation.

[0030] The processing and control unit is communicatively connected to the ultrasonic ranging module, the microwave detection module, and the optical auxiliary module, and is configured to execute:

[0031] The ultrasonic deposition thickness H, the microwave characteristic value, and the dust accumulation profile image are received, and the optical characteristic value is obtained based on the dust accumulation profile image;

[0032] Run the density estimation method to calculate the equivalent density ρ;

[0033] Calculate the deposition equivalent mass M.

[0034] The optical auxiliary module includes a telescopic optical prism that extends into the pipe when it is necessary to acquire a dust accumulation profile image and exits the pipe when it is not necessary to acquire a dust accumulation profile image.

[0035] By adopting the above technical solution, the present invention uses the microwave detection module and the optical auxiliary module to obtain microwave characteristic values ​​and optical characteristic values ​​respectively, uses the density estimation method to fuse the microwave characteristic values ​​and optical characteristic values ​​to obtain the equivalent density ρ, and further calculates the deposition equivalent mass M based on the equivalent density ρ. The deposition equivalent mass M is a quantitative indicator that can more accurately reflect the dust accumulation conditions. For example, the conditions it can reflect are as follows.

[0036] In addition, the depositional equivalent mass M can also provide a basis for determining the following special operating conditions:

[0037] (1) If the value of M rises rapidly but the dust thickness value changes little, it can be judged that the dust density has increased and there is a risk of clumping. Effective cleaning should be arranged immediately.

[0038] (2) If the value of M is stable but the dust accumulation thickness increases, it can be determined that the dust accumulation is loose and can be effectively cleaned by blowing or other methods.

[0039] (3) An abnormally high M / ρ value indicates uneven deposition, which is a precursor to local blockage. Effective cleaning should be arranged immediately.

[0040] In summary, through multi-module monitoring and fusion calculation, this invention can accurately quantify dust accumulation information to adapt to complex dust conveying conditions, facilitating precise risk assessment and providing maintenance recommendations. Furthermore, the invention employs multi-mode monitoring, enabling self-diagnosis and self-calibration through monitoring data from different modules, thus forming mutual verification and compensation. This further helps avoid situations where a single method is affected by material characteristics, environmental changes, or other factors influencing dust accumulation information feedback. Attached Figure Description

[0041] The following figures are intended only to illustrate and explain the present invention and do not limit the scope of the invention. Wherein:

[0042] Figure 1 This is a logic diagram of the monitoring operation of this invention. Detailed Implementation

[0043] The present invention will be further described below with reference to the accompanying drawings and embodiments. In the following detailed description, only certain exemplary embodiments of the invention are described by way of illustration. Undoubtedly, those skilled in the art will recognize that various modifications can be made to the described embodiments without departing from the spirit and scope of the invention. Therefore, the drawings and description are illustrative in nature and not intended to limit the scope of the claims.

[0044] like Figure 1As shown, the multimodal monitoring method for dust accumulation in dust removal pipelines includes using a microwave detection module to measure microwave characteristic values, using an optical auxiliary module to obtain optical characteristic values, using a density estimation method based on the microwave characteristic values ​​and the optical characteristic values ​​to estimate the equivalent density ρ, and calculating the deposition equivalent mass M based on the equivalent density ρ. The deposition equivalent mass M is used to reflect the deposition conditions.

[0045] The density estimation method includes fusing the microwave feature value and the optical feature value to construct a feature vector, inputting the feature vector into a trained Gaussian process regression model to obtain the density prediction mean; applying physical constraints to the density prediction mean to limit it to a range that is not lower than the air density and not higher than the maximum possible density determined by the spectral characteristics of the material, and finally outputting the equivalent density ρ.

[0046] The deposition equivalent mass M is calculated using the following formula:

[0047] M = k•ρ•S_cross;

[0048] in:

[0049] k is the pipe cross-sectional shape correction factor, which is 0.85 to 0.95 for circular pipes;

[0050] ρ is the equivalent density;

[0051] S_cross is the deposition cross-sectional area, obtained using the optical auxiliary module.

[0052] The equivalent density ρ is obtained by first fusing the microwave and optical characteristic values ​​mentioned above. The deposition equivalent mass M can then be calculated. The deposition equivalent mass M is a quantitative indicator that can more accurately reflect the dust accumulation conditions. For example, the conditions it can reflect are as follows.

[0053] In addition, the depositional equivalent mass M can also provide a basis for determining the following special operating conditions:

[0054] (1) If the value of M rises rapidly but the dust thickness value changes little, it can be judged that the dust density has increased and there is a risk of clumping. Effective cleaning should be arranged immediately.

[0055] (2) If the value of M is stable but the dust accumulation thickness increases, it can be determined that the dust accumulation is loose and can be effectively cleaned by blowing or other methods.

[0056] (3) An abnormally high M / ρ value indicates uneven deposition, which is a precursor to local blockage. Effective cleaning should be arranged immediately.

[0057] In summary, by fully utilizing the spatial information of multimodal data for fusion calculation, it is possible to accurately quantify dust accumulation information in complex dust transportation conditions, which is conducive to accurately achieving risk assessment and providing maintenance recommendations.

[0058] The microwave characteristic values ​​include microwave attenuation A and phase shift Φ, and the optical characteristic values ​​include optical deposition thickness H_opt, surface roughness Ra, and spectral characteristics.

[0059] Specifically, the microwave detection module is a penetration-type detection module. The transmitted signal is emitted from the transmitting antenna in the form of a linear frequency modulated continuous wave (FMCW) with a center frequency of f0. After penetrating the pipe wall δ1, the deposition layer δ2, the air δ3, and the pipe wall δ4 in sequence, it is received by the receiving antenna. The extraction process of the microwave feature values ​​is as follows.

[0060] Transmitted signal S_T(t) = cos[2π(f0t + 0.5αt)] 2 )], where α is the frequency modulation frequency; the transmitted signal is a continuous wave with a frequency starting from f0 and increasing linearly with a slope of α. This type of signal can simultaneously possess the advantages of continuous wave (high ranging accuracy) and pulse (good resolution).

[0061] The received signal S_R(t) = A´•cos[2π(f0(t-τ)+0.5α(t-τ)²+φ0)], where A´ is the total amplitude of the received signal, τ is the total time delay of the signal on the transmission path, and φ0 is the phase offset correction term.

[0062] The original signal and the received signal are orthogonally downconverted and digitized to obtain the baseband signal I(t) + jQ(t) = A0•exp[j(2πf_b•t+φ_total)], where A0 is the complex amplitude of the baseband signal, f_b is the beat frequency, and φ_total is the total phase of the baseband signal.

[0063] The microwave attenuation A = 20•log 10 (│FFT{I+jQ}│_peak / P_ref), where │FFT{I+jQ}│_peak is the amplitude of the peak value of the spectrum, and P_ref is the system reference frequency.

[0064] The phase offset value Φ = unwrap[angle(FFT{I+jQ}_peak)].

[0065] In the above extraction process, the transmitted signal experiences a time delay τ after passing through the dust, which is related to the deposition thickness and dielectric properties. After orthogonal down-conversion and digitization, the signal is converted into a baseband signal I(t) + jQ(t). A Fast Fourier Transform is performed on this baseband signal, and the microwave attenuation A and phase shift Φ are extracted at the peak of the spectrum at the beat frequency f_b. Of course, other derived features can be further obtained through the microwave detection module, such as obtaining the dispersion slope dA / df by analyzing the rate of change of signal attenuation with frequency within the swept bandwidth, and obtaining the polarization ratio (VV / HH) by comparing the attenuation of the microwave signal under different polarization directions (vertical and horizontal). The microwave attenuation A comprehensively reflects the dielectric loss and scattering loss of the dust, the phase shift Φ is mainly related to the real part of the dust dielectric constant, the dispersion slope dA / df indicates the moisture content of the dust, and the polarization ratio (VV / HH) reflects the anisotropy of particle shape.

[0066] The optical auxiliary module employs an endoscopic prism, refracting or reflecting images of the internal cross-section through the prism, which are then captured by an image acquisition device located outside the tube. Image acquisition methods include, but are not limited to, triggering a structured light projector during detection to project an coded grating onto the deposition surface, simultaneously acquiring deformed grating images (multispectral mode), and reconstructing the 3D cross-section using a phase unwrapping algorithm. Image analysis then extracts ground truth parameters such as the optical deposition thickness H_opt, surface roughness Ra, and spectral characteristics. It also obtains the deposition cross-sectional area S_cross, and further acquires the reference density ρ_opt_ref, deposition uniformity index, and thickness variation index. The principle behind acquiring this data using this optical auxiliary module is well-known and will not be elaborated upon here.

[0067] The present invention also provides a specific method for calculating the deposition equivalent mass M, including the following calculation steps.

[0068] DS1. Mesh the pipe cross-section as Δx_i×Δy_j.

[0069] DS2. Based on the density estimation method, assign density values ​​ρ_ij to each grid cell.

[0070] DS3, Integrate over the sedimentary equivalent mass M:

[0071] M=k•Σ_iΣ_jρ_ij•H_ij•Δx_i•Δy_j≈k•ρ•S_cross.

[0072] The density assignment of grid cells in this calculation method is also based on the density estimation method. This is equivalent to using the same extracted features, the same Gaussian process regression model, and the same physical constraints as the overall density estimation in the density estimation method. Therefore, the calculation of the final depositional equivalent mass M can be equivalent to using the above formula "k•ρ•S_cross", which can achieve the purpose of quickly and accurately calculating the depositional equivalent mass M.

[0073] Preferably, the method for calculating the deposition equivalent mass M further includes DS4, correcting the pipe inclination, M_actual=M / cosθ, where θ is the inclination angle.

[0074] The acquisition of characteristic values ​​and the calculation of depositional equivalent mass M in the above monitoring methods are explicit. This invention only provides the following specific examples to illustrate the density estimation method in more detail.

[0075] In this example, the microwave detection module obtains the following microwave characteristic values.

[0076] The optical auxiliary module obtains the following optical feature values.

[0077] According to the density estimation method, a feature vector is constructed, and based on the above source data, a 28-dimensional feature vector is constructed. This 28-dimensional feature vector includes 5 dimensions of microwave features, 8 dimensions of optical features, and 15 dimensions of interaction features between the two.

[0078] Among them, the five dimensions of microwave characteristics are microwave attenuation A, phase shift Φ, A / Φ ratio, dispersion slope dA / df, and polarization ratio (VV / HH). The source code example is as follows.

[0079] Python

[0080] import numpy as np

[0081] # 1. Microwave characteristics (5 dimensions)

[0082] microwave_raw = np.array([

[0083] -28.5, # A_dB

[0084] 1.75, # Φ_rad

[0085] -28.5 / 1.75, # A / Φ ratio

[0086] 0.12, # Dispersion slope

[0087] 1.05 # Polarization ratio

[0088] ]).

[0089] The 8-dimensional microwave feature includes five spectral features, surface roughness Ra, deposition uniformity index, and thickness variation coefficient. An example of the source code is shown below.

[0090] # 2. Optical Derived Features (8 Dimensions)

[0091] optical_features = np.array([

[0092] 0.15, 0.22, 0.18, 0.25, 0.20, #Spectral characteristics λ1-λ5 (5 dimensions)

[0093] 12.5 / 50, # Surface roughness Ra (After normalization, 12.5 / 50=0.25, assuming a maximum roughness of 50μm)

[0094] 0.78, # Depositional homogeneity index

[0095] 0.15 # Thickness variation coefficient

[0096] ]).

[0097] In the interactive features, microwave features are 5-dimensional, and optical features include spectral features, both of which are 5-dimensional. The complete combination results in 25-dimensional interactive features, which is prone to the curse of dimensionality. Therefore, in constructing the interactive features, based on physical mechanism correlation and sensitivity analysis results, the first two core features—microwave attenuation A and phase shift Φ—are selected from the microwave features, and the first three spectral features—λ1, λ2, and λ3—are selected from the optical features. These are then combined in a targeted manner to retain key information while avoiding the curse of dimensionality. Specifically, microwave attenuation A is directly related to the dielectric properties of the material, reflecting the overall electrical properties of the material, and its sensitivity to density changes is as high as 8.2 dB / (100 kg / m³). 3 The phase shift value Φ is strongly correlated with the material density, and its sensitivity to density changes is as high as 0.12 rad / (100 kg / m³). 3 The chemical composition characteristics of the material surface layer corresponding to λ1-λ3 (450-650nm band) are empirically correlated with density, with a correlation coefficient of up to 0.78.

[0098] The specific feature selection and interactive source code examples are as follows.

[0099] interaction_features = []

[0100] microwave_subset = microwave_raw[:2] # A_dB and Φ_rad (core sensitive features)

[0101] optical_subset = optical_features[:3] #λ1, λ2, λ3 (near ultraviolet-visible band, with the strongest correlation to density)

[0102] # Targeted interaction rules (ensure the dimensions are precisely 15):

[0103] # Rule 1: Microwave features × Optical features (multiplicative interaction, 2 × 3 = 6 dimensions)

[0104] for m in microwave_subset:

[0105] for o in optical_subset:

[0106] interaction_features.append(m * o)

[0107] # Rule 2: Microwave features - optical features (difference interaction, 2×3=6 dimensions)

[0108] for m in microwave_subset:

[0109] for o in optical_subset:

[0110] interaction_features.append(abs(m - o))

[0111] # Rule 3: Microwave features / (Optical features + 1) (Normalized division interaction, 2×3=6 dimensions) → Filter core 3-dimensional supplement

[0112] interaction_candidates = [m / (o+1) for m in microwave_subset for o inoptical_subset]

[0113] interaction_features.extend(interaction_candidates[:3]) # Filter the 3 dimensions with the highest relevance.

[0114] The above interactive features include 6 dimensions of multiplication interaction, 6 dimensions of difference interaction, and 3 core dimensions of normalized division interaction, totaling 15 dimensions. This completes the construction of the 28-dimensional feature vector.

[0115] In the Gaussian process regression model, the training example is based on a dataset containing 500 sets of standard materials, covering three typical dust removal materials: coal powder, mineral powder, and fiber dust, with a deposition density range of 1.2–2500 kg / m³. 3 The deposition thickness is 0.001 to 0.1 m. The true density of each group of standard materials is obtained by sampling and weighing (accuracy ±0.1 kg) to ensure the generalization ability of the model.

[0116] The dataset is structured as follows.

[0117] Material type distribution: coal powder (60%), mineral powder (30%), fiber dust (10%).

[0118] Density distribution: Logarithmic uniform sampling was used to cover three types of working conditions: low density (<100 kg / m³), medium density (100~1000 kg / m³), and high density (>1000 kg / m³).

[0119] Deposition thickness distribution: uniformly distributed in the range of 0.001 to 0.1 m.

[0120] Environmental conditions: Temperature range -10℃ to 60℃, humidity range 10%-90% RH.

[0121] Noise processing is performed on various source data separately. For example, microwave data is filtered using a moving average filter (window size = 5) to reduce random noise, optical data is filtered using a median filter (3×3 window) to remove image noise, and ultrasound data is denoised using a threshold denoising method based on wavelet transform.

[0122] Each set of standard materials contains an input 28-dimensional feature vector (constructed as above).

[0123] The following is an example of the key source code for training a Gaussian process regression model.

[0124] Python

[0125] from sklearn.gaussian_process import GaussianProcessRegressor

[0126] from sklearn.gaussian_process.kernels import RBF, WhiteKernel,ConstantKernel

[0127] # Use composite kernel functions to improve fitting ability

[0128] kernel=ConstantKernel(1.0)*RBF(length_scale=1.0)+WhiteKernel(noise_level=0.1)

[0129] # Initialize Gaussian process regressor

[0130] gp = GaussianProcessRegressor(

[0131] kernel = kernel

[0132] n_restarts_optimizer=10,

[0133] alpha=1e-4 # Increase numerical stability )

[0135] # Assume we already have training data X_train (500×28) and y_train (500×1)

[0136] gp.fit(X_train, y_train)

[0137] print(f"Optimized kernel function: {gp.kernel_}")

[0138] print(f"Logarithmic marginal likelihood: {gp.log_marginal_likelihood_value_}") .

[0139] The following is a sample of key source code for predicting a new 28-dimensional feature vector.

[0140] # Make predictions for the new sample X (X is the 28-dimensional vector constructed above).

[0141] X_test = X.reshape(1, -1) # Convert to a 1×28 matrix

[0142] # Predicted mean and standard deviation

[0143] rho_mean, rho_std = gp.predict(X_test, return_std=True)

[0144] print(f"Predicted density mean: {rho_mean[0]:.1f} kg / m³")

[0145] print(f"Prediction standard deviation: {rho_std[0]:.1f} kg / m³").

[0146] Assume the model predicts a mean output of 685.3 kg / m³. 3 The predicted standard deviation output is 42.7 kg / m³. 3 .

[0147] Physical constraints are applied based on the density estimation method described above. Specifically, a minimum air density of 1.2 kg / m³ is required. 3 It is well known that the method for determining the maximum possible density of a material whose required density is not higher than a certain spectral characteristic is well known. Here, we will only briefly introduce the determination process and results based on the above example, as follows.

[0148] The weighting coefficients of five spectral features were determined based on multiple regression analysis of 500 sets of standard materials. For example, the weighting coefficients of λ1-λ5 are 0.4, 0.3, 0.1, 0.1, and 0.1, respectively.

[0149] Spectral effect = 0.15×0.4 + 0.22×0.3 + 0.18×0.1 + 0.25×0.1 + 0.20×0.1 = 0.189.

[0150] For pulverized coal, the basic density is determined to be 1000 kg / m³ based on the typical true density value of pulverized coal in GB / T212-2008 "Industrial Analysis Methods for Coal". For mineral powder and fiber dust, the density is adjusted to 1500 kg / m³ respectively. 3 and 800 kg / m 3 .

[0151] The spectral effect coefficient was obtained as 4500 by fitting the experimental data using the least squares method. For pulverized coal, the maximum possible density = 1000 + 0.189 × 4500 = 1850.5 kg / m³ 3 .

[0152] Therefore, the physical constraints of the upper and lower limits of the equivalent density ρ are 1850.5 kg / m³. 3 and 1.2 kg / m 3 After constraint, the output equivalent density ρ is 685.3 kg / m³. 3 (Not exceeding the limit, remaining unchanged), standard deviation is 42.7 kg / m 3 (Unadjusted).

[0153] Given the above equivalent density ρ, select the pipe cross-section shape correction coefficient k, such as 0.9 for a circular pipe, and then obtain the deposition cross-section S_cross obtained by the optical auxiliary module. The deposition equivalent mass M can then be calculated according to the formula M=k•ρ•S_cross.

[0154] If the deposition equivalent mass M = 2.864 kg / m, according to the following working condition correspondence table, it can be determined that it belongs to light deposition. Based on this, the monitoring frequency can be increased and planned cleaning can be carried out.

[0155] Preferably, such as Figure 1 As shown, this monitoring method also includes a density estimation verification step, which includes density value comparison and confidence level calculation.

[0156] The density value comparison involves comparing the equivalent density ρ obtained by the density estimation method above with the true density value ρ_true obtained by the sampling and weighing method. For example, if the error between the equivalent density ρ and the true density value ρ_true is within ±15%, the calculated result of the equivalent density ρ can be determined to meet the requirements.

[0157] The confidence level calculation is based on the standard deviation of the equivalent density ρ calculation output and the signal quality indices of the microwave and optical eigenvalues. This calculation comprehensively considers the impact of uncertainties in the density estimation method and the quality of the input data on the calculation. It is known that the signal quality indices of the microwave eigenvalues ​​are obtained by calculating the peak signal-to-noise ratio (SNR) and peak sharpness based on the microwave signal's spectrum, and the signal quality indices of the optical eigenvalues ​​are obtained by calculating the global sharpness and deposition area contrast based on the dust accumulation profile image.

[0158] The confidence level is calculated using the following formula:

[0159] Con=[1 / (1+rho_std / Z)]×(Q_opt×0.6+Q_microwave×0.4);

[0160] in:

[0161] Con represents the confidence level;

[0162] rho_std is the standard deviation of the density estimation;

[0163] Q_opt is the signal quality index of the optical eigenvalues;

[0164] Q_microwave is a signal quality index for microwave characteristic values;

[0165] Z is a correction factor, calibrated based on a standard material sample.

[0166] If the calculated confidence level is within the specified range, such as within the range of (0.7, 1), then the confidence level meets the requirements.

[0167] Through the density estimation verification steps described above, the system can determine in real time whether the density estimation results are affected by sensor contamination, extreme operating conditions, etc., resulting in data degradation, thus achieving self-diagnosis and ensuring the accuracy of monitoring results.

[0168] Preferably, the monitoring method further includes a model calibration step, which includes an initial calibration stage and an online calibration stage.

[0169] Specifically, the initial calibration phase includes the following steps.

[0170] BS1: In the clean state of the pipeline, the optical auxiliary module acquires the zero-deposition reference surface.

[0171] BS2: Inject a standard material of known thickness, and collect data using an ultrasonic ranging module, the microwave detection module, and the optical auxiliary module. The ultrasonic ranging module collects data including the ultrasonic time-of-flight T, the microwave detection module collects data including the microwave attenuation A, and the optical auxiliary module collects data including the optical deposition thickness H_opt.

[0172] BS3: Establish initial mapping relationships for data collected after at least two injections of standard materials. For example, establish initial mapping relationships between ultrasonic time of flight T and optical dust accumulation thickness H_opt, and between microwave attenuation A and dielectric constant ε_r. The dielectric constant ε_r of the material can be calculated from the optical reference density ρ_opt_ref, and it is known that this calculation can be performed using a lookup table relationship established through previous experiments.

[0173] The initial calibration described above calibrates the measurements of the ultrasonic ranging module, the microwave detection module, and the optical auxiliary module to ensure the accuracy of the acquired source data.

[0174] The online calibration phase includes: measuring the ultrasonic dust accumulation thickness H using the ultrasonic ranging module, obtaining the optical deposition thickness H_opt using the optical auxiliary module, and periodically, or when the deviation between the optical deposition thickness H_opt and the ultrasonic dust accumulation thickness H exceeds a threshold, fitting a new calibration curve using the historically accumulated optical deposition thickness H_opt and the ultrasonic flight time T, and updating the calibration coefficients of the ultrasonic ranging module. This online calibration uses optical true value data to check whether there is a significant deviation in the monitoring of the ultrasonic ranging module, and corrects its monitoring error and drift after a significant deviation is found, which also helps to ensure the long-term accuracy of the source data. The new calibration curve can be achieved using binomial fitting. After the above calibration, the real-time monitoring of the ultrasonic ranging module is accurate at least in the near term, and it can verify the optical deposition thickness H_opt obtained by the optical auxiliary module within this period, thereby reflecting whether there is dust contamination on the prism or lens of the optical auxiliary module, so as to arrange cleaning in a timely manner. This achieves the purpose of mutual calibration and diagnosis between the ultrasonic ranging module and the optical auxiliary module, which helps to ensure the long-term accuracy of dust accumulation information feedback.

[0175] Preferably, the online calibration stage further includes: obtaining a reference density ρ_opt_ref using the optical auxiliary module, and fine-tuning the parameters of the Gaussian process regression model using the reference density ρ_opt_ref to adapt the density estimation method to changes in material properties. The trained Gaussian process regression model is trained using monitoring data of an initial batch of standard materials, and it can reflect the complex mapping relationship between source data (such as microwave eigenvalues) and equivalent density ρ. However, the characteristics of dust materials in the pipeline may change due to variations in composition, moisture content, etc., at which point the equivalent density ρ obtained using the Gaussian process regression model will be inaccurate. Accordingly, online calibration fine-tunes the model parameters using the reference density ρ_opt_ref to adapt to changes in material properties, ensuring the accuracy of the equivalent density ρ calculation, and achieving the effect of calibrating and updating the model using optical reference information.

[0176] Specifically, when the optical reference density ρ_opt_ref and its corresponding microwave eigenvalues ​​are obtained, the system uses this data pair as a new training sample and performs a gradient descent update on the kernel function hyperparameters of the trained Gaussian process regression model using the maximum a posteriori probability estimation method. Simultaneously, the system can set a sliding time window to retain only a certain amount of recent historical data for updates, enabling the model to track and adapt to changes in material properties while avoiding forgetting early fundamental patterns.

[0177] The present invention also provides a multimodal monitoring system for dust accumulation in dust removal pipelines, using the aforementioned multimodal monitoring method for dust accumulation in dust removal pipelines, and comprising:

[0178] Ultrasonic ranging module, used to measure ultrasonic deposition thickness H;

[0179] Microwave detection module, used to generate microwave characteristic values;

[0180] An optical auxiliary module is used to acquire cross-sectional images of dust accumulation.

[0181] The processing and control unit is communicatively connected to the ultrasonic ranging module, the microwave detection module, and the optical auxiliary module, and is configured to execute:

[0182] The ultrasonic deposition thickness H, the microwave characteristic value, and the dust accumulation profile image are received, and the optical characteristic value is obtained based on the dust accumulation profile image;

[0183] Run the density estimation method to calculate the equivalent density ρ;

[0184] Calculate the deposition equivalent mass M.

[0185] Of course, based on the foregoing, the processing and control unit is also configured to perform density estimation verification steps and model calibration steps.

[0186] Preferably, the optical auxiliary module includes a telescopic optical prism that extends into the pipe when it is necessary to acquire a dust profile image and retracts from the pipe when it is not necessary to acquire a dust profile image, thereby significantly reducing the impact of the dust environment inside the pipe on optical monitoring.

[0187] In summary, as Figure 1 As shown, the conventional monitoring of this invention uses the ultrasonic ranging module to routinely monitor dust accumulation thickness. The microwave detection module and the optical auxiliary module are periodically activated, such as every 4 hours. After activation, the equivalent density ρ and deposition equivalent mass M are calculated to accurately reflect the dust accumulation condition. Based on the feedback, risk assessment and maintenance suggestions are provided. Density estimation verification is performed during each calculation to ensure that the calculation results exclude the influence of data degradation, thus ensuring the accuracy of the monitoring results. After the microwave detection module and the optical auxiliary module are periodically activated, or when the deviation between the optical deposition thickness H_opt and the ultrasonic dust accumulation thickness H exceeds a threshold, online calibration is initiated to update the calibration coefficients of the ultrasonic ranging module and the Gaussian process regression model, ensuring the long-term accuracy of the dust accumulation information feedback.

[0188] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of this invention is defined by the appended claims and their equivalents.

Claims

1. A multi-modal monitoring method for dust accumulation in dust removal pipelines, characterized in that: The process includes using a microwave detection module to measure microwave characteristic values, using an optical auxiliary module to obtain optical characteristic values, using a density estimation method based on the microwave characteristic values ​​and the optical characteristic values ​​to estimate the equivalent density ρ, and calculating the deposition equivalent mass M based on the equivalent density ρ, wherein the deposition equivalent mass M is used to reflect the deposition conditions. The density estimation method includes fusing the microwave and optical feature values ​​into a feature vector, which includes microwave features, optical features, and interaction features between the two. The microwave features are microwave attenuation A, phase shift Φ, A / Φ ratio, dispersion slope dA / df, and polarization ratio VV / HH. The dispersion slope dA / df is obtained by analyzing the rate of change of signal attenuation with frequency within the swept bandwidth, and the polarization ratio VV / HH is obtained by comparing the attenuation of the microwave signal in the vertical and horizontal polarization directions. The optical features include five spectral features, surface roughness Ra, deposition uniformity index, and thickness variation coefficient. The spectral features are reflectance values ​​after normalization at a specified wavelength in a specified band. In the construction of interactive features, microwave attenuation A and phase shift Φ are selected as microwave features, and the three spectral features with the strongest correlation to density are selected as optical features; the interactive features include multiplication interaction, difference interaction and normalized division interaction between the selected microwave features and the selected spectral features. The feature vector is input into a trained Gaussian process regression model to obtain the density prediction mean; the training of the Gaussian process regression model is based on multiple sets of standard materials, and each set of standard materials contains the feature vector used for input. Physical constraints are applied to the predicted density mean, limiting it to a range that is no less than the air density and no more than the maximum possible density determined by the spectral characteristics of the material, and finally the equivalent density ρ is output. The deposition equivalent mass M is calculated using the following formula: M = k•ρ•S_cross; in: k is the pipe cross-sectional shape correction factor; ρ is the equivalent density; S_cross is the deposition cross-sectional area, obtained using the optical auxiliary module.

2. The multi-modal monitoring method for dust accumulation in dust removal pipelines as described in claim 1, characterized in that: It also includes a density estimation verification step, which includes density value comparison and confidence level calculation.

3. The multi-modal monitoring method for dust accumulation in dust removal pipelines as described in claim 1, characterized in that: It also includes a model calibration step, which includes an initial calibration phase and an online calibration phase.

4. The multi-modal monitoring method for dust accumulation in dust removal pipelines as described in claim 3, characterized in that, The online calibration phase includes: measuring the ultrasonic dust thickness H using the ultrasonic ranging module, obtaining the optical deposition thickness H_opt using the optical auxiliary module, and periodically or when the deviation between the optical deposition thickness H_opt and the ultrasonic dust thickness H exceeds a threshold, fitting a new calibration curve with the historically accumulated optical deposition thickness H_opt and the ultrasonic flight time T, and updating the calibration coefficient of the ultrasonic ranging module.

5. The multi-modal monitoring method for dust accumulation in dust removal pipelines as described in claim 4, characterized in that, The online calibration phase also includes: using the optical auxiliary module to obtain a reference density ρ_opt_ref, and using the reference density ρ_opt_ref to fine-tune the parameters of the Gaussian process regression model so that the density estimation method can adapt to changes in material properties.

6. The multi-modal monitoring method for dust accumulation in dust removal pipelines as described in claim 3, characterized in that, The initial calibration phase includes the following steps: BS1: Under clean pipeline conditions, the optical auxiliary module acquires a zero-deposition reference surface; BS2: Inject a standard material of known thickness, and collect data using the ultrasonic ranging module, the microwave detection module, and the optical auxiliary module respectively; BS3: Establish an initial mapping relationship for data collected after at least two injections of standard material.

7. A multi-modal monitoring system for dust accumulation in dust removal pipelines, characterized in that, The method for multimodal monitoring of dust accumulation in dust removal pipelines as described in any one of claims 1 to 6 is used, and includes: Ultrasonic ranging module, used to measure ultrasonic deposition thickness H; Microwave detection module, used to generate microwave characteristic values; An optical auxiliary module is used to acquire cross-sectional images of dust accumulation. The processing and control unit is communicatively connected to the ultrasonic ranging module, the microwave detection module, and the optical auxiliary module, and is configured to execute: The ultrasonic deposition thickness H, the microwave characteristic value, and the dust accumulation profile image are received, and the optical characteristic value is obtained based on the dust accumulation profile image; Run the density estimation method to calculate the equivalent density ρ; Calculate the deposition equivalent mass M.

8. The multi-modal monitoring system for dust accumulation in dust removal pipelines as described in claim 7, characterized in that: The optical auxiliary module includes a telescopic optical prism that extends into the pipe when it is necessary to acquire a dust accumulation profile image and exits the pipe when it is not necessary to acquire a dust accumulation profile image.

Images