Method for predicting concentration of paint mist, method for adjusting recovery negative pressure intensity, and recovery device

Abstract

The present application relates to the technical field of material recycling, in particular to a paint mist concentration prediction method, a recovery negative pressure intensity adjustment method and a recovery device. The present application comprises the following steps: S1, establishing a three-dimensional space coordinate system in a paint spraying operation area and obtaining basic parameters of the paint spraying operation area, the basic parameters including the number of paint mist sources, three-dimensional space coordinate points of distribution of each paint mist source and the intensity of each paint mist source; S2, obtaining the paint mist basic diffusion coefficient of each paint mist source through wind tunnel experiment and actual paint spraying working condition test, and correcting the basic diffusion coefficient through the interference coefficient between each paint mist source to obtain the corrected actual diffusion coefficient of each paint mist source. The present application combines the paint spraying operation environment and effectively utilizes the key data in the paint spraying process, thereby accurately predicting the change trend of the paint mist concentration in advance, providing technical support on the software for the recovery device to actively adjust the negative pressure intensity in advance, thereby stabilizing the paint mist concentration and improving the recovery effect.

Description

Methods for predicting paint mist concentration, methods for adjusting negative pressure intensity during recovery, and recovery devices. Technical Field

[0001] This invention relates to the field of material recycling technology, specifically to a method for predicting paint mist concentration, a method for adjusting negative pressure intensity during recycling, and a recycling device. Background Technology

[0002] Currently, when commonly used spray painting equipment coats workpiece surfaces, the sprayed paint is in a mist-like, diffusing, and volatile state, with paint particles and harmful gases permeating the entire work area. In the shipbuilding industry, this spray painting process is a crucial step in ensuring the service life and appearance quality of ships.

[0003] Because spray painting often results in paint waste, it is necessary to recover the paint mist generated during the process. While common water curtain recovery technology can capture some paint mist, it suffers from significant water consumption, high levels of impurities in the recovered paint, and high subsequent processing costs. Therefore, current paint mist recovery primarily utilizes dry filtration technology. In practice, this method often employs techniques such as the negative pressure adsorption method described in Chinese Patent Publication No. CN202270607U, entitled "A Negative Pressure Spray Painting Recovery Device," to capture diffused paint mist.

[0004] However, it is clear that the concentration of paint mist generated during painting operations often varies due to the demands of the work and environmental factors. If a fixed negative pressure power is used to capture the paint mist, the concentration will inevitably fluctuate significantly. Firstly, in LNG cargo tank scenarios with stringent explosion-proof restrictions, achieving good paint mist capture at high concentrations makes it difficult to control the concentration within the explosion-proof threshold. Secondly, at low concentrations, there may be excessive negative pressure power. This excessive negative pressure, resulting in excessive adsorption, leads to the adsorption of more impurities from the environment. Since the dust holding capacity of dry filtration recovery technology is limited, the increased adsorption of impurities reduces its filtration efficiency, thus increasing the frequency of filter maintenance. Furthermore, in practice, excessive negative pressure adsorption may cause paint mist that should adhere to the workpiece surface to separate from the surface under negative pressure, affecting the painting effect on the workpiece surface.

[0005] In practice, while paint mist concentration can be predicted, it is mostly based on a rough estimate using paint flow rate, resulting in low accuracy, especially under conditions with significant environmental influences and multiple paint mist sources. Furthermore, some methods use concentration sensors to monitor concentration and regulate negative pressure adsorption intensity. However, since these sensors can only detect changes after the paint mist concentration reaches a certain level, they cannot predict trends in paint mist intensity in advance. This makes the adjustment of the negative pressure intensity of the recovery device rather passive, and therefore urgently needs to be addressed. Summary of the Invention

[0006] To avoid and overcome the technical problems existing in the prior art, this invention provides a method for predicting the concentration of paint mist, a method for adjusting the negative pressure intensity of recovery, and a recovery device. By combining the painting operation environment and making effective use of key data during the painting process, the method can accurately predict the changing trend of paint mist concentration in advance, and provide software technical support for the recovery device to actively adjust the negative pressure intensity in advance, thereby stabilizing the paint mist concentration and improving the recovery effect.

[0007] To achieve the above objectives, the present invention provides the following technical solution:

[0008] The method for predicting paint mist concentration includes the following steps:

[0009] S1. Establish a three-dimensional spatial coordinate system in the spray painting operation area and obtain the basic parameters of the spray painting operation area. These basic parameters include the number of paint mist sources, the three-dimensional spatial coordinate points of each paint mist source distribution, and the intensity of each paint mist source.

[0010] S2. Obtain the basic diffusion coefficient of each paint mist source through wind tunnel experiments and actual painting conditions, and correct the basic diffusion coefficient through the interference coefficient between each paint mist source to obtain the corrected actual diffusion coefficient of each paint mist source.

[0011] S3. Input the acquired basic parameters and the actual diffusion coefficients of each paint mist source into the optimized diffusion model to obtain the basic diffusion paint mist concentration at the three-dimensional spatial coordinate point to be predicted under the superposition of paint mist diffusion from each paint mist source. Then, introduce the concentration change law of the paint mist source in the time series and the environmental benchmark concentration of paint mist in the painting operation area into the basic diffusion paint mist concentration. Finally, correct the basic diffusion paint mist concentration to the predicted paint mist concentration.

[0012] Compared with the prior art, the beneficial effects of the present invention are:

[0013] 1. This application utilizes precise three-dimensional spatial positioning, accurate diffusion coefficient acquisition and correction, and comprehensive diffusion model optimization and concentration prediction, while comprehensively considering various complex conditions in the painting operation area, including the number, distribution, intensity, mutual interference of paint mist sources, and environmental factors. This comprehensive analysis and processing capability enables the system to better adapt to different painting processes and environmental conditions, giving the predictions of this application strong adaptability, flexibility, and accuracy. By accurately predicting the changing trend of paint mist concentration in advance, this application provides software technical support for the recovery device to proactively adjust the negative pressure intensity in advance, thereby stabilizing the paint mist concentration and improving the recovery effect.

[0014] 2. The optimized diffusion model of this application fully considers the environmental factors and physical characteristics of paint mist as it diffuses from the source point to the surrounding area, and corrects the diffusion coefficient based on the distribution of each paint mist source. At the same time, it introduces the periodic change in the time dimension to simulate the fluctuation of paint mist concentration over time due to factors such as equipment working rhythm and paint supply stability during the painting process, which effectively ensures the accuracy of the diffusion model prediction results.

[0015] 3. This application uses mean squared error as the loss function and employs stochastic gradient descent algorithm for parameter optimization training. The learning rate during the optimization training process is adjusted using the adaptive learning rate adjustment algorithm Adagrad, which ultimately obtains the optimized diffusion model efficiently, further ensuring the accuracy of the diffusion model's prediction results.

[0016] 4. Compared to traditional methods that require waiting for a change in paint mist concentration to be detected by the sensor before initiating the adjustment process, which introduces a certain reaction time lag, this application can shorten or even eliminate this time lag, making negative pressure adjustment more timely and better adaptable to rapid changes in paint mist concentration. Specifically, the negative pressure adjustment for paint mist recovery in this application is based on the prediction of upcoming concentration changes, and the negative pressure intensity is adjusted specifically at each point of significant fluctuation. This method of anticipating the trend of paint mist concentration changes in advance and preparing for changes by proactively adjusting the negative pressure intensity ensures that paint mist can be recovered in a timely and effective manner, avoiding paint mist diffusion caused by sudden increases in concentration, and avoiding the adjustment delay caused by sensor detection lag in traditional methods.

[0017] 5. The recovery device of this application adopts at least three stages of filtration, including a filter screen, an electrostatic filter, and a membrane filtration unit, which effectively reduces the impurity content of the recovered paint mist. Attached Figure Description

[0018] Figure 1 is a flowchart of the workflow of the paint mist concentration prediction method in this invention.

[0019] Figure 2 is a flowchart of the workflow of the negative pressure intensity adjustment method for paint mist recovery in this invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] For ease of understanding, the specific structure and operation of the present invention will be further described below with reference to the accompanying drawings:

[0022] The present invention mainly includes a method for predicting the concentration of paint mist in a jetting operation area, a method for adjusting the negative pressure intensity of paint mist recovery applied to the paint mist concentration prediction method, and a negative pressure recovery device applied to the negative pressure intensity adjustment method of paint mist.

[0023] The method for predicting paint mist concentration, as shown in Figure 1, includes the following steps:

[0024] S1. Establish a three-dimensional spatial coordinate system in the spray painting operation area and obtain the basic parameters of the spray painting operation area. These basic parameters include the number of paint mist sources, the three-dimensional spatial coordinate points of each paint mist source distribution, and the intensity of each paint mist source.

[0025] S2. Obtain the basic diffusion coefficient of each paint mist source through wind tunnel experiments and actual painting conditions, and correct the basic diffusion coefficient through the interference coefficient between each paint mist source to obtain the corrected actual diffusion coefficient of each paint mist source.

[0026] S3. Input the acquired basic parameters and the actual diffusion coefficients of each paint mist source into the optimized diffusion model to obtain the basic diffusion paint mist concentration at the three-dimensional spatial coordinate point to be predicted under the superposition of paint mist diffusion from each paint mist source. Then, introduce the concentration change law of the paint mist source in the time series and the environmental benchmark concentration of paint mist in the painting operation area into the basic diffusion paint mist concentration. Finally, correct the basic diffusion paint mist concentration to the predicted paint mist concentration.

[0027] This application utilizes precise three-dimensional spatial positioning, accurate diffusion coefficient acquisition and correction, and comprehensive diffusion model optimization and concentration prediction, while also taking into account various complex conditions in the painting operation area, including the number, distribution, intensity, mutual interference of paint mist sources, and environmental factors. This comprehensive analysis and processing capability enables the system to better adapt to different painting processes and environmental conditions, giving the predictions of this application strong adaptability, flexibility, and accuracy. By accurately predicting the changing trend of paint mist concentration in advance, this application provides software technical support for the recovery device to proactively adjust the negative pressure intensity in advance, thereby stabilizing the paint mist concentration and improving the recovery effect.

[0028] Specifically, the optimized diffusion model in step S3 is as follows:

[0029]

[0030] In the formula, For the three-dimensional spatial coordinates of the points to be predicted ( Predicted paint mist concentration at () location, in mg / m³ 3 .

[0031] The environmental baseline concentration of paint mist in the painting operation area, in mg / m³. 3 This is used to eliminate the interference of background environmental factors on paint mist concentration, making the model prediction more consistent with the actual painting scenario.

[0032] Furthermore, in this diffusion model, Partly based on the Gaussian distribution principle, this method describes the diffusion pattern of paint mist in space. Ideally, paint mist diffuses from its source outwards, with the concentration decreasing exponentially with increasing distance. As the diffusion coefficient, it directly determines the shape of the Gaussian distribution curve, and its value reflects the diffusion range of paint mist in the painting operation area.

[0033] , , These are the x-coordinate, y-coordinate, and vertical coordinates of the three-dimensional spatial coordinate point to be predicted.

[0034] , , Let x, y, and y be the x, y, and y coordinates of the i-th paint mist source in three-dimensional space.

[0035] The quantity of paint mist sources is dimensionless.

[0036] Let be the intensity of the i-th paint mist source, in g / min, where the intensity of any paint mist source is ;

[0037]

[0038] In the formula, The intensity of the paint mist source is the mass of paint mist that is sprayed from the spray gun and effectively participates in diffusion per unit time, expressed in g / min.

[0039] The flow rate of paint sprayed by the spray gun is measured in ml / min.

[0040] This refers to the density of the paint, expressed in g / ml.

[0041] The percentage of non-volatile solid particles in the paint, dimensionless;

[0042] It is the proportion of paint mist that can be captured after being atomized by a spray gun, and is dimensionless.

[0043] During data acquisition, flow sensors and pressure sensors are installed on the spray gun of the paint mist source to obtain the paint flow parameters, and the intensity of the paint mist source is obtained by inputting them into the above-mentioned formula for calculating the intensity of the paint mist source.

[0044] Let be the actual diffusion coefficient after correction for the i-th paint mist source. It is dimensionless and is obtained by correcting the basic diffusion coefficient of paint mist.

[0045] Specifically, the basic diffusion coefficient of paint mist from each paint mist source is as follows:

[0046]

[0047] The formula consists of two parts:

[0048] It is a linear regression structure, supported by the core theoretical assumption of multivariate linear association and Gauss-Markov theorem, adapting to the "linear influence of environmental factors on the diffusion coefficient". This structure is based on the physical law that wind speed, temperature, and humidity have a linear relationship with the diffusion coefficient, and uses "multivariate linear regression" to describe the effect of environmental factors on the diffusion coefficient. The influence of this can accurately reflect the actual relationship and simplify the model structure.

[0049] It is a nonlinear correction term, based on the physical principles of particle dynamics and aerosol diffusion, and the characteristic variables of paint mist (average particle size). ,density The influence of the diffusion coefficient stems from the "resistance law of particle motion in fluid," and its physical mechanism is inherently nonlinear. Through experimental fitting, the square root form of the norm can approximately match the average particle size. ,density and Nonlinear correlation.

[0050] In the formula, Let be the basic diffusion coefficient of the i-th paint mist source, dimensionless, representing the baseline value of the diffusion coefficient under conditions of no environmental interference and standard paint mist characteristics;

[0051] This is the fundamental diffusion coefficient constant term, which is dimensionless.

[0052] The environmental factor influence coefficient is dimensionless.

[0053] Represents environmental factors; dimensionless.

[0054] The number of environmental factors is dimensionless.

[0055] Let j be the j-th environmental factor variable, dimensionless, where j takes the values ​​1, 2, or 3. The wind speed is the maximum wind speed, which causes the paint mist to diffuse faster and has a significant positive impact on the diffusion coefficient. Temperature is a factor; as temperature increases, air density decreases, reducing the resistance to the movement of paint mist particles and promoting diffusion. Humidity is a factor; increased humidity may cause paint mist particles to absorb moisture, affecting their diffusion ability, and is generally negatively correlated.

[0056] is the coefficient for the influence of paint mist characteristics, dimensionless;

[0057] Let k represent the k-th paint mist characteristic variable, which is dimensionless and k takes the value 1 or 2. The average particle size of the paint mist is the smaller the particle size, the more obvious the Brownian motion and the stronger the diffusion ability. This refers to the density of the paint mist; the lower the density, the easier it is to diffuse in the air.

[0058] This is a dimensionless random error term used to characterize other minor influencing factors not included in the model, as well as measurement errors.

[0059] Traditional Gaussian models assume that "each spray gun diffuses independently without mutual influence." However, in actual painting, the paint mist sprayed by multiple spray guns will create airflow disturbances (such as jet superposition and eddies), which will cause the diffusion range of the target spray gun to expand. The stronger the interference, the larger the diffusion range.

[0060] The degree of interference is positively correlated with two factors: the distance between spray guns (the closer the distance, the better). The ratio of the source strength to the larger the diffusion coefficient is, the more accurate the diffusion coefficient becomes.

[0061] The actual diffusion coefficients after correction for each paint mist source are as follows:

[0062]

[0063] In the formula, The actual diffusion coefficient after correction for paint mist source is dimensionless.

[0064] Let be the basic diffusion coefficient of the paint mist from the i-th paint mist source, which is dimensionless;

[0065] This represents the interference coefficient between the i-th paint mist source and the r-th paint mist source, which is dimensionless.

[0066] and , This represents the distance between the i-th and r-th paint mist sources, in meters (m). L represents the jet influence characteristic length, also in meters, and is taken as 0.3 times the spray gun range. The angle between the spraying directions of the i-th paint mist source and the r-th paint mist source is expressed in degrees.

[0067] The intensity of the i-th paint mist source is expressed in g / min.

[0068] Let be the intensity of the r-th paint mist source, expressed in g / min.

[0069] In this diffusion model, Some studies introduce periodic changes in the time dimension, that is, the concentration change pattern of the paint mist source over time, to simulate the fluctuation of paint mist concentration over time caused by factors such as equipment working rhythm and paint supply stability during the painting process. Let be the initial phase of the i-th paint mist source, in rad; ωi represents the angular frequency of the concentration change at the i-th paint mist source, in rad / s, which determines the period of concentration change; t represents time, in seconds. By adjusting the parameters... and It can accurately reflect the concentration change patterns of different paint mist sources over time.

[0070] Specifically, It reflects the rate of periodic fluctuation of paint mist concentration over time. The core reason for its existence is the periodic operating conditions of the painting operation, including: the periodicity of equipment operation, such as: the spray gun flow rate may fluctuate periodically due to pump pulsation, pressure regulation, etc. (such as the reciprocating frequency of a high-pressure airless spray pump); or the ventilation system fan may have periodic airflow disturbances (such as the vortex period generated by the blade rotation).

[0071] The periodicity of the operation process, such as the repetitive nature of the worker's arm swing and gun changing frequency when spraying paint; or the periodic changes in paint intensity caused by shift changes and equipment maintenance.

[0072] In the actual forecasting process, Actual measurements show that when continuous data on paint mist concentration over time is available, a concentration-time series is obtained. Random noise is removed by filtering (such as moving average or wavelet filtering), retaining the periodic component. The time difference between two adjacent peaks (or troughs) is measured, which is the period T. This period T is then calculated using the formula... It can be calculated that, among which For frequency.

[0073] The parameter describing the initial state of the periodic change in paint mist concentration, mathematically speaking, is: a sine function. initial angle, initial phase This determines the starting position of the concentration fluctuation at time t=0, for example, when At t=0, This indicates that the initial concentration is at the peak of the fluctuation. At t=0, This indicates that the initial concentration was at the zero point of fluctuation.

[0074] At the physical level, this means reflecting the initial state difference when a paint mist source starts up or switches operating conditions; for example, if a paint mist source's spray gun is just turned on at t=0, it is in the concentration rising phase, and the initial phase... (Corresponding to the rising edge of the sine curve); the spray gun from another paint mist source has been working stably for a period of time, and at t=0, the concentration is at the peak of the fluctuation, the initial phase (Approaching peak).

[0075] angular frequency of the concentration change of the i-th paint mist source and the initial phase of the i-th paint mist source The data is primarily based on statistical analysis of historical painting operation data. Data on the variation of paint mist concentration over time under different types of ships and different painting processes was collected. Signal processing methods such as Fourier transform were used to extract the periodic characteristics of the concentration changes, thereby determining... and The initial value is then dynamically adjusted based on real-time monitoring data in practical applications.

[0076] Based on the above, the optimization process of the optimized diffusion model in step S2 is as follows:

[0077] S21. Collect paint mist data, environmental data, and painting condition data within the painting operation area at each time point in the painting operation sequence. Specifically, this includes collecting key data such as paint mist concentration, particle size distribution, ambient temperature and humidity, and equipment operating parameters at a second-level frequency, and creating a 3D spatial model of the painting operation area to provide a more accurate spatial reference for the model. Within the ship's hull painting operation area, deploy a monitoring network composed of sensors to collect relevant parameters in real time. This can be implemented using the following methods:

[0078] A laser particle size analyzer was used to collect paint mist particle size distribution data at a frequency of 10 times per second, with a measurement range covering 0.1-100 micrometers;

[0079] The paint mist concentration data is acquired in real time using a high-precision concentration sensor, and the detection range for acquiring the paint mist concentration data is 0-2000 mg / m³.

[0080] The ambient temperature and humidity are recorded in real time by a temperature and humidity sensor. The measurement range of ambient temperature is -20℃ to 150℃.

[0081] S22. Clean the data collected in step S1 to remove outliers and missing values. For missing values, use linear interpolation or a weighted average method based on neighboring data to fill them in; for outliers, identify and remove them using statistical methods such as box plots to ensure data quality; and divide the cleaned data into training set, validation set and test set along the time series in a 7:1:2 ratio.

[0082] In practice, data cleaning can be carried out in the following ways:

[0083] The collected raw data is transmitted to the edge computing node via industrial Ethernet, and preprocessed using the Kalman filter algorithm to remove data noise. The specific formula is as follows:

[0084]

[0085] in, This is the optimal estimate at time k;

[0086] Let be the prior estimate at time k;

[0087] Kalman gain;

[0088] The measurement value at time k;

[0089] This is the observation matrix.

[0090] After data preprocessing, the accuracy and stability of the data can be significantly improved, providing a reliable foundation for subsequent analysis.

[0091] S23. Input the parameters from the training set into the diffusion model, using the mean squared error (MSE) as the loss function. The loss function formula is: ,in, The paint mist concentration predicted by the model. The actual measured paint mist concentration is N, where N is the number of samples, and the stochastic gradient descent algorithm is used for parameter optimization training.

[0092] During training, the learning rate is a key parameter affecting the model's convergence speed and performance. To avoid the learning rate being too large, causing the model to skip the optimal solution, or too small, slowing down the training process, this application uses the adaptive learning rate adjustment algorithm Adagrad to optimize the learning rate during training, ultimately obtaining an optimized diffusion model. This algorithm dynamically adjusts the learning rate based on the parameter update history; for parameters that are updated frequently, the learning rate is reduced; for parameters that are updated less frequently, the learning rate is increased, thereby accelerating the model's convergence speed and improving training efficiency.

[0093] Based on the above, the further step S23 to optimize the training process includes:

[0094] S231. Initialize parameters: Initialize all parameters in the model, including... , , , Then, perform random initialization to provide a starting point for the training process.

[0095] S232, Forward Propagation: Input the data in the training set into the diffusion model, and calculate the predicted value of the diffusion model based on the current parameters.

[0096] S233. Calculate the loss: Substitute the predicted value and the true value into the mean square error formula to calculate the loss value MSE of the loss function under the current parameters.

[0097] S234. Backpropagation: Based on the gradient of the loss function with respect to each parameter, update the parameters in the opposite direction of the gradient to reduce the loss value. In practice, during the backpropagation process, the chain rule can be used to calculate the gradient of each parameter to ensure the accuracy of parameter updates.

[0098] S235. Repeat steps S231-S234 to iterate the training until the loss value converges to the preset structure or reaches the preset maximum number of iterations. During each iteration, the model performance is evaluated using a validation set to prevent overfitting.

[0099] S236. After training is completed, use the test set to perform a final evaluation of the model, calculate the model's prediction error, accuracy and other indicators. If the model performance does not meet expectations, adjust the model parameters or optimize the training algorithm, and repeat steps S231-S235 until the model meets the design requirements and the prediction error is controlled within ±5%.

[0100] In practical applications, the above-mentioned method for predicting paint mist concentration is as follows:

[0101] Example 1: For the enclosed painting operation of a cargo hold section (18m long × 10m wide × 7m high) of a 50,000-ton bulk carrier, four fixed spray guns were used to operate simultaneously. The concentration prediction method of paint mist in this application was applied to predict the concentration. The experimental period was 48 hours.

[0102] Sensor deployment: Sixteen concentration sensors (accuracy 0.05 mg / m³) are deployed within the cargo hold sections, covering the area around the spray guns (0.5-3 m away), the corners of the bulkhead, and the top area; simultaneously, the flow rate (150-200 ml / min), spray angle (30°-60°), and environmental parameters (temperature 25±2℃, humidity 55±5% RH) of the four spray guns are collected. and the initial phase of the i-th paint mist source .

[0103] It should be noted that the coordinates of the four spray guns are (1,2,2), (3,4,2), (15,8,2), and (17,2,2).

[0104] In the model parameter settings: the environmental baseline concentration is 0.8 mg / m³, and the paint mist source intensity of the four fixed spray guns is... - The corrected diffusion coefficients are 120 g / min, 115 g / min, 125 g / min, and 110 g / min, respectively. - The measurements are 0.95m, 0.92m, 0.98m, and 0.90m respectively. - The measured value is 0.785 rad / s, with a period of 8 s and an initial phase. - They are π / 6, π / 4, π / 3, and π / 5 respectively.

[0105] when The prediction results are compared in Table 1 below:

[0106] Table 1 Comparison of First Prediction Results

[0107]

[0108] Example 1 uses a prediction model corrected for multi-source interference to reduce the average concentration prediction error from 38.16% to 3.40%, providing accurate data support for subsequent negative pressure intensity regulation.

[0109] It should be noted that the traditional scheme described in this embodiment refers to a single-source prediction model based on Gaussian diffusion, which assumes that the paint mist diffuses uniformly in a spherical shape from the spray gun, and does not consider the influence of airflow interference between multiple spray guns or structural obstruction on the diffusion path. The core formula is:

[0110]

[0111] in, The paint mist concentration is predicted by a traditional prediction model, in mg / m³. The paint flow rate of the spray gun from the paint mist source is expressed in ml / min. The distance between the monitoring point and the spray gun is expressed as a straight line, in meters (m). u represents the ambient wind speed, in m / min. The diffusion coefficient is fixed and dimensionless. Furthermore, the measured values ​​in the above embodiments refer to results obtained through a concentration sensor.

[0112] Example 2: For the explosion-proof painting operation of the cargo tank (a cylindrical enclosed space with a diameter of 36m and a length of 25m) of a 174,000 cubic meter liquefied natural gas (LNG) carrier, two explosion-proof spray guns (model EX-900) were used to operate in different areas. The paint mist concentration inside the tank was required to be strictly controlled below 10mg / m³ (explosion-proof threshold). The paint mist concentration prediction method of this application was applied, and the experimental period was 60 hours.

[0113] Sensor deployment: Twenty intrinsically safe concentration sensors are evenly distributed on the inner wall of the liquid cargo tank, covering the top dome, the middle cylindrical section and the bottom area; the spray gun flow rate (80-120ml / min, within the explosion-proof limit), spray angle (45°-90°) and the micro-positive pressure environment parameters inside the tank (pressure 50-100Pa, temperature 22±1℃) are collected simultaneously.

[0114] It should be noted that two spray guns are used in this embodiment, with coordinates as follows: spray gun 1 (3,8,2) and spray gun 2 (30,20,2).

[0115] In the model parameter settings: the environmental baseline concentration is 0.5 mg / m³, and the paint mist source intensity of the spray gun is... and The corrected diffusion coefficients are 90 g / min and 85 g / min, respectively. and The angular frequencies are 0.82m and 0.78m respectively. and The measured value is 0.523 rad / s, with a period of 12 s and an initial phase. and They are π / 4 and π / 6 respectively.

[0116] when The prediction results are compared in Table 2 below:

[0117] Table 2 Comparison of Second Prediction Results

[0118]

[0119] It should be noted that the traditional solution in Example 2 refers to an empirical estimate, which is essentially a rough judgment made manually based on a limited set of parameters. The estimation formula used in this example is:

[0120]

[0121] in, This is an empirical estimate of paint mist concentration, in mg / m³. 3 , This is a fixed empirical coefficient, dimensionless, with values ​​ranging from 0.5 to 0.8. The paint flow rate of the spray gun from the paint mist source is expressed in ml / min. This represents the straight-line distance between the monitoring point and the spray gun, in meters (m).

[0122] In addition, the measured values ​​in the above embodiments refer to the results measured by the concentration sensor.

[0123] In Example 2, in the scenario of LNG ship cargo tanks with strict explosion-proof restrictions, the prediction model of this solution reduced the average error from 43.59% to 2.85%, ensuring that the concentration is controlled within the explosion-proof threshold, and also providing accurate data support for subsequent negative pressure intensity control.

[0124] The method for adjusting the negative pressure intensity for paint mist recovery, as shown in Figure 2, includes the following steps:

[0125] K1. Obtain the concentration change curve of the predicted paint mist concentration at the predicted three-dimensional spatial coordinate points of each negative pressure recovery area during the painting operation time by using the paint mist concentration prediction method.

[0126] K2. Along the time-increasing direction of the concentration-time curve, starting from the origin, select the time node interval in which the concentration change in each negative pressure recovery area first exceeds the preset concentration fluctuation threshold, and record it as the first time node interval.

[0127] K3. Starting from the end of the first time node interval, continue along the time growth direction of the concentration change curve to select the next time node interval in each negative pressure recovery area where the concentration change exceeds the preset concentration fluctuation threshold, and record it as the second time node interval. Repeat this step until each negative pressure recovery area reaches the preset end condition, that is, the cycle of the painting operation ends or there is no time node interval where the concentration fluctuation is greater than the threshold.

[0128] K4. Obtain the concentration change within each time node interval of each negative pressure recovery zone;

[0129] K5. Set each negative pressure recovery zone with the end point of its corresponding time node as an adjustment node. Input the concentration change of each time node interval into the negative pressure adjustment model. Based on the negative pressure intensity of the current adsorption port in each time node interval, obtain the adsorption port negative pressure intensity that the corresponding adjustment node of each negative pressure recovery zone needs to adjust. Control each negative pressure recovery zone to adjust the adsorption port negative pressure intensity to the adsorption port negative pressure intensity that the corresponding adjustment node needs to adjust at the end point of the corresponding time node.

[0130] Compared to traditional methods that require waiting for a change in paint mist concentration to be detected by a sensor before initiating the adjustment process, which introduces a certain reaction time lag, this application can shorten or even eliminate this time lag, making negative pressure adjustment more timely and better adaptable to rapid changes in paint mist concentration. This application predicts upcoming changes in concentration and adjusts the negative pressure intensity specifically at each point of significant fluctuation. This proactive approach, which anticipates changes in paint mist concentration and prepares for changes before they occur, ensures that paint mist is effectively and promptly recovered, preventing paint mist diffusion caused by sudden increases in concentration and avoiding the adjustment delays caused by sensor detection lag in traditional methods.

[0131] Specifically, the negative pressure regulation model in step K3 is as follows:

[0132]

[0133] In the formula, The negative pressure intensity at the adsorption port of the a-th negative pressure recovery zone after adjustment at the b-th adjustment node;

[0134] The negative pressure intensity of the current adsorption port of the a-th negative pressure recovery zone at the b-th adjustment node; the negative pressure intensity of the adsorption port when the negative pressure mechanism is started is set according to the value of the preset concentration fluctuation threshold, and the negative pressure intensity of the current adsorption port of the first adjustment node at this point is the negative pressure intensity of the adsorption port when it is started.

[0135] This is the adjustment coefficient;

[0136] This represents the concentration change in the a-th negative pressure recovery zone during the time interval corresponding to the b-th adjustment node.

[0137] This formula is based on proportional feedback control theory and the concentration-negative pressure correlation law in fluid mechanics.

[0138] The foundation and core logic of feedback control systems is that "the control quantity (negative pressure) and the deviation quantity are linearly proportional." That is, the larger the deviation, the larger the adjustment range of the control quantity, so as to quickly correct the deviation. The concentration-negative pressure correlation law is based on the principle of "negative pressure driving airflow" in fluid mechanics. The stronger the negative pressure, the greater the airflow speed, the higher the paint mist capture efficiency, and the greater the concentration decrease. The two are positively correlated.

[0139] For example, when the concentration change in the second negative pressure recovery zone increases by 50 mg / m³ in the time interval corresponding to the sixth adjustment node, and the negative pressure intensity P0 of the current adsorption port of the second negative pressure recovery zone at the sixth adjustment node is -300 Pa, the adjustment coefficient... hour;

[0140] The adjusted negative pressure intensity Pa, thereby improving the paint mist collection efficiency in this area.

[0141] The paint mist recovery device includes a negative pressure channel connected to the negative pressure inlet of the negative pressure generating mechanism, and a filter screen, an electrostatic filter, and a membrane filter unit arranged sequentially along the negative pressure adsorption direction within the negative pressure channel. The filter screen has a side length of 100 micrometers. The filter unit has 100-micron diamond-shaped pores and a filter thickness of 2 mm. The membrane filter unit is made of polyvinylidene fluoride microporous membrane as the filter substrate.

[0142] Clearly, this paint mist recovery device is a multi-stage composite filtration system consisting of a filter screen, an electrostatic filter, and a membrane filtration unit. During implementation, pressure sensors are used to monitor pressure changes at each node of the multi-stage filtration unit.

[0143] In the specific implementation of the first-stage filter:

[0144] Employing a woven metal mesh structure, the filter is made of high-strength, corrosion-resistant 304 stainless steel wire, manufactured through a precision weaving process, resulting in a regularly distributed diamond-shaped mesh. The mesh size is designed to be 100 microns, ensuring a large ventilation area, reducing air resistance, and effectively intercepting large particles and some larger-diameter paint mist. The filter is 2mm thick, possessing good mechanical strength and capable of withstanding long-term impact from paint mist airflow without deformation.

[0145] When the gas containing paint mist enters the pre-filter, large particles and larger paint mist particles cannot pass smoothly through the mesh due to inertia and will collide with the metal wire mesh, thus being intercepted. At the same time, some smaller particles will be adsorbed onto the surface of the wire mesh due to intermolecular forces such as van der Waals forces when they come into contact with the metal wire, achieving preliminary filtration of paint mist particles.

[0146] In practical implementation, the second-stage electrostatic filter is as follows:

[0147] The electrostatic filter consists of a discharge electrode and a dust collection electrode. The discharge electrode is made of barbed stainless steel, which can generate a strong corona discharge at a low voltage, ionizing the surrounding air to form an ion region. The dust collection electrode is made of flat aluminum plate with an anodized surface to enhance corrosion resistance and dust collection performance.

[0148] When paint mist-containing gas passes through an electrostatic filter, under the influence of the strong electric field generated by the discharge electrode, the free electrons and ions in the gas gain sufficient energy to collide with the paint mist particles, causing the particles to become charged.

[0149] In practical implementation, the third-stage membrane filtration unit:

[0150] Polyvinylidene fluoride (PVDF) microporous membranes are used as the core filtration material. These membranes possess advantages such as good chemical stability, high mechanical strength, and moderate hydrophilicity. The membrane module adopts a spiral wound structure, with the membrane sheet, support material, and spacer sequentially wound onto the central water collection pipe and distributed within the negative pressure channel to form a compact filtration unit. The membrane pore size is precisely controlled, with an average pore size of 0.1 micrometers, effectively intercepting tiny particles and solvent molecules.

[0151] Based on Darcy's law, the pressure difference By controlling the pressure below 0.3 MPa, efficient separation of fine particles and solvents can be achieved, reducing the impurity content of the recovered paint to below 0.5%.

[0152] It should be noted that in addition to the three filtration methods mentioned above, gas filtration methods can also be added or used, such as activated carbon adsorption filtration, deep fiber filtration, and precision ceramic filtration. Activated carbon adsorption filtration utilizes the porous structure and strong adsorption capacity of activated carbon to remove residual odors and organic pollutants; deep fiber filtration uses ultra-fine fiber materials to further intercept tiny particles and improve filtration accuracy; precision ceramic filtration utilizes the high hardness and uniform microporous structure of ceramic materials to perform final fine filtration of paint, ensuring that the recycled paint meets industrial reuse standards.

[0153] Of course, those skilled in the art will recognize that the present invention is not limited to the details of the exemplary embodiments described above, but also includes the same or similar structures that can be implemented in other specific forms without departing from the spirit or essential characteristics of the invention. Therefore, the embodiments should be considered illustrative and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

[0154] Furthermore, it should be understood that although this specification describes embodiments, not every embodiment contains only one independent technical solution. This narrative style is merely for clarity. Those skilled in the art should consider the specification as a whole, and the technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.

[0155] The technologies, shapes, and structures not described in detail in this invention are all known technologies.

Claims

1. A method for predicting the concentration of paint mist, characterized in that, Includes the following steps: S1. Establish a three-dimensional spatial coordinate system in the painting operation area and obtain the basic parameters of the painting operation area, including the number of paint mist sources, the three-dimensional spatial coordinates of each paint mist source distribution, and the intensity of each paint mist source; S2. Obtain the basic diffusion coefficient of each paint mist source through wind tunnel experiments and actual painting operation tests, and correct the basic diffusion coefficient through the interference coefficient between each paint mist source to obtain the corrected actual diffusion coefficient of each paint mist source; S3. Input the obtained basic parameters and the actual diffusion coefficient of each paint mist source into the optimized diffusion model to obtain the basic diffusion paint mist concentration at the required predicted three-dimensional spatial coordinate points under the superposition of paint mist diffusion from each paint mist source, and introduce the concentration change law of the paint mist source in the time series and the environmental benchmark concentration of paint mist in the painting operation area into the basic diffusion paint mist concentration, and finally correct the basic diffusion paint mist concentration to the predicted paint mist concentration; The optimized diffusion model in step S3 is as follows: In the formula, For the three-dimensional spatial coordinates of the points to be predicted ( Predicted paint mist concentration at () location, in mg / m³ 3 ; The environmental baseline concentration of paint mist in the painting operation area, in mg / m³. 3 ; The intensity of the i-th paint mist source is expressed in g / min. 、 、 These are the x-coordinate, y-coordinate, and vertical coordinates of the three-dimensional spatial coordinate point to be predicted; 、 、 Let x, y, and y be the x, y, and y coordinates of the i-th paint mist source in three-dimensional space. Let be the actual diffusion coefficient after correction for the i-th paint mist source, which is dimensionless; The quantity of paint mist sources is dimensionless. Let be the angular frequency of the concentration change of the i-th paint mist source, in rad / s; Let be the initial phase of the i-th paint mist source, in rad; The time unit is seconds (s); the basic diffusion coefficient of paint mist for each paint mist source in step S2 is as follows: In the formula, Let be the basic diffusion coefficient of the paint mist from the i-th paint mist source, which is dimensionless; This is the fundamental diffusion coefficient constant term, which is dimensionless. The environmental factor influence coefficient is dimensionless. Represents environmental factors; dimensionless. The number of environmental factors is dimensionless. Let j be the j-th environmental factor variable, dimensionless, where j takes the values ​​1, 2, or 3. For wind speed, For temperature, Humidity; is the coefficient for the influence of paint mist characteristics, dimensionless; This represents the k-th paint mist characteristic variable, where k takes the value 1 or 2. The average particle size of the paint mist. The density of the paint mist is dimensionless. The random error term is dimensionless; the actual diffusion coefficients after correction for each paint mist source in step S2 are as follows: In the formula, The actual diffusion coefficient after correction for paint mist source is dimensionless. Let be the basic diffusion coefficient of the paint mist from the i-th paint mist source, which is dimensionless; Let represent the interference coefficient between the i-th paint mist source and the r-th paint mist source, which is dimensionless and ， This represents the distance between the i-th and r-th paint mist sources, in meters (m). L represents the jet influence characteristic length, also in meters, and is taken as 0.3 times the spray gun range. The angle between the spraying directions of the i-th paint mist source and the r-th paint mist source is expressed in degrees. The intensity of the i-th paint mist source is expressed in g / min. Let be the intensity of the r-th paint mist source, expressed in g / min.

2. The method for predicting paint mist concentration according to claim 1, characterized in that, The specific intensity of the paint mist source is; In the formula, The intensity of the paint mist source is expressed in g / min. The flow rate of paint sprayed by the spray gun is measured in ml / min. This refers to the density of the paint, expressed in g / ml. The percentage of non-volatile solid particles in the paint, dimensionless; It is the proportion of paint mist that can be captured after being atomized by a spray gun, and is dimensionless.

3. The method for predicting paint mist concentration according to claim 1, characterized in that, The optimization process of the diffusion model in step S2 is as follows: S21, collect paint mist data, environmental data and painting condition data in the painting operation area at each time period of the painting operation sequence, and perform three-dimensional spatial modeling of the painting operation area. S22. Clean the data collected in step S1, and divide the cleaned data into training set, validation set and test set according to the time series in a ratio of 7:1:2; S23. Input the parameters in the training set into the diffusion model, use the mean squared error as the loss function, and use the stochastic gradient descent algorithm to optimize the parameters during training. Then, use the adaptive learning rate adjustment algorithm Adagrad to adjust the learning rate during the optimization training process to finally obtain the optimized diffusion model.

4. The method for predicting paint mist concentration according to claim 3, characterized in that, The optimization training process in step S23 includes: S231, randomly initializing all parameters in the model; S232, inputting the data from the training set into the diffusion model and calculating the predicted values ​​of the diffusion model based on the current parameters; S233, substituting the predicted values ​​and the true values ​​into the mean squared error formula to calculate the loss value of the loss function under the current parameters; S234, updating the parameters in the opposite direction of the gradient according to the gradient of the loss function to each parameter, so as to reduce the loss value; S235, repeating steps S231-S234, iterating the training continuously until the loss value converges to the preset structure or reaches the preset maximum number of iterations, and using the validation set to evaluate the model performance during each iteration to prevent the model from overfitting; S236, after training is completed, using the test set to perform a final evaluation of the model, calculating the model's prediction error, accuracy, and other indicators. If the model performance does not meet expectations, adjust the model parameters or optimize the training algorithm, and repeat steps S231-S235 again until the model meets the design requirements and the prediction error is controlled within ±5%.

5. A method for adjusting the intensity of recoverable negative pressure, wherein the method for adjusting the intensity of recoverable negative pressure is applied to the method for predicting the concentration of paint mist as described in any one of claims 1-4, characterized in that, Includes the following steps: K1. Obtain the concentration change curve of the predicted paint mist concentration at the predicted three-dimensional spatial coordinate points of each negative pressure recovery area during the painting operation time by using the paint mist concentration prediction method. K2. Starting from the origin, along the time-increasing direction of the concentration-time curve, select the time interval in which the concentration change in each negative pressure recovery area first exceeds the preset concentration fluctuation threshold, and record it as the first time interval. K3. Starting from the end point of the first time interval, continue along the time-increasing direction of the concentration change curve to select the next time interval in which the concentration change in each negative pressure recovery area exceeds the preset concentration fluctuation threshold, and record it as the second time interval. Repeat this step until each negative pressure recovery area reaches the preset end condition. K4. Obtain the concentration change in each time interval of each negative pressure recovery area. K5. Set each negative pressure recovery area with the end point of its corresponding time node as an adjustment node. Input the concentration change in each time interval into the negative pressure adjustment model, and according to the current negative pressure intensity of the adsorption port in each time interval, obtain the adsorption port negative pressure intensity that the corresponding adjustment node of each negative pressure recovery area needs to adjust, and control each negative pressure recovery area to adjust the adsorption port negative pressure intensity to the adsorption port negative pressure intensity that the corresponding adjustment node needs to adjust at the end point of the corresponding time node.

6. The method for adjusting the negative pressure intensity of recovery according to claim 5, characterized in that, The negative pressure regulation model in step K3 is as follows: In the formula, The negative pressure intensity at the adsorption port of the a-th negative pressure recovery zone after adjustment at the b-th adjustment node is expressed in Pa. The negative pressure intensity of the current adsorption port of the a-th negative pressure recovery zone at the b-th adjustment node is expressed in Pa. The adjustment coefficient is dimensionless. The concentration change in the a-th negative pressure recovery zone during the time interval corresponding to the b-th adjustment node is expressed in mg / m³.

7. A recovery device, wherein the recovery device uses the negative pressure intensity adjustment method described in claim 5 to recover paint mist, characterized in that, The system includes a negative pressure channel connected to the negative pressure inlet of the negative pressure generating mechanism, and a filter screen, an electrostatic filter, and a membrane filter unit arranged sequentially along the negative pressure adsorption direction within the negative pressure channel. The filter screen has a side length of 100 micrometers. The filter unit has 100-micron diamond-shaped pores and a filter thickness of 2 mm. The membrane filter unit is made of polyvinylidene fluoride microporous membrane as the filter substrate.

Images