A painting plant environment quality prediction method and a prediction model establishment method

Abstract

The application discloses a kind of painting workshop environment quality prediction method and prediction model establishment method, the present application selects volatile organic compound concentration as painting workshop environment quality reference index, selects volatile organic compound concentration as dependent variable, spraying time, spraying distance, air flow rate as independent variable, establishes the multiple linear regression model of painting workshop environment quality, establishes neural network model, the fitting value of multiple regression model is used as the input of neural network model together with independent variable, constructs multiple linear regression-neural network combination model, according to painting working condition, the air quality in workshop is predicted using combination model, when volatile organic compound concentration in painting workshop reaches certain threshold value, it is necessary to carry out alarm and forced ventilation.The present application can accurately and quickly calculate volatile organic compound concentration in workshop by painting workshop environment quality prediction model, so as to adjust ventilation volume in time, protect construction personnel, save energy consumption.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of environmental monitoring, and relates to a coating workshop environment quality judgment technology, in particular to a coating workshop environment quality prediction method and a prediction model establishment method. BACKGROUND

[0002] The pollution generated in the coating workshop is one of the most serious stages of pollution in the entire industrial construction process, and has the characteristics of universality and diversity of emission. The most important one is atmospheric pollutants, such as volatile organic compounds (VOCs). Volatile organic compounds can cause harm to human health in many ways, for example, some VOCs with irritating odor can cause headache and nausea, and some VOCs, such as toluene and benzene, are carcinogenic and can pose a serious threat to human health. VOCs not only harm human health in many ways, but also have a serious negative impact on the atmospheric environment. Industrial painting will produce over-spraying paint mist, and the VOCs generated by painting are important precursors of particulate matter and one of the sources of particulate matter formation. VOCs and nitrogen oxides in the atmosphere undergo chemical reactions under the action of light to generate photochemical oxidants, such as ozone, peroxynitro acyl and other substances, which have oxidizing properties, and photochemical smog, and VOCs can also participate in the formation of secondary sols, increasing the probability of haze weather.

[0003] In the painting process, the emission of VOCs is mainly derived from various painting and drying and curing processes, in addition, a small amount of VOCs will be unorganizedly dispersed in the process of transportation, storage and blending of paint. Taking spraying as an example, in the spraying process, the paint attached to the surface of the workpiece volatilizes VOCs gas, and the over-spraying paint mist formed during spraying also volatilizes VOCs gas. In the drying and curing process, the solvent in the paint on the workpiece will volatilize into the air in the form of VOCs gas. For example, if a 300,000-ton super large oil tanker uses 72% high solid primer in the ballast tank, the theoretical consumption of general primer in the ballast tank of the whole ship is about 167,000 liters, and the VOCs generated is 43 tons. Therefore, it is necessary to monitor and predict the volatile organic compounds (VOCs) in the painting process, which is of great significance to the safe operation of related coating enterprises and the physical and mental health of the operating personnel.

[0004] The VOCs concentration on-site monitoring method in the prior art is not mature. On the one hand, the relatively mature method used in the laboratory stage is gas chromatography analysis. Although the accuracy of the technology in detecting the components and concentration of VOCs is very high, the operation process is complex and time-consuming. When the detection result is obtained, a long time has passed, real-time monitoring cannot be performed, and the professional level of the operator is required to be high. Therefore, real-time and accurate monitoring of the VOCs concentration in the painting workshop cannot be realized. On the other hand, the gas chromatograph is relatively expensive, and the actual VOCs concentration that can be monitored at different monitoring points in the painting workshop is quite different, and cannot reflect the VOCs level in the whole painting workshop, that is, it is difficult to effectively evaluate the impact on the production personnel. In addition, the VOCs concentration in the painting workshop is mainly affected by two factors, namely, the painting speed and the ventilation speed. Generally, the optimal control level is to control the ventilation speed according to the VOCs concentration level in the painting workshop, and to design a ventilation volume self-adaptive control system. In this way, energy is not wasted, and the painting workshop can be controlled in safe production. However, real-time monitoring of the comprehensive VOCs concentration in the painting workshop becomes a technical bottleneck for designing the self-adaptive ventilation system.

[0005] Therefore, it is extremely important to detect the VOCs concentration of the fixed sampling point in the workshop under different environments by using the gas chromatography method under the premise that other working conditions remain normal, analyze the rules therein, and establish a prediction model of the VOCs concentration changing with the environment.

[0006] Patent CN103884780A discloses a modeling and prediction method for VOC concentration in a furniture paint workshop. In a furniture paint finishing workshop, a plurality of concentration sampling points are set according to the finishing process arrangement characteristics of the paint workshop, and a test point with high VOC concentration in the workshop is obtained. A plurality of height samples are set at the selected test point with high concentration by using a tower scale, and VOC concentrations at different heights at the same time are obtained. Four parameters of sampling temperature, humidity, air pressure and paint spraying amount in the sampling time are obtained. A VOC concentration height prediction model is established according to the influence of temperature, humidity, air pressure and paint spraying amount on VOC concentration. The specific way of the VOC concentration height prediction model is to use the stepwise regression of the SPSS data processing software, to exclude factors with little correlation, to establish a multiple linear equation, to solve the parameters of the multiple linear equation by using known measurement data, and to obtain the VOC concentration height prediction model. In the solving process, because there is no obvious linear relationship between the dependent variable and the independent variable, the natural logarithm of both sides of the formula is taken, and then linear regression is performed. Whether or not the transformation is performed, only the linear relationship between the dependent variable and the independent variable is considered in the technology. Although the model can predict to a certain extent, the prediction accuracy of the model is poor in practice because only the linear relationship is considered. SUMMARY

[0007] The purpose of the present application is to provide a painting workshop environment quality prediction method and a prediction model establishment method, based on a large amount of obtained painting workshop environment monitoring data, the relationship between volatile organic compounds (VOCs) in the painting workshop, spraying time, spraying distance, and air flow rate is studied by multivariate linear regression analysis method, and an environment quality prediction model is established to ensure the safe and low-consumption operation of related painting enterprises and the physical and mental health of the operating personnel.

[0008] The technical solution adopted by the present application to achieve the above purpose is as follows:

[0009] In one aspect, the present application provides a painting workshop environment quality prediction model establishment method, comprising the following steps:

[0010] Step 1, parameter selection, selecting volatile organic compound concentration as the painting workshop environment quality characterization parameter, and selecting painting process parameters and environmental parameters as independent variables;

[0011] Step 2, selection of monitoring points, according to the field layout of the painting workshop, N explosion-proof integrated monitoring points are installed in the painting room, long-term painting room environment tracking monitoring is carried out, painting process parameters and environmental parameters are recorded, and each group of painting process parameters and environmental parameters of a monitoring point constitute a sample;

[0012] Step 3, data arrangement, according to the distribution of each explosion-proof integrated monitoring point, check whether the data is abnormal, and eliminate or supplement the abnormal data;

[0013] Step 4, screening the processed data for independent variables, eliminating independent variables with small correlation, and finally obtaining K independent variables for establishing a prediction model;

[0014] Step 5, according to the screened independent variables, a multivariate linear regression-neural network combined model is established, specifically as follows:

[0015] Step 5.1, the multivariate linear regression model for initial prediction is as follows:

[0016] f(x)=b+ω1X1+ω2X2+…+ω m X m +εEquation 1

[0017] Wherein: b is a constant; f(x) represents the concentration of volatile organic compounds; X m represents the mth independent variable, m is a positive integer with a value of 1-K; ω m represents the partial regression coefficient of the mth independent variable, and ε is a constant term;

[0018] Step 5.2, construct a composite sample, use a multiple linear regression model to fit each sample data to obtain a fitting value, and re-construct a composite sample by taking the fitting value and the corresponding sample as input and taking the predicted volatile organic compound concentration as output;

[0019] Step 5.3, establish a neural network model, the neural network model includes an input layer, an intermediate layer and an output layer, and each layer of neurons is fully connected with adjacent neurons; determine the number of input layer neurons M, the number of intermediate layer P, the number of neurons in each layer, the number of output layer neurons Q, the activation function, the optimizer function and the learning rate; continuously optimize the neural network model through the composite sample to obtain a multiple linear regression-neural network combined model.

[0020] As a preferred technical solution, in step 1, the coating process parameters include paint type, spraying time and spraying distance, and the environmental parameters include particulate matter concentration, environmental temperature, environmental humidity, air flow rate and air pressure.

[0021] As a preferred technical solution, in step 2, the number N of explosion-proof integrated monitoring points is 2-8.

[0022] As a preferred technical solution, the coating used is solvent-based paint.

[0023] As a preferred technical solution, in step 3, the data is normalized after being sorted.

[0024] As a preferred technical solution, in step 5.3, in the process of layer-by-layer processing of the intermediate layer of the neural network model, the state of each neuron in each layer only affects the state of the next layer of neurons, and the influence formula is as follows:

[0025]

[0026] Each neuron in each layer receives information transmitted by all neurons in the previous layer, and the input information xi transmitted by the i-th neuron in the previous layer corresponds to an input weight β i , which represents the strength of the transmitted signal; the weighted sum of the input signals of each neuron in the previous layer is used as the total input value of a neuron in the current layer, and the total input value is compared with the threshold value θ in the neuron, and only when the sum exceeds θ, the total input value is transmitted as transmission information to the next layer.

[0027] As a preferred technical solution, in step 5.3, the number of input layer neurons M is 3-6, the number of intermediate layers P is 1-2, and the number of output layer neurons Q is 1-3.

[0028] As a preferred technical solution, in step 5.3, the activation function is one of Sigmoid, Tanh, ReLU, LeakReLU, etc., the loss function is one of mean squared error loss function, cross-entropy error loss function, and exponential loss function, etc., the optimizer function is one of GradientDescent, Adadelta, Momentum, and Adam, etc., and the learning rate ranges from 0.001 to 0.1.

[0029] On the other hand, the present invention provides a method for predicting the environmental quality of a painting workshop, which uses a multiple linear regression-neural network combined model established by any of the above methods, and includes the following steps:

[0030] K independent variables are required to collect the coating process parameters and environmental parameters in the coating workshop. Each set of coating process parameters and environmental parameters of a monitoring point constitute a sample.

[0031] Check the data for anomalies and remove or supplement any abnormal data.

[0032] The processed data is input into a multivariate linear regression-neural network combined model, and the output is the concentration of volatile organic compounds that characterizes the environmental quality of the painting workshop.

[0033] The beneficial effects of this invention are:

[0034] (1) Based on monitoring data from multiple integrated monitoring points, this invention establishes a combined analytical model for the concentration of volatile organic compounds (VOCs) in solvent-based paints, considering the effects of spraying time, spraying distance, and air velocity. The fitting model has high accuracy and can effectively reflect the influence of spraying time, spraying distance, and air velocity on VOCs, providing a basis for monitoring pollutant emissions in painting workshops.

[0035] (2) The prediction model of the present invention can meet the needs of safe production management, improve the ability of painting workshop managers to respond to and handle accidents, and can be combined with the workshop control system. When pollutants are about to exceed the standard, the workshop dust removal and ventilation device and organic gas purification device can be intelligently controlled to control the pollutant level in the workshop in a timely manner, and protect the physical and mental health of workshop operators working in a high-hazard environment. Attached Figure Description

[0036] Figure 1 This is a schematic diagram of the combined multiple linear regression and neural network model.

[0037] Figure 2 The process for constructing a combined multiple linear regression-neural network model.

[0038] Figure 3Influence of spray time on VOCs concentration.

[0039] Figure 4 Influence of spray distance on VOCs concentration.

[0040] Figure 5 Influence of air flow rate on VOCs concentration.

[0041] Figure 6 Distribution of VOCs concentration measured by toluene sensor at a horizontal distance of 1 m from the spray port.

[0042] Figure 7 Distribution of VOCs concentration measured by xylene sensor at a horizontal distance of 1 m from the spray port. DETAILED DESCRIPTION

[0043] The embodiments of the present application will be further described in detail below with reference to the accompanying drawings and examples. The following examples are used to illustrate the present application, but cannot be used to limit the scope of the present application.

[0044] As shown in Figure 2 , the present application provides a method for establishing a coating workshop environment quality prediction model, and the specific steps are as follows:

[0045] S1, parameter selection, selecting volatile organic compound concentration (VOCs) as the coating workshop environment quality characterization parameter, and selecting coating process parameters and environmental parameters as independent variables;

[0046] S2, selection of monitoring points, according to the field layout of the coating workshop, installing N explosion-proof integrated monitoring points in the coating room, conducting long-term coating room environment tracking monitoring, recording coating process parameters and environmental parameters, and each group of coating process parameters and environmental parameters of a monitoring point constitute a sample; the coating process parameters at least include paint type, spraying time and spraying distance, and the environmental parameters include particulate matter concentration, environmental temperature, environmental humidity, air flow rate and air pressure.

[0047] The present application selects two integrated monitoring points of the coating process room as test objects, and the sensors used for testing by the explosion-proof integrated monitoring equipment are shown in Table 1.

[0048] Table 1 Test sensor

[0049]

[0050] S3, data arrangement, according to the distribution of each explosion-proof integrated monitoring point, checking whether the data is abnormal, rejecting or supplementing the abnormal data, and normalizing the data;

[0051] S4, the processed data is subjected to independent variable screening, and independent variables with small correlation are removed, and finally K (K=3) independent variables are obtained for establishing a prediction model; the details are as follows:

[0052] First step: the concentration of volatile organic compounds (VOCs) is plotted against the spraying time, spraying distance and air flow rate respectively to observe the distribution trend, as shown in Figures 2-5 ;

[0053] Second step: the model variables are screened by stepwise regression method, the concentration of volatile organic compounds is selected as the dependent variable, and the spraying time, spraying distance and air flow rate are selected as the independent variables, and the results are shown in the following table:

[0054] Table 2 Analysis of model variables

[0055]

[0056]

[0057] Step 5, according to the screened independent variables, a multiple linear regression-neural network combined model is established, and the details are as follows:

[0058] Step 5.1, the multiple linear regression model for initial prediction is established as follows:

[0059] f(x) = b + ω1X1 + ω2X2 + … + ω m X m + ε Formula 1

[0060] Where: b is a constant, when X1, X2, … X m are all 0, this constant represents the value of f(x), which is called intercept; f(x) represents the concentration of volatile organic compounds; X m represents the mth independent variable, m is a positive integer with a value of 1-K; ω m represents the partial regression coefficient of the mth independent variable, ω T = (ω1, ω2, … ω m ) is the partial regression coefficient matrix; ε is a constant term, which is a random error generated after removing the influence of all independent variables on the model, also known as residual error;

[0061] The correlation coefficient test and t test are performed on the fitted multiple linear regression model, as shown in Tables 3 and 4:

[0062] Table 3 Correlation coefficient test

[0063] R [R 2 ]]> Sig Lag regression coefficient 0.764 0.782 0.000 0.1

[0064] Table 4 t test

[0065] Model Non-standardized coefficient Standardized coefficient t Significance Constant 398.212 - 6.972 0.00 Paint spraying time 17.386 -0.649 2.981 0.00 Spraying distance -19.733 -0.557 4.851 0.00 Air flow rate -32.142 -0.285 -9.215 0.00

[0066] Step 5: Based on the above tests, the following model can be obtained:

[0067] f(x)=17.386x1-19.733x2-32.142x3+189.75

[0068] Step 5.2: Construct a composite sample. Use a multiple linear regression model to fit the data of each sample to obtain the fitted value. Use the fitted value and the corresponding sample as input, and use the predicted volatile organic compound concentration as output to reconstruct the composite sample. In this embodiment, the input samples are the fitted value of the multiple linear regression model, the painting time, the painting distance, and the air velocity.

[0069] Step 5.3: Establish a neural network model, such as... Figure 1 As shown, the neural network model includes an input layer, intermediate layers, and an output layer. The neurons in each layer are fully connected only to their adjacent neurons. The number of neurons M in the input layer, the number of intermediate layers P, the number of neurons in each layer, the number of neurons Q in the output layer, the activation function, the optimizer function, and the learning rate are determined. The neural network model is continuously learned and optimized through composite samples to obtain a combined multiple linear regression-neural network model.

[0070] In this embodiment, a BP neural network is used as the neural network model. The fitted value of the multiple linear regression model, the painting time, the painting distance, and the air velocity are selected as the input layer neurons of the combined model. The output layer neurons are the volatile organic compound concentration. There are 9 hidden layer neurons and 1 intermediate layer. The activation function of the combined model is Sigmoid, the loss function is the mean squared error loss function, the optimizer function is the Adam Optimizer function, and the learning rate is 0.01.

[0071] A comparative analysis of the multiple linear regression-neural network combined model is shown in Table 5:

[0072] Table 5 Results Analysis

[0073] Serial number Actual value Simulated value Relative error Working condition 1 247.3 220.81 9.91％ Working condition 2 82.9 87.16 5.14％ Working condition 3 114.2 123.41 8.07％

[0074] Multiple linear regression models offer good interpretability and low computational complexity, but their fitting performance is relatively weak. Nonlinear neural network models, on the other hand, exhibit strong fitting performance but are prone to overfitting and have high computational complexity. This invention combines linear and nonlinear models, using the fitting results of a multiple linear regression model along with the independent variables as input neurons in a neural network model. By following the neural network model construction steps, a combined multiple linear regression-neural network model is obtained, achieving the best fitting result.

[0075] Once the combined model of multiple linear regression and neural network is obtained, the environmental quality of the painting workshop can be predicted. The specific method is as follows:

[0076] N monitoring points are set up in the painting room, and each monitoring point is equipped with an explosion-proof integrated monitoring device to monitor environmental parameters. Of course, for the multiple linear regression-neural network combined model established in the above embodiment, the only environmental parameter is air velocity. Therefore, the explosion-proof integrated monitoring device needs to be equipped with at least a wind direction and wind speed sensor. Other sensors can be installed as needed. At least two painting process parameters, spraying time and spraying distance, are recorded. The air velocity, spraying time and spraying distance are combined to form a data collection sample. Each monitoring point is a data sample, and data collection is completed at regular intervals.

[0077] The data collected from N monitoring points are checked, and abnormal data are removed or supplemented to obtain N samples;

[0078] The processed N sample data are input into the multiple linear regression-neural network combined model. The output result is the volatile organic compound concentration that characterizes the environmental quality of the painting workshop. The average value is taken as the current average level of volatile organic compound concentration in the painting workshop.

[0079] The algorithm designed based on the above prediction method is encapsulated in the controller and combined with explosion-proof integrated monitoring equipment to form a painting workshop environmental quality prediction system, which can be produced as a complete set and sold on the market.

[0080] This invention employs a multiple linear regression method, based on monitoring data from multiple integrated monitoring points, to establish a regression analysis model for the concentration of volatile organic compounds (VOCs) in solvent-based paints in relation to spraying time, spraying distance, and air velocity. The fitted model exhibits high accuracy, with clear and concise formulas, effectively reflecting the impact of spraying time, spraying distance, and air velocity on VOCs, thus providing a basis for monitoring pollutant emissions in painting workshops.

[0081] The predictive model of this invention can meet the needs of safe production management, improve the ability of painting workshop managers to respond to and handle accidents, and can be combined with the workshop control system to intelligently control the workshop dust removal and ventilation devices and organic gas purification devices when pollutants are about to exceed the standard, so as to control the pollutant level in the workshop in a timely manner and protect the physical and mental health of workshop operators working in a high-hazard environment.

[0082] The above embodiments are for illustrative purposes only and are not intended to limit the scope of the invention. Although the invention has been described in detail with reference to the embodiments, those skilled in the art should understand that various combinations, modifications, or equivalent substitutions of the technical solutions of the invention do not depart from the spirit and scope of the invention and should be covered within the scope of the claims of the invention.

Claims

1. A method of establishing a paint shop environment quality prediction model, characterized by, Comprise the following steps: Step 1, parameter selection, select volatile organic compounds concentration as the coating workshop environment quality characterization parameter, select coating process parameters and environmental parameters as independent variables; Step 2, selection of monitoring points, according to the field layout of the coating workshop, install N explosion-proof integrated monitoring points in the coating room, carry out long time coating room environment tracking monitoring, record the coating process parameters and environmental parameters, each group of coating process parameters and the environmental parameters of a monitoring point constitute a sample; Step 3, data processing, according to the distribution of each explosion-proof integrated monitoring point, check whether the data is abnormal, and eliminate or supplement the abnormal data; Step 4, the processed data is screened according to the independent variable, the independent variable with small correlation is eliminated, and finally K independent variables are obtained for establishing the prediction model; Step 5, according to the screened independent variables, a multiple linear regression-neural network combined model is established, as follows: Step 5.1, the multiple linear regression model for initial prediction is as follows: Equation One wherein: is a constant; represents the concentration of volatile organic compounds; represents the mth independent variable, m being a positive integer taking values from 1 to K; represents the partial regression coefficient of the mth independent variable, is a constant term; Step 5.2, construct a composite sample, use the multiple linear regression model to fit each sample data to obtain the fitting value, and use the fitting value and the corresponding sample as input, and use the predicted volatile organic compound concentration as output to reconstruct the composite sample; Step 5.3, establish a neural network model, the neural network model includes input layer, intermediate layer and output layer, each layer of neurons is only connected with adjacent neurons; determine the number of input layer neurons M, the number of intermediate layer P, the number of neurons in each layer, the number of output layer neurons Q, the activation function, the optimizer function and the learning rate; through the composite sample, the neural network model is continuously learned and optimized to obtain a multiple linear regression-neural network combined model; In step 5.3, in the process of layer-by-layer processing in the intermediate layer of the neural network model, the state of each layer of neurons only affects the state of the next layer of neurons, and the influence formula is as follows: Formula 2 Each neuron in each layer receives information transmitted from all neurons in the previous layer, and the input information transmitted from the first neuron in the previous layer i x 1 corresponds to an input weight , the size of which represents the strength of the transmitted signal; the weighted sum of the input signals of each neuron in the previous layer is taken as the total input value of a certain neuron in the current layer, and the total input value is compared with the threshold value in the neuron In step 1, the coating process parameters include paint type, spraying time and spraying distance, and the environmental parameters include particulate matter concentration, environmental temperature, environmental humidity, air flow rate and air pressure. In step 2, the number of explosion-proof integrated monitoring points N is 2-8. Only when the sum exceeds the threshold value will the total input value be transmitted as transmitted information to the next layer.​​ 2. The method according to claim 1, wherein: The coating used is solvent-based paint.

3. The method of claim 1, wherein the method further comprises: determining a plurality of environmental quality parameters of the painting plant environment; and determining a plurality of environmental quality parameters of the painting plant environment. In step 3, after data processing, the data is normalized.

4. The method of claim 1, wherein the method further comprises: In step 5.3, the number of input layer neurons M is 3-6, the number of intermediate layers P is 1-2, and the number of output layer neurons Q is 1-3.

5. The method of claim 1, wherein the method further comprises: In step 5.3, the activation function is one of Sigmoid, Tanh, ReLU and LeakRelu, the loss function is one of mean square error loss function, cross entropy error loss function and exponential loss function, the optimizer function is one of GradientDescent, Adadelta, Momentum and Adam, and the learning rate is 0.001-0.

1.

6. The method of claim 1, wherein: Comprise the following steps:

7. The method of claim 1, wherein the method further comprises: determining a plurality of environmental quality parameters of the painting plant environment; and determining a plurality of environmental quality parameters of the painting plant environment. Collect K independent variables required in coating process parameters and environmental parameters of coating workshop, each group of coating process parameters and the environmental parameters of a monitoring point constitute a sample; 8. A method for predicting the quality of a painting plant environment, characterized by using the multiple linear regression-neural network combined model established by any one of claims 1-7. ​ ​ Check whether the data is abnormal, and eliminate or supplement the abnormal data; The processed data is input into the multiple linear regression-neural network combined model, and the output result is the volatile organic compound concentration representing the environment quality of the painting workshop.

Images