Accurate oil distribution method and system for matte spraying

Abstract

The invention relates to the technical field of spraying processes, and discloses an accurate oil distribution method and system for matte spraying, and the method comprises the steps: collecting data through a spectrum analyzer, a color difference meter and an environment temperature and humidity sensor, carrying out the dimension reduction of high-dimensional data to construct a spectral feature vector, and predicting an initial oil distribution ratio through a multi-layer perceptron model; a spraying process rheological characteristic coupling model is constructed based on a measurement result of a dynamic rheometer, oil distribution and spraying parameters are optimized by adopting an improved particle swarm optimization algorithm, and finally, real-time feedback control is realized through a fuzzy PID (Proportion Integration Differentiation) controller. The system comprises a data acquisition module, an oil distribution prediction module, a coupling modeling module, an optimization module and a real-time control module, and all the modules work cooperatively to achieve full-process automation. According to the invention, the precision of matte spraying oil distribution is improved, the environmental adaptability is enhanced, the process synergy is optimized, and the product quality stability and the production efficiency are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of spraying processes, and particularly to a precise oil blending method and system for matte spraying. Background Art

[0002] In modern industrial production, matte coatings are widely used in many fields such as automobiles, furniture, and electronic devices due to their unique appearance effects and excellent properties. However, there are many problems in the oil blending link of current matte spraying, which seriously restrict product quality and production efficiency.

[0003] Traditional matte spraying oil blending methods mostly rely on manual experience. Workers rely on long-term accumulated operating habits and subjective judgments to adjust the proportions of resin, curing agent, matte powder, and solvent. This method is greatly affected by the individual skill levels and working states of workers. The oil blending proportions prepared by different workers vary significantly, resulting in uneven matte effects of products in the same batch, and it is difficult to meet the strict requirements of the market for product quality consistency. Moreover, manual blending is inefficient, and in large-scale production scenarios, it will cause an extended production cycle and increased costs. For example, on the matte spraying production line of automotive parts, due to the instability of manual oil blending, the gloss of the coatings of some parts may not meet the standards and need to be reworked, which not only wastes a large amount of raw materials and labor costs but also affects the smooth progress of the entire production process.

[0004] Some existing automated oil blending systems, although improving the accuracy of oil blending to a certain extent, still have great limitations. These systems often do not fully consider the influence of environmental factors on oil blending and spraying effects. Changes in environmental temperature and humidity will change the physical and chemical properties of the oil blending system, thereby affecting the drying speed, gloss, and adhesion of the coating. For example, in a high-humidity environment, the solvent evaporation speed slows down, which may lead to an extended drying time of the coating and even defects such as sagging and whitening; while at a low temperature, the curing reaction rate decreases, which will affect the final hardness and wear resistance of the coating. However, the existing automated oil blending systems cannot monitor and compensate for the effects brought by these environmental factors in real time, making it difficult for the oil blending plan to adapt to complex and changeable production environments.

[0005] In addition, there are also deficiencies in the collaborative optimization of the oil distribution ratio and spraying process parameters in the prior art. The oil distribution ratio is closely related to parameters such as the atomization pressure of the spraying equipment and the distance between the spray gun. Inappropriate spraying process parameters will lead to poor atomization effect of the oil distribution system, resulting in problems such as particle feeling and uneven thickness of the coating. However, most current oil distribution systems and spraying processes are designed independently of each other, and no effective coupling relationship has been established. In actual production, operators often can only adjust these parameters through multiple trials and errors, lacking scientific theoretical basis and systematic optimization methods. This not only increases production costs but also reduces production efficiency, making it difficult to meet the requirements of modern industry for high-efficiency and precise production.

[0006] In terms of spectral matching degree and rheological stability, it is also difficult to achieve precise control in the existing matte spraying technology. The spectral matching degree determines whether the color and gloss effect of the coating meet the target requirements, while the rheological stability affects the fluidity and uniformity of the oil distribution system during the spraying process. The existing oil distribution and spraying technologies cannot accurately predict and control these two key indicators, resulting in large fluctuations in the quality of products and a high rejection rate in actual production. For example, in the matte spraying of furniture surfaces, if the spectral matching degree is inaccurate, it will cause a deviation between the color of the furniture and the design scheme, affecting the aesthetics of the product; poor rheological stability may lead to defects such as flow marks and orange peel on the coating, reducing the quality of the product. Summary of the Invention

[0007] The purpose of the present invention is to provide a precise oil distribution method and system for matte spraying to solve the problems mentioned in the above background technology.

[0008] To achieve the above purpose, the present invention provides the following technical solution: A precise oil distribution method for matte spraying, the method includes:

[0009] Step 1: Collect the reflection spectral data and chromaticity parameters of the target matte coating through a spectral analyzer and a color difference meter, and obtain the spraying environment parameters in combination with an environmental temperature and humidity sensor;

[0010] Step 2: Based on the spectral characteristics of the target coating, use the high-dimensional data dimensionality reduction method to extract the principal component spectrum and construct the reflection spectral feature vector of the matte coating;

[0011] Step 3: Establish a mapping relationship through the spraying experimental data and the historical formula database, and use a multi-layer perceptron model to predict the initial oil distribution ratio, including the mixing ratio of resin, curing agent, matte powder and solvent;

[0012] Step 4: Based on the dynamic rheometer to measure the rheological properties of the oil distribution system, combined with the atomization pressure and spray gun distance parameters of the spraying equipment, construct a spraying process - rheological property coupling model;

[0013] Step 5: Use the improved particle swarm optimization algorithm to jointly optimize the oil blending ratio and spraying parameters, and generate the optimal oil blending scheme with the spectral matching degree and rheological stability as the constraint conditions;

[0014] Step 6: Dynamically adjust the oil delivery rate and spray gun parameters during the spraying process through a real-time feedback control system to achieve closed-loop precise oil blending.

[0015] Preferably, the high-dimensional data dimensionality reduction method in Step 2 uses the kernel principal component analysis algorithm to perform non-linear dimensionality reduction after mapping the reflection spectral data to a high-dimensional feature space, extract the first k principal components to form a feature vector, and calculate the similarity between samples using the Gaussian kernel function.

[0016] Preferably, the multi-layer perceptron model in Step 3 adopts a double hidden layer structure. The input layer includes the spectral feature vector, environmental parameters, and substrate type, and the output layer is the proportion of each component; prevent overfitting through the Bayesian regularization method, and the loss function is the weighted sum of the mean square error and the L2 regularization term.

[0017] Preferably, the coupling model in Step 4 establishes the latent variable relationship between the spraying pressure, spray gun distance, and rheological characteristic parameters through the partial least squares regression algorithm, and introduces interaction terms to describe the non-linear effect. The model parameters determine the optimal number of latent variables through cross-validation.

[0018] Preferably, the improved particle swarm optimization algorithm in Step 5 adopts an adaptive inertia weight strategy to dynamically adjust the global and local search capabilities according to the number of iterations, and introduces the Cauchy mutation operator to enhance the population diversity and avoid falling into local optimal solutions.

[0019] Preferably, the scale parameter of the Cauchy mutation operator is adaptively adjusted according to the fitness variance of the current population, the mutation probability is negatively correlated with the number of iterations, and the mutation amplitude follows the Cauchy distribution.

[0020] Preferably, the real-time feedback control system in Step 6 uses a fuzzy PID controller. The input is the deviation between the real-time spectral monitoring data and the target value, and the output is the pulse frequency of the oil pump and the adjustment amount of the spray gun pressure. The control rule base is generated through the joint training of expert experience and historical data.

[0021] Preferably, the membership function of the fuzzy PID controller uses a Gaussian function, the defuzzification method is the centroid method, and the rule weights are optimized online through the gradient descent algorithm.

[0022] Preferably, during the optimization process of Step 5, the robustness evaluation of the oil blending scheme is introduced, and the influence of raw material concentration fluctuations on the spectral matching degree is analyzed through Monte Carlo simulation to screen out the optimal solutions that are insensitive to noise.

[0023] Preferably, the present invention further includes a precise oil distribution system for matte spraying, and the system includes:

[0024] A data acquisition module for obtaining spectral data, environmental parameters and rheological properties;

[0025] An oil distribution prediction module for generating an initial oil distribution plan based on a multi-layer perceptron model;

[0026] A coupling modeling module for establishing a spraying process - rheological property coupling model;

[0027] An optimization module for optimizing oil distribution parameters by using an improved particle swarm algorithm;

[0028] A real-time control module for dynamically adjusting spraying equipment parameters through a fuzzy PID controller;

[0029] The modules are connected to an oil distribution pump, a spray gun and sensors through an industrial bus to achieve full-process automatic control.

[0030] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0031] In terms of accuracy, comprehensive data is collected through devices such as a spectral analyzer, a color difference meter and an environmental temperature and humidity sensor, and a precise spectral feature vector is constructed by using a high-dimensional data dimensionality reduction method, providing a reliable basis for subsequent oil distribution prediction. The multi-layer perceptron model is trained based on a large amount of spraying experiment data and a historical formula database, and can accurately predict the initial oil distribution ratio, greatly improving the accuracy of oil distribution compared with traditional manual experience-based oil distribution. Moreover, during the optimization process, with the spectral matching degree and rheological stability as constraint conditions, an improved particle swarm optimization algorithm is used to jointly optimize the oil distribution ratio and spraying parameters, further ensuring the accuracy of the final oil distribution plan. For example, in the matte spraying of an automobile body, the present invention can control the spectral matching degree error of the coating within a very small range, making the body color more uniform and pure, meeting the strict requirements of high-end automobiles for appearance quality.

[0032] In terms of environmental adaptability, the system collects spraying environmental parameters in real time and uses them as model inputs, fully considering the influence of environmental temperature and humidity on oil distribution and spraying effects. Whether in the humid southern region or the dry and cold northern region, the oil distribution plan and spraying parameters can be adjusted according to the real-time environment to ensure the stable performance of the coating. For example, in an environment with high humidity, the system will automatically adjust the proportion of the solvent and the atomization pressure of the spray gun to accelerate the solvent evaporation speed and avoid phenomena such as sagging and whitening of the coating; in a low-temperature environment, the proportion of the curing agent is optimized to promote the curing reaction and ensure that the hardness and wear resistance of the coating meet the standards.

[0033] In terms of process collaborative optimization, the rheological property coupling model of the spraying process constructed by the present invention reveals the internal relationship between the rheological properties of the oil distribution system and the spraying process parameters. Through this model, spraying parameters such as atomization pressure and spray gun distance can be accurately adjusted according to the rheological properties of the oil distribution system, realizing the perfect coordination of the oil distribution and spraying processes. For example, for an oil distribution system with a relatively high viscosity, the atomization pressure is appropriately increased to enable better atomization of the coating material, evenly spraying it on the surface of the workpiece, avoiding problems such as particle sense and uneven thickness, and effectively improving the quality and appearance effect of the coating.

[0034] In terms of quality stability and robustness, the present invention introduces the robustness evaluation of the oil distribution scheme during the optimization process, analyzes the influence of raw material concentration fluctuations on the spectral matching degree through Monte Carlo simulation, and screens out the optimal solution that is insensitive to noise. This means that even when there are certain fluctuations in the raw material concentration, the spectral matching degree and other performance indicators of the coating can be guaranteed to be stable. In actual production, the quality of raw materials may vary to a certain extent. Adopting the technology of the present invention can effectively reduce product quality problems caused by raw material fluctuations, reduce the scrap rate, and improve the stability of product quality. For example, in the matte spraying production of electronic device casings, even if there are slight changes in the concentration of raw materials in different batches, the oil distribution system of the present invention can automatically adjust to ensure that the quality of the casing coating remains consistent, enhancing the market competitiveness of the product.

[0035] In terms of production efficiency, the present invention realizes full-process automatic control. The data acquisition module, oil distribution prediction module, coupling modeling module, optimization module, and real-time control module are connected to the oil distribution pump, spray gun, and sensors through an industrial bus, without frequent manual intervention. From data acquisition to the generation of the oil distribution scheme and then to the real-time adjustment during the spraying process, the entire process is efficient and fast. Compared with the traditional manual oil distribution and trial-and-error parameter adjustment methods, the production cycle is significantly shortened, and the production efficiency is improved. For example, on a large-scale furniture production line, after adopting the technology of the present invention, the product output per unit time has increased significantly, while reducing rework and scrap caused by quality problems, reducing production costs, and bringing considerable economic benefits to the enterprise. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 is the working principle diagram of the precise oil distribution method for matte spraying described in the present invention;

[0037] Figure 2 is the working principle diagram of the improved particle swarm optimization algorithm;

[0038] Figure 3 is the working principle diagram of the Cauchy mutation operator;

[0039] Figure 4 is the step diagram of the fuzzy PID controller parameter optimization. Detailed implementation manners

[0040] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0041] Please refer to Figures 1-4 , the present invention provides a precise oil blending method and system for matte spraying, and its overall implementation scheme is as follows:

[0042] Step 1: Collect the reflection spectrum data and chromaticity parameters of the target matte coating through a spectral analyzer and a color difference meter, and obtain the spraying environment parameters in combination with an environmental temperature and humidity sensor. The spectral analyzer can accurately measure the reflection light intensity of the coating at different wavelengths to obtain the reflection spectrum data; the color difference meter is used to obtain the chromaticity parameters of the coating, such as the lightness, hue, and saturation of the color, etc. The environmental temperature and humidity sensor monitors the temperature and humidity of the spraying environment in real time. These environmental parameters will affect the final effect of the coating, so it is necessary to accurately obtain them.

[0043] Step 2: Based on the spectral characteristics of the target coating, use a high-dimensional data reduction method to extract the principal component spectrum and construct a reflection spectrum feature vector of the matte coating. The high-dimensional data reduction method can extract key information from complex spectral data, reduce the data dimension, and improve the subsequent processing efficiency. By extracting the principal component spectrum, a vector that can represent the spectral characteristics of the target coating is constructed, providing a basis for subsequent analysis and prediction.

[0044] Step 3: Establish a mapping relationship between the spraying experiment data and the historical formula database, and use a multi-layer perceptron model to predict the initial oil blending ratio, including the mixing ratio of resin, curing agent, matte powder, and solvent. A large amount of actual data has been accumulated in the spraying experiment, and the historical formula database contains past successful formula information. The multi-layer perceptron model uses these data to learn the relationship between spectral characteristics, environmental parameters, etc. and the oil blending ratio, so as to predict the initial oil blending ratio.

[0045] Step 4: Based on the dynamic rheometer to measure the rheological properties of the oil blending system, combined with the atomization pressure and spray gun distance parameters of the spraying equipment, construct a spraying process rheological property coupling model. The dynamic rheometer can measure the rheological parameters of the oil blending system under different conditions, such as viscosity, elastic modulus, etc. Combining these rheological properties with the key parameters of the spraying equipment, a coupling model is constructed to describe the influence of spraying process parameters on rheological properties and the effect of rheological properties on coating quality.

[0046] Step 5: Use the improved particle swarm optimization algorithm to jointly optimize the oil blending ratio and spraying parameters. With the spectral matching degree and rheological stability as the constraint conditions, generate the optimal oil blending scheme. The improved particle swarm optimization algorithm can find the optimal solution in a complex parameter space. Taking the spectral matching degree and rheological stability as the constraint conditions can ensure that the optimized oil blending scheme can not only make the spectral characteristics of the coating match the target, but also ensure the rheological stability of the oil blending system during the spraying process.

[0047] Step 6: Dynamically adjust the oil blending delivery rate and spray gun parameters during the spraying process through a real-time feedback control system to achieve closed-loop precise oil blending. The real-time feedback control system dynamically adjusts the oil blending delivery rate and spray gun parameters according to the deviation between the real-time monitoring data of the spectrum and the target value. Through continuous adjustment, the actual spraying effect gradually approaches the target value, realizing precise oil blending and high-quality matte spraying.

[0048] The present invention will be further described below in conjunction with Examples 1 to 5:

[0049] Example 1:

[0050] This example details how to use the kernel principal component analysis algorithm to reduce the dimension of the reflection spectrum data, extract the principal component spectrum, and construct an accurate reflection spectrum feature vector to provide effective data input for subsequent model prediction and analysis.

[0051] In step 2, the kernel principal component analysis (Kernel Principal Component Analysis, KPCA) algorithm is used for high-dimensional data reduction. First, map the reflection spectrum data X to the high-dimensional feature space Φ(X). Assume that the original data X is an n×p matrix, where n is the number of samples and p is the dimension of the spectral data (i.e., the number of wavelengths).

[0052] Use the Gaussian kernel function to calculate the similarity between samples, where x i and x j are two samples in the original data, and σ is the bandwidth parameter of the Gaussian kernel function. Through the Gaussian kernel function, the original data is nonlinearly mapped in the high-dimensional feature space.

[0053] Then, perform principal component analysis on the mapped data in the high-dimensional feature space. Calculate the kernel matrix K, whose elements are K ij = K(x i , x j ). Centralize the kernel matrix K to obtain

[0054] Then, solve the eigenvalues λ m and eigenvectors vm (m = 1, 2, …, n). Select the eigenvectors v1, v2, …, v corresponding to the top k largest eigenvalues. k , and construct the eigenvector matrix V = [v1, v2, …, v k .

[0055] Finally, take the projection coefficients of the original data in the high-dimensional feature space as the principal component spectra to form the reflectance spectral feature vector. For a new sample x, its projection on the principal components is thus obtaining the k-dimensional reflectance spectral feature vector [α1, α2, …, α k . In this way, the non-linear dimensionality reduction of the high-dimensional reflectance spectral data is achieved, effectively extracting the key information and providing a more concise and effective data representation for subsequent analysis and processing.

[0056] Example 2:

[0057] In step 3, the multi-layer perceptron model adopts a double hidden layer structure. The input layer includes the spectral feature vector, environmental parameters (temperature, humidity), and substrate type. The spectral feature vector is obtained through the kernel principal component analysis algorithm in step 2, which represents the spectral characteristics of the target coating; the environmental parameters reflect the environmental conditions during spraying, and these factors will affect the formation and performance of the coating; different substrate types have different requirements for the oil mixing ratio.

[0058] Let the number of neurons in the input layer be n in , which is jointly determined by the dimension of the spectral feature vector, the number of environmental parameters, and the encoding number of the substrate type.

[0059] The number of neurons in the first hidden layer is n h1 , the number of neurons in the second hidden layer is n h2 , and the number of neurons in the output layer is n out , corresponding to the mixing ratios of resin, curing agent, matte powder, and solvent, that is, n out = 4.

[0060] The activation functions of the hidden layer and the output layer are selected as the ReLU function and the linear function respectively. The expression of the ReLU function is f(x) = max(0, x), which can effectively alleviate the gradient disappearance problem and improve the training efficiency of the model.

[0061] The Bayesian regularization method is used to prevent overfitting. The loss function is defined as the weighted sum of the mean square error (MSE) and the L2 regularization term, that is:

[0062]

[0063] where, N is the number of training samples, y iis the actual oil distribution ratio, is the oil distribution ratio predicted by the model; λ is the regularization coefficient used to balance the weights of the mean square error and the regularization term; W is the set of all trainable parameters (weights and biases) in the model.

[0064] During the training process, the Stochastic Gradient Descent (SGD) algorithm is used to update the model parameters. By continuously adjusting the weights and biases of the model, the loss function is gradually reduced, enabling the model to learn the complex relationships between spectral features, environmental parameters, etc. and the oil distribution ratio, and achieving accurate prediction of the initial oil distribution ratio.

[0065] Example 3:

[0066] This example details how to establish a coupling model between spray process parameters (atomization pressure, spray gun distance) and rheological characteristic parameters of the oil distribution system through the partial least squares regression algorithm, considering the nonlinear effect, to provide an accurate model basis for subsequent parameter optimization.

[0067] In step 4, the Partial Least Squares Regression (PLSR) algorithm is used to establish a coupling model for the rheological characteristics of the spray process. Let the spray pressure be x1, the spray gun distance be x2, and the rheological characteristic parameter (such as viscosity) be y.

[0068] First, the independent variables x1 and x2 and the dependent variable y are standardized so that their means are 0 and variances are 1.

[0069] Then, the partial least squares regression algorithm is used to find the latent variable relationships between the independent and dependent variables. Assume there are latent variables t1 and t2, which are linear combinations of x1 and x2 respectively, i.e., t1 = w 11 x1 + w 12 x2, t2 = w 21 x1 + w 22 x2, where w ij are the weight coefficients.

[0070] At the same time, the dependent variable y can also be expressed as a linear combination of the latent variables y = b0 + b1t1 + b2t2 +, where b0 is the intercept, b1 and b2 are the regression coefficients, and is the error term.

[0071] To describe the nonlinear effect, the interaction term x1x2 is introduced. The interaction term is added to the model, and the partial least squares regression analysis is performed again to determine the new latent variable relationships and regression coefficients.

[0072] The model parameters determine the optimal number of latent variables through cross-validation. The training data is divided into k subsets. Each time, one of the subsets is selected as the test set, and the remaining subsets are used as the training set. By comparing the prediction errors (such as the root mean square error RMSE) of the model on the test set under different numbers of latent variables, the number of latent variables that minimizes the prediction error is selected as the optimal value. The established coupling model can more accurately reflect the influence of spraying process parameters on rheological properties, providing a reliable basis for subsequent parameter optimization.

[0073] Example 4:

[0074] In step 5, the improved particle swarm optimization algorithm is used to jointly optimize the oil distribution ratio and spraying parameters. In the particle swarm optimization algorithm, each particle represents a set of oil distribution ratios and spraying parameters. Let the position vector of particle i be X i =(x i1 ,x i2 ,…,x in ), and the velocity vector be V i =(v i1 ,v i2 ,…,v in ), where n is the number of optimization parameters.

[0075] An adaptive inertia weight strategy is adopted. The inertia weight w is dynamically adjusted according to the iteration number t, and the formula is:

[0076]

[0077] where, w max and w min are the maximum and minimum values of the inertia weight respectively, and T is the maximum number of iterations. In the initial stage of iteration, a larger inertia weight is beneficial for the particles to conduct global search and quickly explore the entire solution space; as the number of iterations increases, the inertia weight gradually decreases, enabling the particles to pay more attention to local search and finely adjust the quality of the solution.

[0078] The Cauchy mutation operator is introduced to enhance the population diversity. The scale parameter of the Cauchy mutation operator is adaptively adjusted according to the fitness variance σ 2 of the current population, and the formula is:

[0079]

[0080] where, τ is a constant threshold used to prevent the scale parameter from being too small. The mutation probability P m is negatively correlated with the iteration number t and can be expressed as where P m0 and P m1 are the initial mutation probability and the final mutation probability respectively.

[0081] When the particles need to mutate, the mutation amplitude follows the Cauchy distribution. For the j-th dimension x of particle i ij , the mutated position x i ′ j is:

[0082] x i ′ j = x ij +γC(0,1)

[0083] where C(0,1) is a random number following the standard Cauchy distribution. Through this adaptive adjustment and mutation operation, the improved particle swarm optimization algorithm can better balance the global search and local search capabilities during the search process, avoid falling into local optimal solutions, and thus find the optimal oil blending scheme and spraying parameters that meet the requirements of spectral matching degree and rheological stability.

[0084] Example 5:

[0085] This example elaborates in detail the design of a fuzzy PID controller in a real-time feedback control system, including input and output settings, generation of a control rule base, selection of membership functions and defuzzification methods, as well as the specific implementation method of introducing the robustness evaluation of the oil blending scheme, ensuring that the oil blending and spray gun parameters can be dynamically adjusted according to the actual situation during the spraying process to achieve closed-loop precise oil blending.

[0086] In step 6, the real-time feedback control system adopts a fuzzy PID controller. The inputs are the deviation e between the real-time spectral monitoring data and the target value and the deviation change rate The real-time spectral monitoring data is obtained in real time through a spectral analyzer, compared with the target value to obtain the deviation e, and the deviation change rate is obtained by performing a difference calculation on the deviation

[0087] The outputs are the pulse frequency u1 of the oil delivery pump and the spray gun pressure adjustment amount u2.

[0088] The membership functions of the fuzzy PID controller adopt Gaussian functions. For the input variables e and and the output variables u1 and u2, the corresponding Gaussian membership functions are defined respectively. Taking the input variable e as an example, its membership function is:

[0089]

[0090] where A i is the fuzzy subset, c i is the center of the membership function, and σ i is the width of the membership function.

[0091] The defuzzification method is the centroid method, which converts the fuzzy output obtained from fuzzy inference into an accurate control quantity through this method.

[0092] The control rule base is generated by jointly training expert experience and historical data. Experts formulate initial control rules based on long-term practical experience, and then use historical data to optimize and adjust the rules.

[0093] In the optimization process of step 5, the robustness evaluation of the oil distribution scheme is introduced. The influence of raw material concentration fluctuation on spectral matching degree is analyzed through Monte Carlo simulation. Assuming that the raw material concentration follows a certain probability distribution (such as normal distribution), in each simulation, a random fluctuation value of the raw material concentration is generated, and the corresponding spectral matching degree is calculated according to the oil distribution scheme. After a large number of simulation experiments (such as N times), the distribution of the spectral matching degree is statistically analyzed. The oil distribution scheme with less change in spectral matching degree under the fluctuation of raw material concentration is selected as the optimal solution insensitive to noise. In this way, during the actual spraying process, even if there is a certain fluctuation in the raw material concentration, the spectral characteristics of the coating can be ensured to be close to the target value, realizing stable and accurate matte spraying.

[0094] It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not expressly listed, or further includes elements inherent to such process, method, article or device.

[0095] Although the embodiments of the present invention have been shown and described, those of ordinary skill in the art can understand that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A precise oil dispensing method for matte spraying, characterized in that: The following steps are involved: Step 1: Collect the reflectance spectrum data and chromaticity parameters of the target matte coating through a spectrum analyzer and a colorimeter, and obtain the spraying environment parameters in combination with an ambient temperature and humidity sensor; Step 2: Based on the spectral characteristics of the target coating, the principal component spectrum is extracted using the high-dimensional data dimensionality reduction method to construct the reflectance spectrum feature vector of the matte coating; Step 3: Establish a mapping relationship between the spraying experimental data and the historical formula database, and use the multi-layer perceptron model to predict the initial oil mixing ratio, including the mixing ratio of resin, curing agent, matte powder and solvent; Step 4: Based on the dynamic rheometer to measure the rheological properties of the oil distribution system, combined with the atomization pressure and spray gun distance parameters of the spray equipment, a spray process-rheological property coupling model is constructed; Step 5: The improved particle swarm optimization algorithm is used to jointly optimize the oil distribution ratio and spraying parameters, and the optimal oil distribution scheme is generated with spectral matching and rheological stability as constraints; Step 6: Dynamically adjust the oil delivery rate and spray gun parameters during the spraying process through a real-time feedback control system to achieve closed-loop precise oil distribution.

2. The precise oil dispensing method for matte spraying according to claim 1, characterized in that: The high-dimensional data dimensionality reduction method described in step 2 adopts the kernel principal component analysis algorithm, maps the reflectance spectrum data to the high-dimensional feature space, performs nonlinear dimensionality reduction, extracts the first k principal components to form the feature vector, and uses the Gaussian kernel function to calculate the similarity between samples.

3. The precise oil dispensing method for matte spraying according to claim 1, characterized in that: The multilayer perceptron model described in step 3 adopts a double hidden layer structure, the input layer includes spectral feature vectors, environmental parameters and substrate types, and the output layer is the proportion of each component; the Bayesian regularization method is used to prevent overfitting, and the loss function is the weighted sum of the mean square error and the L2 regularization term.

4. The precise oil dispensing method for matte spraying according to claim 1, characterized in that: The coupling model described in step 4 establishes the hidden variable relationship between spraying pressure, spray gun distance and rheological characteristic parameters through the partial least squares regression algorithm, and introduces interaction terms to describe nonlinear effects. The model parameters are cross-validated to determine the optimal number of hidden variables.

5. The precise oil dispensing method for matte spraying according to claim 1, characterized in that: The improved particle swarm optimization algorithm described in step 5 adopts an adaptive inertia weight strategy, dynamically adjusts the global and local search capabilities according to the number of iterations, and introduces the Cauchy mutation operator to enhance population diversity and avoid falling into the local optimal solution.

6. The precise oil dispensing method for matte spraying according to claim 5, characterized in that: The scale parameter of the Cauchy mutation operator is adaptively adjusted according to the current population fitness variance, the mutation probability is negatively correlated with the number of iterations, and the mutation amplitude obeys the Cauchy distribution.

7. The precise oil dispensing method for matte spraying according to claim 1, characterized in that: The real-time feedback control system described in step 6 adopts a fuzzy PID controller, the input of which is the deviation between the real-time spectral monitoring data and the target value, and the output is the pulse frequency of the oil distribution pump and the spray gun pressure adjustment amount. The control rule base is generated through joint training of expert experience and historical data.

8. The precise oil dispensing method for matte spraying according to claim 7, characterized in that: The membership function of the fuzzy PID controller adopts a Gaussian function, the defuzzification method is the centroid method, and the rule weight is optimized online through a gradient descent algorithm.

9. The precise oil dispensing method for matte spraying according to claim 1, characterized in that: In the optimization process of step 5, the robustness evaluation of the oil distribution scheme is introduced. The influence of raw material concentration fluctuation on the spectral matching is analyzed through Monte Carlo simulation, and the optimal solution that is insensitive to noise is screened out.

10. A precise oil distribution system for matte spraying, characterized in that: The system comprises: Data acquisition module, used to obtain spectral data, environmental parameters and rheological properties; The oil distribution prediction module generates the initial oil distribution plan based on the multi-layer perceptron model; Coupling modeling module to establish the spraying process-rheological characteristics coupling model; Optimization module, using improved particle swarm algorithm to optimize oil distribution parameters; Real-time control module, dynamically adjusts spray equipment parameters through fuzzy PID controller; The module is connected with the oil distribution pump, spray gun and sensor through the industrial bus to realize full-process automatic control.

Images