Spraying parameter determination method and device, and electronic equipment

Abstract

This application discloses a method, apparatus, and electronic device for determining spraying parameters. The method includes: determining the target color system and target color value for spraying a target vehicle; analyzing the target color system and target color value using a spraying color model to obtain spraying parameters corresponding to the target vehicle. The spraying color model includes a first sub-model for modeling the nonlinear relationship between the target color system and target color value and the corresponding spraying parameters; a second sub-model for enhancing the robustness against noise and outliers; a third sub-model for mining data patterns between the target color value and the corresponding spraying parameters; and a fourth sub-model for extracting the local similarity between the target color system and target color value. This application solves the technical problem that traditional spraying parameter design often relies on repeated manual adjustments of multiple spraying parameters, resulting in long debugging cycles and low spraying efficiency.

Description

Technical Field

[0001] This application relates to the field of automotive painting processes, and more specifically, to a method, apparatus, and electronic device for determining spraying parameters. Background Technology

[0002] As the automotive industry accelerates its transformation towards digital and intelligent manufacturing, the painting process, as a core link that determines the overall appearance quality and production efficiency of a vehicle, has become a key path for manufacturing to reduce costs and increase efficiency through the accumulation and reuse of its process knowledge.

[0003] In the design of color spraying processes in traditional automotive painting, the initial setting of spraying parameters for color spraying of newly developed models or the introduction of new batches of paint of the same color relies heavily on the personal experience of senior engineers. Engineers need to comprehensively evaluate variables such as particle size distribution, mixing uniformity, and slight color differences between batches of paint raw materials, and manually intervene and adjust five key process parameters of the spraying equipment, including flow rate, rotation speed, voltage, internal flow rate of forming air, and external flow rate of forming air, in conjunction with the real-time operating conditions of the spraying equipment. This process lacks unified quantitative standards and systematic methods, and parameter matching is entirely limited by the operator's technical level and historical trial-and-error experience. Each round of color debugging usually requires several or even more than ten rounds of repeated spraying verification, resulting in a lengthy overall debugging cycle. This not only easily leads to high costs for test paint, energy waste, and equipment downtime losses, but also seriously restricts the flexible production rhythm of modern automobile manufacturing in the painting process.

[0004] There is currently no effective solution to the above problems. Summary of the Invention

[0005] This application provides a method, apparatus, and electronic device for determining spraying parameters, which at least solves the technical problem that traditional spraying parameter design often relies on repeated manual trials to adjust multiple spraying parameters, resulting in long debugging cycles and low spraying efficiency.

[0006] According to one aspect of the embodiments of this application, a method for determining spraying parameters is provided, comprising: determining a target color system and a target color value for spraying a target vehicle, wherein the target color system is used to represent the category to which the target color belongs, the target color value is used to represent the three-dimensional coordinate value of the target color in the color space, and the target color is the desired spraying color of the target vehicle; and analyzing the target color system and the target color value using a spraying color model to obtain spraying parameters corresponding to the target vehicle, wherein the spraying color model includes a first sub-model for modeling the nonlinear relationship between the target color system and the target color value and the corresponding spraying parameters, a second sub-model for enhancing the robustness of noise and outliers, a third sub-model for mining the data patterns between the target color value and the corresponding spraying parameters, and a fourth sub-model for extracting the local similarity between the target color system and the target color value.

[0007] Optionally, the spray color model is trained in the following way: acquiring historical spray data; independently training the first sub-model, the second sub-model, the third sub-model, and the fourth sub-model based on the historical spray data to obtain the first prediction result, the second prediction result, the third prediction result, and the fourth prediction result; determining the reconstructed dataset based on the first prediction result, the second prediction result, the third prediction result, the fourth prediction result, and the historical spray data; and training the multilayer perceptron model based on the reconstructed dataset to obtain the spray color model.

[0008] Optionally, acquiring historical spraying data includes: acquiring historical spraying parameters from the database, and determining the historical spraying pigment information, historical target color value, and historical color measurement value corresponding to the historical spraying parameters; using the historical spraying pigment information, historical target color value, and historical color measurement value as feature fields, and using the historical spraying parameters as data labels to obtain structured historical spraying data; normalizing the historical spraying data, and dividing the normalized historical spraying data into a training set and a test set according to a preset ratio.

[0009] Optionally, the method further includes: a first sub-model employing supervised regression training to minimize the mean squared error between the first prediction result and historical spraying parameters, and optimizing the first model parameters of the first sub-model using a Bayesian optimization algorithm, wherein the first model parameters include at least one of the following: learning rate, maximum depth, number of weak learners, and regularization coefficient; and a second sub-model employing parallel ensemble training to minimize the mean squared error between the second prediction result and historical spraying parameters, and optimizing the second model parameters of the second sub-model using a Bayesian optimization algorithm, wherein the second model parameters include at least one of the following: decision tree depth, decision tree height, ... The third sub-model is trained using gradient boosting, aiming to optimize the mean square error between the third prediction result and the historical spraying parameters round by round. The parameters of the third sub-model are optimized using a Bayesian optimization algorithm. The parameters of the third sub-model include at least one of the following: learning rate, maximum depth, number of weak learners, and regularization coefficient. The fourth sub-model is trained using a nearest neighbor search strategy, aiming to minimize the mean square error between the fourth prediction result and the historical spraying parameters. The optimal number of nearest neighbor samples corresponding to the fourth sub-model is determined by iteratively searching within a preset K value range using a Bayesian optimization algorithm.

[0010] Optionally, the multilayer perceptron model is trained based on the reconstructed dataset to obtain a spraying color model, including: using the first, second, third, and fourth prediction results as derived features, fusing the derived features and historical spraying parameters to obtain a reconstructed dataset; training the multilayer perceptron model using a preset optimizer based on the reconstructed dataset, and smoothly decaying the initial learning rate of the multilayer perceptron model to a preset value according to a cosine function within a preset training period; obtaining the target prediction result output by the multilayer perceptron model, determining the mean square error between the target prediction result and the historical spraying parameters, and adjusting the model weights corresponding to the first, second, third, and fourth sub-models respectively based on the mean square error; and stopping training when the mean square error does not decrease for a consecutive preset number of rounds to obtain the spraying color model.

[0011] Optionally, the method further includes: determining the color measurement value of the target vehicle after spraying according to the spraying parameters; comparing the color measurement value and the target color value to obtain a comparison result; if the comparison result indicates that the difference between the color measurement value and the target color value is less than a preset difference value, determining the spraying parameters as the target spraying parameters of the target vehicle; if the comparison result indicates that the difference between the color measurement value and the target color value is greater than or equal to the preset difference value, using a spraying color model to analyze the target color system, the target color value, and the color measurement value to obtain the target spraying parameters of the target vehicle.

[0012] Optionally, the method further includes: obtaining hard constraints for painting the target vehicle, wherein the hard constraints are one or more combinations of painting parameters set based on actual needs; and based on the hard constraints, using a painting color model to analyze the target color system and target color value to obtain painting parameters corresponding to the target vehicle, wherein the painting parameters include at least the paint flow rate, atomization speed, painting voltage, internal flow rate of forming air, and external flow rate of forming air of the painting equipment.

[0013] According to another aspect of the embodiments of this application, a spraying parameter determination device is also provided, comprising: a determination module, configured to determine a target color system and a target color value for spraying a target vehicle, wherein the target color system represents the category to which the target color belongs, the target color value represents the three-dimensional coordinate value of the target color in the color space, and the target color is the desired spraying color of the target vehicle; and an analysis module, configured to analyze the target color system and the target color value using a spraying color model to obtain spraying parameters corresponding to the target vehicle, wherein the spraying color model includes a first sub-model for modeling the nonlinear relationship between the target color system and the target color value and the corresponding spraying parameters, a second sub-model for enhancing the robustness of noise and outliers, a third sub-model for mining the data patterns between the target color value and the corresponding spraying parameters, and a fourth sub-model for extracting the local similarity between the target color system and the target color value.

[0014] According to another aspect of the embodiments of this application, an electronic device is also provided, including: a memory and a processor, wherein the memory is used to store program instructions; and the processor is connected to the memory and used to execute the above-described method for determining spraying parameters.

[0015] According to another aspect of the embodiments of this application, a non-volatile storage medium is also provided, the non-volatile storage medium including a stored computer program, wherein the device containing the non-volatile storage medium executes the above-described spraying parameter determination method by running the computer program.

[0016] According to another aspect of the embodiments of this application, a computer program product is also provided, including computer instructions, which, when executed by a processor, implement the above-described method for determining spraying parameters.

[0017] In this embodiment, by determining the target color system and target color value for spraying the target vehicle, where the target color system represents the category of the target color and the target color value represents the three-dimensional coordinate value of the target color in the color space, and the target color is the desired spraying color of the target vehicle, a spraying color model is used to analyze the target color system and target color value to obtain the spraying parameters corresponding to the target vehicle. The spraying color model includes a first sub-model for modeling the nonlinear relationship between the target color system and target color value and the corresponding spraying parameters, a second sub-model for enhancing the robustness of noise and outliers, a third sub-model for mining the data patterns between the target color value and the corresponding spraying parameters, and a fourth sub-model for extracting the local similarity between the target color system and target color value. This achieves the goal of reverse intelligent derivation of multi-dimensional spraying equipment parameters using the target color system and the three-dimensional coordinates of the color space as input, thereby significantly reducing the debugging cycle and trial and error costs and achieving the technical effect of highly flexible automated and accurate color matching. This solves the technical problem that traditional spraying parameter design often relies on repeated manual trials to adjust multiple spraying parameters, resulting in long debugging cycles and low spraying efficiency. Attached Figure Description

[0018] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0019] Figure 1 This is a hardware structure diagram of a computer terminal for implementing a spraying parameter determination method according to an embodiment of this application;

[0020] Figure 2 This is a flowchart of a method for determining spraying parameters according to an embodiment of this application;

[0021] Figure 3 This is a flowchart of another method for determining spraying parameters according to an embodiment of this application;

[0022] Figure 4 This is a structural diagram of a spray color model according to an embodiment of this application.

[0023] Figure 5 This is a structural diagram of a spraying parameter determination system according to an embodiment of this application.

[0024] Figure 6 This is a structural diagram of a spraying parameter determining device according to an embodiment of this application. Detailed Implementation

[0025] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0026] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0027] First, some nouns or terms that appear in the explanation of the embodiments of this application shall be interpreted as follows:

[0028] CIE L a b Color space (International Commission on Illumination L) a b Color Space (LAB Color Space, International Commission on Illumination): A color measurement system for perceiving uniformity. Indicates brightness (0 for black, 100 for white), a Indicates the positive and negative values ​​of the red and green axes, b This indicates the positive and negative values ​​of the yellow and blue axes.

[0029] XGBoost (eXtreme Gradient Boosting): A high-efficiency machine learning algorithm based on gradient boosting decision trees. By introducing L1 / L2 regularization terms, sparse-aware splitting, and column sampling mechanisms, it significantly improves the prediction accuracy and training efficiency of regression tasks.

[0030] RF (Random Forest): An ensemble learning method based on the Bagging idea, which constructs multiple independent decision trees through bootstrapping and outputs the results by voting or averaging. This algorithm is highly robust to noise and outliers in the original data.

[0031] Gradient Boosting (GB) is a sequential reinforcement learning method that gradually reduces the overall error by training weak learners in each round to fit the prediction residuals of the previous model. In this scheme, it is responsible for uncovering deep implicit patterns in coating process data and improving generalization ability under complex working conditions.

[0032] KNN (K-Nearest Neighbors): An instance-based nonparametric machine learning algorithm that calculates the distance between the sample to be predicted and historical samples in a multidimensional feature space, and selects the K nearest samples for weighted regression.

[0033] MLP (Multi-Layer Perceptron): A fully connected feedforward neural network consisting of an input layer, one or more hidden layers, and an output layer. In this application, it serves as the meta-model in a two-stage stacked architecture, receiving independent predictions from four base models, learning the optimal fusion weights, and outputting recommended values ​​for the final spraying parameters.

[0034] Bayesian Optimization (BO) is a global hyperparameter optimization method based on a probabilistic surrogate model (usually a Gaussian process). It approaches the optimal parameter combination with the fewest iterations by balancing exploration and utilization through a sampling function.

[0035] This application provides a method for determining spraying parameters, which can be run in... Figure 1 The computer terminal shown is described below.

[0036] The spraying parameter determination method provided in this application can be executed on a mobile terminal, computer terminal, or similar computing device. Figure 1 A hardware block diagram of a computer terminal for implementing a method for determining spraying parameters is shown. Figure 1As shown, the computer terminal 10 may include one or more processors (shown as 102a, 102b, ..., 102n in the figure) (the processor may include, but is not limited to, a microprocessor MCU or a programmable logic device FPGA, etc.), a memory 104 for storing data, and a transmission module 106 for communication functions connected via wired and / or wireless networks. In addition, it may also include: a display, a keyboard, a cursor control device, an input / output interface (I / O interface), a universal serial bus (USB) port (which may be included as one of the ports of the I / O interface), a network interface, and a BUS bus. Those skilled in the art will understand that... Figure 1 The structure shown is for illustrative purposes only and does not limit the structure of the aforementioned electronic device. For example, computer terminal 10 may also include... Figure 1 The more or fewer components shown, or having the same Figure 1 The different configurations shown.

[0037] It should be noted that the aforementioned one or more processors and / or other data processing circuits are generally referred to herein as "data processing circuits". These data processing circuits may be embodied, in whole or in part, in software, hardware, firmware, or any other combination thereof. Furthermore, the data processing circuits may be a single, independent processing module, or may be integrated, in whole or in part, into any other element within the computer terminal 10. As involved in the embodiments of this application, the data processing circuits serve as a processor control mechanism (e.g., selection of a variable resistor termination path connected to an interface).

[0038] The memory 104 can be used to store software programs and modules of application software, such as the program instructions / data storage device corresponding to the spraying parameter determination method in this embodiment. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory 104, thereby realizing the above-mentioned spraying parameter determination method. The memory 104 may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include memory remotely located relative to the processor, and these remote memories can be connected to the computer terminal 10 via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0039] The transmission module 106 is used to receive or send data via a network. Specific examples of the network described above may include a wireless network provided by the communication provider of the computer terminal 10. In one example, the transmission module 106 includes a network interface controller (NIC), which can connect to other network devices via a base station to communicate with the Internet. In another example, the transmission module 106 may be a radio frequency (RF) module, used for wireless communication with the Internet.

[0040] The display can be, for example, a touchscreen liquid crystal display (LCD) that allows the user to interact with the user interface of the computer terminal 10.

[0041] It should be noted here that, in some optional embodiments, the above... Figure 1 The computer terminal shown may include hardware elements (including circuitry), software elements (including computer code stored on a computer-readable medium), or a combination of both hardware and software elements. It should be noted that... Figure 1 This is only one instance of a specific particular instance, and is intended to illustrate the types of components that may exist in the aforementioned computer terminal.

[0042] In the above operating environment, this application provides an embodiment of a method for determining spraying parameters. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0043] Figure 2 This is a flowchart of a method for determining spraying parameters according to an embodiment of this application, such as... Figure 2 As shown, the method includes the following steps:

[0044] Step S202: Determine the target color system and target color value for the target vehicle to be painted. The target color system is used to indicate the category to which the target color belongs, and the target color value is used to indicate the three-dimensional coordinate value of the target color in the color space. The target color is the desired paint color of the target vehicle.

[0045] Step S204: The spraying color model is used to analyze the target color system and target color value to obtain the spraying parameters corresponding to the target vehicle. The spraying color model includes a first sub-model for modeling the nonlinear relationship between the target color system and target color value and the corresponding spraying parameters, a second sub-model for enhancing the robustness of noise and outliers, a third sub-model for mining the data patterns between the target color value and the corresponding spraying parameters, and a fourth sub-model for extracting the local similarity between the target color system and the target color value.

[0046] Through steps S202 to S204 above, the goal of intelligently deriving multi-dimensional spraying equipment parameters using the target color system and three-dimensional coordinates of the color space as input is achieved. This significantly reduces the debugging cycle and trial-and-error costs, and achieves a highly flexible, automated, and precise color matching effect. It also solves the technical problem of long debugging cycles and low spraying efficiency caused by traditional spraying parameter design often relying on repeated manual adjustments of multiple spraying parameters. The following is a detailed explanation.

[0047] Figure 3 This is a flowchart of another method for determining spraying parameters according to an embodiment of this application, including a pre-training process for the spraying color model (steps S302 to S306) and an actual recommendation process for the spraying parameters (step S308, including the above steps S202 to S204). The specific process analysis is as follows:

[0048] Step S302: Acquisition and preprocessing of spray color data.

[0049] In this embodiment of the application, obtaining historical spraying data includes: obtaining historical spraying parameters from the database, and determining historical spraying pigment information, historical target color values, and historical color measurement values ​​corresponding to the historical spraying parameters; using the historical spraying pigment information, historical target color values, and historical color measurement values ​​as feature fields, and using the historical spraying parameters as data labels to obtain structured historical spraying data; normalizing the historical spraying data, and dividing the normalized historical spraying data into training set and test set according to a preset ratio.

[0050] Specifically, based on existing accumulated painting process data and professional design requirements, historical painting data for automotive colors can be obtained from the original database. This includes historical painting parameters (flow rate, rotation speed, voltage, internal and external flow rates of forming air) in the LAB color space, historical painting pigment information, historical target color values ​​(three-dimensional coordinate representation in the LAB color space), and historical color measurements after painting. The historical painting data is processed into structured data, with feature fields including historical painting pigment information, historical target color values, and historical color measurements after painting. The data label is the historical painting parameters, resulting in a basic data sample. Subsequently, the Min-Max linear scaling method is used to normalize all sample data, mapping the original values ​​of different dimensions to the [0,1] interval to eliminate gradient-dominated bias. Finally, the normalized dataset is randomly divided into a training set and an independent test set according to a preset ratio (e.g., 80% and 20%), forming a color painting process database adapted for regression modeling.

[0051] In this step, the scale differences and numerical dominance bias of the original process data are effectively eliminated through structured feature mapping, cross-dimensional normalization and independent dataset partitioning. This provides a standardized, high-quality training base for the spraying color model, which can significantly improve the model's convergence efficiency, generalization ability and the prediction accuracy of the final spraying parameters.

[0052] Step S304: Selecting the spray color model (overall training logic).

[0053] Optionally, the spray color model is trained in the following way: the first sub-model, the second sub-model, the third sub-model, and the fourth sub-model are trained independently based on historical spray data to obtain the first prediction result, the second prediction result, the third prediction result, and the fourth prediction result; the reconstructed dataset is determined based on the first prediction result, the second prediction result, the third prediction result, the fourth prediction result, and the historical spray data; the multilayer perceptron model is trained based on the reconstructed dataset to obtain the spray color model.

[0054] In this embodiment, the design of the spray color model needs to meet the basic requirements of the regression strategy. Therefore, multiple models that perform well in regression scenarios and have strong fitting ability for nonlinear relationships can be selected for preliminary experiments. Based on the validation results, four models—XGBoost, Random Forest (RF), Gradient Boosting (GB), and K-Nearest Neighbors—are selected as sub-models for ensemble learning. The prediction results output by each sub-model are used as new input features and combined with historical spray data (data labels in the training set, i.e., historical spray parameters) to form a reconstructed dataset. A multilayer perceptron (MLP) model is then trained to learn how to optimally combine the outputs of each sub-model, resulting in the final spray color model used for spray parameter recommendation. The overall framework is as follows: Figure 4As shown in the figure, XGBoost (the first sub-model) excels at handling complex nonlinear relationships and has strong resistance to overfitting; RF (the second sub-model) has good robustness to noise and outliers; GB (the third sub-model) is suitable for discovering deep patterns in data; and K-Nearest Neighbors (the fourth sub-model) can predict based on local similarity and effectively capture local patterns. The four models complement each other, effectively improving the overall generalization and stability of the model.

[0055] It should be noted that the four sub-models mentioned above exist only as base models at the first level and do not directly participate in the final process parameter output. Their output results only serve as feature inputs to the meta-model and do not have independent decision-making power. The MLP meta-model, on the other hand, is located at the second level and has absolute control over the secondary abstraction, cross-validation, and weight redistribution of the base model outputs. Together, they constitute a master-slave collaborative relationship for feature extraction and intelligent fusion, ensuring that the spraying color model has adaptive optimization capabilities when faced with unknown color formulas or new batches of raw materials, rather than relying on fixed rules hard-coded.

[0056] In this embodiment, the four sub-models can be constructed based on a two-level stacked ensemble learning architecture. Their combination strictly follows the progressive logic of "base model layer independently generates derived features → meta-model layer adaptively learns and fuses weights". The specific process analysis is as follows:

[0057] 1. Data Source Consistency and Training Decoupling: The four sub-models share the same normalized historical spraying dataset as their input source, but are completely physically isolated and do not communicate with each other during the model training phase. Each sub-model independently performs forward propagation and hyperparameter optimization based on its own algorithmic characteristics, ensuring that the learning process is not disturbed by the gradients or parameter updates of other models, and maintaining the independent inductive bias of each base learner.

[0058] 2. Parallel Output and Feature Reconstruction: After training, the four sub-models take the same historical spraying data as input and output predicted vectors of five-dimensional spraying parameters in parallel. These four sets of prediction results are not used as the final process instructions, but are extracted into high-dimensional derived feature columns, which are then horizontally concatenated with the original input features (historical spraying parameters) in the feature dimension to form a comprehensive representation space that includes global nonlinear mapping, noise robust response, deep pattern mining, and local similarity matching, i.e., reconstructing the dataset.

[0059] 3. Meta-model supervised fusion: The MLP, as a meta-model, receives the reconstructed dataset and automatically learns the nonlinear interaction relationship and dynamic combination weights between the outputs of the four sub-models through end-to-end supervised training. Finally, it outputs the optimal spraying parameter recommendation value, realizing adaptive distillation and integrated gain of cross-algorithm knowledge.

[0060] In this step, the parallel prediction of heterogeneous base models and the adaptive feature fusion of meta-models effectively avoid the overfitting and generalization bottlenecks of single algorithms when dealing with strong nonlinear coupling relationships, and significantly improve the accuracy, stability and adaptability of five-dimensional spraying parameter reverse recommendation across batches of raw materials.

[0061] Step S306: Spraying color model training and generation.

[0062] In this embodiment, the four sub-models are trained as follows: The first sub-model is trained using supervised regression to minimize the mean squared error between the first prediction result and the historical spraying parameters, and the first model parameters of the first sub-model are optimized using a Bayesian optimization algorithm, wherein the first model parameters include at least one of the following: learning rate, maximum depth, number of weak learners, and regularization coefficient; The second sub-model is trained using parallel ensemble training to minimize the mean squared error between the second prediction result and the historical spraying parameters, and the second model parameters of the second sub-model are optimized using a Bayesian optimization algorithm, wherein the second model parameters include at least one of the following: The decision tree depth, number of decision trees, and maximum feature sampling number per node are considered. The third sub-model is trained using gradient boosting, aiming to optimize the mean squared error between the third prediction result and historical spraying parameters round by round. The parameters of the third sub-model are optimized using a Bayesian optimization algorithm. The parameters of the third sub-model include at least one of the following: learning rate, maximum depth, number of weak learners, and regularization coefficient. The fourth sub-model is trained using a nearest neighbor search strategy, aiming to minimize the mean squared error between the fourth prediction result and historical spraying parameters. The optimal number of nearest neighbor samples corresponding to the fourth sub-model is determined by iteratively searching within a preset K value range using a Bayesian optimization algorithm.

[0063] Specifically, the four sub-models are trained independently in parallel, using the same historical spraying data as input. Each model learns independently and is then fused at the output stage.

[0064] The first sub-model is built based on the XGBoost algorithm and adopts a supervised regression training framework, with the core optimization objective being to minimize the mean square error between the predicted results and historical spraying parameters. This model accelerates convergence by introducing Hessian second derivative information and incorporates L1 and L2 regularization terms to effectively suppress overfitting. During the hyperparameter optimization phase, the system calls a Bayesian optimization algorithm to jointly iteratively search for the learning rate, maximum tree depth, number of weak learners, and regularization coefficients (setting the maximum number of training iterations to 1000 rounds and the Bayesian optimization iterations to 50 rounds), giving it excellent capabilities in handling strongly nonlinear mappings and high-dimensional feature interactions, making it the mainstay for fitting globally complex surfaces.

[0065] The second sub-model is built upon the random forest algorithm and employs a parallel ensemble training strategy. Its core mechanism lies in significantly reducing model variance through bootstrapping and random subspace partitioning of node features. During training, the model is also guided by minimizing the mean squared error, utilizing Bayesian optimization to precisely determine key configurations such as decision tree depth, total number of trees, and maximum feature sampling count per node (maximum 500 iterations, 30 Bayesian optimizations). This model exhibits strong robustness to batch fluctuations in paint raw materials, sensor measurement noise, and outliers, primarily serving to smooth data and resist interference.

[0066] The third sub-model employs a gradient boosting algorithm architecture, following a sequential residual fitting logic. The model continuously corrects accumulated biases by calculating the negative gradient direction between the current predicted value and the actual label in each iteration, thereby gradually approximating the deep-seated process evolution patterns under multivariate coupling conditions. Its hyperparameter optimization process is consistent with the first sub-model, focusing on tuning the learning rate, maximum depth, number of weak learners, and regularization coefficients, with a maximum training iteration count set at 800 rounds. This design enhances the model's depth of understanding of implicit nonlinear relationships, compensating for the fitting blind spots of a single boosting algorithm under complex working conditions.

[0067] The fourth sub-model is implemented based on a K-nearest neighbor instance-based nonparametric learning method. It does not rely on explicit loss function iterations, but instead captures local similarity patterns using distance metrics in a multi-dimensional feature space. During the training phase, the goal is also to minimize the mean squared error. The Bayesian optimization algorithm iterates through the preset discrete range of K values ​​(40 iterations), automatically determining the optimal number of neighbor samples K based on the validation set accuracy. This model focuses on responding to micro-parameter transitions under specific paint batches or equipment fine-tuning conditions, providing high-precision local region matching support.

[0068] In the above process, the training mechanism of independent decoupling of the four sub-models ensures that the errors of each model do not spread to each other. The unified mean square error target and dimensional standardization output lay the mathematical foundation for the adaptive weight allocation of the subsequent meta-model. Through knowledge distillation and error decoupling of heterogeneous algorithms, this architecture effectively breaks through the generalization bottleneck of a single model when dealing with strong nonlinear coupling relationships in coating, and significantly improves the accuracy, stability and cross-batch raw material adaptability of the five-dimensional spraying parameter reverse recommendation.

[0069] In this embodiment, a multilayer perceptron model is trained based on a reconstructed dataset to obtain a spraying color model. This includes: using the first, second, third, and fourth prediction results as derived features, fusing the derived features with historical spraying parameters to obtain a reconstructed dataset; training the multilayer perceptron model using a preset optimizer based on the reconstructed dataset, and smoothly decaying the initial learning rate of the multilayer perceptron model to a preset value using a cosine function within a preset training period; obtaining the target prediction result output by the multilayer perceptron model, determining the mean square error between the target prediction result and the historical spraying parameters, and adjusting the model weights corresponding to the first, second, third, and fourth sub-models based on the mean square error; and stopping training when the mean square error does not decrease for a consecutive preset number of rounds, thus obtaining the spraying color model. The specific process analysis is as follows:

[0070] Training begins with the structured reorganization of multidimensional information. First, the first to fourth prediction results output by the four independent sub-models are extracted into high-dimensional derived feature columns, and then tensor-concatenated with the data label field of the original historical spraying parameters to construct a reconstructed dataset with expanded dimensions. This breaks down the feature silos of traditional single models, enabling the meta-model to directly observe the prediction distribution, residual structure, and complementary boundaries of each heterogeneous algorithm under strongly nonlinear mapping, providing a complete information foundation for subsequent intelligent decision-making.

[0071] In the parameter update phase, the MLP model employs the Adam adaptive momentum optimizer to drive forward and backward propagation. Addressing the inherent multi-scale fluctuations and gradient sparsity of coating process data, the training process strictly adheres to a preset period and incorporates a built-in cosine annealing learning rate decay mechanism. Specifically, the network starts with an initial learning rate of 0.001, which smoothly decays to a preset lower limit along a cosine function curve as training epochs progress. This avoids gradient oscillations or getting trapped in local minima caused by a fixed learning rate, ensuring rapid convergence of network weights in the early stages of training and fine-tuning in later stages, significantly improving the stability of parameter optimization.

[0072] During model training, mean squared error (MSE) is used as the core supervision signal. The gradient backpropagation path is calculated by comparing the target prediction results output by the multilayer perceptron (MLP) with historical actual spraying parameters in real time. In this process, the MLP does not perform static arithmetic averaging or hard-coded weighting. Instead, it autonomously discovers the cross-terms and interaction patterns between the four prediction results through the nonlinear activation functions of the hidden layers and the weight matrix of the fully connected layers. Based on the gradient feedback of the MSE, the system dynamically adjusts the connection weights of internal neurons to achieve adaptive calibration of the output contributions of the first to fourth sub-models.

[0073] Finally, to prevent memory-based overfitting of the meta-model on the reconstructed dataset, an early stopping regularization strategy is forcibly embedded in the training process. The system continuously monitors the mean squared error trend on the validation set and sets a threshold for determining whether the loss value has not decreased for 20 consecutive training epochs. Once this condition is triggered, the training engine immediately interrupts backpropagation, automatically rolls back, and loads the model snapshot with the lowest loss on the validation set as the final state. This eliminates unnecessary computational power consumption, ensures the reliability of the model's generalization boundary, and guarantees that it can still output robust process parameter recommendations when faced with unknown car model color schemes or new batches of painting materials.

[0074] In the above process, by using derived feature reconstruction and fusion, cosine annealing learning rate smooth decay, adaptive weight allocation based on mean square error, and early stopping mechanism, the risks of gradient oscillation and model overfitting in the later stage of training are effectively avoided. Flexible weighted fusion of heterogeneous base model output is realized, which significantly improves the convergence stability, generalization accuracy and cross-batch raw material adaptability of spraying parameter recommendation.

[0075] Step S308: Recommendation and compensation of spraying parameters.

[0076] Corresponding to steps S202 to S204 above, in practical application, the process includes: determining the target color system and target color value for spraying the target vehicle, and using a spraying color model to analyze the target color system and target color value to obtain the spraying parameters corresponding to the target vehicle.

[0077] Furthermore, the process of spraying according to the spraying parameters also includes: determining the color measurement value of the target vehicle after spraying according to the spraying parameters; comparing the color measurement value and the target color value to obtain a comparison result; if the comparison result indicates that the difference between the color measurement value and the target color value is less than a preset difference value, determining the spraying parameters as the target spraying parameters of the target vehicle; if the comparison result indicates that the difference between the color measurement value and the target color value is greater than or equal to the preset difference value, using a spraying color model to analyze the target color system, target color value, and color measurement value to obtain the target spraying parameters of the target vehicle.

[0078] Specifically, in practice, the target vehicle series and target color (such as pearl white) LAB value can be selected through the interactive terminal. The deployed painting color model is then invoked for forward inference, outputting five-dimensional initial recommended parameters including flow rate, rotation speed, voltage, and internal / external flow rate of the forming air. Subsequently, the physical execution layer is entered. After the actual painting and drying curing of the vehicle body is completed, a high-precision colorimeter is used to collect color measurement values ​​(measured LAB values) in a designated area of ​​the vehicle body, forming a closed-loop verification node.

[0079] When the color difference between the measured LAB value and the target LAB value is lower than the preset difference value, the process window is considered to have converged, and the current parameters are directly locked as the target painting parameters for the target vehicle, and the process terminates. If the color difference exceeds the preset difference value, a secondary inference mechanism is automatically triggered: the original target color system, the target LAB value, and the measured LAB value are all used as composite input features and re-injected into the painting color model. Based on the "target-measured deviation mapping law" learned from historical data, the model automatically compensates for systematic deviations caused by differences in particle distribution of raw material batches, fluctuations in environmental temperature and humidity, or minor equipment drift, and generates a corrected combination of painting parameters. This process supports multiple rounds of execution until the color difference meets the process requirements, achieving rapid convergence of prediction, verification, and compensation.

[0080] In the above process, by constructing a closed-loop iterative mechanism of prediction, measurement, and feedback compensation, the five-dimensional equipment parameters can be dynamically corrected according to the actual color difference. This effectively avoids reliance on human experience and multiple rounds of physical trial and error. While ensuring high accuracy of color matching, it significantly reduces the process debugging cycle and improves the system's adaptive stability to raw material batch fluctuations.

[0081] In this embodiment, the user can also fix 1 to 4 parameters in the spraying parameters based on experience as hard constraints, requiring the spraying color model to only provide recommendations for other variable parameters. The specific implementation is as follows: Obtain the hard constraints for spraying the target vehicle, where the hard constraints are one or more combinations of spraying parameters set based on actual needs; based on the hard constraints, analyze the target color system and target color values ​​using the spraying color model to obtain the spraying parameters corresponding to the target vehicle, where the spraying parameters include at least the paint flow rate, atomization speed, spraying voltage, internal flow rate of forming air, and external flow rate of forming air of the spraying equipment.

[0082] By explicitly injecting actual process boundaries and equipment limits into the model inference chain, intelligent dimensionality reduction of the search space and mandatory constraints on the physical feasible region are achieved. While ensuring the accuracy of color matching, the risk of parameter out-of-bounds errors is eliminated, significantly improving the feasibility, operational safety, and cross-condition generalization adaptability of the model recommendation results.

[0083] Step S310: Optimization and iteration of the spraying color model.

[0084] During the process of using the spray color model to predict spray parameters, users will continuously generate data results corresponding to the spray parameters and the actual spray color, thereby increasing the number of samples required for training the spray color model. After completing a certain amount of time and number of spray color predictions, the newly generated spray color prediction samples will be automatically added to the spray color model training sample library for iterative training, continuously improving the accuracy of the spray color model prediction.

[0085] Figure 5 This is a structural diagram of a spraying parameter determination system according to an embodiment of this application, such as... Figure 5 As shown, the system includes:

[0086] Spraying color data acquisition and preprocessing module 50: is used to perform the above step S302, process the accumulated key parameters of all car colors into structured data, obtain data samples, and normalize all data samples, scaling feature values ​​of different ranges to 0-1 space, thereby forming a color spraying process database.

[0087] Spraying color model selection module 52: is used to perform the above step S304, select multiple models that perform well in regression scenarios and have strong fitting ability to nonlinear relationships for preliminary experiments, select a suitable model algorithm based on the verification results, improve the model prediction accuracy through ensemble learning, and ensure the overall generalization and stability of the model.

[0088] Spraying color model training and generation module 54: This is used to perform the above step S306. With the training dataset as input, multiple selected sub-models are trained independently. The hyperparameters of the sub-models are tuned using the Bayesian optimization method to obtain the optimal combination weights of the model, ensuring that the overall prediction performance is maximized.

[0089] Spraying color parameter recommendation and compensation module 56: Used to perform the above step S308, using the selected color system and LAB value of the desired spraying color, start the prediction to obtain the spraying parameters, verify whether they meet the requirements, if not, input the previously measured color value again to form new spraying parameters, repeat the operation until the ideal spraying parameters are found.

[0090] Spraying color model iteration module 58: is used to execute the above step S310. During the spraying parameter prediction process, new spraying colors and spraying parameters are continuously accumulated. After a certain time and number of spraying color predictions, the newly generated spraying color prediction samples are automatically put into the training sample library to continuously improve the accuracy of spraying color model prediction.

[0091] Overall, this application breaks through the traditional linear trial-and-error mode of automotive paint color matching, which relies heavily on manual experience, requires more than ten rounds of physical prototyping, and has a cycle of several weeks. It pioneers a closed-loop intelligent decision-making architecture that combines heterogeneous base model stacking and fusion with actual state feedback compensation. At the technical level, by constructing a four-way complementary prediction network of XGBoost, random forest, gradient boosting, and K-nearest neighbors, and combining it with the nonlinear weighted fusion of Bayesian optimization and multilayer perceptron model, the strong interaction relationship of five key parameters—paint flow rate, atomization speed, spraying voltage, and internal and external forming air—is effectively decoupled. At the application level, the actual LAB measurement value after drying is innovatively reinjected into the model inference link as a dynamic state variable, enabling the system to automatically identify and compensate for systematic color difference deviations caused by fluctuations in raw material particle distribution, drift in environmental temperature and humidity, or slight changes in equipment. At the same time, it allows process engineers to flexibly lock 1 to 4 parameters for targeted optimization, balancing the accuracy of algorithm recommendation with the boundaries of production line engineering.

[0092] The overall design upgrades open-loop static mapping to closed-loop control with observed variables, completely reconstructing the coating process preparation paradigm. While ensuring high accuracy in color matching, it achieves significant industrial benefits such as reducing the debugging cycle to 10% of the original and achieving 100% automation.

[0093] According to an embodiment of this application, a spraying parameter determining device is provided. It should be noted that the spraying parameter determining device of this application embodiment can be used to execute the spraying parameter determining method provided in the embodiment of this application. The spraying parameter determining device provided in the embodiment of this application is described below.

[0094] Figure 6 This is a structural diagram of a spraying parameter determining device provided according to an embodiment of this application. Figure 6 As shown, the device includes:

[0095] The determination module 60 is used to determine the target color system and target color value for spraying the target vehicle. The target color system is used to indicate the category to which the target color belongs, and the target color value is used to indicate the three-dimensional coordinate value of the target color in the color space. The target color is the desired spraying color of the target vehicle.

[0096] Analysis module 62 is used to analyze the target color system and target color value using a spraying color model to obtain the spraying parameters corresponding to the target vehicle. The spraying color model includes a first sub-model for modeling the nonlinear relationship between the target color system and target color value and the corresponding spraying parameters, a second sub-model for enhancing the robustness of noise and outliers, a third sub-model for mining the data patterns between the target color value and the corresponding spraying parameters, and a fourth sub-model for extracting the local similarity between the target color system and the target color value.

[0097] Through the determination and analysis modules in the above-mentioned spraying parameter determination device, the goal of reverse intelligent derivation of multi-dimensional spraying equipment parameters is achieved by using the target color system and three-dimensional coordinates of the color space as input. This results in a significant reduction in debugging cycle and trial and error costs, and achieves the technical effect of highly flexible, automated, and precise color matching. In turn, it solves the technical problem that traditional spraying parameter design often relies on repeated manual trials and adjustments of multiple spraying parameters, leading to long debugging cycles and low spraying efficiency.

[0098] In the spraying parameter determination device provided in this application embodiment, the analysis module is further used to determine the color measurement value after spraying the target vehicle according to the spraying parameters; compare the color measurement value and the target color value to obtain a comparison result; if the comparison result indicates that the difference between the color measurement value and the target color value is less than a preset difference value, the spraying parameters are determined as the target spraying parameters of the target vehicle; if the comparison result indicates that the difference between the color measurement value and the target color value is greater than or equal to the preset difference value, the target color system, the target color value and the color measurement value are analyzed using a spraying color model to obtain the target spraying parameters of the target vehicle.

[0099] In the spraying parameter determination device provided in this application embodiment, the analysis module is further used to obtain the hard constraints for spraying the target vehicle, wherein the hard constraints are one or more combinations of spraying parameters set based on actual needs; based on the hard constraints, the target color system and target color value are analyzed using a spraying color model to obtain the spraying parameters corresponding to the target vehicle, wherein the spraying parameters include at least the paint flow rate, atomization speed, spraying voltage, internal flow rate of forming air, and external flow rate of forming air of the spraying equipment.

[0100] The spraying parameter determination device provided in this application embodiment further includes a training module 64, which is used to acquire historical spraying data; independently train the first sub-model, the second sub-model, the third sub-model, and the fourth sub-model based on the historical spraying data to obtain a first prediction result, a second prediction result, a third prediction result, and a fourth prediction result; determine a reconstructed dataset based on the first prediction result, the second prediction result, the third prediction result, the fourth prediction result, and the historical spraying data; and train the multilayer perceptron model based on the reconstructed dataset to obtain a spraying color model.

[0101] In the spraying parameter determination device provided in this application embodiment, the training module is further used to obtain historical spraying parameters in the database, and determine the historical spraying pigment information, historical target color value and historical color measurement value corresponding to the historical spraying parameters; use the historical spraying pigment information, historical target color value and historical color measurement value as feature fields, and use the historical spraying parameters as data labels to obtain structured historical spraying data; normalize the historical spraying data, and divide the normalized historical spraying data into training set and test set according to a preset ratio.

[0102] In the spraying parameter determination device provided in this application embodiment, the training module is further used to train sub-models, including: a first sub-model employing supervised regression training with the objective of minimizing the mean square error between the first prediction result and historical spraying parameters, and optimizing the first model parameters of the first sub-model using a Bayesian optimization algorithm, wherein the first model parameters include at least one of the following: learning rate, maximum depth, number of weak learners, and regularization coefficient; a second sub-model employing parallel ensemble training with the objective of minimizing the mean square error between the second prediction result and historical spraying parameters, and optimizing the second model parameters of the second sub-model using a Bayesian optimization algorithm, wherein the second model parameters include At least one of the following: decision tree depth, number of decision trees, maximum number of feature samples per node; the third sub-model is trained using gradient boosting, with the goal of optimizing the mean square error between the third prediction result and the historical spraying parameters round by round, and the third model parameters of the third sub-model are optimized using a Bayesian optimization algorithm, wherein the third model parameters include at least one of the following: learning rate, maximum depth, number of weak learners, and regularization coefficient; the fourth sub-model is trained using a nearest neighbor search strategy, with the goal of minimizing the mean square error between the fourth prediction result and the historical spraying parameters, and the optimal number of nearest neighbor samples corresponding to the fourth sub-model is determined by iteratively searching within a preset K value range using a Bayesian optimization algorithm.

[0103] In the spraying parameter determination device provided in this application embodiment, the training module is further used to take the first prediction result, the second prediction result, the third prediction result, and the fourth prediction result as derived features, fuse the derived features and historical spraying parameters to obtain a reconstructed dataset; based on the reconstructed dataset, a preset optimizer is used to train the multilayer perceptron model, and the initial learning rate of the multilayer perceptron model is smoothly decayed to a preset value according to the cosine function within a preset training period; and the target prediction result output by the multilayer perceptron model is obtained, the mean square error between the target prediction result and the historical spraying parameters is determined, and the model weights corresponding to the first sub-model, the second sub-model, the third sub-model, and the fourth sub-model are adjusted according to the mean square error; and if the mean square error does not decrease for a consecutive preset number of rounds, the training is stopped to obtain the spraying color model.

[0104] This application also provides an electronic device, including: a memory and a processor, wherein the memory is used to store program instructions; and the processor is connected to the memory and used to execute the above-described method for determining spraying parameters.

[0105] It should be noted that the aforementioned electronic equipment is used to perform Figure 2 The method for determining the spraying parameters shown above also applies to this electronic device, and will not be repeated here.

[0106] This application also provides a non-volatile storage medium, which includes a stored computer program, wherein the device containing the non-volatile storage medium executes the above-described spraying parameter determination method by running the computer program.

[0107] It should be noted that the aforementioned non-volatile storage media is used for execution. Figure 2 The method for determining the spraying parameters shown above is also applicable to this non-volatile storage medium, and will not be repeated here.

[0108] This application also provides a computer program product, including computer instructions, which, when executed by a processor, implement the above-described method for determining spraying parameters.

[0109] It should be noted that the above-mentioned computer program product is used to execute Figure 2 The method for determining spraying parameters shown above also applies to this computer program product, and will not be repeated here.

[0110] The sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0111] In the above embodiments of this application, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0112] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units can be a logical functional division, and in actual implementation, there may be other division methods. For instance, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual coupling, direct coupling, or communication connection may be through some interfaces; the indirect coupling or communication connection between units or modules may be electrical or other forms.

[0113] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0114] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0115] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, read-only memory (ROM), random access memory (RAM), portable hard drive, magnetic disk, or optical disk.

[0116] The above description is only a preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A method for determining spraying parameters, characterized in that, include: Determine the target color scheme and target color value for painting the target vehicle, wherein the target color scheme is used to represent the category to which the target color belongs, the target color value is used to represent the three-dimensional coordinate value of the target color in the color space, and the target color is the desired painting color of the target vehicle; A spraying color model is used to analyze the target color system and the target color value to obtain the spraying parameters corresponding to the target vehicle. The spraying color model includes a first sub-model for modeling the nonlinear relationship between the target color system and the target color value and the corresponding spraying parameters, a second sub-model for enhancing the robustness to noise and outliers, a third sub-model for mining the data patterns between the target color value and the corresponding spraying parameters, and a fourth sub-model for extracting the local similarity between the target color system and the target color value.

2. The method according to claim 1, characterized in that, The spray color model was trained in the following way: Obtain historical spraying data; Based on the historical spraying data, the first sub-model, the second sub-model, the third sub-model, and the fourth sub-model are trained independently to obtain the first prediction result, the second prediction result, the third prediction result, and the fourth prediction result; The reconstructed dataset is determined based on the first prediction result, the second prediction result, the third prediction result, the fourth prediction result, and the historical spraying data; The multilayer perceptron model is trained based on the reconstructed dataset to obtain the spray color model.

3. The method according to claim 2, characterized in that, Obtain historical spraying data, including: Obtain historical spraying parameters from the database, and determine the historical spraying pigment information, historical target color value, and historical color measurement value corresponding to the historical spraying parameters; By using the historical sprayed pigment information, the historical target color value, and the historical color measurement value as feature fields, and the historical spraying parameters as data labels, structured historical spraying data is obtained. The historical spraying data is normalized, and the normalized historical spraying data is divided into training set and test set according to a preset ratio.

4. The method according to claim 3, characterized in that, The method further includes: The first sub-model is trained using supervised regression with the goal of minimizing the mean square error between the first prediction result and the historical spraying parameters. The first model parameters of the first sub-model are optimized using a Bayesian optimization algorithm. The first model parameters include at least one of the following: learning rate, maximum depth, number of weak learners, and regularization coefficient. The second sub-model employs parallel ensemble training with the objective of minimizing the mean square error between the second prediction result and the historical spraying parameters, and optimizes the second model parameters of the second sub-model using the Bayesian optimization algorithm, wherein the second model parameters include at least one of the following: decision tree depth, number of decision trees, and maximum number of feature samples per node; The third sub-model is trained using gradient boosting, with the goal of optimizing the mean square error between the third prediction result and the historical spraying parameters round by round. The third model parameters of the third sub-model are optimized using the Bayesian optimization algorithm. The third model parameters include at least one of the following: learning rate, maximum depth, number of weak learners, and regularization coefficient. The fourth sub-model is trained using a nearest neighbor search strategy, with the goal of minimizing the mean square error between the fourth prediction result and the historical spraying parameters. The optimal number of nearest neighbor samples corresponding to the fourth sub-model is determined by iteratively searching within a preset K value range using the Bayesian optimization algorithm.

5. The method according to claim 3, characterized in that, The multilayer perceptron model is trained based on the reconstructed dataset to obtain the spraying color model, including: The first prediction result, the second prediction result, the third prediction result, and the fourth prediction result are used as derived features. The derived features and the historical spraying parameters are then fused to obtain the reconstructed dataset. Based on the reconstructed dataset, the multilayer perceptron model is trained using a preset optimizer, and the initial learning rate of the multilayer perceptron model is smoothly decayed to a preset value according to a cosine function within a preset training period; and Obtain the target prediction result output by the multilayer perceptron model, determine the mean square error between the target prediction result and the historical spraying parameters, and adjust the model weights corresponding to the first sub-model, the second sub-model, the third sub-model, and the fourth sub-model according to the mean square error; and If the mean square error does not decrease for a consecutive preset number of rounds, training is stopped, and the spray color model is obtained.

6. The method according to claim 1, characterized in that, The method further includes: Determine the color measurement value of the target vehicle after spraying according to the spraying parameters; The color measurement value and the target color value are compared to obtain a comparison result; If the comparison result indicates that the difference between the color measurement value and the target color value is less than a preset difference value, the spraying parameters are determined as the target spraying parameters for the target vehicle. If the comparison result indicates that the difference between the color measurement value and the target color value is greater than or equal to the preset difference value, the spraying color model is used to analyze the target color system, the target color value, and the color measurement value to obtain the target spraying parameters of the target vehicle.

7. The method according to claim 1, characterized in that, The method further includes: Obtain the hard constraints for painting the target vehicle, wherein the hard constraints are one or more combinations of painting parameters set based on actual needs; Based on the aforementioned hard constraints, the target color system and target color value are analyzed using the spraying color model to obtain the spraying parameters corresponding to the target vehicle. The spraying parameters include at least the paint flow rate, atomization speed, spraying voltage, internal flow rate of forming air, and external flow rate of forming air of the spraying equipment.

8. A device for determining spraying parameters, characterized in that, include: The determination module is used to determine the target color system and target color value for painting the target vehicle, wherein the target color system is used to represent the category to which the target color belongs, the target color value is used to represent the three-dimensional coordinate value of the target color in the color space, and the target color is the desired painting color of the target vehicle; The analysis module is used to analyze the target color system and the target color value using a spraying color model to obtain the spraying parameters corresponding to the target vehicle. The spraying color model includes a first sub-model for modeling the nonlinear relationship between the target color system and the target color value and the corresponding spraying parameters, a second sub-model for enhancing the robustness to noise and outliers, a third sub-model for mining the data patterns between the target color value and the corresponding spraying parameters, and a fourth sub-model for extracting the local similarity between the target color system and the target color value.

9. An electronic device, characterized in that, include: A memory and a processor, wherein the memory is used to store program instructions; The processor, connected to the memory, is used to execute the spraying parameter determination method according to any one of claims 1 to 7.

10. A non-volatile storage medium, characterized in that, The non-volatile storage medium includes a stored computer program, wherein the device containing the non-volatile storage medium executes the spraying parameter determination method according to any one of claims 1 to 7 by running the computer program.

11. A computer program product comprising computer instructions, characterized in that, When the computer instructions are executed by the processor, they implement the spraying parameter determination method according to any one of claims 1 to 7.

Images