Machine Learning-Based Closed-Loop Control System and Method for Vacuum Thermal Transfer Printing on Mineral Surfaces

Abstract

This application relates to the field of industrial printing and surface treatment technology, and provides a closed-loop control system and method for vacuum thermal transfer printing on mineral surfaces based on machine learning. The method includes: acquiring a target digital image; extracting the color distribution, texture features, and pigment density features of the target digital image; generating initial printing parameters using a machine learning prediction model based on the material's optical properties, ink absorption characteristics, and the equipment's printhead configuration, ink characteristics, and operating parameters; performing an image transfer operation on the material surface according to the initial printing parameters to achieve seamless integrated transfer of the six sides of the stone and its internal pores; acquiring the actual image result after the transfer is completed; comparing the actual image result with the target digital image, calculating color difference and structural similarity, and generating error data; automatically adjusting the printing parameters based on the error data to generate corrected printing parameters; and inputting the error data and corrected printing parameters into the machine learning prediction model to update the model's parameters.

Description

Technical Field

[0001] This application relates to the field of industrial printing and surface treatment technology, and in particular to a closed-loop control system and method for vacuum thermal transfer printing on mineral surfaces based on machine learning. Background Technology

[0002] In the field of mineral material surface treatment, vacuum heat transfer technology is widely used in industries such as architectural decoration and home building materials to reproduce designed patterns on the surfaces of natural stone, quartz, artificial stone, ceramics, and composite mineral materials. Unlike ordinary surface printing, vacuum heat transfer requires the dye to vaporize and diffuse in a vacuum and high-temperature environment, penetrating to a certain depth into the mineral material to ultimately form a textured structure with depth and layers. Currently, the industry generally uses static ICC color profiles combined with manual color matching to control surface color, failing to fully integrate closed-loop control of vacuum pressure, heating temperature, transfer time, and dye penetration depth.

[0003] The aforementioned existing technology has the following insurmountable drawbacks: First, static color curves can only reflect the color mapping relationship between materials and equipment under specific standard conditions. They cannot adapt to the inherent differences in ink absorption, reflectivity, surface texture, etc. of different batches of materials, nor can they compensate for output fluctuations caused by equipment aging and changes in ambient temperature and humidity, resulting in a continuous deviation between digital images and actual transfer results.

[0004] Secondly, manual color matching relies heavily on the experience and subjective judgment of operators, resulting in long cycles, low efficiency, and significant differences in results among different operators, making it difficult to guarantee consistency in mass production. Furthermore, manual color matching cannot quantify color and structural deviations, making precise local corrections impossible.

[0005] Third, existing technologies employ an open-loop control mode, where printing parameters remain unchanged throughout the production process once determined, lacking an automatic feedback correction mechanism based on actual transfer results. When deviations occur, production must be paused for manual adjustments, severely impacting production efficiency.

[0006] Fourth, existing technologies lack self-learning capabilities and cannot accumulate transfer printing experience under different materials, equipment, and environmental conditions. Each time materials or equipment are changed, color curves need to be remade and colors need to be manually adjusted, which increases production and time costs.

[0007] Fifth, even if existing technologies employ machine learning or color correction, they typically only optimize surface color or printing parameters, without predicting and correcting the amount of dye, heating temperature, vacuum pressure, transfer time, and the penetration depth of the dye within the mineral material as an interconnected industrial physical process.

[0008] Sixth, existing technologies lack a constraint mechanism that combines AI prediction results with physical models such as heat conduction, vacuum conditions, dye vaporization and diffusion, and mineral material adsorption and penetration. This makes it easy to remain at the level of ordinary algorithms and difficult to stably control the formation process of mineral surface textures with deep layers.

[0009] Therefore, a method is urgently needed to solve at least one of the above problems. Summary of the Invention

[0010] This application provides a closed-loop control system and method for vacuum thermal transfer printing on mineral surfaces based on machine learning. It aims to solve the problems of existing image transfer technologies that only focus on surface color correction and lack joint prediction and closed-loop correction of vacuum pressure, heating temperature, transfer time and dye penetration depth inside mineral materials, resulting in unstable image reproduction, internal penetration effect and batch production consistency of mineral material surfaces.

[0011] In a first aspect, embodiments of this application provide a closed-loop control method for vacuum thermal transfer on mineral surfaces based on machine learning, the method comprising: Acquire the target digital image; extract the color distribution, texture features, and pigment density features of the target digital image; Material parameters of the mineral material are acquired, including material type, thickness, density, thermal conductivity, dye adsorption / diffusion characteristics, and surface pretreatment parameters. Operating parameters of the vacuum heating equipment, transfer equipment, vacuum pressure sensor, and temperature sensor are also acquired. The image features, material parameters, and equipment operating parameters are input into a machine learning prediction model, which predicts initial process control parameters, including dye amount, heating temperature, vacuum pressure, transfer time, color mapping relationship, and the target penetration depth of the dye within the mineral material. Based on these initial process control parameters, thermal transfer is performed on the surface of the mineral material under vacuum and heating conditions, achieving seamless integrated transfer across all six sides and internal pores of the stone. The dye is vaporized / diffused under heat and penetrates into the mineral material. Actual image results and / or actual penetration effects are collected after the transfer is completed. The actual image results and / or actual penetration effect are compared with the target digital image to calculate color difference, structural similarity, and penetration depth deviation, generating error data. Based on the error data, the dye amount, heating temperature, vacuum pressure, transfer time, and target penetration depth are automatically adjusted to generate corrected process control parameters. The error data and corrected process control parameters are input into the machine learning prediction model to update the parameters of the machine learning prediction model.

[0012] Secondly, this application provides a machine learning-based closed-loop control system for vacuum thermal transfer printing on mineral surfaces, used to implement the method provided in any embodiment of this application, characterized in that it includes: The image acquisition unit is used to acquire the target digital image and extract the color distribution, texture features, and pigment density features of the target digital image. The process prediction and control unit is used to generate initial process control parameters based on mineral material parameters, dye characteristics, vacuum heating equipment parameters, and sensor operating parameters through a machine learning prediction model. The initial process control parameters include dye amount, heating temperature, vacuum pressure, transfer time, color mapping relationship, and target penetration depth. The unit then performs thermal transfer on the surface of the mineral material under vacuum and heating conditions according to the initial process control parameters. The result acquisition unit is used to acquire the actual image results and / or actual penetration effect after the transfer is completed. The model update unit is used to compare the actual image results and / or the actual penetration effect with the target digital image, calculate the color difference, structural similarity and penetration depth deviation, and generate error data; automatically adjust the dye amount, heating temperature, vacuum pressure, transfer time and target penetration depth based on the error data to generate corrected process control parameters; input the error data and corrected process control parameters into the machine learning prediction model to update the parameters of the machine learning prediction model.

[0013] This application achieves closed-loop adaptive control of the entire vacuum thermal transfer process on mineral surfaces. By automatically collecting transfer results and / or penetration effects, calculating color difference and penetration depth deviations, correcting process parameters, and updating the model, deviation correction can be completed without manual intervention, significantly improving the accuracy, depth perception, and batch production consistency of mineral surface image transfer. By employing a machine learning prediction model to generate process control parameters, it can comprehensively learn the influence of mineral material optical properties, thermal conductivity, dye adsorption / diffusion properties, vacuum pressure, heating temperature, transfer time, and equipment status on transfer effects and penetration depth, effectively compensating for differences in mineral materials from different batches and changes in equipment status. Based on the feedback learning mechanism of actual transfer results and actual penetration effects, the model can continuously accumulate experience and optimize parameters, achieving system self-evolution and making it suitable for large-scale industrial production.

[0014] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Attached Figure Description

[0015] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1This is a schematic flowchart illustrating the steps of a closed-loop control method for vacuum thermal transfer on mineral surfaces based on machine learning, provided in one embodiment of this application. Figure 2 This is a schematic block diagram of a closed-loop control system for vacuum thermal transfer of mineral surfaces based on machine learning, provided in one embodiment of this application. Figure 3 This is a schematic block diagram of the structure of a computer device provided in an embodiment of this application.

[0017] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Detailed Implementation

[0018] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0019] The flowchart shown in the attached diagram is for illustrative purposes only and does not necessarily include all content and operations / steps, nor does it necessarily have to be performed in the order described. For example, some operations / steps can be broken down, combined, or partially merged, so the actual execution order may change depending on the actual situation.

[0020] It should be understood that, in order to clearly describe the technical solutions of the embodiments of the present invention, the terms "first" and "second" are used in the embodiments of the present invention to distinguish identical or similar items with essentially the same function and effect. Those skilled in the art will understand that the terms "first" and "second" do not limit the quantity or execution order, and the terms "first" and "second" are not necessarily different.

[0021] It should be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the scope of the application. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0022] It should also be understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0023] In the field of mineral material surface treatment, vacuum heat transfer technology is widely used in industries such as architectural decoration and home building materials to reproduce designed patterns on the surfaces of natural stone, quartz, artificial stone, ceramics, and composite mineral materials. Unlike ordinary surface printing, vacuum heat transfer requires the dye to vaporize and diffuse in a vacuum and high-temperature environment, penetrating to a certain depth into the mineral material to ultimately form a textured structure with depth and layers. Currently, the industry generally uses static ICC color profiles combined with manual color matching to control surface color, failing to fully integrate closed-loop control of vacuum pressure, heating temperature, transfer time, and dye penetration depth.

[0024] The aforementioned existing technology has the following insurmountable drawbacks: First, static color curves can only reflect the color mapping relationship between materials and equipment under specific standard conditions. They cannot adapt to the inherent differences in ink absorption, reflectivity, surface texture, etc. of different batches of materials, nor can they compensate for output fluctuations caused by equipment aging and changes in ambient temperature and humidity, resulting in a continuous deviation between digital images and actual transfer results.

[0025] Secondly, manual color matching relies heavily on the experience and subjective judgment of operators, resulting in long cycles, low efficiency, and significant differences in results among different operators, making it difficult to guarantee consistency in mass production. Furthermore, manual color matching cannot quantify color and structural deviations, making precise local corrections impossible.

[0026] Third, existing technologies employ an open-loop control mode, where printing parameters remain unchanged throughout the production process once determined, lacking an automatic feedback correction mechanism based on actual transfer results. When deviations occur, production must be paused for manual adjustments, severely impacting production efficiency.

[0027] Fourth, existing technologies lack self-learning capabilities and cannot accumulate transfer printing experience under different materials, equipment, and environmental conditions. Each time materials or equipment are changed, color curves need to be remade and colors need to be manually adjusted, which increases production and time costs.

[0028] Fifth, even if existing technologies employ machine learning or color correction, they typically only optimize surface color or printing parameters, without predicting and correcting the amount of dye, heating temperature, vacuum pressure, transfer time, and the penetration depth of the dye within the mineral material as an interconnected industrial physical process.

[0029] Sixth, existing technologies lack a constraint mechanism that combines AI prediction results with physical models such as heat conduction, vacuum conditions, dye vaporization and diffusion, and mineral material adsorption and penetration. This makes it easy to remain at the level of ordinary algorithms and difficult to stably control the formation process of mineral surface textures with deep layers.

[0030] Therefore, a method is urgently needed to solve at least one of the above problems.

[0031] To solve the above problem, please refer to Figure 1 This application provides a machine learning-based closed-loop control method for vacuum thermal transfer printing on mineral surfaces, applied to computer equipment. The computer equipment can be deployed on a single server or server cluster, or it can be deployed in an industrial control computer, vacuum thermal transfer equipment controller, handheld terminal, laptop, wearable device, or robot, etc. The computer equipment is communicatively connected to vacuum heating equipment, transfer equipment, vacuum pressure sensor, temperature sensor, image acquisition equipment, and / or penetration effect detection equipment to achieve closed-loop control combining AI and physical industrial processes.

[0032] The provided machine learning-based closed-loop control method for vacuum thermal transfer on mineral surfaces includes steps S101 to S103. Details are as follows: Step S101. Acquire the target digital image; extract the color distribution, texture features and pigment density features of the target digital image.

[0033] Specifically, the core purpose of this step is to obtain a standard image to be transferred and extract multi-dimensional features that can quantify the requirements for transfer quality, providing a standardized data foundation for the input of subsequent machine learning models.

[0034] The target digital image acquisition process involves receiving the digital image to be transferred, input by the user. This image can be a digital file of any format or a physical image directly captured by an image acquisition device. After acquiring the image, a preliminary integrity check is performed to inspect it for damage, missing parts, or other issues.

[0035] Image preprocessing involves standardizing the acquired raw images, including format conversion, color space conversion, and resizing. Images of different formats are uniformly converted to a bitmap format for easier processing, the color space is uniformly converted to RGB color space, and the image size is adjusted to match the surface dimensions of the workpiece to be transferred.

[0036] Feature extraction involves performing multi-dimensional feature extraction on the preprocessed image, specifically extracting the following three core features: Color distribution features are generated by statistically analyzing the frequency of occurrence, distribution area, and color gradient of different colors in an image to produce color distribution histograms and color heatmaps. Texture features are extracted by using algorithms such as gray-level co-occurrence matrix and wavelet transform to extract features such as texture direction, contrast, roughness and periodicity of the image; Pigment density features are generated by dividing the image into several uniform sub-regions, calculating the average gray value of each sub-region, and generating the pigment density features of each sub-region based on the correspondence between gray value and pigment deposition amount. Feature standardization normalizes all extracted feature data, converting them into values ​​within the range of [0,1], thereby eliminating the influence of different feature units and ensuring that the machine learning model can process them stably.

[0037] Step S102. Obtain the material parameters of the mineral material, including material type, thickness, density, thermal conductivity, dye adsorption / diffusion characteristics, and surface pretreatment parameters. Also obtain the operating parameters of the vacuum heating equipment, transfer equipment, vacuum pressure sensor, and temperature sensor. Input the image features, material parameters, and equipment operating parameters into a machine learning prediction model. The machine learning prediction model predicts initial process control parameters, including dye amount, heating temperature, vacuum pressure, transfer time, color mapping relationship, and the target penetration depth of the dye within the mineral material. Based on the initial process control parameters, perform thermal transfer on the surface of the mineral material under vacuum and heating conditions to achieve seamless integrated transfer of the six sides of the stone and its internal pores. Allow the dye to vaporize / diffuse under heat and penetrate into the mineral material. Collect the actual image results and / or actual penetration effect after the transfer is completed.

[0038] Specifically, this step is the core execution link of the present invention. It uses a machine learning model to intelligently predict process parameters and executes a complete vacuum heat transfer process, ultimately achieving seamless integrated transfer of the six sides and internal pores of the mineral material.

[0039] Multi-source parameter acquisition involves obtaining material parameters of mineral materials, including material type, thickness, density, thermal conductivity, porosity, dye adsorption coefficient, dye diffusion coefficient, moisture content, and surface pretreatment parameters.

[0040] Acquire equipment operating parameters, including the number of printheads, printhead resolution, ink type, and ink viscosity of the transfer equipment; the temperature range, vacuum range, heating rate, and cooling rate of the vacuum heating oven; and the sampling frequency and accuracy parameters of various sensors.

[0041] The initial process parameter prediction involves inputting the image features extracted in step S101, the material parameters obtained above, and the equipment operating parameters into a pre-trained machine learning prediction model. The model, through comprehensive analysis of the input data, outputs the optimal combination of initial process control parameters, including: Pattern output resolution: 200-400 DPI; Printing ink volume: 50%-300% (relative to standard ink volume); Vacuum heating furnace heating temperature: 100-300℃; Vacuum pressure: -0.08MPa to -0.1MPa; Transfer time: 5-60 minutes; Color mapping table: Establishes the correspondence between the RGB values ​​of the target image and the actual ink output volume; Target penetration depth: 0.1mm to 2mm; Preparation before transfer: Based on the initial color mapping table and the amount of printing ink, print the target digital image onto special transfer paper. Place the mineral material to be transferred into a vacuum heating furnace and preheat it according to the initial heating temperature to remove the moisture inside the material. The preheating time is determined according to the material thickness and moisture content, generally 10-30 minutes. Simultaneously dry the printed transfer paper to remove the moisture from the transfer paper.

[0042] The transfer paper lamination and wrapping process involves cutting the dried transfer paper into shapes that match the dimensions of each outer surface and the inner wall of the pores of the mineral material. The cut transfer paper is then individually lamination onto all the outer surfaces and inner walls of the pores of the preheated mineral material, ensuring complete adhesion between the transfer paper and the material surface, without bubbles or wrinkles. A high-temperature resistant silicone mold, perfectly adapted to the three-dimensional shape of the mineral material, is used to completely wrap the mineral material. A pressure of 0.1-0.5 MPa is applied using a uniform pressure device to ensure a tight fit between the transfer paper and each surface and the inner wall of the pores of the mineral material.

[0043] Vacuum heat transfer printing is performed by sending the coated mineral material into a vacuum heating furnace, closing the furnace door, and starting the vacuum system; the vacuum level inside the furnace is evacuated to the initial vacuum pressure, and the furnace temperature is raised to the initial heating temperature; this temperature and pressure condition is maintained for the initial transfer time, allowing the dye to vaporize, diffuse, and penetrate into the mineral material under vacuum and high temperature conditions; during the transfer process, the vacuum pressure inside the furnace is monitored in real time by a vacuum pressure sensor, and the temperature inside the furnace is monitored in real time by a temperature sensor to ensure stable process parameters.

[0044] When a vacuum failure is detected, an audible and visual alarm will be issued immediately and heating will be stopped to prevent transfer failure.

[0045] After the transfer is complete, the heating and vacuum systems are shut off, and inert gas is introduced into the furnace. Once the furnace pressure returns to normal, the mineral material is removed. The mineral material is then allowed to stand at room temperature, and the transfer paper is peeled off. Multiple images of all outer surfaces and the inner walls of the pores of the transferred mineral material are taken using a high-resolution industrial camera to obtain actual image results. The transferred surface is then scanned using a spectral detection device to obtain data on the actual penetration depth of the dye.

[0046] Step S103. Compare the actual image result and / or the actual penetration effect with the target digital image, calculate the color difference, structural similarity and penetration depth deviation, and generate error data; automatically adjust the dye amount, heating temperature, vacuum pressure, transfer time and target penetration depth according to the error data to generate corrected process control parameters; input the error data and corrected process control parameters into the machine learning prediction model to update the parameters of the machine learning prediction model.

[0047] Specifically, this step achieves closed-loop control of the transfer process. By comparing the actual transfer results with the target requirements, the process parameters are automatically adjusted, and the prediction accuracy of the machine learning model is continuously optimized.

[0048] Error calculation is performed by registering and aligning the acquired actual image results with the target digital image pixel by pixel.

[0049] The color difference of the corresponding pixels is calculated in the CIELab color space, and the color difference value of each pixel is calculated using the CIEDE2000 color difference formula.

[0050] The structural similarity (SSIM) between the actual image and the target image is calculated, including three components: brightness similarity, contrast similarity, and structural similarity.

[0051] The actual penetration depth is compared with the target penetration depth to calculate the penetration depth deviation.

[0052] Generate an error distribution heatmap and mark areas where the difference exceeds a preset threshold.

[0053] The automatic adjustment of process parameters establishes a mapping relationship between errors and process parameters based on the error distribution thermogram and the penetration depth deviation.

[0054] For areas with significant color differences, adjust the color mapping values ​​and printing ink volume for the corresponding areas.

[0055] For areas with blurred textures, adjust the heating temperature and transfer time for the corresponding areas.

[0056] For areas with insufficient penetration depth, increase the heating temperature of the corresponding area, extend the transfer time, or increase the vacuum level.

[0057] For areas with excessive penetration, reduce the heating temperature of the corresponding area, shorten the transfer time, or reduce the vacuum level.

[0058] Integrate the parameter adjustment results from all regions to generate the corrected complete process control parameters.

[0059] The machine learning model is updated by using the input data of this transfer (image features, material parameters, equipment parameters), initial process parameters, error data, and corrected process parameters as a new training sample.

[0060] Add new training samples to the model's training dataset.

[0061] Incremental learning is used to retrain the machine learning prediction model.

[0062] During the training process, pre-established heat conduction models, dye diffusion models, and vacuum pressure models are introduced as physical constraints to ensure that the model's prediction results conform to physical laws.

[0063] Update the model parameters to enable the model to achieve higher accuracy in subsequent predictions.

[0064] In some embodiments, acquiring the target digital image includes: receiving a digital image file in a preset image format; converting the received digital image file to a format to generate a target digital image in a uniform format; and performing noise removal and resolution unification processing on the target digital image.

[0065] In this embodiment, the step of acquiring the target digital image specifically includes: It accepts digital image files in preset image formats, including but not limited to common formats such as JPEG, PNG, BMP, TIFF, and PSD.

[0066] The received digital image files are converted to generate target digital images in a uniform format. Specifically, an image format conversion library is called to convert image files of different formats into 24-bit true color bitmap format, ensuring that each pixel contains complete RGB color information.

[0067] Noise removal is performed on the target digital image. An adaptive median filter is used to remove salt-and-pepper noise, and a Gaussian filter is used to remove Gaussian noise. The size of the filter window is automatically adjusted according to the noise level of the image, typically ranging from 3×3 to 7×7 pixels.

[0068] The target digital image undergoes resolution standardization. The image resolution is uniformly adjusted to 300 DPI, a standard resolution commonly used in industrial transfer printing, ensuring the clarity and detail of the transferred image. For images with a resolution lower than 300 DPI, bicubic interpolation is used for upsampling; for images with a resolution higher than 300 DPI, region averaging is used for downsampling.

[0069] This embodiment ensures the consistency and high quality of image data input into the machine learning model through a standardized image acquisition and preprocessing process, laying a solid foundation for subsequent feature extraction and parameter prediction.

[0070] In some embodiments, the extracted color distribution, texture features, and pigment density features of the target digital image include: extracting the red, green, and blue channel values ​​of each pixel in the target digital image; converting them to generate color distribution data for each pixel; extracting texture feature data of the target digital image; performing region segmentation on the target digital image; calculating the pigment deposition density of each segmented region; and forming pigment density feature data.

[0071] This embodiment details the specific implementation method of target digital image feature extraction in step S101.

[0072] In this embodiment, the steps of extracting the color distribution, texture features, and pigment density features of the target digital image specifically include: extracting the red, green, and blue channel values ​​of each pixel in the target digital image. This involves traversing all pixels of the image and obtaining the values ​​of the R, G, and B channels for each pixel, with the values ​​ranging from 0 to 255.

[0073] The process involves converting and generating color distribution data for each pixel. The RGB values ​​of each pixel are converted to CIELab color space values, where L represents luminance, a represents red-green hue, and b represents yellow-blue hue. The distribution of L, a*, and b* values ​​for all pixels in the image is statistically analyzed, generating a three-channel histogram as the color distribution feature.

[0074] Extract texture feature data from the target digital image. The Gray-Level Co-occurrence Matrix (GLCM) algorithm is used to extract texture features. The specific steps are as follows: Convert a color image to a grayscale image; Calculate the gray-level co-occurrence matrix of the gray-level image in the four directions of 0°, 45°, 90°, and 135°; Four statistical measures—energy, contrast, correlation, and entropy—are extracted from each gray-level co-occurrence matrix. The average of the statistics in the four directions is taken as the final texture feature; The target digital image is segmented into regions, and the pigment deposition density of each segmented region is calculated. The K-means clustering algorithm is used to segment the image into K regions (K is typically 5-20), where pixels within each region have similar color and texture features. The average gray value of each region is calculated, and the pigment density of each region is calculated using the formula "pigment density = 255 - average gray value", forming pigment density feature data.

[0075] This embodiment uses a multi-dimensional, multi-algorithm feature extraction method to comprehensively and accurately capture the color, texture, and density information of the target image, providing rich input features for machine learning models.

[0076] In some embodiments, acquiring the material parameters of the mineral material, including material type, thickness, density, thermal conductivity, dye adsorption / diffusion characteristics, and surface pretreatment parameters, and acquiring the operating parameters of the vacuum heating equipment, transfer equipment, vacuum pressure sensor, and temperature sensor, includes: retrieving the material parameters of the corresponding mineral material from a pre-stored mineral material database, the material parameters including the reflectivity, gloss, porosity, thickness, density, thermal conductivity, dye adsorption coefficient, dye diffusion coefficient, target penetration depth range, and surface pretreatment parameters of the material surface; retrieving the equipment parameters of the corresponding vacuum heat transfer equipment from a pre-stored equipment database, the equipment parameters including the printhead configuration of the transfer equipment, dye characteristics, vacuum heating furnace temperature range, vacuum pressure range, transfer time range, heating / cooling curve, and sensor acquisition parameters; inputting the extracted image features and the retrieved material and equipment parameters into a machine learning prediction model; and generating an initial color map, initial dye amount, initial heating temperature, initial vacuum pressure, initial transfer time, and target penetration depth through the machine learning prediction model.

[0077] In this embodiment, the step of obtaining the material parameters of the mineral material and the equipment operating parameters, and generating initial process control parameters specifically includes: Material parameters for the corresponding mineral material are retrieved from a pre-stored mineral material database. This database pre-stores detailed parameters for common mineral materials such as granite, marble, artificial stone, and quartz stone, including: Optical parameters: surface reflectivity, gloss, refractive index; Physical parameters: thickness, density, porosity, moisture content; Thermal parameters: thermal conductivity, specific heat capacity, coefficient of thermal expansion; Chemical parameters: dye adsorption coefficient, dye diffusion coefficient, weather resistance parameters; Process parameters: Recommended heating temperature range, recommended transfer time range, recommended penetration depth range Users only need to input the type and specifications of the mineral material, and the system can automatically retrieve all the corresponding parameters from the database.

[0078] The equipment parameters of the corresponding vacuum heat transfer equipment are retrieved from a pre-stored equipment database. The equipment database pre-stores detailed parameters for different models of vacuum heat transfer equipment, including: Transfer equipment parameters: number of printheads, printhead resolution, maximum printing area, ink type, ink viscosity; Vacuum heating furnace parameters: effective volume, temperature range, temperature uniformity, vacuum range, and pumping rate; Sensor parameters: temperature sensor accuracy, vacuum pressure sensor accuracy, sampling frequency; Users only need to enter the device's serial number, and the system can automatically retrieve all the corresponding parameters from the database.

[0079] The extracted image features are concatenated with the retrieved material and equipment parameters to form a feature vector of dimension N.

[0080] The feature vectors are input into a pre-trained deep neural network prediction model. This model employs a multilayer perceptron architecture, comprising an input layer, three hidden layers, and an output layer. The number of neurons in the input layer equals the dimension N of the feature vectors. Each hidden layer contains 256 neurons, and the output layer contains 7 neurons, corresponding to 7 initial process control parameters.

[0081] The model calculates and outputs the initial color map, initial dye amount, initial heating temperature, initial vacuum pressure, initial transfer time, and target penetration depth through forward propagation.

[0082] This embodiment achieves rapid and accurate parameter acquisition by establishing a standardized material database and equipment database. Combined with a deep neural network model, it enables intelligent prediction of process parameters, greatly improving the efficiency and consistency of the transfer process.

[0083] In some embodiments, the step of performing heat transfer printing on the surface of mineral materials under vacuum and heating conditions according to the initial process control parameters to achieve seamless integrated transfer printing on the six sides and internal pores of the stone includes: converting the color data of the target digital image according to the initial color map table and controlling the dye output according to the initial dye amount; controlling the operation of the vacuum heat transfer printing equipment according to the initial heating temperature, initial vacuum pressure and initial transfer time; performing vacuum heat transfer printing on the surface of the mineral materials, so that the dye vaporizes, diffuses and enters the interior of the mineral materials under vacuum and high temperature environment, completing the image transfer and forming a textured structure with depth and layering.

[0084] This embodiment details the specific implementation method of vacuum heat transfer in step S102, focusing on the key technologies for achieving seamless transfer across 6 sides.

[0085] In this embodiment, the step of performing vacuum thermal transfer according to the initial process control parameters specifically includes: The color data of the target digital image is converted based on the initial color map table. The RGB value of each pixel in the target image is substituted into the color map table to obtain the corresponding output amounts of cyan, magenta, yellow, and black inks.

[0086] The ink output is controlled based on the initial dye quantity. The ink output of each pixel is multiplied by the initial dye quantity coefficient to obtain the final ink output quantity. The printhead of the transfer equipment is then controlled to print the pattern onto the transfer paper according to the final ink output quantity.

[0087] Place the mineral material to be transferred into a vacuum heating furnace for preheating. Set the furnace temperature to 80% of the initial heating temperature. The preheating time is determined according to the material thickness, with 10 minutes of preheating for every 10mm of thickness. During preheating, introduce a small amount of dry air into the furnace to accelerate the evaporation of moisture from the material.

[0088] Simultaneously dry the printed transfer paper. Place the transfer paper in a drying oven and dry it at 60-80℃ for 5-10 minutes to remove moisture and prevent air bubbles from forming during the transfer process.

[0089] Transfer paper application. Cut the dried transfer paper to a shape that perfectly matches the dimensions of all surfaces of the mineral material and the inner walls of the boreholes. For flat surfaces, apply the transfer paper directly and smoothly to the material surface; for sides and edges, use a stretching application method to ensure complete adhesion between the transfer paper and the edges; for the inner walls of the boreholes, use a rolling application method, rolling the transfer paper into a cylinder with the same diameter as the borehole, inserting it into the borehole, and then unrolling it to ensure complete adhesion between the transfer paper and the inner wall of the borehole.

[0090] Silicone mold coating and pressurization. A high-temperature resistant silicone mold (temperature resistance ≥350℃) perfectly matches the three-dimensional shape of the mineral material to completely coat the mineral material. The silicone mold is 3-5mm thick and has good elasticity and thermal conductivity. A hydraulic pressurization device applies uniform pressure to the silicone mold at a pressure of 0.2-0.3MPa to ensure that the transfer paper and all surfaces of the mineral material are tightly adhered without any gaps.

[0091] Vacuum heat transfer. The coated mineral material is placed into the vacuum furnace, and the furnace door is closed. The vacuum system is activated, and the vacuum level inside the furnace is evacuated to -0.095 MPa. The heating system is activated, and the furnace temperature is raised to the initial heating temperature. This temperature and pressure condition is maintained for the initial transfer time. During the transfer process, temperature fluctuations are controlled within ±5℃, and vacuum level fluctuations are controlled within ±0.005 MPa.

[0092] Post-transfer processing: After the transfer time is complete, turn off the heating system and maintain a vacuum until the furnace temperature drops below 100°C. Then, introduce nitrogen into the furnace. Once the furnace pressure returns to atmospheric pressure, open the furnace door and remove the mineral material. Let the mineral material stand at room temperature for 2-4 hours until it is completely cooled. Peel off the transfer paper to complete the transfer process.

[0093] This embodiment successfully achieved seamless transfer of the six sides and internal pores of the mineral material through precise transfer paper bonding process and silicone mold overall covering and pressure technology. The transferred image is seamless, bubble-free, with clear texture and uniform color.

[0094] In some embodiments, acquiring the actual image result and / or actual penetration effect after the transfer is completed includes: taking a picture of the material surface after the transfer is completed to obtain visible light data of the actual image; scanning the material surface after the transfer is completed to obtain spectral data of the actual image; and integrating the visible light data and spectral data into the actual image result and / or actual penetration effect.

[0095] In this embodiment, the steps of acquiring the actual image results and actual penetration effect after the transfer are completed specifically include: A standardized image acquisition platform was established. This platform includes: a high-resolution industrial camera (resolution ≥ 12 megapixels), a standard D65 light source, a multi-angle rotating stage, and a black background. The acquisition environment is a darkroom to avoid interference from external light.

[0096] Actual image acquisition. The transferred mineral material is fixed on a multi-angle rotating stage. The stage is controlled to rotate at a preset angle, and an industrial camera captures images of the mineral material from six directions: front, back, left side, right side, top, and bottom, as well as the axial direction of all holes. Three images are taken from each direction, and the average value is taken as the actual image result for that direction.

[0097] Spectral data acquisition. A portable spectrometer was used to scan the transfer surface point by point. The scanning step size was 1 mm, and the scanning range covered all outer surfaces of the mineral material and the inner walls of the pores. Spectral reflectance data of each scanning point in the visible light range of 400-700 nm were acquired.

[0098] Actual penetration effect data acquisition. A microscopic spectral imaging system was used to examine the cross-section of the mineral material. Thin slices with a thickness of 1 mm were taken from different locations on the mineral material and observed under the microscopic spectral imaging system. The system can automatically identify the boundary of dye penetration, measure the actual penetration depth at each location, and generate a penetration depth distribution map.

[0099] Data integration. The collected multi-angle image data, spectral data, and penetration depth data are registered and fused to generate a transfer result dataset containing complete spatial, color, and penetration information.

[0100] This embodiment obtains comprehensive and accurate transfer quality data through a standardized, multi-dimensional method for collecting transfer results, providing a reliable basis for subsequent error analysis and parameter adjustment.

[0101] In some embodiments, comparing the actual image result and / or the actual penetration effect with the target digital image, calculating color difference, structural similarity, and penetration depth deviation, and generating error data includes: aligning the actual image result with the target digital image pixel by pixel; calculating the color difference value of the corresponding pixel; calculating the brightness similarity, contrast similarity, and structural similarity between the actual image result and the target digital image; generating difference distribution data based on the color difference value and structural similarity, and marking areas where the difference exceeds a preset threshold.

[0102] In this embodiment, the step of comparing the actual image results and actual penetration effect with the target digital image to generate error data specifically includes: Image registration. A feature-point-based image registration algorithm is used to accurately align the actual image with the target image. The specific steps are as follows: The SIFT algorithm is used to extract feature points from the actual image and the target image respectively; The RANSAC algorithm is used to remove feature points that are incorrectly matched. Calculate the homography matrix based on the correctly matched feature points; Geometric transformations are performed on the actual image based on the homography matrix to achieve pixel-by-pixel alignment with the target image.

[0103] Color difference calculation involves calculating the L*, a*, and b* differences for each corresponding pixel in the CIELab color space. The total color difference ΔE00 for each pixel is calculated using the CIEDE2000 color difference formula. The CIEDE2000 color difference formula comprehensively considers differences in brightness, chromaticity, and hue, and is currently the color difference calculation formula that best reflects the characteristics of human vision.

[0104] Structural similarity calculation. The Structural Similarity Calculation (SSIM) algorithm is used to calculate the structural similarity between the actual image and the target image. The SSIM algorithm evaluates image similarity from three aspects: brightness, contrast, and structure. The calculation formula is as follows: SSIM(x,y)=[l(x,y)]^α [c(x,y)]^β [s(x,y)]^γ; Where l(x,y) is the brightness similarity, c(x,y) is the contrast similarity, s(x,y) is the structural similarity, and α, β, and γ are weighting coefficients, generally taken as α=β=γ=1.

[0105] Permeability depth deviation calculation. The actual permeability depth distribution map is compared with the target permeability depth distribution map point by point, and the permeability depth deviation Δd = d_actual - d_target is calculated for each point.

[0106] Error distribution generation. Generate a color difference distribution map, a structural similarity distribution map, and a penetration depth deviation distribution map with the same dimensions as the target image. Set the color difference threshold to ΔE00=2.0, the structural similarity threshold to 0.95, and the penetration depth deviation threshold to 0.1mm. Mark areas with differences exceeding the thresholds with different colors to generate an error distribution heatmap.

[0107] This embodiment adopts internationally recognized color difference and structural similarity evaluation standards, combined with penetration depth deviation calculation, to achieve objective and quantitative evaluation of transfer quality. The generated error distribution heatmap can intuitively show the areas where there are problems with transfer quality.

[0108] In some embodiments, the step of automatically adjusting the dye amount, heating temperature, vacuum pressure, transfer time, and target penetration depth based on error data to generate corrected process control parameters includes: determining the process parameter deviation corresponding to each difference region based on the difference distribution data and penetration depth deviation; adjusting the corresponding value of the color mapping table and the dye amount of the corresponding region for each difference region; adjusting the heating temperature, vacuum pressure, transfer time, and target penetration depth of the corresponding region; and integrating all adjusted parameters to generate corrected process control parameters.

[0109] In this embodiment, the step of automatically adjusting the process control parameters based on error data specifically includes: Establish an error-parameter mapping relationship. Based on extensive experimental data, a quantitative mapping relationship between error and process parameters is established in advance. The relationship between color difference ΔE00 and dye amount adjustment coefficient: Δ dye amount = k1 × ΔE00, where k1 is a proportional coefficient, generally taken as 0.05-0.1; The relationship between structural similarity (SSIM) and heating temperature adjustment is: Δtemperature = k2 × (1 - SSIM), where k2 is a proportionality coefficient, typically taken as 50-100℃. The relationship between penetration depth deviation Δd and transfer time adjustment: Δtime = k3 × Δd, where k3 is a proportionality coefficient, generally taken as 10-20 minutes / mm; Regional parameter adjustment. Based on the error distribution heatmap, individual parameter adjustments are made for each error region: For areas where the color difference exceeds the threshold, adjust the color mapping table value and dye amount for the corresponding area based on the average color difference of that area. If the actual color is too light, increase the dye amount; if the actual color is too dark, decrease the dye amount.

[0110] For regions with structural similarity below a threshold, the heating temperature and transfer time are adjusted based on the average structural similarity of that region. If the texture is blurry, the heating temperature is increased or the transfer time is extended; if the texture is too coarse, the heating temperature is decreased or the transfer time is shortened.

[0111] For areas where the penetration depth deviation exceeds the threshold, the heating temperature, transfer time, and vacuum level of the corresponding area are adjusted based on the average penetration depth deviation of that area. If the penetration depth is insufficient, the heating temperature is increased, the transfer time is extended, or the vacuum level is increased; if the penetration depth is too deep, the heating temperature is decreased, the transfer time is shortened, or the vacuum level is decreased.

[0112] Parameter smoothing. To avoid excessive parameter differences between adjacent areas that could lead to obvious boundaries in the transfer effect, the adjusted parameters are smoothed. A Gaussian filter algorithm is used to filter the parameter distribution map, ensuring that the parameters vary continuously in space.

[0113] Parameter boundary limits. To ensure process safety and normal equipment operation, boundary limits are imposed on the adjusted parameters. For example, the heating temperature must not exceed 300℃, the transfer time must not exceed 60 minutes, and the vacuum degree must not be lower than -0.1MPa.

[0114] Generate revised process control parameters. Integrate the adjusted color map, dye quantity, heating temperature, vacuum pressure, transfer time, and target penetration depth into a complete revised process control parameter file.

[0115] This embodiment achieves refined control of transfer quality through a regional parameter adjustment method based on error-parameter mapping, which can specifically solve transfer quality problems in different areas and significantly improve the consistency of transfer quality.

[0116] In some embodiments, the step of inputting error data and corrected process control parameters into the machine learning prediction model and updating the parameters of the machine learning prediction model includes: using error data and corrected process control parameters as training samples; inputting the training samples into the machine learning prediction model; combining pre-established heat conduction models, vacuum pressure models, dye gasification diffusion models, and mineral material adsorption and permeation models to physically constrain and adjust the weights of the machine learning prediction model; and completing the parameter update of the machine learning prediction model.

[0117] In this embodiment, the step of inputting error data and corrected process control parameters into the machine learning prediction model to update the model parameters specifically includes: Training sample generation. The input data from this transfer process (image features, material parameters, equipment parameters) is used as the model input, the corrected process control parameters are used as the model's expected output, and the error data is used as the sample weight coefficients to generate a new training sample. The sample weight coefficients are proportional to the error magnitude; the larger the error, the higher the sample weight, and the more the model will focus on that sample during training.

[0118] Training set update. Newly generated training samples are added to the model's training dataset. To prevent the training set from becoming too large and slowing down the training process, a sliding window mechanism is used, retaining the most recent 10,000 training samples and deleting the oldest samples.

[0119] Incremental model training. The model is retrained using incremental learning. The specific steps are as follows: Freeze the parameters of the first two layers of the model and train only the subsequent layers; The mini-batch gradient descent algorithm is used for training, with a batch size of 32. The learning rate is set to 1 / 10 of the initial learning rate to avoid catastrophic forgetting in the model; The training iterations are 100 times, and training is stopped early when the loss on the validation set no longer decreases; physical constraints are introduced. During model training, a pre-established physical model is introduced as a constraint to correct the model's output, specifically including: Heat conduction model: Based on Fourier's law of heat conduction, calculate the temperature distribution inside the material to ensure that the heating temperature and transfer time predicted by the model can make the material reach the required temperature; Dye diffusion model: Based on Fick's diffusion law, the diffusion process of dye inside the material is calculated to ensure that the penetration depth predicted by the model conforms to the diffusion law; Vacuum pressure model: Based on the ideal gas law, calculate the change process of vacuum pressure inside the furnace to ensure that the vacuum pressure predicted by the model can be realized; Model evaluation and saving. The updated model is evaluated using an independent test set, and the mean absolute error between the model's predicted parameters and the actual optimal parameters is calculated. If the model's accuracy meets the requirements (mean absolute error less than 5%), the updated model parameters are saved; otherwise, training continues with additional training samples.

[0120] This embodiment uses a model update method that combines incremental learning and physical constraints, enabling the model to continuously learn from actual production data, thereby continuously improving prediction accuracy. At the same time, it ensures the physical rationality of the model's prediction results, greatly improving the reliability and practicality of the model.

[0121] Please see Figure 2 As shown, Figure 2This is a schematic diagram of the structure of a machine learning-based closed-loop control system 200 for vacuum thermal transfer printing of mineral surfaces provided in this application embodiment. This machine learning-based closed-loop control system 200 is used to execute the steps of the machine learning-based closed-loop control method for vacuum thermal transfer printing of mineral surfaces shown in the above embodiments. The machine learning-based closed-loop control system 200 can be a single server or a server cluster, or it can be a terminal, such as a handheld terminal, laptop computer, wearable device, or robot.

[0122] like Figure 2 As shown, the machine learning-based closed-loop control system 200 for vacuum thermal transfer of mineral surfaces includes: Image acquisition unit 201 is used to acquire target digital images and extract color distribution, texture features and pigment density features of target digital images; The process prediction and control unit 202 is used to generate initial process control parameters based on mineral material parameters, dye characteristics, vacuum heating equipment parameters, and sensor operating parameters through a machine learning prediction model. The initial process control parameters include dye amount, heating temperature, vacuum pressure, transfer time, color mapping relationship, and target penetration depth. The unit then performs thermal transfer on the surface of the mineral material under vacuum and heating conditions according to the initial process control parameters. The result acquisition unit is used to acquire the actual image results and / or actual penetration effect after the transfer is completed. The model update unit 203 is used to compare the actual image results and / or the actual penetration effect with the target digital image, calculate the color difference, structural similarity and penetration depth deviation, and generate error data; automatically adjust the dye amount, heating temperature, vacuum pressure, transfer time and target penetration depth according to the error data, and generate corrected process control parameters; input the error data and corrected process control parameters into the machine learning prediction model to update the parameters of the machine learning prediction model.

[0123] It should be noted that those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the above-described machine learning-based closed-loop control system for vacuum thermal transfer of mineral surfaces and its modules can be found in the corresponding embodiments of the above-described machine learning-based closed-loop control method for vacuum thermal transfer of mineral surfaces, and will not be repeated here.

[0124] The aforementioned closed-loop control method for vacuum thermal transfer on mineral surfaces based on machine learning can be implemented as a computer program, which can be used in various applications such as... Figure 2 It runs on the device shown.

[0125] Please see Figure 3 , Figure 3This is a schematic block diagram of the structure of a computer device provided in an embodiment of this application. The computer device includes a processor, a memory, and a network interface connected via a device bus, wherein the memory may include a storage medium and internal memory.

[0126] The storage medium can store operating devices and computer programs. The computer program includes program instructions that, when executed, cause the processor to perform any machine learning-based closed-loop control method for vacuum thermal transfer on mineral surfaces.

[0127] The processor provides computing and control capabilities, supporting the operation of the entire computer device.

[0128] The internal memory provides an environment for the execution of computer programs in non-volatile storage media. When the computer program is executed by the processor, it enables the processor to execute any machine learning-based closed-loop control method for vacuum thermal transfer of mineral surfaces.

[0129] This network interface is used for network communication, such as sending assigned tasks. Those skilled in the art will understand that... Figure 3 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the terminal to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0130] It should be understood that the processor can be a Central Processing Unit (CPU), but it can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among these, a general-purpose processor can be a microprocessor or any conventional processor.

[0131] In one embodiment, the processor is configured to run a computer program stored in memory to perform the following steps: Acquire the target digital image; extract the color distribution, texture features, and pigment density features of the target digital image; Material parameters of the mineral material are acquired, including material type, thickness, density, thermal conductivity, dye adsorption / diffusion characteristics, and surface pretreatment parameters. Operating parameters of the vacuum heating equipment, transfer equipment, vacuum pressure sensor, and temperature sensor are also acquired. The image features, material parameters, and equipment operating parameters are input into a machine learning prediction model, which predicts initial process control parameters, including dye amount, heating temperature, vacuum pressure, transfer time, color mapping relationship, and the target penetration depth of the dye within the mineral material. Based on these initial process control parameters, thermal transfer is performed on the surface of the mineral material under vacuum and heating conditions, achieving seamless integrated transfer across all six sides and internal pores of the stone. The dye is vaporized / diffused under heat and penetrates into the mineral material. Actual image results and / or actual penetration effects are collected after the transfer is completed. The actual image results and / or actual penetration effect are compared with the target digital image to calculate color difference, structural similarity, and penetration depth deviation, generating error data. Based on the error data, the dye amount, heating temperature, vacuum pressure, transfer time, and target penetration depth are automatically adjusted to generate corrected process control parameters. The error data and corrected process control parameters are input into the machine learning prediction model to update the parameters of the machine learning prediction model.

[0132] In some embodiments, acquiring the target digital image includes: receiving a digital image file in a preset image format; converting the received digital image file to a format to generate a target digital image in a uniform format; and performing noise removal and resolution unification processing on the target digital image.

[0133] In some embodiments, the extracted color distribution, texture features, and pigment density features of the target digital image include: extracting the red, green, and blue channel values ​​of each pixel in the target digital image; converting them to generate color distribution data for each pixel; extracting texture feature data of the target digital image; performing region segmentation on the target digital image; calculating the pigment deposition density of each segmented region; and forming pigment density feature data.

[0134] In some embodiments, acquiring the material parameters of the mineral material, including material type, thickness, density, thermal conductivity, dye adsorption / diffusion characteristics, and surface pretreatment parameters, and acquiring the operating parameters of the vacuum heating equipment, transfer equipment, vacuum pressure sensor, and temperature sensor, includes: retrieving the material parameters of the corresponding mineral material from a pre-stored mineral material database, the material parameters including the reflectivity, gloss, porosity, thickness, density, thermal conductivity, dye adsorption coefficient, dye diffusion coefficient, target penetration depth range, and surface pretreatment parameters of the material surface; retrieving the equipment parameters of the corresponding vacuum heat transfer equipment from a pre-stored equipment database, the equipment parameters including the printhead configuration of the transfer equipment, dye characteristics, vacuum heating furnace temperature range, vacuum pressure range, transfer time range, heating / cooling curve, and sensor acquisition parameters; inputting the extracted image features and the retrieved material and equipment parameters into a machine learning prediction model; and generating an initial color map, initial dye amount, initial heating temperature, initial vacuum pressure, initial transfer time, and target penetration depth through the machine learning prediction model.

[0135] In some embodiments, the step of performing heat transfer printing on the surface of mineral materials under vacuum and heating conditions according to the initial process control parameters to achieve seamless integrated transfer printing on the six sides and internal pores of the stone includes: converting the color data of the target digital image according to the initial color map table and controlling the dye output according to the initial dye amount; controlling the operation of the vacuum heat transfer printing equipment according to the initial heating temperature, initial vacuum pressure and initial transfer time; performing vacuum heat transfer printing on the surface of the mineral materials, so that the dye vaporizes, diffuses and enters the interior of the mineral materials under vacuum and high temperature environment, completing the image transfer and forming a textured structure with depth and layering.

[0136] In some embodiments, acquiring the actual image result and / or actual penetration effect after the transfer is completed includes: taking a picture of the material surface after the transfer is completed to obtain visible light data of the actual image; scanning the material surface after the transfer is completed to obtain spectral data of the actual image; and integrating the visible light data and spectral data into the actual image result and / or actual penetration effect.

[0137] In some embodiments, comparing the actual image result and / or the actual penetration effect with the target digital image, calculating color difference, structural similarity, and penetration depth deviation, and generating error data includes: aligning the actual image result with the target digital image pixel by pixel; calculating the color difference value of the corresponding pixel; calculating the brightness similarity, contrast similarity, and structural similarity between the actual image result and the target digital image; generating difference distribution data based on the color difference value and structural similarity, and marking areas where the difference exceeds a preset threshold.

[0138] In some embodiments, the step of automatically adjusting the dye amount, heating temperature, vacuum pressure, transfer time, and target penetration depth based on error data to generate corrected process control parameters includes: determining the process parameter deviation corresponding to each difference region based on the difference distribution data and penetration depth deviation; adjusting the corresponding value of the color mapping table and the dye amount of the corresponding region for each difference region; adjusting the heating temperature, vacuum pressure, transfer time, and target penetration depth of the corresponding region; and integrating all adjusted parameters to generate corrected process control parameters.

[0139] In some embodiments, the step of inputting error data and corrected process control parameters into the machine learning prediction model and updating the parameters of the machine learning prediction model includes: using error data and corrected process control parameters as training samples; inputting the training samples into the machine learning prediction model; combining pre-established heat conduction models, vacuum pressure models, dye gasification diffusion models, and mineral material adsorption and permeation models to physically constrain and adjust the weights of the machine learning prediction model; and completing the parameter update of the machine learning prediction model.

[0140] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, causes the processor to implement the steps of the machine learning-based closed-loop control method for vacuum thermal transfer of mineral surfaces as provided in any embodiment of this application.

[0141] The computer-readable storage medium may be an internal storage unit of the computer device described in the foregoing embodiments, such as the hard disk or memory of the computer device. The computer-readable storage medium may also be an external storage device of the computer device, such as a plug-in hard disk, SmartMedia Card (SMC), Secure Digital (SD) card, or Flash Card equipped on the computer device.

[0142] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and these modifications or substitutions should all be covered within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A closed-loop control method for vacuum thermal transfer on mineral surfaces based on machine learning, characterized in that, include: Acquire the target digital image; extract the color distribution, texture features, and pigment density features of the target digital image; Material parameters of the mineral material are acquired, including material type, thickness, density, thermal conductivity, dye adsorption / diffusion characteristics, and surface pretreatment parameters. Operating parameters of the vacuum heating equipment, transfer equipment, vacuum pressure sensor, and temperature sensor are also acquired. The image features, material parameters, and equipment operating parameters are input into a machine learning prediction model, which predicts initial process control parameters, including dye amount, heating temperature, vacuum pressure, transfer time, color mapping relationship, and the target penetration depth of the dye within the mineral material. Based on these initial process control parameters, thermal transfer is performed on the surface of the mineral material under vacuum and heating conditions, achieving seamless integrated transfer across all six sides and internal pores of the stone. The dye is vaporized / diffused under heat and penetrates into the mineral material. Actual image results and / or actual penetration effects are collected after the transfer is completed. Compare the actual image results and / or actual penetration effect with the target digital image, calculate the color difference, structural similarity and penetration depth deviation, and generate error data; The system automatically adjusts the dye amount, heating temperature, vacuum pressure, transfer time, and target penetration depth based on error data to generate corrected process control parameters. The error data and corrected process control parameters are then input into the machine learning prediction model to update the model's parameters.

2. The method according to claim 1, characterized in that, The acquisition of the target digital image includes: Receives digital image files in a preset image format; The received digital image files are converted to generate target digital images in a uniform format. The target digital image is subjected to noise removal and resolution unification processing.

3. The method according to claim 1, characterized in that, The extracted color distribution, texture features, and pigment density features of the target digital image include: Extract the red, green, and blue channel values ​​of each pixel in the target digital image; convert and generate color distribution data for each pixel; Extract texture feature data from the target digital image; The target digital image is segmented into regions, and the pigment deposition density of each segmented region is calculated to form pigment density feature data.

4. The method according to claim 1, characterized in that, The process involves acquiring material parameters for the mineral material, including material type, thickness, density, thermal conductivity, dye adsorption / diffusion characteristics, and surface pretreatment parameters. It also involves acquiring operating parameters for the vacuum heating equipment, transfer equipment, vacuum pressure sensor, and temperature sensor, including: The material parameters of the corresponding mineral material are retrieved from the pre-stored mineral material database. The material parameters include the reflectivity, gloss, porosity, thickness, density, thermal conductivity, dye adsorption coefficient, dye diffusion coefficient, target penetration depth range, and surface pretreatment parameters of the material surface. The equipment parameters of the corresponding vacuum heat transfer equipment are retrieved from the pre-stored equipment database. The equipment parameters include the nozzle configuration of the transfer equipment, dye characteristics, temperature range of the vacuum heating furnace, vacuum pressure range, transfer time range, heating / cooling curve, and sensor-collected parameters. The extracted image features, along with the retrieved material and equipment parameters, are input into the machine learning prediction model. The initial color map, initial dye amount, initial heating temperature, initial vacuum pressure, initial transfer time, and target penetration depth are generated using a machine learning prediction model.

5. The method according to claim 1, characterized in that, The process involves performing heat transfer printing on the surface of the mineral material under vacuum and heating conditions according to the initial process control parameters, achieving seamless integrated transfer printing on all six sides of the stone and its internal pores, including: The color data of the target digital image is converted according to the initial color map table, and the dye output is controlled according to the initial dye amount. The operation of the vacuum heat transfer equipment is controlled based on the initial heating temperature, initial vacuum pressure, and initial transfer time. Vacuum thermal transfer is performed on the surface of mineral materials, causing the dye to vaporize, diffuse, and penetrate into the interior of the mineral materials under vacuum and high temperature conditions, thus completing the image transfer and forming a textured structure with depth and layers.

6. The method according to claim 1, characterized in that, The actual image results and / or actual penetration effect after the transfer are acquired include: The surface of the transferred material is photographed to obtain visible light data of the actual image; The surface of the transferred material is scanned to obtain the spectral data of the actual image; Integrate visible light data and spectral data into actual image results and / or actual penetration effects.

7. The method according to claim 1, characterized in that, The step of comparing the actual image results and / or the actual penetration effect with the target digital image, calculating color difference, structural similarity, and penetration depth deviation, and generating error data includes: Align the actual image result with the target digital image pixel by pixel; Calculate the color difference value of the corresponding pixels; Calculate the brightness similarity, contrast similarity, and structural similarity between the actual image result and the target digital image; Generate difference distribution data based on color difference values ​​and structural similarity, and mark areas where the difference exceeds a preset threshold.

8. The method according to claim 1, characterized in that, The process of automatically adjusting dye quantity, heating temperature, vacuum pressure, transfer time, and target penetration depth based on error data to generate corrected process control parameters includes: Based on the differential distribution data and penetration depth deviation, determine the process parameter deviation corresponding to each differential region; For each area of ​​difference, adjust the corresponding values ​​in the color map and the amount of dye in that area; Adjust the heating temperature, vacuum pressure, transfer time, and target penetration depth of the corresponding area; Integrate all adjusted parameters to generate revised process control parameters.

9. The method according to claim 1, characterized in that, The step of inputting error data and corrected process control parameters into the machine learning prediction model to update the parameters of the machine learning prediction model includes: The error data and the corrected process control parameters were used as training samples. Input the training samples into the machine learning prediction model; By combining pre-established heat conduction models, vacuum pressure models, dye vaporization and diffusion models, and mineral material adsorption and permeation models, the weights of the machine learning prediction model are physically constrained and adjusted. Complete the parameter update of the machine learning prediction model.

10. A closed-loop control system for vacuum thermal transfer printing on mineral surfaces based on machine learning, used to implement the method as described in any one of claims 1-9, characterized in that, include: The image acquisition unit is used to acquire the target digital image and extract the color distribution, texture features, and pigment density features of the target digital image. The process prediction and control unit is used to generate initial process control parameters based on mineral material parameters, dye characteristics, vacuum heating equipment parameters, and sensor operating parameters through a machine learning prediction model. The initial process control parameters include dye amount, heating temperature, vacuum pressure, transfer time, color mapping relationship, and target penetration depth. The unit then performs thermal transfer on the surface of the mineral material under vacuum and heating conditions according to the initial process control parameters. The result acquisition unit is used to acquire the actual image results and / or actual penetration effect after the transfer is completed; The model update unit is used to compare the actual image results and / or the actual penetration effect with the target digital image, calculate the color difference, structural similarity and penetration depth deviation, and generate error data. The system automatically adjusts the dye amount, heating temperature, vacuum pressure, transfer time, and target penetration depth based on error data to generate corrected process control parameters. The error data and corrected process control parameters are then input into the machine learning prediction model to update the model's parameters.

Images