Skin-touch UV ink generation method based on deep learning and skin-touch UV ink

Abstract

The invention relates to the technical field of deep learning, ink production and the like, and provides a skin-touch UV ink generation method based on deep learning and skin-touch UV ink, raw material composition data and performance parameter data are obtained and then are respectively input into corresponding input layers of a deep learning model, the deep learning model is iteratively trained through a back propagation algorithm, and the skin-touch UV ink is generated. The trained model can output the optimal raw material feeding ratio, the optimal raw material feeding ratio is applied to production, including feeding mixing, dispersion machine stirring and auxiliary agent dropwise adding, UV curing testing is carried out after stirring is completed, the deep learning model is updated again according to test feedback, and the optimal raw material feeding ratio is obtained. And finally, the skin-touch UV ink with expected curing speed, adhesive force and hand feeling performance is generated, so that the efficiency and accuracy of analyzing the multi-parameter coupling relationship of the skin-touch UV ink are improved, and the inconsistency of ink batches and the performance fluctuation are avoided.

Description

Technical Field

[0001] The present invention relates to the technical fields of deep learning, ink production, etc., and particularly relates to a method for generating skin-feel UV ink based on deep learning and skin-feel UV ink. Background Art

[0002] With the continuous development of printing technology, UV ink has been widely used in various high-quality packaging and special printing. However, traditional UV ink formulations often need to balance various resins, additives, and photoinitiators to meet multi-dimensional performance requirements such as curing speed, adhesion, abrasion resistance, and handfeel. Especially in fields such as cigarette packages and outer packages of cosmetics, which have high requirements for both touch and visual effects, additional "skin-feel" function improvements are often required for the ink, including adding specific proportions of matte resin, easy-to-extinguish resin, and touch resin. However, in the existing process, the proportioning of these raw materials mostly relies on manual experience or limited trial-and-error processes, which easily leads to insufficient control of key parameters in complex formulations and makes it difficult to balance fast curing, high adhesion, and comfortable handfeel in a timely manner. Once there is a deviation in a certain formulation ratio, it may lead to insufficient curing, decreased adhesion, or unqualified abrasion resistance, and even problems with poor handfeel. In addition, with the continuous improvement of customer requirements, in more cases, it is necessary to repeatedly adjust or customize in small batches in a short time to adapt to the appearance and function requirements of different products. However, traditional methods are often costly and time-consuming, lacking flexibility and accuracy. Based on the above situation, how to use intelligent means to quickly iterate and optimize the feeding ratios of various raw materials has become a technical problem to be solved urgently. On the one hand, it is necessary to extract feasible feeding rules from a large amount of historical formulation and performance test data to avoid blind manual tests; on the other hand, it is necessary to achieve a comprehensive balance among goals such as curing speed, adhesion, abrasion resistance, and handfeel score to take into account different performance requirements. In the traditional formulation adjustment mode, it is difficult for humans to accurately analyze the multi-parameter coupling relationship in a short time, and it is easy to make misjudgments or over-adjustments, resulting in inconsistent ink batches and performance fluctuations. Summary of the Invention

[0003] Aiming at the deficiencies of the above-mentioned existing technologies, the present invention provides a method for generating skin-feel UV ink based on deep learning to improve the efficiency and accuracy of analyzing the multi-parameter coupling relationship of skin-feel UV ink and avoid inconsistent ink batches and performance fluctuations.

[0004] In the first aspect, the present invention provides a method for generating skin-feel UV ink based on deep learning, including:

[0005] Obtain the raw material composition data and performance parameter data of the skin-friendly UV ink. The raw material composition data includes the weight percentages of polyester matte resin, polyurethane easy-to-extinguish resin, skin-friendly resin, photoinitiator, slip aid, and wax aid. The performance parameter data includes curing speed, adhesion level, abrasion resistance times, and handfeel score;

[0006] Establish a deep learning model of a multi-layer neural network, and input the raw material composition data and the performance parameter data into the corresponding input layer of the deep learning model respectively; perform iterative training on the deep learning model through the backpropagation algorithm so that the trained model can output the optimal raw material feeding ratio;

[0007] Apply the optimal raw material feeding ratio to production, including feeding and mixing, stirring with a disperser, and dropping additives, and conduct a UV curing test after the stirring is completed. Update the deep learning model again according to the test feedback, and finally generate a skin-friendly UV ink with expected curing speed, adhesion, and handfeel performance.

[0008] In a second aspect, the present invention provides a skin-friendly UV ink produced by using the above-mentioned method for generating a skin-friendly UV ink based on deep learning.

[0009] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0010] The present invention provides a method for generating a skin-friendly UV ink and a skin-friendly UV ink based on deep learning. By obtaining the raw material composition data and performance parameter data of the skin-friendly UV ink, the raw material composition data includes the weight percentages of polyester matte resin, polyurethane easy-to-extinguish resin, skin-friendly resin, photoinitiator, slip aid, and wax aid, and the performance parameter data includes curing speed, adhesion level, abrasion resistance times, and handfeel score, establish a deep learning model of a multi-layer neural network, and input the raw material composition data and the performance parameter data into the corresponding input layer of the deep learning model respectively; perform iterative training on the deep learning model through the backpropagation algorithm so that the trained model can output the optimal raw material feeding ratio, apply the optimal raw material feeding ratio to production, including feeding and mixing, stirring with a disperser, and dropping additives, and conduct a UV curing test after the stirring is completed. Update the deep learning model again according to the test feedback, and finally generate a skin-friendly UV ink with expected curing speed, adhesion, and handfeel performance, thereby improving the efficiency and accuracy of analyzing the multi-parameter coupling relationship of the skin-friendly UV ink and avoiding ink batch inconsistency and performance fluctuations. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] The accompanying drawings described herein are used to provide a further understanding of the present invention and form a part of the present invention. The schematic embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. Some specific embodiments of the present invention will be described in detail hereinafter with reference to the accompanying drawings in an exemplary rather than restrictive manner. The same reference numerals in the drawings denote the same or similar components or parts. Those skilled in the art should understand that these drawings are not necessarily drawn to scale. In the drawings:

[0012] Figure 1 is a schematic flow chart of a method for generating skin-friendly UV ink based on deep learning according to an embodiment of the present invention. Detailed implementation manners

[0013] In order to enable those skilled in the art to better understand the solution of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0014] Embodiment 1

[0015] See Figure 1 , this embodiment provides a method for generating skin-friendly UV ink based on deep learning, including the following steps:

[0016] S101. Obtain the raw material composition data and performance parameter data of the skin-friendly UV ink. The raw material composition data includes the weight percentages of polyester matte resin, polyurethane easy-to-extinguish resin, skin-friendly resin, photoinitiator, slip aid, and wax aid. The performance parameter data includes curing speed, adhesion level, abrasion resistance times, and hand feeling score. For example, the weight percentages of polyester matte resin, polyurethane easy-to-extinguish resin, skin-friendly resin, photoinitiator, slip aid, and wax aid meet the range of 40%-60% for polyester matte resin, 10%-25% for polyurethane easy-to-extinguish resin, 20%-30% for skin-friendly resin, 4%-5% for photoinitiator, 2%-5% for slip aid, and 2%-3% for wax aid;

[0017] S102. Establish a deep learning model of a multi-layer neural network, and input the raw material composition data and the performance parameter data into the corresponding input layer of the deep learning model respectively; perform iterative training on the deep learning model through the backpropagation algorithm so that the trained model can output the optimal raw material feeding ratio;

[0018] S103. Apply the optimal raw material feeding ratio in production, including feeding and mixing, stirring with a disperser, and dropping additives, and conduct a UV curing test after stirring is completed. Update the deep learning model again according to the test feedback, and finally generate a skin-friendly UV ink with the expected curing speed, adhesion, and feel performance.

[0019] It should be noted that the method for generating skin-friendly UV ink based on deep learning proposed in this embodiment can realize the automatic prediction and output of the optimal raw material feeding ratio through a multi-layer neural network model in a scenario where the ratios of multiple raw materials (polyester matte resin, polyurethane easy-to-extinguish resin, skin-friendly resin, photoinitiator, slip additive, wax additive) are complex and the performance targets (curing speed, adhesion, abrasion resistance, feel score) affect each other. In traditional formula design, the ratio is often determined only by experience or limited tests, and it is difficult to take into account multiple performance dimensions in a timely manner. In this embodiment, iterative training using a deep learning model can effectively extract the potential rules in a large amount of data. In this embodiment, by obtaining the raw material composition data and performance parameter data and inputting these data into a deep learning model (for example, a neural network), continuous iteration is performed through the backpropagation algorithm until the model can output the optimal formula. After verification through actual production and UV curing tests, it is updated again to achieve the closed-loop optimization of the skin-friendly UV ink. In this embodiment, the entire process from data collection and modeling, to training and output, and then to actual production and feedback solves the formula uncertainty caused by multi-parameter coupling, thereby improving the efficiency and accuracy of formula design.

[0020] In some preferred embodiments, before obtaining the raw material composition data and performance parameter data, it further includes establishing a sample collection subsystem. The sample collection subsystem conducts experimental measurements on multiple batches of skin-friendly UV inks to obtain accurate experimental measurement data. The experimental measurement data includes the curing speed, adhesion level, abrasion resistance times, and handfeel scores corresponding to different ratios of polyester matte resin, polyurethane easy-to-extinguish resin, skin-friendly resin, photoinitiator, slip aid, and wax aid. During the experimental measurement, a unified test standard is used for measurement and the temperature and humidity are stably controlled. The measured data is transmitted to a database for summary and collation, and the sorted experimental measurement data is input into the training set and validation set of the deep learning model to improve the generalization ability and prediction accuracy of the model. After the deep learning model is trained, the model parameters are fixed and the first-round optimal formulation suggestions are output to enter the subsequent actual production verification and retraining stage. It should be noted that in this embodiment, the step of establishing a sample collection subsystem is added before obtaining the raw material composition data and performance parameter data to conduct experimental measurements on multiple batches of skin-friendly UV inks to form a training set and a validation set. The reason for this setting is that the performance of the deep learning model is closely related to the quality of the data it uses, and traditional manual tests often lack systematic collection and management, resulting in incomplete or inaccurate model training data. In this embodiment, the experimental measurement data is input into the model after being collated in a database through a unified test standard and stably controlled temperature and humidity, ensuring the reliability and consistency of the data. At the same time, by fixing the model parameters and outputting the first-round optimal formulation suggestions, it can be achieved that after the initial training is completed, the model has a certain generalization ability and prediction accuracy, forming a preliminary formulation to lay a foundation for subsequent retraining if needed.

[0021] In some preferred embodiments, when establishing a deep learning model of a multi-layer neural network, an input layer is set to receive raw material composition data and performance parameter data, and at least three hidden layers are set to construct a feature extraction and non-linear mapping structure; residual connections are introduced in the hidden layers to avoid the phenomenon of gradient vanishing or exploding when the number of network layers is relatively deep; a multi-objective regression unit is established in the output layer to simultaneously predict the curing speed, adhesion level, abrasion resistance times, and hand feeling score of the skin-friendly UV ink; during the training process, a mean square error loss function is adopted, and L2 regularization is combined to constrain the weights to prevent overfitting; through training, the deep learning model can accurately predict the corresponding performance indicators according to the feeding ratio, and realize the intelligent optimization of the ratio of polyester matte resin, polyurethane easy-to-extinguish resin, skin-friendly resin, photoinitiator, slip aid, and wax aid. It should be noted that in this embodiment, it is clearly specified that at least three hidden layers need to be set when establishing a multi-layer neural network, and residual connections are introduced in the hidden layers, and means such as multi-objective regression units, mean square error loss functions, and L2 regularization are combined. The reason for such settings is as follows: there are often highly non-linear correlations among the various performance indicators of the skin-friendly UV ink, and it is difficult for a simple shallow network model to capture complex coupling relationships. At the same time, residual connections can effectively alleviate the problem of gradient vanishing or exploding in deep networks and enhance the sustainability of training. The design of a multi-objective regression unit in the output layer is because four indicators, namely curing speed, adhesion level, abrasion resistance times, and hand feeling score, need to be predicted simultaneously, and then a comprehensive trade-off is made. Adopting the mean square error loss function and L2 regularization can reduce the risk of overfitting and enable the model to maintain good generalization ability when facing various raw material ratios. Therefore, this embodiment clearly stipulates the network structure, training strategy, and loss form, so as to solve the technical problem of how to extract key features in a high-dimensional and multi-objective ink formula and ensure stable training, and enhance the accuracy and reliability of model prediction.

[0022] In some preferred embodiments, during the stirring process of the feeding mixer and disperser, an on-line monitoring module is added to detect the stirring temperature, viscosity and rotation speed in real time, and the detection results are input into the process monitoring sub-network in the deep learning model; the process monitoring sub-network compares the difference between the predicted viscosity and the actual viscosity in real time. If the deviation exceeds the threshold, the rotation speed of the disperser is automatically adjusted or the dropping speed of the auxiliary agent is changed; after the smoothness auxiliary agent and the wax auxiliary agent are dropped, a layered stirring method is adopted for secondary mixing, and the curing efficiency is monitored by UV curing pre-testing; finally, the measured data is sent back to the deep learning model again for fine-tuning, so that the generated skin-friendly UV ink maintains consistency and controllability during the production process, and avoids formula inaccuracy caused by temperature changes or uneven stirring. It should be noted that in this embodiment, an on-line monitoring module is added, and the information such as the stirring temperature, viscosity and rotation speed detected in real time is input into the process monitoring sub-network in the deep learning model, and the rotation speed of the disperser or the dropping speed of the auxiliary agent is automatically adjusted when the deviation exceeds the threshold. The reason for this setting is that: even under the same mixing ratio conditions, changes in environmental factors such as temperature and humidity at the production site will cause the stirring process to be unstable, thus causing fluctuations in the ink performance. If only relying on off-line detection, it is difficult to detect and correct deviations in time. In this embodiment, the process monitoring sub-network is used to achieve on-line closed-loop adjustment. Once a significant deviation appears in the detection results, the production process parameters can be automatically intervened to keep the formula adjustment within the target range. Through such on-line monitoring and feedback, the quality inconsistency caused by uneven stirring or temperature fluctuations can be significantly reduced, and the measured data is sent back to the deep learning model for fine-tuning, forming a cycle of continuous learning and correction, so as to effectively solve the problem of formula inaccuracy of the deep learning caused by fluctuations in environmental factors at the production site, and significantly improve the industrial feasibility and robustness of the entire method.

[0023] In some preferred embodiments, during the training process of the deep learning model, four sub-goals of curing speed, adhesion level, abrasion resistance times, and feel score are respectively set, and a weighted multi-objective optimization strategy is used, so that each sub-goal has a corresponding weight coefficient; through iterative training, multiple rounds of tests are carried out on different weight coefficient configurations to screen out a comprehensive formula that not only meets fast curing but also ensures high adhesion and good feel; the weighted multi-objective optimization strategy generates a feasible solution set at the output layer of the model, and then selects the optimal solution according to the pre-set objective sorting rules; under this optimal solution, a set of recommended ratios of polyester matte resin, polyurethane easy-to-extinguish resin, skin-friendly resin, photoinitiator, slip aid, and wax aid are obtained in the feeding link to balance the requirements of fast printing and durable feel. It should be noted that for skin-friendly UV inks, indicators such as fast curing and high adhesion often conflict with each other, and it is not easy to balance speed and performance. The multi-objective optimization strategy that takes each indicator as a sub-goal and gives different weights helps to find an optimal solution set on the basis of comprehensive balance, and then screen out the final formula according to the pre-set priority, so as to prevent a certain indicator from being ignored and avoid blindly pursuing a certain goal resulting in a serious decline in other performances. Through the weighted multi-objective optimization strategy and screening out a comprehensive formula that not only meets fast curing but also ensures high adhesion and good feel, this embodiment can solve the problem of the mutual influence between multiple performance indicators, enable the model to simultaneously evaluate and weigh the priorities of different indicators, and thus output a globally optimal or approximately optimal feeding plan, making the deep learning model able to adapt to diverse requirements and conform to the actual industrial production.

[0024] In some preferred embodiments, before outputting the optimal raw material feeding ratio, a simulation verification stage is first carried out. The generated formula parameters are input into a simulation module with physical constraints. The simulation module calculates the theoretical curing efficiency and surface tension according to the known kinetic models of polyester matte resin, polyurethane easy-to-extinguish resin, skin-friendly resin, photoinitiator, slip aid, and wax aid in the chemical reaction and photo-curing process; if the simulation result deviates from the result predicted by the deep learning model, the deviation information is fed back to the model for secondary training until the simulation result coincides with the prediction result within the preset error range. It should be noted that the deep learning model can sometimes only capture the statistical correlations at the data level. When encountering extreme situations or special conditions not covered by the training data, the model prediction may deviate. Introducing a simulation module of chemical and photo-curing kinetics can review the rationality of the formula from the mechanism level. Once the simulation result differs greatly from the model prediction result, it will be fed back to the model for correction. This way of combining data-driven and mechanism models can effectively reduce blind spots, improve the credibility of formula prediction, and at the same time save the losses that may be caused by direct tests at the production site, and solve the problem of deviation out of control that may be brought about by simply relying on statistical models.

[0025] In some preferred embodiments, a handfeel scoring sub-network is set up in the deep learning model to separately extract key features related to the skin-friendly resin content and the types of slip additives; the handfeel scoring sub-network receives handfeel scoring training samples and uses the attention mechanism to weight the main features affecting the handfeel; the output of the handfeel scoring sub-network is combined with the remaining prediction results of the main model of the deep learning model to obtain the final score, so that the feeding ratio can more accurately match the excellent tactile performance; during production verification, a standardized blind test touch evaluation process is adopted, a fixed number of evaluators are invited to score, and the obtained scores are input into the handfeel scoring sub-network again for error backtracking correction to achieve refined control of the skin-friendly effect. It should be noted that compared with quantitatively measurable indicators such as curing speed or wear resistance, handfeel scoring is more subjective and diverse, and the relationship with raw materials such as skin-friendly resins and slip additives is more complex. Through a separate handfeel scoring sub-network, the tactile performance can be more focused, avoiding being overwhelmed by other indicators in the comprehensive model. Its combination with the output of the main model gives the final score, and in production verification, it can be combined with blind test touch evaluation, and the feedback scores are input into the sub-network again for correction, enabling the model to achieve deep learning of the subjective evaluation of the handfeel, making the feeding ratio more precisely match the excellent tactile feel, solving the problem that the handfeel index is often ignored or difficult to quantify under multi-objective prediction, and fully optimizing the tactile experience of the finally produced skin-friendly UV ink.

[0026] In some preferred embodiments, when adjusting the proportion of slip additives and wax additives, a chemical surface energy calculation model is introduced, and the chemical surface energy calculation model is used to estimate the influence of different additive combinations on the surface friction coefficient of cigarette pack prints; in the training samples, the high friction coefficient and the scratch rate are stored correspondingly, and a penalty term is set for the friction coefficient prediction error during training; when the additive combination leads to potential scratching risks, the chemical surface energy calculation model increases the loss value of the additive combination scheme in the output layer, prompting the final formulation to tend to reduce the friction coefficient. It should be noted that in production practice, too high a surface friction coefficient will cause white scratches or abrasions on the finished cigarette pack, affecting the packaging quality. And slip additives and wax additives can determine the surface lubricity and anti-friction performance. By incorporating the chemical surface energy calculation model into the deep learning system, the performance of additive combinations in real usage scenarios can be evaluated more precisely, solving the problem of inappropriate scratch rates caused by relying solely on experience in the past. In addition, penalizing the friction coefficient prediction error can also actively tend to better combinations during training, thereby minimizing the scratching risk to the greatest extent while ensuring wear resistance.

[0027] In some preferred embodiments, when performing backpropagation training on the raw material feeding ratio, an automatic learning rate adjustment mechanism is introduced to dynamically modify the learning rate according to the convergence speed of the loss function during the training process. If the loss value decreases slowly within several iterations, the learning rate is automatically increased to accelerate convergence. If the loss value oscillates or increases, the learning rate is reduced to avoid over-updating. It should be noted that in deep learning, the learning rate determines the step size of parameter updates. If it is too large, it may cause the model training to oscillate or even fail to converge. If it is too small, the training iteration speed is too slow and it may fall into a local minimum. Especially in the chemical formula scenario, the parameter space is wide and there are many performance indicators, and a suitable learning rate strategy is required to balance speed and stability. The automatic learning rate adjustment can adaptively increase or decrease the learning rate according to the training curve, avoiding the cumbersome and inaccurate manual parameter tuning. In this way, when the loss function decreases slowly, increasing the learning rate can accelerate convergence. When the loss value oscillates violently, reducing the learning rate can stabilize the update step size. This can significantly reduce the number of iterations and improve the model generalization ability, solve the overshooting or undershooting problems that may occur in traditional fixed learning rate training, and make the deep learning process of the feeding ratio more efficient and reliable.

[0028] In some preferred embodiments, the method for generating skin-friendly UV ink based on deep learning further includes: configuring a model update module based on time series to perform regular retraining on the long-term collected feeding and printing data and compare it with the current model parameters. If the model obtained after retraining performs better on the validation set, the old parameters are replaced. If the two tend to be consistent, the current parameters are maintained unchanged. It should be noted that in actual production, raw material batches, environmental conditions, and even market demands may change over time, resulting in the gradual disconnection of existing model parameters from reality. In order to maintain high prediction accuracy in the long term, it is necessary to periodically update the model. The model update module based on time series can perform retraining after collecting a certain amount of new production data and determine whether the new model is more suitable for the current environment. If it is better, the new model is adopted; otherwise, the current model continues to be used. This can continuously keep the deep learning model in sync with production practice and avoid parameter aging. At the same time, it can also prevent model degradation caused by outliers or incorrect measurements in the new data through validation set detection.

[0029] In some preferred embodiments, the adhesion level is taken as a key focus during the back propagation process, and when the adhesion level range exceeds a predefined threshold, the deep learning model gives priority to increasing the precision adjustment of the ratio of polyurethane-containing easy-to-matt resin to polyester matte resin; if the adhesion training error is still high, an additional constraint item is introduced to limit the use of high-slip additives and high-wax additives to prevent excessive improvement of the feel and matt effect at the expense of adhesion performance. It should be noted that adhesion is one of the important indicators of whether UV ink can be firmly bonded to a specific substrate, but it often conflicts with smoothness and skin feel. If you pursue super smoothness and matt effects, sometimes it will cause a decrease in adhesion. This embodiment triggers the priority strategy by setting the adhesion level range threshold during the training process. Once the model detects that the adhesion deviation is too large, it will constrain the main influencing factors (such as matte resin or slip additive), increase the focus on adhesion, and ensure stable printing quality.

[0030] In some preferred embodiments, when measuring the curing speed, the high-precision UV energy meter device is linked to upload the actual irradiation energy, illumination time and surface hardness test results to the training platform in real time; by comparing the deviation between the model predicted curing speed and the measured curing speed, the gradient weight of the photoinitiator ratio in the output layer is adjusted successively; if the deviation continues to occur, the deep learning model re-examines the photoinitiator ratio area in the raw material composition data so as to adjust the photoinitiator ratio to a more reasonable range in the next iteration. It should be noted that the curing speed is one of the important properties of UV ink, which not only affects the printing efficiency, but also determines the final performance of the ink layer. If only relying on model prediction, it is easy to have deviations that are inconsistent with the actual UV irradiation conditions. By introducing a high-precision energy meter and surface hardness test, the real curing condition can be more comprehensively reflected, especially under UV lamps of different intensities and wavelength distributions. If it is found that the amount of photoinitiator added does not match the actual curing speed, it is corrected in time during training or inference. This enables the model to adapt to a variety of UV equipment and lighting conditions more flexibly, solves the problem of inconsistent curing speed with expectations, and ensures that the ink curing performance under different workshop environments or light source conditions remains stable.

[0031] In some preferred embodiments, a color matching subsystem specifically for cigarette packages is configured to calculate the possible optimal ink layer thickness based on the chromaticity value and apparent gloss requirements given in the cigarette package design file, combined with the effect of polyester matte resin and polyurethane easy-matting resin on diffuse reflection of light after UV curing; the color matching subsystem fuses the calculated value with the optimal feed ratio predicted by the deep learning model to obtain a formula that can satisfy both the matte appearance and accurate chromaticity; during actual silk-screen printing, the CI E Lab value of the sample is measured and compared with the target value, and the difference is fed back to the color matching subsystem to continue to correct the fusion result and achieve accurate restoration of the cigarette package color. It should be noted that cigarette package printing usually requires not only stable performance, but also matching specific designs in appearance, such as the chromaticity and glossiness of the brand logo. Matte resin and easy-matting resin will affect the light reflection characteristics, thereby affecting the color visual effect. This embodiment uses a separate color matching subsystem to compare color spaces (such as CIE Lab values) to help the model meet visual needs. If a large difference from the target value is detected during actual silk screen printing, feedback is given to the subsystem, and the fusion result is then corrected, thereby maintaining the matte effect and accurately presenting the specified color, solving the problem of the coupling effect of pigments and resins on appearance and color tone, so that the printed product can obtain consistent and expected color performance.

[0032] In some preferred embodiments, all collected raw material composition data and performance parameter data are subjected to feature selection and dimensionality reduction processing to remove those redundant features that have very low impact on the four indicators of curing speed, adhesion level, wear resistance, and feel score; in this process, a feature selection algorithm based on information gain is used, combined with a dimensionality reduction algorithm based on principal component analysis, and the simplified key features are input into the deep learning model. It should be noted that although deep learning can process high-dimensional data, too many redundant or noisy features will lead to a decrease in training efficiency and even interfere with model convergence. At the same time, a large number of irrelevant or minimally influential features may increase the risk of overfitting. In this embodiment, by providing a feature selection + dimensionality reduction process, the model maintains a high prediction accuracy and stability under a more streamlined data dimension. In this way, the computational efficiency can be improved, and the learning ability of the model for truly key variables (such as resin ratio, photoinitiator dosage) can be strengthened, and the negative effects of data expansion can be avoided, which helps to achieve a better balance between iteration speed and stability in the industrial field.

[0033] In some preferred embodiments, a linkage process of a programmable logic controller is established between a disperser and a UV curing device. The programmable logic controller automatically performs operations such as feeding materials, stirring, and adding additives according to the feeding ratio and the disperser rotation speed suggestion instruction output by the deep learning model. After the addition is completed, the programmable logic controller triggers the UV curing device to perform a preliminary curing test. If the test results show that the curing speed or the adhesion level does not meet the standard, the iterative update program of the model is called again to adjust the feeding ratio and re-execute the feeding and stirring processes. It should be noted that even if the model can give the most accurate feeding ratio, if there is a lack of automated linkage in the production process, the actual execution is likely to deviate due to human errors or lagged operations. In this embodiment, through the programmable logic controller (PLC), the model output can be seamlessly connected with the device instructions, enabling a closed-loop mechanism for feeding, mixing, and curing tests. When the actual test results do not meet the standard, the PLC immediately triggers the iterative update of the model, eliminating the need for manual waiting or repeated operations. The entire process significantly improves production efficiency and consistency, realizing full-process automation from algorithms to industrial equipment and providing an operable path for large-scale industrial applications.

[0034] Embodiment 2

[0035] See Figure 1 , this embodiment provides a skin-feel UV ink. The skin-feel UV ink is produced using the method for generating a skin-feel UV ink based on deep learning in any of the above embodiments. By obtaining the raw material composition data and performance parameter data of the skin-feel UV ink, the raw material composition data includes the weight percentages of polyester matte resin, polyurethane easy-to-extinguish resin, skin-feel resin, photoinitiator, slip aid, and wax aid, and the performance parameter data includes curing speed, adhesion level, abrasion resistance times, and handfeel score. A deep learning model of a multi-layer neural network is established, and the raw material composition data and the performance parameter data are respectively input into the corresponding input layer of the deep learning model. The deep learning model is iteratively trained through the backpropagation algorithm so that the trained model can output the optimal raw material feeding ratio. The optimal raw material feeding ratio is applied to production, including feeding and mixing, disperser stirring, and adding additives, and a UV curing test is performed after the stirring is completed. The deep learning model is updated again according to the test feedback, and finally, a skin-feel UV ink with expected curing speed, adhesion, and handfeel performance is generated, thereby improving the efficiency and accuracy of analyzing the multi-parameter coupling relationship of the skin-feel UV ink and avoiding ink batch inconsistency and performance fluctuations.

[0036] It should be noted that the above embodiments are only preferred specific embodiments of the present invention, and the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed by the present invention should be covered within the protection scope of the present invention. The protection scope of the present invention should be subject to the protection scope of the claims.

Claims

1. A method for generating skin - feeling UV ink based on deep learning, characterized in that, Including: Obtain the raw material composition data and performance parameter data of the skin-feel UV ink. The raw material composition data includes the weight percentages of polyester matte resin, polyurethane easy-to-extinguish resin, skin-feel resin, photoinitiator, slip aid, and wax aid. The performance parameter data includes curing speed, adhesion level, abrasion resistance times, and handfeel score; Establish a deep learning model of a multi-layer neural network, and input the raw material composition data and the performance parameter data into the corresponding input layer of the deep learning model respectively; perform iterative training on the deep learning model through the backpropagation algorithm so that the trained model can output the optimal raw material feeding ratio; Apply the optimal raw material feeding ratio to production, including feeding and mixing, stirring with a disperser, and adding additives dropwise, and conduct a UV curing test after completion of stirring. Update the deep learning model again according to the test feedback, and finally generate a skin-feel UV ink with expected curing speed, adhesion, and handfeel performance.

2. The method for generating a feel UV ink based on deep learning according to claim 1, characterized in that: Before obtaining the raw material composition data and performance parameter data, it also includes establishing a sample collection subsystem. The sample collection subsystem conducts experimental measurements on multiple batches of skin-feel UV ink to obtain accurate experimental measurement data. The experimental measurement data includes the curing speed, adhesion level, abrasion resistance times, and handfeel score corresponding to different ratios of polyester matte resin, polyurethane easy-to-extinguish resin, skin-feel resin, photoinitiator, slip aid, and wax aid; in the experimental measurement, a unified test standard is used for measurement and the temperature and humidity are stably controlled, and the measured data is transmitted to the database for summary and collation. The obtained collated experimental measurement data is input into the training set and validation set of the deep learning model to improve the generalization ability and prediction accuracy of the model; after the training of the deep learning model is completed, the model parameters are fixed and the first-round optimal formulation suggestion is output to enter the subsequent actual production verification and retraining stage.

3. The method for generating a skin-feel UV ink based on deep learning according to claim 1, characterized in that: When establishing the deep learning model of the multi-layer neural network, set the input layer to receive the raw material composition data and performance parameter data, and set at least three hidden layers to construct a feature extraction and non-linear mapping structure; introduce residual connections in the hidden layer to avoid the phenomenon of gradient vanishing or exploding when the number of network layers is deep; set up a multi-objective regression unit in the output layer to simultaneously predict the curing speed, adhesion level, abrasion resistance times, and handfeel score of the skin-feel UV ink; use the mean square error loss function in the training process and combine L2 regularization to constrain the weights to prevent overfitting; through training, enable the deep learning model to accurately predict the corresponding performance indicators according to the feeding ratio, and realize the intelligent optimization of the ratios of polyester matte resin, polyurethane easy-to-extinguish resin, skin-feel resin, photoinitiator, slip aid, and wax aid.

4. The method for generating skin-friendly UV ink based on deep learning according to claim 1, wherein: During the stirring process of the feeding mixer and disperser, an on-line monitoring module is added to detect the stirring temperature, viscosity and rotation speed in real time, and the detection results are input into the process monitoring sub-network in the deep learning model; the process monitoring sub-network compares the difference between the predicted viscosity and the actual viscosity in real time. If the deviation exceeds the threshold, the rotation speed of the disperser is automatically adjusted or the dropping speed of the auxiliary agent is changed; after adding the smoothness auxiliary agent and the wax auxiliary agent, a layered stirring method is adopted for secondary mixing, and the curing efficiency is monitored by UV curing pre-testing; finally, the measured data is fed back to the deep learning model again for fine-tuning, so that the generated skin-friendly UV ink can maintain consistency and controllability during the production process, and avoid formula inaccuracy caused by temperature changes or uneven stirring.

5. The method for generating a feel UV ink based on deep learning according to claim 1, characterized in that: During the training process of the deep learning model, four sub-goals of curing speed, adhesion level, abrasion resistance times and hand feeling score are set respectively, and a weighted multi-objective optimization strategy is used to make each sub-goal have a corresponding weight coefficient; through iterative training, multiple rounds of tests are carried out on different weight coefficient configurations to screen out a comprehensive formula that not only meets fast curing but also ensures high adhesion and good hand feeling; the weighted multi-objective optimization strategy generates a feasible solution set at the output layer of the model, and then selects the optimal solution according to the pre-set target sorting rules; under this optimal solution, a set of recommended ratios of polyester matte resin, polyurethane easy-to-extinguish resin, skin-friendly resin, photoinitiator, smoothness auxiliary agent and wax auxiliary agent are obtained in the feeding link to balance the requirements of fast printing and durable hand feeling.

6. The method for generating a skin-feel UV ink based on deep learning according to claim 1, characterized in that: Before outputting the optimal raw material feeding ratio, a simulation verification stage is carried out first. The generated formula parameters are input into a simulation module with physical constraints. The simulation module calculates the theoretical curing efficiency and surface tension according to the known kinetic models of polyester matte resin, polyurethane easy-to-extinguish resin, skin-friendly resin, photoinitiator, smoothness auxiliary agent and wax auxiliary agent in the chemical reaction and photo-curing process; if the simulation result deviates from the result predicted by the deep learning model, the deviation information is fed back to the model for secondary training until the simulation result coincides with the prediction result within the preset error range.

7. The method for generating a skin-feel UV ink based on deep learning according to claim 1, characterized in that: A hand feeling evaluation sub-network is set in the deep learning model to separately extract the key features related to the content of the skin-friendly resin and the type of the smoothness auxiliary agent; the hand feeling evaluation sub-network receives the hand feeling score training samples and uses the attention mechanism to weight the main features affecting the hand feeling; the output of the hand feeling evaluation sub-network is combined with the remaining prediction results of the main model of the deep learning model to obtain the final score, so that the feeding ratio can more accurately match the excellent tactile performance; during the production verification, a standardized blind test touch evaluation process is adopted, a fixed number of evaluators are invited to score, and the obtained scores are input into the hand feeling evaluation sub-network for error backtracking correction again to achieve refined control of the skin-friendly effect.

8. The method for generating a skin-feel UV ink based on deep learning according to claim 1, wherein: When adjusting the proportion of the smoothness auxiliary agent and the wax auxiliary agent, a chemical surface energy calculation model is introduced, and the influence of different auxiliary agent combinations on the surface friction coefficient of the cigarette package printed matter is estimated through the chemical surface energy calculation model; In the training samples, the high friction coefficient is stored corresponding to the scratch mark rate, and a penalty term is set for the friction coefficient prediction error during training; when the additive combination leads to potential scratching risks, the chemical surface energy calculation model increases the loss value of the additive combination scheme in the output layer, prompting the final formulation to tend to reduce the friction coefficient.

9. The method for generating a feel UV ink based on deep learning according to claim 1, characterized in that: When performing backpropagation training on the raw material feeding ratio, an automatic learning rate adjustment mechanism is introduced to dynamically modify the learning rate according to the convergence speed of the loss function during training; if the loss value decreases slowly within several iterations, the learning rate is automatically increased to accelerate convergence; if the loss value oscillates or increases, the learning rate is reduced to avoid over-updating.

10. A skin-feel UV ink, characterized in that, The skin-friendly UV ink is produced by using the method for producing skin-friendly UV ink based on deep learning according to any one of claims 1-9.

Images