Near-end strategy optimization reinforcement learning method for optimizing label key indexes

Abstract

The invention discloses a near-end strategy optimization reinforcement learning method for optimizing label key indexes. The method comprises the following steps: preparing a resin coating on one surface of a base material, and heating to form a marking layer; a second step in which a resin coating material having a different color from the resin coating material used in the first step and the resin coating material used in the second step is disposed on the surface of the first colored layer and then heated to form a second colored layer; the outer wall of the base layer is coated with an adhesive, the formula and the coating mode of the adhesive are optimized by applying a reinforcement learning algorithm, the adhesion and the overall structural stability of the label are improved, finally, the base material is stripped from the marking layer and the first coloring layer, and the curing shrinkage rate of the first coloring layer is larger than that of the marking layer. By means of the above method, the present invention can provide a label having excellent durability in which peeling or cracking does not occur between the layers of the label regardless of the thickness of the layers of the label, and a method for manufacturing the label.

Description

Technical Field

[0001] The present invention relates to the application of artificial intelligence technology in a label manufacturing method, and in particular to a proximal strategy optimization reinforcement learning method for optimizing key indicators of a label. Background Art

[0002] Labels mainly refer to simple signs used to indicate the name, weight, volume, purpose and other information of items. They can be used to mark product targets and classifications or contents, similar to keywords for determining targets, and are tools for facilitating the search and positioning of one's own targets. In short, they are printed materials used to identify the relevant instructions of one's own products, and most of them have adhesive on the back, but some are printed without adhesive, which can also be called labels. The labels currently used still have certain defects: most of the existing labels have simple structures and poor waterproof properties. In actual use, some labels need to be attached to some equipment that generates a certain temperature during operation. When the temperature is high, it is easy to cause the label body to be damaged and fall off, and long-term use will cause side flipping, which seriously affects the use effect. Based on this, a proximal strategy optimization reinforcement learning method for optimizing key indicators of labels is proposed to solve the above problems. Summary of the invention

[0003] The present invention aims to provide a proximal strategy optimization reinforcement learning method for optimizing key label indicators to solve the technical problems raised in the background technology.

[0004] To achieve the above purpose, the specific technical solution of the present invention is as follows: A proximal strategy optimization reinforcement learning method for optimizing key label indicators comprises the following steps:

[0005] The first step is to apply a resin coating on one surface of the substrate and then heat it to form a marking layer;

[0006] The second step is to place a resin coating having a different color from the resin coating used in the first step on one surface of the substrate and on the surface of the marking layer, and then heat the resin coating to form a first colored layer.

[0007] The third step is to arrange a resin coating having a different color from the resin coating used in the first step and the resin coating used in the second step on the surface of the first coloring layer, and then heat the resin coating to form a second coloring layer;

[0008] Step 4: Apply adhesive to the outer wall of the base layer, apply the reinforcement learning algorithm to optimize the formula and coating method of the adhesive, improve the adhesion and the overall structural stability of the label, adhere the thermal insulation layer to the base layer, and apply adhesive to the outer wall of the thermal insulation layer and adhere it to the bottom of the inner wall of the waterproof wrapping layer;

[0009] Step 5: Use ultraviolet light to irradiate each adhesive layer to solidify it;

[0010] Step 6: Peeling the substrate off the marking layer and the first coloring layer, wherein the curing shrinkage rate of the first coloring layer is greater than the curing shrinkage rate of the marking layer.

[0011] Furthermore, in the fourth step, the reinforcement learning algorithm is applied to optimize the formulation and coating method of the adhesive to improve the adhesion and the overall structural stability of the label, including the following steps:

[0012] Step A: Environmental Modeling

[0013] 1. Data collection: During the environmental modeling process, key parameter data needs to be collected from the actual production environment. These data will be used to train and optimize the model:

[0014] D(t)={T(t),H(t),V(t),I UV (t)}

[0015] D(t) is the time series data set, T is the temperature during coating, H is the coating thickness, V is the coating speed, I UV is the ultraviolet intensity, t is the time;

[0016] 2. Simulation platform construction: Based on the above collected data, a simulation platform is constructed. The goal of the simulation platform is to simulate the performance of the thermal insulation layer and the waterproof wrapping layer under different parameter settings:

[0017] P ad (t) = f ad (T(t),H(t),V(t),I UV (t))

[0018] P res (t) = f res (T(t),H(t),V(t),I UV (t))

[0019] P water (t) = f water (T(t),H(t),V(t),I UV (t))

[0020] P ad is the adhesion between the thermal insulation layer and the waterproof wrapping layer, P res For high temperature resistance, P water (t) is the waterproof performance index at time t, f ad () is the simulation function for calculating adhesion performance, f res () is the simulation function for calculating high temperature resistance performance, f water () is the simulation function for calculating the waterproof performance, T is the temperature during coating, H is the coating thickness, V is the coating speed, IUV is the ultraviolet intensity, t is the time;

[0021] 3. Through the lizard optimization algorithm, the input parameter D(t) in the simulation platform is optimized so that the output performance index reaches the best state:

[0022] P total (t) = w 1 ·P ad (t)+w 2 ·P res (t)+w 3 ·P water (t)

[0023] P total (t) is the overall performance index, w 1 、w 2 、w 3 is the weight coefficient of each performance index, P ad (t) is the adhesion performance index at time t, P res (t) is the high temperature resistance index at time t, P water (t) is the waterproof performance index at time t, where t is time;

[0024] Through the optimization algorithm, we can find total (t) The parameter combination that maximizes:

[0025] D * (t) = argmax D(t) P total (t)

[0026] D * (t) is the optimal parameter combination after optimization, argmax D(t) is the maximization operator, D(t) is the set of input parameters, P total (t) is the overall performance index;

[0027] Step B: Proximal Strategy Optimization PPO Model

[0028] 2. Strategic Network Design

[0029] The policy network is trained by the PPO algorithm. The purpose is to select the optimal operation strategy to maximize the adhesion effect under given input parameters. The PPO algorithm updates the network parameter θ through the clipped policy gradient. The specific formula is:

[0030]

[0031] L PPO (θ) is the loss function of the algorithm, is the expected value operator, min is the minimum value operator, r t (θ) is the strategy probability ratio, is the advantage estimate, clip(r t (θ),1-∈,1+∈) is the clipping function;

[0032] 2. Reward Function Design

[0033] In order to adjust the policy network towards optimizing the adhesion between the thermal insulation layer and the waterproof wrapping layer, a reward function R is calculated. t , which gives feedback based on the label's waterproofness, anti-warping, high temperature resistance, and adhesion strength indicators. The reward function R t It can be defined as the weighted sum of these performance indicators:

[0034] R t =w ad ·P ad (t)+w res ·P res (t)+w water ·P water (t)+w edge ·P edge (t)

[0035] R t is the reward function value at time t, w ad P is the adhesion performance index ad The weight coefficient of (t), P ad (t) is the adhesion performance index at time t, w res P is the high temperature resistance performance index res The weight coefficient of (t), P res (t) is the high temperature resistance index at time t, w water P is the waterproof performance index water The weight coefficient of (t), P water (t) is the waterproof performance index at time t, w edge P is the anti-warping performance index edge The weight coefficient of (t), P edge (t) is the anti-warping performance index at time t, where t is time;

[0036] Step C: Selective HRL algorithm for gating mechanism

[0037] 1. Hierarchical task decomposition

[0038] When implementing the selected hierarchical reinforcement learning algorithm, the coating process is decomposed into multiple subtasks, each of which corresponds to a sub-strategy, including:

[0039] T1 :Selection of resin coating

[0040] T 2 :Adjustment of coating speed

[0041] T 3 : Optimization of heating time

[0042] T 4 :Control of UV radiation intensity

[0043] Each subtask T i Corresponding to a sub-strategy π i , these strategies form a hierarchical structure, and the overall strategy consists of a high-level strategy π H Control, which is responsible for selecting the current optimal subtask and its corresponding substrategy;

[0044] 2. Application of gating mechanism

[0045] High-level strategy H (s t ) through the gating function g(s t ) to select the most suitable sub-strategy π i (s t ), the gating mechanism formula is expressed as:

[0046]

[0047] π H (s t ) is a high-level strategy, s t is the current state, is the sum symbol, N is the total number of sub-strategies, i is the index of the sub-strategy, g i (s t ) is the i-th output of the gating function, π i (s t ) is the i-th sub-strategy;

[0048] Step D: Application of the Frilled Lizard Optimization Algorithm

[0049] 1. Parameter optimization: The goal of the lizard optimization algorithm is to find the optimal combination of these hyperparameters to maximize the performance indicators of the policy network. The optimization problem can be expressed as:

[0050] {α * ,γ * ,∈ * ,λ *}=argmax α,γ,∈,λ F(α,γ,∈,λ)

[0051] α * is the optimal policy network learning rate, γ* is the optimal discount factor, ∈ * is the optimal exploration rate, λ * is the optimal policy weight, argmax is the operator, α is the learning rate of the policy network, γ is the discount factor, ∈ is the exploration rate, λ is the policy weight, and F(α,γ,∈,λ) is the objective function;

[0052] 2. Global search and local fine-tuning: During the optimization process, the lizard optimization algorithm first performs a global search to explore the entire parameter space to find a near-optimal area, and then refines the parameter selection through local fine-tuning to ensure that the global optimal solution is found. The process of the lizard optimization algorithm can be divided into the following steps:

[0053] Initialize the population: Set a population of frilled lizards, where each individual represents a hyperparameter combination {α i ,γ i ,∈ i ,λ i};

[0054] Fitness evaluation: Calculate the fitness of each individual frilled lizard, that is, the model performance F(α i ,γ i ,∈ i ,λ i );

[0055] Global search: By simulating the group behavior of frilled lizards, we conduct a global exploration of the parameter space and find multiple areas with high fitness.

[0056] Local fine-tuning: In the found high fitness area, make slight adjustments to each hyperparameter to further improve the fitness and eventually converge to the optimal solution;

[0057] Update and select: Update the population according to the fitness value and select the individual with the highest fitness as the optimal solution;

[0058] Through the above optimization process, the umbrella lizard optimization algorithm can find the hyperparameter combination that optimizes the model performance, thereby improving the key performance indicators of the label's waterproofness and high temperature resistance.

[0059] Furthermore, it also includes step E implementation and verification:

[0060] The optimal strategy combination {α * ,γ * ,∈ * ,λ *}, conduct multiple experiments and record the key performance indicators in each experiment, including waterproofness, anti-warping and high temperature resistance. By comparing the experimental results of traditional methods and those after intelligent algorithm optimization, the actual effect of algorithm optimization is evaluated;

[0061]

[0062] ΔP ad (t) is the improvement of adhesion performance, ΔP res (t) is the improvement of high temperature resistance, ΔP water (t) is the improvement in waterproof performance, is the adhesion performance index obtained by the optimized strategy at time t, is the high temperature resistance index obtained by the optimized strategy at time t, is the waterproof performance index obtained by the optimized strategy at time t, is the adhesion performance index obtained using the traditional method at time t, is the high temperature resistance index obtained by traditional method at time t, is the waterproof performance index obtained using the traditional method at time t;

[0063] Apply the optimized strategy to actual production and monitor the performance of labels in real time to ensure that their waterproofness, anti-warping and high temperature resistance meet the expected standards;

[0064]

[0065]

[0066] is the adhesion performance index measured at time t in actual production, To optimize the adhesion performance index under the strategy, is the high temperature resistance index measured at time t in actual production, To optimize the high temperature performance index under the strategy, is the waterproof performance index measured at time t in actual production, It is the waterproof performance index under the optimization strategy.

[0067] Furthermore, the label includes an outermost waterproof wrapping layer, the bottom of the inner wall of the waterproof wrapping layer is coated with an insulation layer, the top of the insulation layer is provided with a base layer, the top of the base layer is provided with a second coloring layer, the top of the second coloring layer is provided with a first coloring layer, and the top inner wall of the first coloring layer is provided with a marking layer.

[0068] Furthermore, the colors of the first coloring layer and the second coloring layer are different; the marking layer is embedded in the inner cavity of the first coloring layer, and the exposed area ratio of the marking layer is not less than 1% and not more than 50%; the waterproof wrapping layer, the first coloring layer, the second coloring layer, the base layer and the heat insulation layer are bonded by an adhesive; the adhesive is a free radical polymerizable ultraviolet curable composition.

[0069] The proximal strategy optimization reinforcement learning method for optimizing label key indicators of the present invention has the following advantages:

[0070] 1. This method configures a resin coating having a color different from that used in the marking layer forming step and the resin coating used in the first coloring layer forming step on the surface of the first coloring layer, and then heats it to form a second coloring layer. In the peeling step of peeling the substrate from the marking layer and the first coloring layer, the first coloring layer clamps the marking layer through curing, and the adhesion between the first coloring layer and the marking layer can be improved regardless of the thickness of the first coloring layer or the marking layer; peeling or cracking between the first coloring layer and the marking layer can be suppressed, and a label with excellent durability can be provided. Through the above method, the present invention can provide a label with excellent durability in which peeling or cracking does not occur between the layers of the label regardless of the thickness of the label layer, and a method for manufacturing the label.

[0071] 2. This method has good waterproof properties by adding a waterproof wrapping layer, and can achieve good heat insulation and heat dissipation effects by adding a heat insulation layer, thereby avoiding damage or even falling off of the label caused by high temperature or humid environment. The waterproof wrapping layer can effectively fix both sides of the label body to avoid side flipping caused by long-term use, thereby ensuring the use effect.

[0072] 3. The waterproof and anti-warping high temperature resistant label and manufacturing method can provide a label manufacturing method and a label that can reliably cut the adhesive. It can also provide a label manufacturing method and a label that can prevent the adhesive from adhering to the cutting edge and also prevent the adhesive from seeping out from the edge of the label.

[0073] 4. The present invention adopts a proximal strategy optimization reinforcement learning method for optimizing key indicators of labels, which significantly improves the stability and durability of labels in extreme environments by adopting a multi-layer structure design and an optimized manufacturing process. In particular, in the ultraviolet irradiation curing process, the parameter configuration optimized by the intelligent algorithm makes the curing effect of each layer of coating more uniform, and the difference in curing shrinkage rate is accurately controlled, thereby enhancing the adhesion strength between layers. Even under harsh conditions of high temperature and high humidity, the label can still maintain the integrity of each layer structure to prevent delamination, warping or performance degradation. This design not only extends the service life of the label, but also improves its reliability and safety in various industrial applications, meeting the stringent requirements for high-quality labels.

[0074] The present invention uses a proximal strategy optimization reinforcement learning method for optimizing key indicators of labels, which combines the unique configuration of multi-layer resin coatings with advanced process technology, and significantly improves the overall performance of the label. By adding a second coloring layer of a different color between the marking layer and the first coloring layer, and performing precise heating and curing treatments, the adhesion between the layers is effectively improved, and the occurrence of interlayer peeling or cracking is avoided. At the same time, the design of the waterproof wrapping layer and the heat insulation layer ensures the stability of the label in a high temperature and high humidity environment, and prevents the label from being damaged or falling off due to harsh environments. The ultraviolet curing process optimized by intelligent algorithms allows the label to maintain its functionality under extreme conditions, further enhancing its reliability and safety in industrial applications. These design innovations and process optimizations make the present invention outstanding in durability, applicability and high temperature protection. BRIEF DESCRIPTION OF THE DRAWINGS

[0075] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0076] Figure 1 It is a schematic diagram of the overall structure of the present invention;

[0077] Figure 2 It is a schematic diagram of the method flow of the present invention;

[0078] Figure 3 Optimizing the manufacturing process of high-performance labels driven by the intelligent algorithm of the present invention;

[0079] Explanation of the markings in the figure: 1. Label; 11. Waterproof wrapping layer; 12. Marking layer; 13. First coloring layer; 14. Second coloring layer; 15. Base layer; 16. Thermal insulation layer. DETAILED DESCRIPTION

[0080] In the following, only some exemplary embodiments are briefly described. As those skilled in the art will appreciate, the described embodiments may be modified in various ways without departing from the spirit or scope of the embodiments of the present invention. Therefore, the drawings and descriptions are considered to be exemplary and non-restrictive in nature.

[0081] In the description of the embodiments of the present invention, it should be understood that the terms "length", "vertical", "horizontal", "top", "bottom", etc. indicating the orientation or positional relationship are based on the orientation or positional relationship shown in the accompanying drawings, and are only for the convenience of describing the embodiments of the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the embodiments of the present invention.

[0082] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of the embodiments of the present invention, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0083] In the embodiments of the present invention, unless otherwise clearly specified and limited, the terms "installed", "connected", "connected", "fixed" and the like should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection, an electrical connection, or a communication; it can be a direct connection, or an indirect connection through an intermediate medium, and it can be the internal connection of two elements or the interaction relationship between two elements. For ordinary technicians in this field, the specific meanings of the above terms in the embodiments of the present invention can be understood according to specific circumstances.

[0084] The disclosure below provides many different embodiments or examples to implement different structures of the embodiments of the present invention. In order to simplify the disclosure of the embodiments of the present invention, the parts and settings of specific examples are described below. Of course, they are only examples, and the purpose is not to limit the embodiments of the present invention. In addition, the embodiments of the present invention can repeat reference numbers and / or reference letters in different examples, and this repetition is for the purpose of simplification and clarity, which itself does not indicate the relationship between the various embodiments and / or settings discussed.

[0085] A proximal policy optimization reinforcement learning method for optimizing label key indicators includes the following steps:

[0086] The first step is to apply a resin coating on one surface of the substrate and then heat it to form a marking layer 12;

[0087] Step 2: A resin coating having a different color from the resin coating used in the first step is disposed on one surface of the substrate and the surface of the marking layer, and then heated to form a first colored layer 13;

[0088] Step 3: After a resin coating having a color different from that used in the first step and the second step is disposed on the surface of the first coloring layer 13, it is heated to form a second coloring layer 14;

[0089] Step 4: Apply adhesive to the outer wall of the base layer, apply the reinforcement learning algorithm to optimize the formula and coating method of the adhesive, improve the adhesion and the overall structural stability of the label, adhere the thermal insulation layer to the base layer, and apply adhesive to the outer wall of the thermal insulation layer and adhere it to the bottom of the inner wall of the waterproof wrapping layer;

[0090] Step 5: Use ultraviolet light to irradiate each adhesive layer to solidify it;

[0091] Step 6: Peeling the substrate off the marking layer 12 and the first coloring layer 13 , wherein the curing shrinkage rate of the first coloring layer 13 is greater than that of the marking layer.

[0092] like Figure 3 , a proximal strategy optimization reinforcement learning method for optimizing key label indicators, can significantly improve the intelligence level of the label manufacturing process and ensure that the product reaches the industry-leading level in both performance and production efficiency:

[0093] Input: Process parameters to be optimized, including:

[0094] T: Heating temperature (unit: Celsius)

[0095] H 1 ,H 2 : Thickness of the first and second coloring layers (unit: mm)

[0096] I UV : Ultraviolet radiation intensity (unit: milliwatt / square centimeter)

[0097] t UV :UV irradiation time (unit: seconds)

[0098] Output: Optimized process parameter combination that maximizes the following performance indicators:

[0099] P ad :Adhesion performance (unit: Newton / square meter)

[0100] P water : Waterproof performance (unit: ml / hour)

[0101] P res :High temperature resistance (unit: degrees Celsius)

[0102] Numerical case study

[0103] Under the initial setting, we optimized the following process parameters:

[0104] Initial heating temperature T 0 =150℃

[0105] Initial first and second coloring layer thickness

[0106] Initial UV irradiation intensity

[0107] Initial UV exposure time

[0108] Through global search and local fine-tuning of the umbrella lizard optimization algorithm, the following optimization results were finally obtained:

[0109] Optimized heating temperature T * =160℃

[0110] Optimized first and second coloring layer thickness

[0111] Optimized UV irradiation intensity

[0112] Optimized UV exposure time

[0113] According to the optimized process parameters, the performance of the label is significantly improved, as shown in the figure:

[0114] name Before optimization After optimization <![CDATA[P ad ]]> <![CDATA[1200N / m 2 ]]> <![CDATA[1500N / m 2 ]]> <![CDATA[P water ]]> 0.5ml / h 0.3ml / h <![CDATA[P res ]]> 200°С 220°С

[0115] Adhesion performance: from the initial Upgrade to

[0116] Waterproof performance: from the initial Reduce to

[0117] High temperature resistance: from the initial Upgrade to

[0118] like Figure 1-2As shown, a waterproof and anti-warping high-temperature resistant label of the present invention comprises a label 1, wherein the label 1 comprises a waterproof wrapping layer 11 as the outermost layer, the bottom of the inner wall of the waterproof wrapping layer 11 is coated with a heat insulating layer 16, a base layer 15 is arranged on the top of the heat insulating layer 16, a second coloring layer 14 is arranged on the top of the base layer 15, a first coloring layer 13 is arranged on the top of the second coloring layer 14, and a marking layer 12 is arranged on the inner wall of the top of the first coloring layer 13; the colors of the first coloring layer 13 and the second coloring layer 14 are different; the marking layer 12 is embedded in the inner cavity of the first coloring layer 13; the exposed area ratio of the marking layer 12 is not less than 1% and not more than 50%; the waterproof wrapping layer 11, the first coloring layer 13, the second coloring layer 14, the base layer 15 and the heat insulating layer 16 are bonded by an adhesive; the adhesive is a free radical polymerizable ultraviolet curable composition.

[0119] The working principle of the waterproof and anti-warping high temperature resistant label and the manufacturing method: the label manufacturing method of the present invention comprises: applying adhesive to a substrate sheet to form an adhesive coating, partially applying an acrylic-based adhesive to the adhesive to cure the adhesive to form a curing, and cutting the curing; after the cutting step, a waste stripping step for removing unnecessary parts can be provided, and the curing step allows the adhesive to form a curing, thereby ensuring the cutting, and this configuration prevents the label from being lifted, eliminating the source of reduced productivity and capacity loss; before curing, the cutting blade does not contact the adhesive, inhibiting the adhesive from adhering to the cutting blade, and preventing the cutting blade performance from being reduced and the stain on the label caused by the adhesive adhering to the cutting blade;

[0120] The labels are coated with an adhesive, the adhesive is layered on a substrate sheet, and portions of the adhesive are cured to form a cure, and the cure is by an acrylic-based adhesive, each label is a label sheet, the label sheet is separated from a partition portion, and the labels are generally traded as a label continuum in which the labels are continuously attached to the partition portion;

[0121] The types and materials of the substrate sheet are not limited to specific types and materials, such as paper and synthetic resin films, and can be common materials used as adhesive paper; for example, as the material of the substrate sheet, paper-based materials such as pure paper, coated paper, and coated paper, synthetic resin films made of PET polyethylene terephthalate, PE polyethylene, PP polypropylene, or PS polystyrene, sheets made of a combination of multiple of the aforementioned synthetic resins, and composite sheets made of a combination of synthetic resin films and paper can also be used. Thermal paper and thermal paper are also allowed to be used; for example, the adhesive can be an emulsion-type adhesive that diffuses in water, a solvent-type adhesive that dissolves in a solvent, or a hot-melt-type adhesive that utilizes its thermoplasticity, etc.; from the material of the adhesive, adhesives based on synthetic rubber, natural rubber, acrylic resin, polyvinyl ether resin, polyurethane resin, or silicone resin can be used, and the adhesive force of the adhesive can be arbitrarily selected;

[0122] The marking layer 12 is a layer showing a mark such as text, a symbol or a graphic, and is a layer whose color is different from the colors of the waterproof wrapping layer 11 and the marking layer 12; the color of the marking layer 12 is different from the colors of the first coloring layer 13 and the second coloring layer 14, and also includes a transparent color; when the color of the marking layer 12 is a transparent color, it can be marked inconspicuously, and in this case, it is suitable for marking a watermark, for example; the marking layer 12 is formed of a thermosetting resin coating containing a resin, a cross-linking agent, a colorant and an organic solvent; it may contain a stabilizer, a flame retardant, an antioxidant, an antistatic agent, a mildew preventer, a lubricant, a filler, etc. as needed;

[0123] As the resin used for the resin coating, for example, acrylic resin, epoxy resin, polyurethane resin, etc. can be listed; in addition, as the cross-linking agent of the resin coating for the marking layer 12, it is preferably a cross-linking agent that makes the curing shrinkage of the marking layer 12 smaller than the curing shrinkage of the first coloring layer 13, and for example, isocyanate cross-linking agents, epoxy cross-linking agents, anhydride cross-linking agents, etc. can be listed. From the viewpoint of improving the adhesion between the second coloring layer 14 and the first coloring layer 13 and the adhesion or durability between the first coloring layer 13 and the marking layer 12, isocyanate-based crosslinking agents are preferred. Isocyanate-based crosslinking agents have an isocyanate group as a functional group. The isocyanate group reacts with a functional group containing active hydrogen, such as a hydroxyl group or an amino group, of the resin contained in the marking layer 12 or the first coloring layer 13, thereby undergoing a crosslinking reaction. In principle, there is no molecular separation in this reaction, and chemical bonds are generated by the reaction between the respective functional groups of the crosslinking agent and the resin, and curing shrinkage occurs due to the decrease in the distance between molecules. Compared with resins containing melamine-based crosslinking agents, benzoguanamine-based crosslinking agents, urea-based crosslinking agents, and metal chelate crosslinking agents, the curing shrinkage rate is smaller. In the case where the resin contains a carboxyl group as a functional group, or when water is present as an impurity, the isocyanate group reacts with these and generates carbon dioxide. In this case, carbon dioxide is separated, which increases the curing shrinkage rate. However, even in this case, the curing shrinkage is generally smaller than that of resins containing melamine-based crosslinking agents, benzoguanamine-based crosslinking agents, urea-based crosslinking agents, and metal chelate crosslinking agents;

[0124] As the acrylic adhesive, an active energy ray-curable composition mainly containing methacrylate can be used, and examples of active energy rays include electron rays, visible light rays, and ultraviolet rays. The adhesive may be acryl-based, and may further include additives such as photoinitiators, inhibitors, thickeners, plasticizers, and colorants. It should be noted that since the acrylic adhesive is coated on the adhesive layer and cured, the acrylic adhesive used in the present invention is preferably a free-radical polymerizable ultraviolet curable composition that is easy to handle;

[0125] Embodiment 1:

[0126] On a PET film serving as a substrate, a pattern of text or graphics is printed by gravure printing using the prepared resin coating, and the area ratio of the exposed surface of the marking layer is made to be 10% relative to the surface area of ​​one side of the label, thereby obtaining a marking layer 12 having a thickness of 2 microns. Next, the prepared resin coating is applied to a portion of the substrate and the marking layer 12, and heated at 150°C for 3 minutes to form a first coloring layer 13 having a thickness of 15 microns. Thereafter, the prepared resin coating is applied on the first coloring layer 13 in such a manner that the thickness after heating becomes 60 microns, and the coating is heated at 140°C for 3 minutes and then at 150°C for 2 minutes to obtain a second coloring layer 14. The crosslinking agent used is a melamine-based crosslinking agent, and finally the substrate is peeled off to produce a label.

[0127] Embodiment 2:

[0128] On a PET film serving as a substrate, a pattern of text or graphics is printed by gravure printing using the prepared resin coating, and the area ratio of the exposed surface of the marking layer is made 20% relative to the surface area of ​​one side of the label, thereby obtaining a marking layer 12 having a thickness of 2 microns. Next, the prepared resin coating is applied to a portion of the substrate and the marking layer 12, and heated at 150°C for 3 minutes to form a first coloring layer 13 having a thickness of 15 microns. Thereafter, the prepared resin coating is applied on the first coloring layer 13 in such a manner that the thickness after heating becomes 60 microns, and heated at 140°C for 3 minutes, and then heated at 150°C for 2 minutes to obtain a second coloring layer 14. The crosslinking agent used is a benzoguanamine-based crosslinking agent, and finally the substrate is peeled off to produce a label.

[0129] Embodiment 3:

[0130] On a PET film serving as a substrate, a pattern of text or graphics is printed by gravure printing using the prepared resin coating, and the area ratio of the exposed surface of the marking layer is made 30% relative to the surface area of ​​one side of the label, thereby obtaining a marking layer 12 having a thickness of 2 microns. Next, the prepared resin coating is applied to a portion of the substrate and the marking layer 12, and heated at 150°C for 3 minutes to form a first coloring layer 13 having a thickness of 15 microns. Thereafter, the prepared resin coating is applied on the first coloring layer 13 in such a manner that the thickness after heating becomes 60 microns, and the coating is heated at 140°C for 3 minutes and then at 150°C for 2 minutes to obtain a second coloring layer 14. The crosslinking agent used is a urea-based crosslinking agent, and finally the substrate is peeled off to produce a label.

[0131] Embodiment 4:

[0132] On a PET film serving as a substrate, a pattern of text or graphics is printed by gravure printing using the prepared resin coating, and the area ratio of the exposed surface of the marking layer is made 40% relative to the surface area of ​​one side of the label, thereby obtaining a marking layer 12 having a thickness of 2 microns. Next, the prepared resin coating is applied to a portion of the substrate and the marking layer 12, and heated at 150°C for 3 minutes to form a first coloring layer 13 having a thickness of 15 microns. Thereafter, the prepared resin coating is applied on the first coloring layer 13 in such a manner that the thickness after heating becomes 60 microns, and the coating is heated at 140°C for 3 minutes and then at 150°C for 2 minutes to obtain a second coloring layer 14. The crosslinking agent used is a metal chelate crosslinking agent, and finally the substrate is peeled off to produce a label.

[0133] Comparative Example 1:

[0134] On a PET film serving as a substrate, a pattern of text or graphics is printed by gravure printing using the prepared resin coating, and the area ratio of the exposed surface of the marking layer is made to be 10% relative to the surface area of ​​one side of the label, thereby obtaining a marking layer 12 having a thickness of 2 μm. Next, the prepared resin coating is applied to a portion of the substrate and the marking layer 12, and heated at 150°C for 3 minutes to form a first coloring layer 13 having a thickness of 15 μm. Thereafter, the prepared resin coating is applied on the first coloring layer 13 in such a manner that the thickness after heating becomes 60 μm, and heated at 140°C for 3 minutes, and then heated at 150°C for 2 minutes to obtain a second coloring layer 14. The cross-linking agent used is changed to a benzoguanamine-based cross-linking agent, and finally the substrate is peeled off to produce a label, as shown in Table 1.

[0135] Table 1

[0136]

[0137]

[0138] The adhesion in Table 1 was evaluated by the following method: the produced label was cut into a square with a side length of 7.5 cm as a test piece, and the test piece was placed in an environment of 150° C. for 1000 hours, and then immersed in a 0.1% rhodamine aqueous solution for 24 hours, and then taken out from the rhodamine aqueous solution, and the excess aqueous solution attached to the test piece was wiped off with a cloth, and then the coloring of rhodamine caused by the penetration of the rhodamine aqueous solution between the marking layer 12 and the first coloring layer 13 was visually observed;

[0139] The case where there was coloration due to rhodamine was judged as having a gap, the case where there was no gap at all was judged as "√", and the case where there was a significant gap was judged as "×";

[0140] In addition, the cracking in the above Table 1 was evaluated by the following method. After the label to be manufactured was placed in an environment of 150°C for 1000 hours, the surface of the marking layer 12 side of the first coloring layer 13 was visually observed to see if there was any cracking; the case where there was no cracking on the surface of the marking layer 12 side of the first coloring layer 13 was judged as "√", and the case where there was cracking was judged as "×"; by comparison, it can be seen that compared with the label of comparative example 1 using a melamine-based crosslinking agent, the label using a melamine-based crosslinking agent, a benzoguanamine-based crosslinking agent, and a urea-based crosslinking agent as the crosslinking agent of the first coloring layer 13 has higher interlayer adhesion and is less likely to crack; as shown in Table 1, since the curing shrinkage rate of the first coloring layer 13 is greater than the curing shrinkage rate of the marking layer 12, the marking layer 12 is more firmly held by the first coloring layer 13.

[0141] It is to be understood that the present invention is described by some embodiments, and it is known to those skilled in the art that various changes or equivalent substitutions may be made to these features and embodiments without departing from the spirit and scope of the present invention. In addition, under the teachings of the present invention, these features and embodiments may be modified to adapt to specific circumstances and materials without departing from the spirit and scope of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed herein, and all embodiments falling within the scope of the claims of this application are within the scope of protection of the present invention.

Claims

1. A proximal strategy optimization reinforcement learning method for optimizing label key indicators, characterized by: The following steps are involved: The first step is to apply a resin coating on one surface of the substrate and then heat it to form a marking layer (12); The second step is to place a resin coating having a different color from the resin coating used in the first step on one surface of the substrate and on the surface of the marking layer, and then heat the coating to form a first colored layer (13); The third step is to arrange a resin coating having a color different from that of the resin coating used in the first step and the resin coating used in the second step on the surface of the first coloring layer (13), and then heat the resin coating to form a second coloring layer (14); Step 4: applying adhesive to the outer wall of the base layer (15), optimizing the formulation and coating method of the adhesive by using a reinforcement learning algorithm, improving the adhesiveness and the overall structural stability of the label, adhering the heat insulating layer (16) to the base layer (15), and applying adhesive to the outer wall of the heat insulating layer (16) and adhering it to the bottom of the inner wall of the waterproof wrapping layer (11); Step 5: Use ultraviolet light to irradiate each adhesive layer to solidify it; Step 6: Peeling the substrate off the marking layer (12) and the first coloring layer (13), wherein the curing shrinkage rate of the first coloring layer (13) is greater than the curing shrinkage rate of the marking layer.

2. According to claim 1, a proximal strategy optimization reinforcement learning method for optimizing label key indicators is characterized by: In the fourth step, the reinforcement learning algorithm is applied to optimize the adhesive formulation and coating method to improve adhesion and the overall structural stability of the label. The following steps are included: Step A: Environmental Modeling 1. Data collection: During the environmental modeling process, key parameter data needs to be collected from the actual production environment. These data will be used to train and optimize the model: D(t)={T(t),H(t),V(t),I UV (t)} D(t) is the time series data set, T is the temperature during coating, H is the coating thickness, V is the coating speed, I UV is the ultraviolet intensity, t is the time; 2. Simulation platform construction: Based on the above collected data, a simulation platform is constructed. The goal of the simulation platform is to simulate the performance of the thermal insulation layer and the waterproof wrapping layer under different parameter settings: P ad (t)=f ad (T(t),H(t),V(t),I UV (t)) P res (t)=f res (T(t),H(t),V(t),I UV (t)) P water (t)=f water (T(t),H(t),V(t),I UV (t)) P ad is the adhesion between the thermal insulation layer and the waterproof wrapping layer, P res For high temperature resistance, P water (t) is the waterproof performance index at time t, f ad () is the simulation function for calculating adhesion performance, f res () is the simulation function for calculating high temperature resistance performance, f water () is the simulation function for calculating the waterproof performance, T is the temperature during coating, H is the coating thickness, V is the coating speed, I UV is the ultraviolet intensity, t is the time; 3. Through the lizard optimization algorithm, the input parameter D(t) in the simulation platform is optimized so that the output performance index reaches the best state: P total (t)=w1·P ad (t)+w2·P res (t)+w3·P water (t) P total (t) is the total performance index, w1, w2, w3 are the weight coefficients of each performance index, P ad (t) is the adhesion performance index at time t, P res (t) is the high temperature resistance index at time t, P water (t) is the waterproof performance index at time t, where t is time; Through the optimization algorithm, we can find total (t) The parameter combination that maximizes: D * (t)=argmax D(t) P total (t) D * (t) is the optimal parameter combination after optimization, argmax D(t) is the maximization operator, D(t) is the set of input parameters, P total (t) is the overall performance index; Step B: Proximal Strategy Optimization PPO Model 1. Strategic network design The policy network is trained by the PPO algorithm. The purpose is to select the optimal operation strategy to maximize the adhesion effect under given input parameters. The PPO algorithm updates the network parameter θ through the clipped policy gradient. The specific formula is: L PPO (θ) is the loss function of the algorithm, is the expected value operator, min is the minimum value operator, r t (θ) is the strategy probability ratio, is the advantage estimate, clip(r t (θ),1-∈,1+∈) is the clipping function; 2. Reward Function Design In order to adjust the policy network towards optimizing the adhesion between the thermal insulation layer and the waterproof wrapping layer, a reward function R is calculated. t , which gives feedback based on the label's waterproofness, anti-warping, high temperature resistance, and adhesion strength indicators. The reward function R t It can be defined as the weighted sum of these performance indicators: R t =w ad ·P ad (t)+w res ·P res (t)+w water ·P water (t)+w edge ·P edge (t) R t is the reward function value at time t, w ad P is the adhesion performance index ad The weight coefficient of (t), P ad (t) is the adhesion performance index at time t, w res P is the high temperature resistance performance index res The weight coefficient of (t), P res (t) is the high temperature resistance index at time t, w water P is the waterproof performance index water The weight coefficient of (t), P water (t) is the waterproof performance index at time t, w edge P is the anti-warping performance index edge The weight coefficient of (t), P edge (t) is the anti-warping performance index at time t, where t is time; Step C: Selective HRL algorithm for gating mechanism 1. Hierarchical task decomposition When implementing the selected hierarchical reinforcement learning algorithm, the coating process is decomposed into multiple subtasks, each of which corresponds to a sub-strategy, including: T1: Selection of resin coating T2: Adjustment of coating speed T3: Optimization of heating time T4: Control of UV radiation intensity Each subtask T i Corresponding to a sub-strategy π i , these strategies form a hierarchical structure, and the overall strategy consists of a high-level strategy π H Control, which is responsible for selecting the current optimal subtask and its corresponding substrategy; 2. Application of gating mechanism High-level strategy H (s t ) through the gating function g(s t ) to select the most suitable sub-strategy π i (s t ), the gating mechanism formula is expressed as: π H (s t ) is a high-level strategy, s t is the current state, is the sum symbol, N is the total number of sub-strategies, i is the index of the sub-strategy, g i (s t ) is the i-th output of the gating function, π i (s t ) is the i-th sub-strategy; Step D: Application of the Frilled Lizard Optimization Algorithm 1. Parameter optimization: The goal of the lizard optimization algorithm is to find the optimal combination of these hyperparameters to maximize the performance indicators of the policy network. The optimization problem can be expressed as: {a * ,c * ,∈ * ,l * }=argmax α,γ,∈,λ F(α,γ,∈,λ) α * is the optimal policy network learning rate, γ * is the optimal discount factor, ∈ * is the optimal exploration rate, λ * is the optimal policy weight, argmax is the operator, α is the learning rate of the policy network, γ is the discount factor, ∈ is the exploration rate, λ is the policy weight, and F(α,γ,∈,λ) is the objective function; 2. Global search and local fine-tuning: During the optimization process, the lizard optimization algorithm first performs a global search to explore the entire parameter space to find a near-optimal area, and then refines the parameter selection through local fine-tuning to ensure that the global optimal solution is found. The process of the lizard optimization algorithm can be divided into the following steps: Initialize the population: Set a population of frilled lizards, where each individual represents a hyperparameter combination {α i ,γ i ,∈ i ,λ i }; Fitness evaluation: Calculate the fitness of each individual frilled lizard, that is, the model performance F(α i ,γ i ,∈ i ,λ i ); Global search: By simulating the group behavior of frilled lizards, we conduct a global exploration of the parameter space and find multiple areas with high fitness. Local fine-tuning: In the found high fitness area, make slight adjustments to each hyperparameter to further improve the fitness and eventually converge to the optimal solution; Update and select: Update the population according to the fitness value and select the individual with the highest fitness as the optimal solution; Through the above optimization process, the umbrella lizard optimization algorithm can find the hyperparameter combination that optimizes the model performance, thereby improving the key performance indicators of the label's waterproofness and high temperature resistance.

3. The proximal strategy optimization reinforcement learning method for optimizing label key indicators according to claim 2 is characterized in that: It also includes step E implementation and verification: The optimal strategy combination {α * ,γ * ,∈ * ,λ * }, conduct multiple experiments and record the key performance indicators in each experiment, including waterproofness, anti-warping and high temperature resistance. By comparing the experimental results of traditional methods and those after intelligent algorithm optimization, the actual effect of algorithm optimization is evaluated; ΔP ad (t) is the improvement of adhesion performance, ΔP res (t) is the improvement of high temperature resistance, ΔP water (t) is the improvement in waterproof performance, is the adhesion performance index obtained by the optimized strategy at time t, is the high temperature resistance index obtained by the optimized strategy at time t, is the waterproof performance index obtained by the optimized strategy at time t, is the adhesion performance index obtained using the traditional method at time t, is the high temperature resistance index obtained by traditional method at time t, is the waterproof performance index obtained using the traditional method at time t; Apply the optimized strategy to actual production and monitor the performance of labels in real time to ensure that their waterproofness, anti-warping and high temperature resistance meet the expected standards; is the adhesion performance index measured at time t in actual production, To optimize the adhesion performance index under the strategy, is the high temperature resistance index measured at time t in actual production, To optimize the high temperature performance index under the strategy, is the waterproof performance index measured at time t in actual production, It is the waterproof performance index under the optimization strategy.

4. The proximal strategy optimization reinforcement learning method for optimizing label key indicators according to claim 1 is characterized in that: The label (1) comprises a waterproof wrapping layer (11) as the outermost layer, the bottom of the inner wall of the waterproof wrapping layer (11) is coated with a heat insulating layer (16), a base layer (15) is arranged on the top of the heat insulating layer (16), a second coloring layer (14) is arranged on the top of the base layer (15), a first coloring layer (13) is arranged on the top of the second coloring layer (14), and a marking layer (12) is arranged on the inner wall of the top of the first coloring layer (13).

5. According to claim 4, a proximal strategy optimization reinforcement learning method for optimizing label key indicators is characterized by: The first coloring layer (13) and the second coloring layer (14) have different colors; the marking layer (12) is embedded in the inner cavity of the first coloring layer (13), and the exposed area ratio of the marking layer (12) is not less than 1% and not more than 50%; the waterproof wrapping layer (11), the first coloring layer (13), the second coloring layer (14), the base layer (15) and the heat insulation layer (16) are bonded by an adhesive; the adhesive is a free radical polymerizable ultraviolet curable composition.

Images