A packaging box cutter design system based on parameter intelligent generation

Abstract

The present invention relates to the field of computer-aided design for personnel, and particularly to a box blade design system based on intelligent parameter generation. A box blade design system based on intelligent parameter generation includes: a box blade design requirement document acquisition module, a population set construction module, a fitness calculation module, a population iteration module, and a box blade design strategy output module. By obtaining the design requirement document of the user for the box, the present invention determines the box blade design requirement items of the user for the box, and in the subsequent process of simulating and updating the box blade design parameters through the genetic algorithm, the recombination and mutation of the box blade design parameter simulation set are guided by the box blade design requirement items of the user for the box, thereby improving the convergence speed of simulating and updating the box blade design parameters, and further improving the efficiency of box blade design.

Description

Technical Field

[0001] The present invention relates to the field of computer-aided design for personnel, and particularly relates to a box blade template design system based on intelligent parameter generation. Background Art

[0002] The design of box blade templates is a crucial link in the packaging industry, and its quality directly affects the production efficiency of boxes, the material utilization rate, and the aesthetics of the finished products. However, the traditional design of box blade templates mainly relies on the experience and manual operations of designers. Designers need to gradually draw the blade template patterns according to the size, shape, functional requirements, and material characteristics of the box, and ensure the feasibility of the design through multiple tests and adjustments. This design method is not only time-consuming and laborious, but also easily restricted by human factors, resulting in low design efficiency. Especially when facing complex or personalized packaging requirements, designers often need to repeatedly modify and verify, resulting in a long design cycle and slow response speed. Summary of the Invention

[0003] The present invention determines the box blade template design requirement items for the box by obtaining the design requirement document of the box from the user, and provides guidance for the recombination and mutation of the box blade template design parameter simulation set through the box blade template design requirement items for the box during the subsequent process of simulating and updating the box blade template design parameters by using the genetic algorithm, so as to improve the convergence speed of simulating and updating the box blade template design parameters, and further improve the design efficiency of the box blade template.

[0004] A box blade template design system based on intelligent parameter generation, comprising:

[0005] A box blade template design requirement document acquisition module, configured to acquire the box blade template design requirement document of the box from the user. The box blade template design requirement document includes H box blade template design requirement items for the box. Based on the box blade template design requirement document, a box blade template design requirement item weight distribution set is acquired. The box blade template design requirement item weight distribution set includes the box blade template design requirement item weights corresponding to the box blade template design requirement items, and the box blade template design requirement item weight distribution set is used for fitness calculation during subsequent box simulation design;

[0006] The population set construction module is used to construct N simulation sets of the die-cutting plate design parameters for packaging boxes. Each simulation set of the die-cutting plate design parameters for packaging boxes includes the die-cutting plate design parameter values corresponding to M die-cutting plate design parameters for packaging boxes. The construction method of the simulation set of the die-cutting plate design parameters for packaging boxes is as follows: for each die-cutting plate design parameter value, set a random value within the corresponding value range of the die-cutting plate design parameter value, and form a simulation set of the die-cutting plate design parameters by combining all the die-cutting plate design parameter values with the set random values; form a population set by combining N simulation sets of the die-cutting plate design parameters for packaging boxes;

[0007] The fitness calculation module is used to calculate the fitness corresponding to the simulation set of the die-cutting plate design parameters for packaging boxes based on the weight distribution set of the die-cutting plate design requirement items;

[0008] The population iteration module is used to iteratively update the population set through the genetic algorithm. The iterative update specifically includes the following steps: randomly select 0.5N simulation sets of the die-cutting plate design parameters for packaging boxes based on the fitness corresponding to the simulation set of the die-cutting plate design parameters for packaging boxes to form a parent set, and form a cross-candidate set by combining the remaining 0.5N simulation sets of the die-cutting plate design parameters for packaging boxes that are not selected; perform a recombination operation based on the parent set and the cross-candidate set to construct a temporary set, and in the recombination operation, adjust the crossover probability of each die-cutting plate design parameter based on the weight distribution set of the die-cutting plate design requirement items; perform a mutation operation on all the simulation sets of the die-cutting plate design parameters in the temporary set to construct a new population set, and in the mutation operation, adjust the mutation probability of each die-cutting plate design parameter based on the weight distribution set of the die-cutting plate design requirement items; and the population iteration module cooperates with the fitness calculation module to continuously iteratively update the population set;

[0009] The die-cutting plate design strategy output module is used to output the simulation set of the die-cutting plate design parameters with the highest fitness as the die-cutting plate design strategy for packaging boxes after the number of iterations of the iterative update of the population set by the population iteration module in cooperation with the fitness calculation module reaches the maximum number of iterations.

[0010] As a preferred aspect, in the die-cutting plate design requirement document acquisition module, to obtain the weight distribution set of the die-cutting plate design requirement items based on the die-cutting plate design requirement document, it specifically includes the following steps:

[0011] Perform a preprocessing operation on the die-cutting plate design requirement document to construct a die-cutting plate design requirement word set, and send the die-cutting plate design requirement word set into the trained named entity extraction model for processing, and perform die-cutting plate design requirement item annotation and priority word annotation on each die-cutting plate design requirement word in the die-cutting plate design requirement word set;

[0012] For each die-cutting template design requirement item of the packaging box, all the die-cutting template design requirement words marked as die-cutting template design requirement items of the packaging box are combined into a set of marked words. Traverse the set of marked words. For each die-cutting template design requirement word in the set of marked words, obtain the priority word closest to the die-cutting template design requirement word, and obtain the priority weight corresponding to the priority word closest to the die-cutting template design requirement word. Calculate the priority average value U of the priority weights corresponding to the priority words of all die-cutting template design requirement words in the set of marked words. Calculate the weight R of the die-cutting template design requirement item to be processed corresponding to the die-cutting template design requirement item of the packaging box through the following formula: R = α 1 (A / B)+α 2 U, where α 1 and α 2 are the word frequency weight and the priority comprehensive weight respectively, and satisfy α 1 +α 2 = 1, A is the total number of die-cutting template design requirement words in the set of marked words, and B is the total number of die-cutting template design requirement words in the set of die-cutting template design requirement words;

[0013] After normalizing the weight R of the die-cutting template design requirement item to be processed corresponding to all die-cutting template design requirement items, the weight of the die-cutting template design requirement item corresponding to the die-cutting template design requirement item of the packaging box is obtained, and the weights of the die-cutting template design requirement items corresponding to all die-cutting template design requirement items are combined into a set of die-cutting template design requirement item weight distributions.

[0014] As a preferred aspect, in the fitness calculation module, calculate the fitness corresponding to the die-cutting template design parameter simulation set based on the set of die-cutting template design requirement item weights, specifically including the following steps:

[0015] Send the die-cutting template design parameter simulation set into the trained packaging box design requirement prediction model for processing, and output the packaging box design requirement data set. The packaging box design requirement data set includes the packaging box design requirement data corresponding to the die-cutting template design requirement item of the packaging box, and traverse the packaging box design requirement data set to perform fitness maximization processing to construct the packaging box design requirement data set to be calculated. The fitness maximization processing is specifically to perform reciprocal processing on the packaging box design requirement data in the packaging box design requirement data set that is opposite to the direction of improving the fitness; calculate the fitness δ corresponding to the die-cutting template design parameter simulation set through the following formula;

[0016] ;

[0017] where β i is the weight of the i-th die-cutting template design requirement item in the set of die-cutting template design requirement item weights, Q iIt is the i-th data of the packaging box design requirement data set to be calculated, where i = 1, 2, 3, …, I, and I is the total number of packaging box die-cutting design requirement items.

[0018] As a preferred aspect, in the population iteration module, a recombination operation is performed based on the parent set and the cross-candidate set to construct a temporary set. And in the recombination operation, the cross probability of each packaging box die-cutting design parameter is adjusted based on the weight distribution set of the packaging box die-cutting design requirement items, which specifically includes the following steps:

[0019] Step T1: Adjust the cross probability of each packaging box die-cutting design parameter based on the weight distribution set of the packaging box die-cutting design requirement items;

[0020] Step T1.1: In the first iteration, traverse H packaging box die-cutting design requirement items, and perform the following operations for the h-th packaging box die-cutting design requirement item to obtain the packaging box die-cutting design parameter association set corresponding to the h-th packaging box die-cutting design requirement item. The packaging box die-cutting design parameter association set includes the packaging box die-cutting design parameters related to the h-th packaging box die-cutting design requirement item, and calculate the to-be-processed cross probability P of the packaging box die-cutting design parameters in the packaging box die-cutting design parameter association set m =M -1 W h , where m = 1, 2, 3, …, M, and W h is the weight of the packaging box die-cutting design requirement item corresponding to the h-th packaging box die-cutting design requirement item, and enter step T1.2; in the subsequent iteration processes except the first iteration, traverse all packaging box die-cutting design parameters, and perform the following operations for the m-th packaging box die-cutting design parameter to calculate the first mean square error ζ of the packaging box die-cutting design parameter value corresponding to the m-th packaging box die-cutting design parameter in the simulation set of all packaging box die-cutting design parameters in the population set m , and judge whether ζ m > D holds, where D is the convergence threshold. If ζ m > D holds, then calculate the to-be-processed cross probability of the m-th packaging box die-cutting design parameter , where e is the natural constant, and enter step T1.2. If ζ m > D does not hold, no operation is performed;

[0021] Step T1.2: Normalize the to-be-processed cross probabilities P of all packaging box die-cutting design parameters m to construct the cross probability of each packaging box die-cutting design parameter;

[0022] Step T2: Randomly select a set of simulated packaging box die-cutting design parameters from the parent set and the crossover candidate set respectively, and the selected set of simulated packaging box die-cutting design parameters will not be selected again. For the two selected sets of simulated packaging box die-cutting design parameters, based on the crossover probability of the packaging box die-cutting design parameters, select K packaging box die-cutting design parameters through the roulette wheel algorithm, and exchange the corresponding packaging box die-cutting design parameter values of the K packaging box die-cutting design parameters in the two selected sets of simulated packaging box die-cutting design parameters to achieve the recombination operation until all the sets of simulated packaging box die-cutting design parameters in the parent set are selected. Combine all the sets of simulated packaging box die-cutting design parameters that have completed the recombination operation with the parent set to form a temporary set.

[0023] As a preferred aspect, in the population iteration module, perform a mutation operation on all the sets of simulated packaging box die-cutting design parameters in the temporary set to construct a new population set. And in the mutation operation, adjust the mutation probability of each packaging box die-cutting design parameter based on the weight distribution set of the packaging box die-cutting design requirement items. The specific steps are as follows:

[0024] Step E1: Adjust the mutation probability of each packaging box die-cutting design parameter based on the weight distribution set of the packaging box die-cutting design requirement items;

[0025] Step E1.1: When performing the first iteration, traverse H packaging box die-cutting design requirement items, and perform the following operations for the hth packaging box die-cutting design requirement item to obtain the associated set of packaging box die-cutting design parameters corresponding to the hth packaging box die-cutting design requirement item. The associated set of packaging box die-cutting design parameters includes the packaging box die-cutting design parameters related to the hth packaging box die-cutting design requirement item, and calculate the to-be-processed mutation probability F of the packaging box die-cutting design parameters in the associated set of packaging box die-cutting design parameters m =M -1 W h , where m = 1, 2, 3,..., M, and W h is the weight of the packaging box die-cutting design requirement item corresponding to the hth packaging box die-cutting design requirement item, and enter step E1.2; in the subsequent iteration processes other than the first iteration, traverse all the packaging box die-cutting design parameters, and perform the following operations for the mth packaging box die-cutting design parameter to obtain the second mean square error η of the packaging box die-cutting design parameter value corresponding to the mth packaging box die-cutting design parameter in all the sets of simulated packaging box die-cutting design parameters in the temporary set m , and judge whether η m > D holds. If η m > D holds, then calculate the to-be-processed mutation probability of the mth packaging box die-cutting design parameter , and enter step E1.2. If η m > D does not hold, no operation is performed;

[0026] Step E1.2: Traverse the simulated set of box die-cutting plate design parameters in the temporary storage set, and perform the following operations for each simulated set of box die-cutting plate design parameters in the temporary storage set. Generate a random value μ between 0 and 1 through a random function, and determine whether μ > Pc holds, where Pc is the mutation threshold. If μ > Pc does not hold, no mutation operation is performed on the selected simulated set of box die-cutting plate design parameters. If μ > Pc holds, based on the crossover probability of the box die-cutting plate design parameters, select L box die-cutting plate design parameters through the roulette wheel algorithm, and perform mutation operations on the box die-cutting plate design parameter values corresponding to the L box die-cutting plate design parameters in the selected simulated set of box die-cutting plate design parameters; until all the simulated sets of box die-cutting plate design parameters in the temporary storage set are traversed, and form a new population set with all the simulated sets of box die-cutting plate design parameters that have not undergone mutation operations and those that have undergone mutation operations.

[0027] As a preferred aspect, in the population iteration module, obtaining the associated set of box die-cutting plate design parameters corresponding to the h-th box die-cutting plate design requirement item specifically includes the following steps:

[0028] Match the h-th box die-cutting plate design requirement item with the box die-cutting plate design requirement items in the box die-cutting plate design parameter association library, output all the box die-cutting plate design parameters corresponding to the h-th box die-cutting plate design requirement item, and form an associated set of box die-cutting plate design parameters with all the box die-cutting plate design parameters corresponding to the h-th box die-cutting plate design requirement item. The box die-cutting plate design parameter association library includes box die-cutting plate design requirement items and their corresponding several box die-cutting plate design parameters;

[0029] The construction method of the box die-cutting plate design parameter association library is: obtain the trained box design requirement prediction model. For each box die-cutting plate design requirement item, obtain the gradient value Y of the m-th box die-cutting plate design parameter in the box design requirement prediction model for the box die-cutting plate design requirement item m and determine whether Y m > X holds, where X is the gradient threshold. If Y m > X holds, establish a corresponding relationship between the m-th box die-cutting plate design parameter and the box die-cutting plate design requirement item. If Y m > X does not hold, no operation is performed; then form a box die-cutting plate design parameter association library with all the box die-cutting plate design requirement items and their corresponding several box die-cutting plate design parameters.

[0030] As a preferred aspect, the named entity extraction model in the box die-cutting plate design requirement document acquisition module is trained through the following steps:

[0031] Obtain several training samples for the die-cutting plate design requirements of the packaging box. The training samples for the die-cutting plate design requirements of the packaging box include a set of words for the die-cutting plate design requirements of the packaging box, and label the training samples for the die-cutting plate design requirements of the packaging box with die-cutting plate design requirement items and priority words. Combine all the labeled training samples for the die-cutting plate design requirements of the packaging box into a training set for the die-cutting plate design requirements of the packaging box. Train the named entity extraction model with the training set for the die-cutting plate design requirements of the packaging box. During the training, use the labels of the die-cutting plate design requirement items and priority words as the target, calculate the training loss value for the die-cutting plate design requirements of the packaging box, and determine whether the training loss value for the die-cutting plate design requirements of the packaging box is within the confidence range for the die-cutting plate design requirements of the packaging box. If the training loss value for the die-cutting plate design requirements of the packaging box is within the confidence range for the die-cutting plate design requirements of the packaging box, output the trained named entity extraction model; otherwise, continue to train the named entity extraction model with the training set for the die-cutting plate design requirements of the packaging box.

[0032] As a preferred aspect, the packaging box design requirement prediction model in the fitness calculation module is trained through the following steps:

[0033] Obtain the training samples for the packaging box design requirement prediction. The training samples for the packaging box design requirement prediction include a data set of die-cutting plate design parameters for the packaging box, and the storage form of the data set of die-cutting plate design parameters for the packaging box is the same as that of the simulation set of die-cutting plate design parameters for the packaging box. Label the training samples for the packaging box design requirement prediction with the packaging box design requirement data set. Combine all the labeled training samples for the packaging box design requirement prediction into a training set for the packaging box design requirement prediction. Train the packaging box design requirement prediction model with the training set for the packaging box design requirement prediction. During the training, use the packaging box design requirement data set as the target, calculate the prediction loss value for the packaging box design requirement, and determine whether the prediction loss value for the packaging box design requirement is within the confidence range for the packaging box design requirement prediction. If the prediction loss value for the packaging box design requirement is within the confidence range for the packaging box design requirement prediction, output the trained packaging box design requirement prediction model; otherwise, continue to train the packaging box design requirement prediction model with the training set for the packaging box design requirement prediction.

[0034] The present invention has the following advantages:

[0035] The present invention determines the die-cutting plate design requirement items for the packaging box by obtaining the design requirement document of the user for the packaging box, and provides guidance for the recombination and mutation of the simulation set of die-cutting plate design parameters for the packaging box through the die-cutting plate design requirement items of the user for the packaging box during the subsequent process of simulating and updating the die-cutting plate design parameters for the packaging box by means of the genetic algorithm, thereby improving the convergence speed of simulating and updating the die-cutting plate design parameters for the packaging box, and further improving the efficiency of die-cutting plate design for the packaging box. Brief Description of the Drawings

[0036] Figure 1 This is a schematic structural diagram of a box knife template design system based on intelligent parameter generation according to an embodiment of the present invention. Specific implementation manners

[0037] In order to enable those skilled in the art to better understand the technical solutions in 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.

[0038] Embodiment 1, a box knife template design system based on intelligent parameter generation, as Figure 1 shown, includes:

[0039] A box knife template design requirement document acquisition module, which is used to acquire the box knife template design requirement document for the box by the user. The box knife template design requirement document includes H box knife template design requirement items for the box, such as material utilization rate, material strength, and production cost, etc. The box knife template design requirement document is, for example, "the box should save materials as much as possible, while ensuring sufficient strength, and the production cost should be controlled within a reasonable range". Based on the box knife template design requirement document, a box knife template design requirement item weight distribution set is obtained. The box knife template design requirement item weight distribution set includes the box knife template design requirement item weights corresponding to the box knife template design requirement items, and the box knife template design requirement item weight distribution set is used for fitness calculation during subsequent box simulation design. The box knife template design requirement item weight describes the emphasis direction of various requirements when the user designs the box knife template, such as emphasizing reducing production cost and emphasizing higher material strength;

[0040] In the box knife template design requirement document acquisition module, obtaining the box knife template design requirement item weight distribution set based on the box knife template design requirement document specifically includes the following steps:

[0041] Perform preprocessing operations on the box knife template design requirement document to construct a box knife template design requirement word set. The preprocessing operations include word segmentation and stop word deletion, etc., and send the box knife template design requirement word set into a trained named entity extraction model for processing. Perform box knife template design requirement item annotation and priority word annotation on each box knife template design requirement word in the box knife template design requirement word set. It should be noted that the named entity extraction model is designed based on the prior art and generally includes a Bi-LSTM layer and a CRF layer. The box knife template design requirement item annotation refers to judging which box knife template design requirement item the box knife template design requirement word belongs to, and the priority word annotation refers to whether the box knife template design requirement word belongs to priority words, such as "the most important", "as much as possible", "reasonable", and "secondary", etc.;

[0042] For each die-cutting design requirement item of the packaging box, all die-cutting design requirement words marked as die-cutting design requirement items of the packaging box are grouped into a set of marked words. Traverse the set of marked words. For each die-cutting design requirement word in the set of marked words, obtain the priority word closest to the die-cutting design requirement word. For example, in "the packaging box should save materials as much as possible", the priority word corresponding to "save materials" is "as much as possible", and obtain the priority weight corresponding to the priority word closest to the die-cutting design requirement word. It should be noted that these priority weights are set in advance by developers. For example, the priority weight corresponding to "the most important" is 0.7, and the priority weight corresponding to "as much as possible" is 0.5. Calculate the priority average value U of the priority weights corresponding to all die-cutting design requirement words in the set of marked words. The priority average value can reflect the user's emotional tendency towards different goals of packaging box design requirements. Calculate the weight R of the die-cutting design requirement item to be processed corresponding to the die-cutting design requirement item of the packaging box through the following formula: R = α 1 (A / B) + α( 2 U, where α( 1 and α( 2 are the word frequency weight and the priority comprehensive weight respectively, which are determined by developers, generally 0.45 and 0.55, and satisfy α( 1 + α( 2 = 1, A is the total number of die-cutting design requirement words in the set of marked words, and B is the total number of die-cutting design requirement words in the set of die-cutting design requirement words of the packaging box;( (

[0043] After normalizing the weights R of the die-cutting design requirement items to be processed corresponding to all die-cutting design requirement items of the packaging box, obtain the weights of the die-cutting design requirement items corresponding to the die-cutting design requirement items of the packaging box, and form a set of weight distributions of the die-cutting design requirement items corresponding to all die-cutting design requirement items of the packaging box;( (

[0044] Population set construction module, which is used to construct N simulation sets of die-cutting plate design parameters for packaging boxes. Each simulation set of die-cutting plate design parameters for packaging boxes includes the values of die-cutting plate design parameters corresponding to M die-cutting plate design parameters for packaging boxes, such as material thickness, packaging box size (if there is a packaging box size requirement in the user's needs, the packaging box size is not adjusted as a simulation parameter but directly used as a fixed parameter), cutting line layout, and material type, etc. It should be noted that the material thickness and packaging box size, etc. can be directly set with numerical values, while those that cannot be directly described by numerical values, such as cutting line layout and material type, etc., can be set with corresponding codes in the way of word embedding; the construction method of the simulation set of die-cutting plate design parameters for packaging boxes is as follows: for each value of the die-cutting plate design parameter for the packaging box, set a random value within the corresponding value range of the die-cutting plate design parameter value. It should be noted that setting random values here also includes randomly selecting codes, and all the die-cutting plate design parameter values with set random values are composed into a simulation set of die-cutting plate design parameters for packaging boxes; the N simulation sets of die-cutting plate design parameters for packaging boxes are composed into a population set;

[0045] Fitness calculation module, which is used to calculate the fitness corresponding to the simulation set of die-cutting plate design parameters for packaging boxes based on the weight distribution set of die-cutting plate design requirements for packaging boxes;

[0046] In the fitness calculation module, calculating the fitness corresponding to the simulation set of die-cutting plate design parameters for packaging boxes based on the weight distribution set of die-cutting plate design requirements for packaging boxes specifically includes the following steps: sending the simulation set of die-cutting plate design parameters for packaging boxes into the trained packaging box design requirement prediction model for processing, and outputting a packaging box design requirement data set. The packaging box design requirement data set includes the packaging box design requirement data corresponding to the die-cutting plate design requirements for packaging boxes, that is, the specific values corresponding to material utilization rate, material strength, production cost, etc., and traversing the packaging box design requirement data set to perform fitness maximization processing to construct a packaging box design requirement data set to be calculated. The fitness maximization processing is specifically to take the reciprocal of the packaging box design requirement data in the packaging box design requirement data set that is opposite to the direction of improving fitness. For example, the lower the production cost, the better, and the higher the corresponding fitness, so the data corresponding to the production cost should be taken as the reciprocal; calculate the fitness δ corresponding to the simulation set of die-cutting plate design parameters for packaging boxes through the following formula: the packaging box design requirement prediction model is established based on the BP neural network. Referring to the existing BP neural network, it also has an input layer, a hidden layer, and an output layer;

[0047] ;

[0048] where β i is the weight of the i-th die-cutting plate design requirement item in the weight distribution set of die-cutting plate design requirements for packaging boxes, and Q iIt is the i-th packaging box design requirement data to be calculated in the packaging box design requirement dataset to be calculated, where i = 1, 2, 3, …, I, and I is the total number of packaging box die-cutting plate design requirement items;

[0049] The population iteration module is used to iteratively update the population set through the genetic algorithm. The iterative update specifically includes the following steps: Randomly select 0.5N packaging box die-cutting plate design parameter simulation sets from the fitness corresponding to the packaging box die-cutting plate design parameter simulation set to form the parent set, and form the cross candidate set with the unselected 0.5N packaging box die-cutting plate design parameter simulation sets. It should be noted that in the process of randomly selecting 0.5N packaging box die-cutting plate design parameter simulation sets to form the parent set, the roulette wheel selection algorithm can be used; Perform recombination operations based on the parent set and the cross candidate set to construct a temporary set. And in the recombination operation, adjust the crossover probability of each packaging box die-cutting plate design parameter based on the packaging box die-cutting plate design requirement item weight distribution set. It should be noted that the crossover probability here refers to the probability that the packaging box die-cutting plate design parameter values corresponding to the packaging box die-cutting plate design parameters are exchanged during the recombination operation. Adjusting the crossover probability of each packaging box die-cutting plate design parameter based on the packaging box die-cutting plate design requirement item weight distribution set can guide the search of the packaging box die-cutting plate design parameter simulation set in the direction that meets the user's needs during the iteration process and improve the convergence speed; Perform mutation operations on all the packaging box die-cutting plate design parameter simulation sets in the temporary set to construct a new population set. And in the mutation operation, adjust the mutation probability of each packaging box die-cutting plate design parameter based on the packaging box die-cutting plate design requirement item weight distribution set. It should be noted that the mutation probability refers to the probability that the packaging box die-cutting plate design parameter values corresponding to the packaging box die-cutting plate design parameters mutate during the recombination operation. Adjusting the mutation probability of each packaging box die-cutting plate design parameter based on the packaging box die-cutting plate design requirement item weight distribution set can guide the search of the packaging box die-cutting plate design parameter simulation set in the direction that meets the user's needs during the iteration process and further improve the convergence speed; And the population iteration module cooperates with the fitness calculation module to continuously iteratively update the population set;

[0050] In the population iteration module, perform recombination operations based on the parent set and the cross candidate set to construct a temporary set. And in the recombination operation, adjust the crossover probability of each packaging box die-cutting plate design parameter based on the packaging box die-cutting plate design requirement item weight distribution set, which specifically includes the following steps:

[0051] Step T1: Adjust the crossover probability of each packaging box die-cutting plate design parameter based on the packaging box die-cutting plate design requirement item weight distribution set;

[0052] Step T1.1: In the first iteration, traverse H packaging box die-cutting plate design requirement items, and perform the following operations for the h-th packaging box die-cutting plate design requirement item to obtain the packaging box die-cutting plate design parameter association set corresponding to the h-th packaging box die-cutting plate design requirement item. The packaging box die-cutting plate design parameter association set includes the packaging box die-cutting plate design parameters related to the h-th packaging box die-cutting plate design requirement item, and calculate the to-be-processed crossover probability P of the packaging box die-cutting plate design parameters in the packaging box die-cutting plate design parameter association set m =M -1 W h , where m = 1, 2, 3,..., M. It should be noted that here, m in the to-be-processed crossover probability P m is the serial number of the packaging box die-cutting plate design parameter, and W h is the weight of the packaging box die-cutting plate design requirement item corresponding to the h-th packaging box die-cutting plate design requirement item. Enter step T1.2. By the user's preference for the design requirement, increasing the crossover probability of the packaging box die-cutting plate design parameters corresponding to the design requirement with the user's preference can enable the packaging box die-cutting plate design parameter values corresponding to these packaging box die-cutting plate design parameters with the user's preference to inherit from the individuals with excellent performance, improving the convergence speed; in the subsequent iteration process except for the first iteration, traverse all packaging box die-cutting plate design parameters, and perform the following operations for the m-th packaging box die-cutting plate design parameter. Calculate the first mean square error ζ of the packaging box die-cutting plate design parameter value corresponding to the m-th packaging box die-cutting plate design parameter in all packaging box die-cutting plate design parameter simulation sets in the population set m , and judge whether ζ m > D holds. D is the convergence threshold, determined by the developer. If ζ m > D holds, it means that the packaging box die-cutting plate design parameter value corresponding to the m-th packaging box die-cutting plate design parameter is already close to the optimal solution. At this time, it is necessary to reduce the crossover probability corresponding to the m-th packaging box die-cutting plate design parameter and allocate more computing resources to the search of other packaging box die-cutting plate design parameters. Then calculate the to-be-processed crossover probability of the m-th packaging box die-cutting plate design parameter , where e is the natural constant. Enter step T1.2. If ζ m > D does not hold, it means that the packaging box die-cutting plate design parameter value corresponding to the m-th packaging box die-cutting plate design parameter is not close to the optimal solution, and no operation is performed;

[0053] Step T1.2: Normalize the to-be-processed crossover probabilities P of all packaging box die-cutting plate design parameters m to construct the crossover probability of each packaging box die-cutting plate design parameter;

[0054] Step T2: Randomly select a set of packaging box die-cutting plate design parameter simulations from the parent set and the cross-candidate set respectively, and the selected set of packaging box die-cutting plate design parameter simulations will not be selected again. For the two selected sets of packaging box die-cutting plate design parameter simulations, based on the crossover probability of the packaging box die-cutting plate design parameters, select K packaging box die-cutting plate design parameters through the roulette wheel algorithm. K is determined by the developer and is generally 1. When multi-point crossover is required, K can be selected as a value greater than 1. Exchange the values of the packaging box die-cutting plate design parameters corresponding to the K packaging box die-cutting plate design parameters in the two selected sets of packaging box die-cutting plate design parameter simulations to achieve the recombination operation until all the sets of packaging box die-cutting plate design parameter simulations in the parent set are selected. Combine all the sets of packaging box die-cutting plate design parameter simulations that have completed the recombination operation with the parent set to form a temporary set;

[0055] In the population iteration module, perform mutation operations on all the sets of packaging box die-cutting plate design parameter simulations in the temporary set to construct a new population set. And in the mutation operation, adjust the mutation probability of each packaging box die-cutting plate design parameter based on the packaging box die-cutting plate design requirement item weight distribution set. The specific steps are as follows:

[0056] Step E1: Adjust the mutation probability of each packaging box die-cutting plate design parameter based on the packaging box die-cutting plate design requirement item weight distribution set;

[0057] Step E1.1: When performing the first iteration, traverse H packaging box die-cutting plate design requirement items, and perform the following operations for the h-th packaging box die-cutting plate design requirement item to obtain the set of packaging box die-cutting plate design parameter associations corresponding to the h-th packaging box die-cutting plate design requirement item. The set of packaging box die-cutting plate design parameter associations includes the packaging box die-cutting plate design parameters related to the h-th packaging box die-cutting plate design requirement item, and calculate the to-be-processed mutation probability F of the packaging box die-cutting plate design parameters in the set of packaging box die-cutting plate design parameter associations m =M -1 W h , where m = 1, 2, 3,..., M. It should be noted that here the to-be-processed mutation probability F m in which m is the serial number of the packaging box die-cutting plate design parameter, W his the weight of the die-cutting plate design requirement item corresponding to the h-th packaging box die-cutting plate design requirement item. Enter step E1.2. By the user's preference for design requirements, the mutation probability of the die-cutting plate design parameters corresponding to the design requirements that match the user's preference is increased, which can expand the search space corresponding to these die-cutting plate design parameters that match the user's preference and avoid falling into a local optimal solution. In the subsequent iteration process except for the first iteration, traverse all die-cutting plate design parameters, and perform the following operations on the m-th die-cutting plate design parameter to obtain the second mean square error η of the die-cutting plate design parameter value corresponding to the m-th die-cutting plate design parameter in all die-cutting plate design parameter simulation sets in the temporary set. m , determine whether η m > D holds. If η m > D holds, it means that the die-cutting plate design parameter value corresponding to the m-th die-cutting plate design parameter is already close to the optimal solution. At this time, the mutation probability corresponding to the m-th die-cutting plate design parameter needs to be reduced, and more computing resources are allocated to the search of other die-cutting plate design parameters. Then calculate the mutation probability to be processed for the m-th die-cutting plate design parameter , enter step E1.2. If η m > D does not hold, it means that the die-cutting plate design parameter value corresponding to the m-th die-cutting plate design parameter is not close to the optimal solution, and no operation is performed;

[0058] Step E1.2: Traverse the die-cutting plate design parameter simulation sets in the temporary set, and perform the following operations on each die-cutting plate design parameter simulation set in the temporary set. Generate a random value μ between 0 and 1 through a random function, and determine whether μ > Pc holds. Pc is the mutation threshold, generally 0.7. If μ > Pc does not hold, no mutation operation is performed on the selected die-cutting plate design parameter simulation set. If μ > Pc holds, based on the crossover probability of the die-cutting plate design parameters, select L die-cutting plate design parameters through the roulette wheel algorithm. L is determined by the developer. L is generally 1. When multi-point mutation is required, L can be selected as a value greater than 1. Mutate the die-cutting plate design parameter values corresponding to the L die-cutting plate design parameters in the selected die-cutting plate design parameter simulation set, that is, set random values within the corresponding value range of the die-cutting plate design parameter values; until all die-cutting plate design parameter simulation sets in the temporary set are traversed, form a new population set with all die-cutting plate design parameter simulation sets that have not performed mutation operations and those that have performed mutation operations;

[0059] The die-cutting plate design strategy output module is used to output the die-cutting plate design parameter simulation set with the highest fitness as the die-cutting plate design strategy after the number of iterations of the population iteration module cooperating with the fitness calculation module to iteratively update the population set reaches the maximum number of iterations.

[0060] In this application, by obtaining the design requirement document of the packaging box for the user, the design requirement items of the die-cutting plate for the packaging box are determined for the user. In the subsequent process of simulating and updating the die-cutting plate design parameters of the packaging box through the genetic algorithm, the recombination and mutation of the die-cutting plate design parameter simulation set for the packaging box are guided by the design requirement items of the die-cutting plate for the packaging box, so as to improve the convergence speed of simulating and updating the die-cutting plate design parameters of the packaging box, and further improve the design efficiency of the die-cutting plate for the packaging box.

[0061] In the population iteration module, obtaining the die-cutting plate design parameter association set corresponding to the h-th die-cutting plate design requirement item for the packaging box specifically includes the following steps:

[0062] Match the h-th die-cutting plate design requirement item for the packaging box with the die-cutting plate design requirement items in the die-cutting plate design parameter association library for the packaging box, output all die-cutting plate design parameters corresponding to the h-th die-cutting plate design requirement item for the packaging box, and form the die-cutting plate design parameter association set with all die-cutting plate design parameters corresponding to the h-th die-cutting plate design requirement item for the packaging box. The die-cutting plate design parameter association library for the packaging box includes die-cutting plate design requirement items and several corresponding die-cutting plate design parameters for each of them;

[0063] The construction method of the die-cutting plate design parameter association library for the packaging box is as follows: Obtain the trained packaging box design requirement prediction model. For each die-cutting plate design requirement item, obtain the gradient value Y of the m-th die-cutting plate design parameter in the packaging box design requirement prediction model for the die-cutting plate design requirement item m , and the larger the gradient value Y m , the stronger the correlation between the m-th die-cutting plate design parameter and the die-cutting plate design requirement item. Then judge whether Y m > X holds, where X is the gradient threshold determined by the developer. If Y m > X holds, it indicates that the correlation between the m-th die-cutting plate design parameter and the die-cutting plate design requirement item is relatively strong, and establish a corresponding relationship between the m-th die-cutting plate design parameter and the die-cutting plate design requirement item. If Y m > X does not hold, it indicates that the correlation between the m-th die-cutting plate design parameter and the die-cutting plate design requirement item is not strong, and no operation is performed. Then form the die-cutting plate design parameter association library with all die-cutting plate design requirement items and their corresponding several die-cutting plate design parameters.

[0064] The named entity extraction model in the packaging box die-cutting plate design requirement document acquisition module is trained through the following steps:

[0065] Obtain several training samples for the die-cutting template design requirements of the packaging box. The training samples for the die-cutting template design requirements of the packaging box include a set of words for the die-cutting template design requirements of the packaging box. It should be noted that the training samples for the die-cutting template design requirements of the packaging box are constructed based on the actual die-cutting template design requirements document of the packaging box, and the training samples for the die-cutting template design requirements of the packaging box are labeled with die-cutting template design requirement items and priority words. All the labeled training samples for the die-cutting template design requirements of the packaging box are composed into a training set for the die-cutting template design requirements of the packaging box. Use the training set for the die-cutting template design requirements of the packaging box to train the named entity extraction model. During the training, use the labels of the die-cutting template design requirement items and priority words as the target, calculate the training loss value for the die-cutting template design requirements of the packaging box, and determine whether the training loss value for the die-cutting template design requirements of the packaging box is within the confidence range for the die-cutting template design requirements of the packaging box. The confidence range for the die-cutting template design requirements of the packaging box is used to characterize the accuracy of the named entity extraction model. If the training loss value for the die-cutting template design requirements of the packaging box is within the confidence range for the die-cutting template design requirements of the packaging box, output the trained named entity extraction model; otherwise, continue to train the named entity extraction model with the training set for the die-cutting template design requirements of the packaging box.

[0066] The packaging box design requirement prediction model in the fitness calculation module is trained through the following steps:

[0067] Obtain the training samples for the packaging box design requirement prediction. The training samples for the packaging box design requirement prediction include a data set of die-cutting template design parameters for the packaging box, and the storage form of the data set of die-cutting template design parameters for the packaging box is the same as that of the simulation set of die-cutting template design parameters for the packaging box, and it also includes the die-cutting template design parameter values corresponding to M die-cutting template design parameters for the packaging box. It should be noted that the data set of die-cutting template design parameters for the packaging box is obtained by developers based on the already designed packaging box. Label the training samples for the packaging box design requirement prediction with the packaging box design requirement data set. The packaging box design requirement data set here is also obtained by developers based on the already designed packaging box, and there is a one-to-one correspondence between the training samples for the packaging box design requirement prediction and the packaging box design requirement data set, and they all come from the same designed packaging box. All the labeled training samples for the packaging box design requirement prediction are composed into a training set for the packaging box design requirement prediction. Use the training set for the packaging box design requirement prediction to train the packaging box design requirement prediction model. During the training, use the packaging box design requirement data set as the target, calculate the prediction loss value for the packaging box design requirement prediction, and determine whether the prediction loss value for the packaging box design requirement prediction is within the confidence range for the packaging box design requirement prediction. The confidence range for the packaging box design requirement prediction is used to characterize the accuracy of the packaging box design requirement prediction model. If the prediction loss value for the packaging box design requirement prediction is within the confidence range for the packaging box design requirement prediction, output the trained packaging box design requirement prediction model; otherwise, continue to train the packaging box design requirement prediction model with the training set for the packaging box design requirement prediction.

[0068] It should be understood that those of ordinary skill in the art can make improvements or modifications according to the above description, and all such improvements and modifications shall fall within the protection scope of the appended claims of the present invention. The parts not described in detail in this specification belong to the prior art well known to those of ordinary skill in the art.

Claims

1. A system for intelligently generating packaging box die plate design based on parameters, characterized in that: include: A packaging box cutter design requirement document acquisition module is used to acquire a user's packaging box cutter design requirement document for a packaging box. The packaging box cutter design requirement document includes H user's packaging box cutter design requirement items for a packaging box. Based on the packaging box cutter design requirement document, a packaging box cutter design requirement item weight distribution set is acquired. The packaging box cutter design requirement item weight distribution set includes the packaging box cutter design requirement item weights corresponding to the packaging box cutter design requirement items. The packaging box cutter design requirement item weight distribution set is used for fitness calculation in subsequent packaging box simulation design. A population set construction module is used to construct N packaging box cutter design parameter simulation sets, wherein the packaging box cutter design parameter simulation sets include packaging box cutter design parameter values ​​corresponding to M packaging box cutter design parameters, and the packaging box cutter design parameter simulation sets are constructed in the following manner: a random value within a corresponding value range of the packaging box cutter design parameter value is set for each packaging box cutter design parameter value, and all packaging box cutter design parameter values ​​with set random values ​​are grouped into a packaging box cutter design parameter simulation set; and N packaging box cutter design parameter simulation sets are grouped into a population set; A fitness calculation module is used to calculate the fitness corresponding to the packaging box cutter design parameter simulation set based on the weight distribution set of the packaging box cutter design requirement items; The population iteration module is used to iteratively update the population set through a genetic algorithm, and the iterative update specifically includes the following steps: randomly selecting 0.5N packaging box blade design parameter simulation sets to form a parent set based on the fitness corresponding to the packaging box blade design parameter simulation set, and forming the unselected 0.5N packaging box blade design parameter simulation sets into a crossover candidate set; performing a recombination operation based on the parent set and the crossover candidate set to construct a temporary set, and in the recombination operation, adjusting the crossover probability of each packaging box blade design parameter based on the packaging box blade design requirement item weight distribution set; performing a mutation operation on all packaging box blade design parameter simulation sets in the temporary set to construct a new population set, and in the mutation operation, adjusting the mutation probability of each packaging box blade design parameter based on the packaging box blade design requirement item weight distribution set; and the population iteration module cooperates with the fitness calculation module to continuously iteratively update the population set; The packaging box cutter design strategy output module is used to output the packaging box cutter design parameter simulation set with the highest fitness as the packaging box cutter design strategy after the number of iterations of the population iteration module cooperating with the fitness calculation module to iteratively update the population set reaches the maximum number of iterations.

2. According to the parameter-based intelligent generation of packaging box die design system according to claim 1, it is characterized in that: In the module for obtaining the packaging box cutter design requirement document, a weight distribution set of packaging box cutter design requirement items is obtained based on the packaging box cutter design requirement document, specifically including the following steps: Preprocess the packaging box cutter design requirement document, construct a packaging box cutter design requirement word set, and send the packaging box cutter design requirement word set to the trained named entity extraction model for processing, and label each packaging box cutter design requirement word in the packaging box cutter design requirement word set with packaging box cutter design requirement item and priority word labeling; For each packaging box cutting board design requirement item, all packaging box cutting board design requirement words marked as packaging box cutting board design requirement items are formed into a marked word set, and the marked word set is traversed. For each packaging box cutting board design requirement word in the marked word set, the priority word closest to the packaging box cutting board design requirement word is obtained, and the priority weight corresponding to the priority word closest to the packaging box cutting board design requirement word is obtained, and the priority average U of the priority weights corresponding to the priority words corresponding to all packaging box cutting board design requirement words in the marked word set is calculated, and the weight R of the pending packaging box cutting board design requirement item corresponding to the packaging box cutting board design requirement item is calculated by the following formula, R=α1(A / B)+α2U, where α1 and α2 are the word frequency weight and the priority comprehensive weight respectively, and satisfy α1+α2=1, A is the total number of packaging box cutting board design requirement words in the marked word set, and B is the total number of packaging box cutting board design requirement words in the packaging box cutting board design requirement word set; The weights R of the pending package box cutter design requirement items corresponding to all package box cutter design requirement items are normalized to obtain the package box cutter design requirement item weights corresponding to the package box cutter design requirement items, and the package box cutter design requirement item weights corresponding to all package box cutter design requirement items are combined into a package box cutter design requirement item weight distribution set.

3. The system for intelligently generating packaging box die plate design based on parameters according to claim 2, characterized in that: In the fitness calculation module, the fitness corresponding to the packaging box cutter design parameter simulation set is calculated based on the weight distribution set of the packaging box cutter design requirement items, which specifically includes the following steps: The packaging box blade design parameter simulation set is sent to the trained packaging box design demand prediction model for processing, and the packaging box design demand data set is output. The packaging box design demand data set includes the packaging box design demand data corresponding to the packaging box blade design demand items, and the packaging box design demand data set is traversed to perform fitness maximization processing, and the packaging box design demand data set to be calculated is constructed. The fitness maximization processing is specifically to perform inverse processing on the packaging box design demand data in the packaging box design demand data set that is opposite to the direction of improving fitness; the fitness δ corresponding to the packaging box blade design parameter simulation set is calculated by the following formula; ； where β i is the weight of the i-th packaging box cutter design requirement item in the packaging box cutter design requirement item weight distribution set, Q i is the i-th packaging box design requirement data to be calculated in the packaging box design requirement data set to be calculated, i=1, 2, 3, ..., I, I is the total number of packaging box knife design requirement items.

4. The system for intelligently generating packaging box die plate design based on parameters according to claim 3 is characterized in that: In the population iteration module, a recombination operation is performed based on the parent set and the crossover candidate set to construct a temporary set. In the recombination operation, the crossover probability of each packaging box cutter design parameter is adjusted based on the weight distribution set of the packaging box cutter design requirement items. Specifically, the steps include: Step T1: adjusting the crossover probability of each packaging box cutter design parameter based on the weight distribution set of packaging box cutter design requirement items; Step T1.1: When performing the first iteration, traverse H packaging box cutter design requirement items, and perform the following operations for the h-th packaging box cutter design requirement item to obtain the packaging box cutter design parameter association set corresponding to the h-th packaging box cutter design requirement item, the packaging box cutter design parameter association set includes the packaging box cutter design parameters related to the h-th packaging box cutter design requirement item, and calculate the pending crossover probability P of the packaging box cutter design parameters in the packaging box cutter design parameter association set m =M -1 W h , where m=1, 2, 3, …, M, W h is the weight of the packaging box cutter design requirement item corresponding to the h-th packaging box cutter design requirement item, and enters step T1.2; in the subsequent iterations except the first iteration, traverse all packaging box cutter design parameters, and perform the following operations for the m-th packaging box cutter design parameter to calculate the first mean square error ζ of the packaging box cutter design parameter value corresponding to the m-th packaging box cutter design parameter of the simulation set of all packaging box cutter design parameters in the population set m , judge m >D is true, D is the convergence threshold, if ζ m >D holds true, then calculate the pending crossover probability of the mth packaging box blade design parameters , where e is a natural constant, go to step T1.

2. If ζ m >D is not true, no operation; Step T1.2: Set the pending crossover probability P of all packaging box blade design parameters m Perform normalization processing to construct the crossover probability of each packaging box blade design parameter; Step T2: randomly select a packaging box cutter design parameter simulation set from the parent set and the crossover candidate set respectively, and the selected packaging box cutter design parameter simulation set will no longer be selected, for the two selected packaging box cutter design parameter simulation sets, based on the crossover probability of the packaging box cutter design parameters, select K packaging box cutter design parameters through the roulette algorithm, and exchange the packaging box cutter design parameter values ​​corresponding to the K packaging box cutter design parameters in the two selected packaging box cutter design parameter simulation sets to implement the reorganization operation until all the packaging box cutter design parameter simulation sets in the parent set are selected, and all the packaging box cutter design parameter simulation sets that have completed the reorganization operation and the parent set form a temporary storage set.

5. The system for intelligently generating packaging box die plate design based on parameters according to claim 4, characterized in that: In the population iteration module, a mutation operation is performed on all the simulation sets of packaging box die design parameters in the temporary set to construct a new population set. In the mutation operation, the mutation probability of each packaging box die design parameter is adjusted based on the weight distribution set of the packaging box die design requirement items. Specifically, the following steps are included: Step E1: adjusting the variation probability of each packaging box cutter design parameter based on the weight distribution set of packaging box cutter design requirement items; Step E1.1: When performing the first iteration, traverse H packaging box cutter design requirement items, and perform the following operations for the h-th packaging box cutter design requirement item to obtain the packaging box cutter design parameter association set corresponding to the h-th packaging box cutter design requirement item, the packaging box cutter design parameter association set includes the packaging box cutter design parameters related to the h-th packaging box cutter design requirement item, and calculate the pending mutation probability F of the packaging box cutter design parameters in the packaging box cutter design parameter association set m =M -1 W h , where m=1, 2, 3, …, M, W h is the weight of the packaging box cutter design requirement item corresponding to the h-th packaging box cutter design requirement item, and enters step E1.2; in the subsequent iterations except the first iteration, traverse all packaging box cutter design parameters, and perform the following operations for the m-th packaging box cutter design parameter to obtain the mean square error η of the packaging box cutter design parameter value corresponding to the m-th packaging box cutter design parameter of the simulation set of all packaging box cutter design parameters in the temporary collection m , judge η m >D is true, if η m >D holds true, then calculate the probability of variation to be processed for the design parameters of the mth packaging box blade , go to step E1.2, if η m >D is not true, no operation; Step E1.2: Traverse the packaging box cutter design parameter simulation sets in the temporary collection, and perform the following operations for each packaging box cutter design parameter simulation set in the temporary collection, generate a random value μ between 0 and 1 through a random function, and determine whether μ>Pc is established, Pc is the mutation threshold, if μ>Pc is not established, do not perform mutation operation on the selected packaging box cutter design parameter simulation set, if μ>Pc is established, select L packaging box cutter design parameters through the roulette algorithm based on the crossover probability of the packaging box cutter design parameters, and perform mutation operation on the packaging box cutter design parameter values ​​corresponding to the L packaging box cutter design parameters in the selected packaging box cutter design parameter simulation set, until all the packaging box cutter design parameter simulation sets in the temporary collection are traversed, and all the packaging box cutter design parameter simulation sets that have not performed mutation operation and those that have performed mutation operation are formed into a new population set.

6. The system for intelligently generating packaging box die plate design based on parameters according to claim 5, characterized in that: In the population iteration module, the associated set of packaging box cutter design parameters corresponding to the h-th packaging box cutter design requirement item is obtained, which specifically includes the following steps: Match the h-th packaging box cutting board design requirement item with the packaging box cutting board design requirement item in the packaging box cutting board design parameter association library, output all the packaging box cutting board design parameters corresponding to the h-th packaging box cutting board design requirement item, and form all the packaging box cutting board design parameters corresponding to the h-th packaging box cutting board design requirement item into a packaging box cutting board design parameter association set, and the packaging box cutting board design parameter association library includes the packaging box cutting board design requirement item and several corresponding packaging box cutting board design parameters; The method of constructing the packaging box cutter design parameter association library is as follows: obtain the trained packaging box design demand prediction model, and for each packaging box cutter design demand item, obtain the gradient value Y of the mth packaging box cutter design parameter in the packaging box design demand prediction model for the packaging box cutter design demand item. m , and judge Y m > Is X true? X is the gradient threshold. If Y m >X is established, the mth packaging box blade design parameters are corresponding to the packaging box blade design requirements. If Y m >X is not true, no operation; then all packaging box cutter design requirement items and their corresponding packaging box cutter design parameters are combined into a packaging box cutter design parameter association library.

7. The system for intelligently generating packaging box die plate design based on parameters according to claim 6, characterized in that: The named entity extraction model in the packaging box cutting design requirement document acquisition module is trained through the following steps: obtaining several packaging box cutting design requirement training samples, which include a packaging box cutting design requirement word set, and annotating the packaging box cutting design requirement training samples with packaging box cutting design requirement items and priority words, and forming all the annotated packaging box cutting design requirement training samples into a packaging box cutting design requirement training set, and training the named entity extraction model with the packaging box cutting design requirement training set. During the training, the annotation of packaging box cutting design requirement items and priority words is used as the target, and the packaging box cutting design requirement training loss value is calculated to determine whether the packaging box cutting design requirement training loss value is within the packaging box cutting design requirement confidence range. If the packaging box cutting design requirement training loss value is within the packaging box cutting design requirement confidence range, output the trained named entity extraction model; otherwise, continue to train the named entity extraction model with the packaging box cutting design requirement training set.

8. The system for intelligently generating packaging box die plate design based on parameters according to claim 7, characterized in that: The packaging box design demand prediction model in the fitness calculation module is trained through the following steps: obtaining packaging box design demand prediction training samples, the packaging box design demand prediction training samples include a packaging box blade design parameter data set, and the storage form of the packaging box blade design parameter data set is consistent with the storage form of the packaging box blade design parameter simulation set, labeling the packaging box design demand prediction training samples through the packaging box design demand data set, and forming all labeled packaging box design demand prediction training samples into a packaging box design demand prediction training set, and training the packaging box design demand prediction model through the packaging box design demand prediction training set. During the training, the packaging box design demand data set is used as the target, the packaging box design demand prediction loss value is calculated, and it is determined whether the packaging box design demand prediction loss value is within the packaging box design demand prediction confidence range. If the packaging box design demand prediction loss value is within the packaging box design demand prediction confidence range, the trained packaging box design demand prediction model is output; otherwise, the packaging box design demand prediction model continues to be trained through the packaging box design demand prediction training set.

Images