A building mold optimization system based on NSGA2

Abstract

The invention relates to the technical field of building mold production, and specifically discloses a building mold optimization system based on NSGA2, which is used to solve the problem that the existing precast concrete wall mold design is difficult to optimize and meet the cost, strength, material utilization rate and customer personalized needs at the same time, resulting in a disconnection between design and production and failure to efficiently meet the needs. The system comprises a mold design module and a mold production module, wherein the input end of the mold design is connected to a personalized demand extraction module, the mold design module and the mold production module are connected via a multi-objective optimization module, the mold production module transmits actual production data to the multi-objective optimization module via a feedback adjustment module, and the personalized demand extraction module generates target material type, target mold thickness and target size parameters from a customer order through an intelligent algorithm combined with a priori knowledge graph; the invention realizes the coordinated optimization of mold design and production through an NSGA2 algorithm and a preset intelligent algorithm, and automatically generates an optimal design scheme.

Description

Technical Field

[0001] The present invention relates to the technical field of building mold production, and more specifically, to a building mold optimization system based on NSGA2. Background Art

[0002] Building molds play an important role in modern construction. The quality of their design and production directly affects the shape accuracy, construction cost, construction period and resource utilization efficiency of building components. In particular, building molds for precast concrete walls are particularly important in current frame structures and prefabricated buildings. However, the design and optimization of building molds for precast concrete walls usually involve multiple conflicting goals, especially in terms of production cost, mold strength and material utilization, which poses certain technical challenges. In the mold production process, multiple design goals such as mold material thickness, geometry and material selection often conflict with each other. Existing mold design and production methods make it difficult to simultaneously and collaboratively optimize these multiple conflicting goals. In addition, when faced with personalized needs in customer orders, the existing design and production processes make it difficult to flexibly adjust the design parameters of the mold to meet the specific requirements of different projects. NSGA2 (Non-dominated Sorting Genetic Algorithm 2) is an evolutionary algorithm model for solving multi-objective optimization problems. It is used to deal with the problem of optimizing multiple conflicting objectives at the same time. It merges the parent and child generations to form a temporary population twice the size of the original population, and then selects the individuals of the next generation through non-dominated sorting. If the individuals at a certain non-dominated level exceed the remaining population capacity, individuals are selected according to the crowding sorting to retain diversity. Compared with the original NSGA algorithm, the amount of calculation is significantly reduced by introducing crowding sorting and fast non-dominated sorting. By calculating the crowding and giving priority to diverse solutions, the distribution of the Pareto optimal solution set is guaranteed to be uniform, which is helpful for solving the multi-objective optimization of the production process of concrete wall molds. How to extract the specific production requirements of precast concrete building molds from the personalized needs of customers based on NSGA2, break the barriers between mold design and production based on multi-objective optimization technology, avoid information isolation in the traditional design process, and realize multi-objective collaborative optimization of mold design and production has become an urgent need of the current industry.

[0003] In the construction molds of precast concrete walls, the steel molds used have high strength and durability, and are suitable for large-scale construction companies and infrastructure projects (such as bridges and high-rise buildings). Aluminum alloy molds are light and easy to carry, and are suitable for customers with low mold strength requirements and lightweight requirements. Composite molds (such as glass fiber reinforced plastics) have good corrosion resistance and lightweight characteristics, and are suitable for chemical plants, coastal areas or customers who need strong corrosion resistance. Standardized rectangular molds are suitable for standard walls in residential or commercial buildings, while concrete walls with curvature, decoration or special functions are suitable for architectural designers or high-end customized artistic buildings, museums and other places, used to emphasize the diversity of visual effects and functions. The thickness of the construction mold material determines the strength and rigidity of the mold. Increasing the material thickness can improve the rigidity and bearing capacity of the mold, which is suitable for contractors of large-scale infrastructure projects. Reducing the material thickness can reduce the mold weight and material cost, which is suitable for small and medium-sized construction projects and residential developers. Therefore, different construction molds have different personalized demand groups, and customers are not completely professional in the production and design knowledge of construction molds. Some customers can only write their corresponding needs in the order information, and cannot accurately describe the construction mold materials, mold shapes and mold thickness information they need. Therefore, how to extract the customer's personalized production and design needs for this type of mold from the customer's order has become a problem that needs to be solved urgently. Summary of the invention

[0004] In order to overcome the above-mentioned defects of the prior art, the present invention provides a building mold optimization system based on NSGA2, which realizes multi-objective collaborative optimization of mold design and production through a multi-objective genetic algorithm (NSGA2), automatically generates the optimal design solution, improves production efficiency, reduces costs, and meets diverse customer needs.

[0005] To achieve the above-mentioned purpose, the present invention provides the following technical solutions: a building mold optimization system based on NSGA2, including a mold design module and a mold production module, the input end of the mold design is connected to a personalized demand extraction module, the mold design module and the mold production module are connected through a multi-objective optimization module, the mold production module transmits actual production data to the multi-objective optimization module through a feedback adjustment module, the personalized demand extraction module generates target material type, target mold thickness and target size parameters from customer orders through an intelligent algorithm combined with a priori knowledge graph, the multi-objective optimization module inputs the target material type, target mold thickness and target size parameters, generates an initial population of design solutions for mold material type, thickness and size parameters, calculates the objective function values ​​of each individual production cost objective function, mold strength objective function, durability objective function, material utilization objective function and construction convenience objective function, and uses the non-dominated sorting mechanism in NSGA2 to sort the target function of each solution according to the objective function. The target function value is sorted, and the crowding distance of each solution is calculated. Through the binary tournament selection method, based on the non-dominated sorting and crowding distance, suitable individuals are selected from the current population for the next reproduction. A new population is generated through crossover and mutation operations. The objective function, non-dominated sorting and crowding calculation are performed on the new population, and it is merged with the old population. The new generation population is selected through the elite strategy. When the specified number of iterations is reached or the Pareto front converges, the optimization process is stopped, and the Pareto optimal solution is obtained through the user-defined priority. The mold design module generates CAD drawings and 3D models based on the selected Pareto optimal solution. The mold production module adjusts the execution parameters of the precast concrete wall mold production equipment based on the selected Pareto optimal solution. The new data generated in the production process is transmitted to the multi-objective optimization module through the feedback adjustment module. The multi-objective optimization module performs dynamic optimization based on the new feedback data to further improve the mold production process.

[0006] As a further solution of the present invention, in the mold production module, the adjusted execution parameters of the precast concrete wall mold production equipment include but are not limited to the spindle speed, tool type, tool replacement frequency of the CNC processing equipment, the molding equipment type, molding parameters, the coating type and thickness, spraying pressure and time of the surface treatment equipment.

[0007] As a further scheme of the present invention, in the mold production module, the spindle speed of the CNC machining equipment is positively correlated with the cutting speed and negatively correlated with the tool diameter. The tool type of the CNC machining equipment is adapted according to the mold material and geometric shape. The tool replacement frequency of the CNC machining equipment is selected based on the wear limit of the tool and the cutting speed ratio. When producing metal molds, the molding equipment is a die-casting equipment. When producing plastic and composite molds, the molding equipment is an injection molding equipment. The coating type and thickness of the surface treatment equipment are adaptively adjusted according to the use environment of the mold. The spraying pressure and time are adaptively selected according to the hardness of the mold material. The spraying time is adaptively adjusted according to the surface area and spraying thickness of the mold.

[0008] As a further solution of the present invention, in the personalized demand extraction module, the specific implementation process of generating target material type, target mold thickness and target size parameters from customer orders through intelligent algorithms combined with prior knowledge graphs includes:

[0009] Step 1, triggering the recognition condition, and performing the first semantic analysis model analysis: segmenting the customer order by word segmentation, based on the number of words and the number of target words, and considering the degree of support of the information of each target word in the context for the order content, evaluating the degree of matching between the customer order and the precast concrete mold, and obtaining the target material type, target mold thickness and target size parameters of the precast concrete wall mold in the customer order based on the first semantic analysis model, otherwise executing step 2;

[0010] Step 2, performing a second semantic analysis model analysis: obtaining the environmental requirements, material requirements, size requirements and customer required functions of the precast concrete wall mold in the customer order based on the second semantic analysis model;

[0011] Step 3, first knowledge graph matching and retrieval: Use the first knowledge graph containing historical data, standardization requirements, and industry historical practice data related to the mold design to match and screen environmental requirements, material requirements, and size requirements to generate pre-selected material types, mold thickness ranges, and size parameter ranges;

[0012] Step 4, second knowledge graph matching and retrieval: using the second knowledge graph containing actual construction data, production feedback and cost-benefit analysis information in the industry, match and screen environmental requirements, material requirements, size requirements and other special functions to generate a production cost requirement range and a production difficulty requirement range;

[0013] Step 5, third knowledge graph matching and retrieval: Using the third knowledge graph containing technical specifications, material application guidelines, construction processes and international building standards, the pre-selected material types, mold thickness ranges and size parameter ranges are searched and matched under the constraints of the production cost requirement range and the production difficulty requirement range to generate the target material type, target mold thickness and target size parameters.

[0014] In step 1, the text representation and relationship of the document and topic are combined to make full use of context information. After the combination, a multi-layer perceptron and a linear layer are used to predict the labels of the segmented vocabulary. The labels of the segmented vocabulary include material type, mold thickness, mold size, surface treatment, environmental requirements, functional requirements, cost constraints, unrecognized vocabulary, size units, and material units. The processing flow of the multi-layer perceptron and linear layer includes:

[0015] Step 21, input text representation: customer orders are processed by a word segmentation tool to generate word segmentation results, and each word is converted into a vector representation through an embedding layer (BERT). The embedding vector of each word not only contains the semantic information of the word, but also combines the context to capture the semantic association between words;

[0016] Step 22: Contextual information integration: Use representations based on context-aware models, such as the BERT model, to ensure that the meaning of each word in context is fully understood;

[0017] Step 23, multi-layer perceptron processing: Process the embedded vector and pass it to the multi-layer perceptron. The multi-layer perceptron further transforms the input vocabulary vector into a higher-level semantic representation through multi-layer nonlinear mapping to extract more complex features.

[0018] Step 24, linear layer output: After passing through the multi-layer perceptron, the new features are passed to the final linear layer to predict the label of each word. After passing through the linear layer, the softmax function is used to convert the result into a probability distribution, and finally the label of each word is predicted. According to the label classification results of each word, the target material type, target thickness and size parameters of the mold are extracted.

[0019] In step 1, the total score in the trigger condition for triggering the first semantic analysis model is the ratio of the number of target words to the total number of words in the customer order, the weighted sum of the relevance score of the target words and the context information score. The formula for the total score in the trigger condition is:

[0020] ;

[0021] Where: is the total score, which is used to decide whether to trigger the first semantic analysis model. is the target vocabulary size, is the total number of words in the customer's order, Score the relevance of the target word, indicating the relevance between the word in the order and the precast concrete wall mold (a value between 0 and 1, 1 being completely relevant), Score the context information, indicating the support degree of the target vocabulary context information in the order to the order content (a value between 0 and 1), , , is the weight parameter, which is adjusted through analysis of historical order data.

[0022] in, , is the index of the target vocabulary, For the target vocabulary, For the The similarity between the target vocabulary and the precast concrete domain is calculated using the domain vocabulary similarity database;

[0023] , where For the The semantic similarity score between a target word and its context is calculated using the contextual language model.

[0024] As a further scheme of the present invention, in the multi-objective optimization module, the production cost objective function is the sum of material cost, processing cost, surface treatment cost and energy consumption processing cost, the mold strength objective function is the ratio of the yield strength of the material to the maximum working stress of the mold during use, the durability objective function is the tested service life of the material plus the extended service life of the surface treatment minus the duration of durability weakening due to environmental factors, the material utilization objective function is the ratio of the volume of effectively used materials to the total material volume, and the construction convenience objective function is the reciprocal of the sum of the standardized values ​​of the total weight of the mold, the mold installation time and the mold disassembly time.

[0025] As a further solution of the present invention, in the multi-objective optimization module, when the non-dominated sorting layers are the same, the solution with a larger crowding distance between two solutions at the same non-dominated level is preferentially selected, and the formula is:

[0026] exist hour, ;

[0027] Where: , are two solutions in the same non-dominated hierarchy obtained by NSGA2, , The solution is , The corresponding crowding distance is ,in is the target function index, , The solution is The neighboring solutions on this objective function are: , The solution is In the The objective function values ​​corresponding to adjacent solutions on the objective function are: , Respectively The maximum and minimum values ​​of the objective function, The same method , has precedence over the operator.

[0028] As a further solution of the present invention, the mold design module includes a CAD geometric design and modeling module and a finite element analysis and structural optimization module. The CAD geometric design and modeling module automatically generates the geometric shape of the mold according to the Pareto optimal solution, adds preset holes, grooves and connection ports according to the mold design requirements, uses Boolean operations to add and subtract different geometric shapes, obtains the hollowing and depressions of the mold, automatically sets the mold thickness and material properties, and transmits the generated model to the finite element analysis and structural optimization module. The finite element analysis and structural optimization module performs stress analysis and deformation analysis based on the finite element analysis mold, adjusts the support structure and thickness of the model according to the analysis results, generates an optimized precast concrete mold model, and outputs the model as a CAD drawing and a 3D model.

[0029] As a further solution of the present invention, in the multi-objective optimization module, the crossover operation of NSGA2 is a single-point crossover operation, and the mutation operation is to generate diversified solutions by changing individual decision variables by a preset amplitude.

[0030] As a further solution of the present invention, in the multi-objective optimization module, the specific implementation process of generating diversified solutions by changing individual decision variables by preset amplitudes is as follows:

[0031] Step 11, define decision variables: select decision variables from material type, mold thickness, size parameters, surface treatment type and production process parameters;

[0032] Step 22, select the mutation method according to the selected decision variables: for discrete variables, random reselection is used to achieve mutation. If the mutated value is not in the specified set, a value that meets the specified set is reselected to replace it. For continuous variables, Gaussian perturbation is used to perform Gaussian random perturbation on the value of the current solution. During the mutation process, if the mutated value exceeds the domain of definition, the value is corrected to the selected boundary value.

[0033] Compared with the prior art, the technical effects and advantages of the present invention are as follows: the personalized demand extraction module of the present invention automatically generates the target material type, mold thickness and size parameters based on the information in the customer order, and uses the NSGA2 algorithm to achieve multi-objective optimization of mold design, ensuring that the customized needs of different customers are met. By dynamically adjusting the mold design and production equipment parameters, the coordinated optimization of the mold design and production process is achieved, avoiding the problem of information fragmentation in the traditional process. The Pareto optimal solution finally generated is used for the mold design module, and CAD drawings and 3D models are automatically generated, which significantly improves the efficiency of mold design and the adaptability of production, while reducing material waste and production costs, meeting the diverse requirements of large projects and customers with special needs, improving the flexibility and reliability of the production process, and solving the trade-off between multi-objective conflicts in the design and production of precast concrete wall building molds. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 This is a flow chart of the NSGA2 algorithm concept in the system proposed by the present invention;

[0035] Figure 2 Schematic diagram of the congestion distance calculation for each solution in the system proposed by the present invention. DETAILED DESCRIPTION

[0036] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0037] like Figure 1 to Figure 2 As shown, the present invention proposes a building mold optimization system based on NSGA2. The personalized demand extraction module in the system automatically generates the target material type, mold thickness and size parameters based on the information in the customer order, and uses the NSGA2 algorithm to achieve multi-objective optimization of the mold design to ensure that the customized needs of different customers are met. By dynamically adjusting the mold design and production equipment parameters, the coordinated optimization of the mold design and production process is achieved, avoiding the problem of information fragmentation in the traditional process. The Pareto optimal solution finally generated is used for the mold design module to automatically generate CAD drawings and 3D models, which significantly improves the efficiency of mold design and the adaptability of production, while reducing material waste and production costs, meeting the diverse requirements of large projects and customers with special needs, improving the flexibility and reliability of the production process, and solving the trade-off between the multi-objective conflict problems in the design and production of precast concrete wall building molds.

[0038] like Figure 1 As shown in the figure, the NSGA2 algorithm generates a temporary optimal population from the parent population Pt through a tournament mechanism. Then, crossover and mutation operations are performed to generate a child population Qt. The parent population Pt and the child population Qt are recombined into a larger population of size 2N. Each individual in the population is divided into several non-inferior frontiers through a fast non-dominated sort. Finally, NSGA2 maintains the diversity of the population by calculating the crowding distance of each individual. For a multi-objective evolutionary algorithm, it is not only hoped that it can quickly converge to the non-dominated optimal solution of the desired Pareto frontier set, but also that the non-dominated optimal solution can maintain good uniformity. The NSGA2 algorithm can maintain the diversity of the population by comparing the crowding distances of individuals. Before calculating the crowding distance, the population is divided into several layers according to the objective function value through a fast non-dominated sort. Then, the infinite crowding distance d=∞ between the five objective function solutions is assigned to ensure that five solutions on the boundary can be selected for the next generation each time.

[0039] The system proposed in the present invention includes: a mold design module and a mold production module. The input end of the mold design is connected to a personalized demand extraction module. The mold design module and the mold production module are connected through a multi-objective optimization module. The mold production module transmits actual production data to the multi-objective optimization module through a feedback adjustment module. The personalized demand extraction module generates target material type, target mold thickness and target size parameters from customer orders through an intelligent algorithm combined with a priori knowledge graph. The multi-objective optimization module inputs the target material type, target mold thickness and target size parameters, generates an initial population of design solutions for mold material type, thickness and size parameters, calculates the objective function values ​​of each individual production cost objective function, mold strength objective function, durability objective function, material utilization objective function and construction convenience objective function, and sorts each solution according to its objective function value through the non-dominated sorting mechanism in NSGA2 to calculate the ownership of each solution. Crowding distance, through the binary tournament selection method, based on non-dominated sorting and crowding distance, select suitable individuals from the current population for the next reproduction, generate a new population through crossover and mutation operations, calculate the objective function, non-dominated sorting and crowding degree of the new population, and merge it with the old population, select a new generation of population through the elite strategy, and stop the optimization process when the specified number of iterations is reached or the Pareto front converges. The Pareto optimal solution is obtained through the user-defined priority. The mold design module generates CAD drawings and 3D models based on the selected Pareto optimal solution. The mold production module adjusts the execution parameters of the precast concrete wall mold production equipment based on the selected Pareto optimal solution. The new data generated in the production process is transmitted to the multi-objective optimization module through the feedback adjustment module. The multi-objective optimization module performs dynamic optimization based on the new feedback data to further improve the mold production process.

[0040] It should be noted that in the mold production module, the adjusted execution parameters of the precast concrete wall mold production equipment include but are not limited to the spindle speed, tool type, tool replacement frequency of the CNC processing equipment, the type and molding parameters of the molding equipment, the coating type and thickness, spraying pressure and time of the surface treatment equipment.

[0041] It should be noted that in the mold production module, the spindle speed of the CNC machining equipment is positively correlated with the cutting speed and negatively correlated with the tool diameter. The tool type of the CNC machining equipment is adapted according to the mold material and geometry. The tool replacement frequency of the CNC machining equipment is selected based on the tool wear limit and the cutting speed ratio. When producing metal molds, the molding equipment is a die-casting equipment, and when producing plastic and composite molds, the molding equipment is an injection molding equipment. The coating type and thickness of the surface treatment equipment are adaptively adjusted according to the use environment of the mold, the spraying pressure and time are adaptively selected according to the hardness of the mold material, and the spraying time is adaptively adjusted according to the surface area and spraying thickness of the mold.

[0042] It should be noted that in the personalized demand extraction module, the specific implementation process of generating target material type, target mold thickness and target size parameters from customer orders through intelligent algorithms combined with prior knowledge graphs includes:

[0043] Step 1, triggering the recognition condition, and performing the first semantic analysis model analysis: segmenting the customer order by word segmentation, based on the number of words and the number of target words, and considering the degree of support of the information of each target word in the context for the order content, evaluating the degree of matching between the customer order and the precast concrete mold, and obtaining the target material type, target mold thickness and target size parameters of the precast concrete wall mold in the customer order based on the first semantic analysis model, otherwise executing step 2;

[0044] Step 2, performing a second semantic analysis model analysis: obtaining the environmental requirements, material requirements, size requirements and customer required functions of the precast concrete wall mold in the customer order based on the second semantic analysis model;

[0045] Step 3, first knowledge graph matching and retrieval: Use the first knowledge graph containing historical data, standardization requirements, and industry historical practice data related to the mold design to match and screen environmental requirements, material requirements, and size requirements to generate pre-selected material types, mold thickness ranges, and size parameter ranges;

[0046] Step 4, second knowledge graph matching and retrieval: using the second knowledge graph containing actual construction data, production feedback and cost-benefit analysis information in the industry, match and screen environmental requirements, material requirements, size requirements and other special functions to generate a production cost requirement range and a production difficulty requirement range;

[0047] Step 5, third knowledge graph matching and retrieval: Using the third knowledge graph containing technical specifications, material application guidelines, construction processes and international building standards, the pre-selected material types, mold thickness ranges and size parameter ranges are searched and matched under the constraints of the production cost requirement range and the production difficulty requirement range to generate the target material type, target mold thickness and target size parameters.

[0048] In step 1, the text representation and relationship of the document and topic are combined to make full use of context information. After the combination, a multi-layer perceptron and a linear layer are used to predict the labels of the segmented vocabulary. The labels of the segmented vocabulary include material type, mold thickness, mold size, surface treatment, environmental requirements, functional requirements, cost constraints, unrecognized vocabulary, size units, and material units. The processing flow of the multi-layer perceptron and linear layer includes:

[0049] Step 21, input text representation: customer orders are processed by a word segmentation tool to generate word segmentation results, and each word is converted into a vector representation through an embedding layer (BERT). The embedding vector of each word not only contains the semantic information of the word, but also combines the context to capture the semantic association between words;

[0050] Step 22: Contextual information integration: Use representations based on context-aware models, such as the BERT model, to ensure that the meaning of each word in context is fully understood;

[0051] Step 23, multi-layer perceptron processing: Process the embedded vector and pass it to the multi-layer perceptron. The multi-layer perceptron further transforms the input vocabulary vector into a higher-level semantic representation through multi-layer nonlinear mapping to extract more complex features.

[0052] Step 24, linear layer output: After passing through the multi-layer perceptron, the new features are passed to the final linear layer to predict the label of each word. After passing through the linear layer, the softmax function is used to convert the result into a probability distribution, and finally the label of each word is predicted. According to the label classification results of each word, the target material type, target thickness and size parameters of the mold are extracted.

[0053] In step 1, the total score in the trigger condition for triggering the first semantic analysis model is the ratio of the number of target words to the total number of words in the customer order, the weighted sum of the relevance score of the target words and the context information score. The formula for the total score in the trigger condition is:

[0054] ;

[0055] Where: is the total score, which is used to decide whether to trigger the first semantic analysis model. is the target vocabulary size, is the total number of words in the customer's order, Score the relevance of the target word, indicating the relevance between the word in the order and the precast concrete wall mold (a value between 0 and 1, 1 being completely relevant), Score the context information, indicating the support degree of the target vocabulary context information in the order to the order content (a value between 0 and 1), , , is the weight parameter, which is adjusted through analysis of historical order data.

[0056] in, , is the index of the target vocabulary, For the target vocabulary, For the The similarity between the target vocabulary and the precast concrete domain is calculated using the domain vocabulary similarity database;

[0057] , where For the The semantic similarity score between a target word and its context is calculated using the contextual language model.

[0058] By calculating the sum of the distances between the objective function values ​​of each two adjacent individuals, the crowding distance of the middle individual is obtained. Figure 2 As shown, through The crowding distance is calculated based on the calculation principle of

[0059] It should be noted that in the multi-objective optimization module, the production cost objective function is the sum of material cost, processing cost, surface treatment cost and energy consumption processing cost; the mold strength objective function is the ratio of the yield strength of the material to the maximum working stress of the mold during use; the durability objective function is the tested service life of the material plus the extended service life of the surface treatment minus the duration of durability weakening due to environmental factors; the material utilization objective function is the ratio of the volume of effectively used materials to the total material volume; and the construction convenience objective function is the reciprocal of the sum of the standardized values ​​of the total weight of the mold, the mold installation time and the mold disassembly time.

[0060] It should be noted that in the multi-objective optimization module, when the non-dominated sorting layers are the same, the solution with the larger crowding distance between the two solutions at the same non-dominated level is preferred. The formula is:

[0061] exist hour, ;

[0062] Where: , are two solutions in the same non-dominated hierarchy obtained by NSGA2, , The solution is , The corresponding crowding distance is ,in is the target function index, , The solution is The neighboring solutions on this objective function are: , The solution is In the The objective function values ​​corresponding to adjacent solutions on the objective function are: , Respectively The maximum and minimum values ​​of the objective function, The same method , has precedence over the operator.

[0063] It should be noted that the mold design module includes a CAD geometric design and modeling module and a finite element analysis and structural optimization module. The CAD geometric design and modeling module automatically generates the geometric shape of the mold according to the Pareto optimal solution, adds preset holes, grooves and connectors according to the mold design requirements, uses Boolean operations to add and subtract different geometric shapes, obtains the hollowing and depressions of the mold, automatically sets the mold thickness and material properties, and transmits the generated model to the finite element analysis and structural optimization module. The finite element analysis and structural optimization module performs stress analysis and deformation analysis based on the finite element analysis mold, adjusts the support structure and thickness of the model according to the analysis results, generates an optimized precast concrete mold model, and outputs the model as a CAD drawing and a 3D model.

[0064] It should be noted that in the multi-objective optimization module, the crossover operation of NSGA2 is a single-point crossover operation, and the mutation operation is to generate diversified solutions by changing the individual decision variables by a preset amplitude.

[0065] It should be noted that in the multi-objective optimization module, the specific implementation process of generating diversified solutions by changing individual decision variables by preset amplitudes is as follows:

[0066] Step 11, define decision variables: select decision variables from material type, mold thickness, size parameters, surface treatment type and production process parameters;

[0067] Step 22, select the mutation method according to the selected decision variables: for discrete variables, random reselection is used to achieve mutation. If the mutated value is not in the specified set, a value that meets the specified set is reselected to replace it. For continuous variables, Gaussian perturbation is used to perform Gaussian random perturbation on the value of the current solution. During the mutation process, if the mutated value exceeds the domain of definition, the value is corrected to the selected boundary value.

[0068] In order to clearly illustrate the technical effect of the above technical solution, it is now described and explained in detail through embodiments.

[0069] Example. Background: A construction company needs to produce precast concrete wall molds for high-rise buildings. The customer's requirements include high strength, durability, adaptability to different environmental conditions, and the desire to reduce production costs and material waste. The customer's order does not clearly state the material type and mold thickness, but only states that the mold needs to withstand high loads, environmental adaptability, and minimize costs.

[0070] The personalized demand extraction module automatically extracts and generates the following preliminary target information by analyzing customer orders:

[0071] Target material type: The customer order does not specify the mold material type. The personalized demand extraction module combines the prior knowledge graph to infer the material type from the customer's environmental requirements and strength requirements. Based on the high strength requirement, stainless steel is recommended as the primary material type.

[0072] Target mold thickness: Based on the high load bearing capacity mentioned in the order and combined with the prior knowledge graph, the system recommends a mold thickness of 15mm to ensure the strength of the mold.

[0073] Target size parameters: The order specifies the basic size of the mold. The length, width and height of the mold are automatically extracted from the order, which are: 3000mm×1000mm 200mm.

[0074] After receiving the target material type, mold thickness and size parameters, the multi-objective optimization module generates an initial population and calculates the following objective functions:

[0075] Production cost objective function: Calculated based on material cost, processing cost, surface treatment cost and energy consumption cost:

[0076] 𝑓1=Material Cost+Processing Cost+Surface Treatment Cost+Energy Cost;

[0077] The production cost of each design was estimated based on the cost of stainless steel, CNC machining and surface treatment, combined with energy consumption.

[0078] Mold strength objective function: Based on the ratio of material yield strength to mold working stress,

[0079] 𝑓2=Yield Strength / Max Working Stress

[0080] The strength performance of each design solution is calculated based on the yield strength of stainless steel and the maximum working stress in the mold use environment.

[0081] Durability objective function: Extend the life according to the material service life and surface treatment, minus the impact of environmental factors,

[0082] 𝑓3=Base Life+Surface Treatment Life−Environmental Degradation

[0083] The tested service life of the stainless steel material is combined with the extended life after surface treatment, minus the life loss in high humidity or corrosive environments.

[0084] Material utilization objective function: the ratio of the effective material volume to the total material volume,

[0085] 𝑓4=Effective Material Volume / Total Material Volume

[0086] Optimize material utilization and reduce waste for each design solution.

[0087] Construction convenience objective function: the inverse of the sum of the standardized values ​​of mold weight, installation time and disassembly time,

[0088] 𝑓5=1 / (TotalWeight+Installation Time+Disassembly Time)

[0089] The total weight of the mold affects the ease of handling and installation. The construction ease is calculated by the standardized total weight and installation / disassembly time.

[0090] NSGA2 optimization process

[0091] The above objective function is optimized through the multi-objective optimization of the NSGA2 algorithm:

[0092] Generate initial population: Generate multiple design schemes based on the initially extracted mold material, thickness and size parameters.

[0093] Non-dominated sorting and crowding distance calculation: Perform non-dominated sorting on the individuals in the initial population, calculate the crowding distance, and select the scheme with greater diversity.

[0094] Binary tournament selection, crossover, and mutation:

[0095] Use single-point crossover operations to generate new design solutions, and generate diversified solutions by slightly changing the mold thickness or material parameters (such as performing Gaussian perturbation on the thickness of 15mm to change it to 14.8mm or 15.2mm).

[0096] Perform multiple rounds of iterations, calculate the objective function value of the new population, merge it with the old population, and select the next generation of population through the elite strategy.

[0097] Pareto optimal solution selection

[0098] After several rounds of iterations, the NSGA2 algorithm converges to the Pareto frontier. The system selects the Pareto optimal solution based on the customer's priorities (such as production cost and mold strength) and generates the final design solution:

[0099] Material Type: Stainless Steel.

[0100] Mould thickness: 15mm.

[0101] Mould size: 3000mm × 1000mm × 200mm.

[0102] Mould design and production module

[0103] CAD geometry design and modeling module: Based on the Pareto optimal solution, automatically generate the geometry of the mold, including preset holes, grooves and connections. Use Boolean operations to add and subtract geometric structures and set the thickness and material properties of the mold.

[0104] Finite element analysis and structural optimization module: perform stress analysis and deformation analysis, fine-tune the mold support structure and thickness, and output the optimized mold model as CAD drawings and 3D models.

[0105] In the mold production module, CNC processing equipment parameter adjustment: the system automatically adjusts the spindle speed of the CNC equipment according to the optimal solution (positively correlated with the cutting speed and negatively correlated with the tool diameter), selects the tool type suitable for stainless steel materials, and adjusts the tool replacement frequency according to the cutting speed.

[0106] Molding equipment and surface treatment equipment: The system selects die-casting molding equipment for stainless steel molds and adaptively adjusts molding parameters; the surface treatment equipment selects corrosion-resistant coatings and adjusts spraying thickness and pressure according to the use environment of stainless steel materials.

[0107] Feedback adjustment and dynamic optimization

[0108] During the production process, the system obtains actual production data, such as material consumption and processing time, through the feedback adjustment module, and transmits it back to the multi-objective optimization module in real time for dynamic optimization, further improving the mold production process to improve production efficiency and quality.

[0109] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0110] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A building mold optimization system based on NSGA2, including a mold design module and a mold production module, characterized in that: The input end of the mold design is connected to the personalized demand extraction module. The mold design module is connected to the mold production module through the multi-objective optimization module. The mold production module transmits the actual production data to the multi-objective optimization module through the feedback adjustment module. The personalized demand extraction module generates the target material type, target mold thickness and target size parameters from the customer order through the intelligent algorithm combined with the prior knowledge graph. The multi-objective optimization module inputs the target material type, target mold thickness and target size parameters to generate the initial population of design solutions for mold material type, thickness and size parameters, calculates the objective function values ​​of each individual production cost objective function, mold strength objective function, durability objective function, material utilization objective function, and construction convenience objective function, and sorts each solution according to its objective function value through the non-dominated sorting mechanism in NSGA2 to calculate each solution. Crowding distance, through the binary tournament selection method, select suitable individuals from the current population for the next reproduction, generate a new population through crossover and mutation operations, calculate the objective function, non-dominated sorting and crowding degree of the new population, and merge it with the old population, select a new generation of population through the elite strategy, and stop the optimization process when the specified number of iterations is reached or the Pareto front converges. The Pareto optimal solution is obtained through the user-defined priority. The mold design module generates CAD drawings and 3D models based on the selected Pareto optimal solution. The mold production module adjusts the execution parameters of the precast concrete wall mold production equipment based on the selected Pareto optimal solution. The new data generated in the production process is transmitted to the multi-objective optimization module through the feedback adjustment module, and the multi-objective optimization module performs dynamic optimization based on the new feedback data.

2. The construction mold optimization system based on NSGA2 according to claim 1 is characterized in that: In the mold production module, the adjusted execution parameters of the precast concrete wall mold production equipment include the spindle speed, tool type, tool replacement frequency of the CNC processing equipment, the type and molding parameters of the molding equipment, the coating type and thickness, spraying pressure and time of the surface treatment equipment.

3. The construction mold optimization system based on NSGA2 according to claim 2 is characterized in that: In the mold production module, the spindle speed of the CNC machining equipment is positively correlated with the cutting speed and negatively correlated with the tool diameter. The tool type of the CNC machining equipment is adapted according to the mold material and geometry. The tool replacement frequency of the CNC machining equipment is selected based on the tool wear limit and the cutting speed ratio.

4. The construction mold optimization system based on NSGA2 according to claim 1 is characterized in that: In the personalized demand extraction module, the process of generating target material type, target mold thickness and target size parameters from customer orders through intelligent algorithms combined with prior knowledge graphs includes: Step 1, triggering the recognition condition, and performing the first semantic analysis model analysis: segmenting the customer order by word segmentation, based on the number of words and the number of target words, and considering the degree of support of the information of each target word in the context for the order content, evaluating the degree of matching between the customer order and the precast concrete mold, and obtaining the target material type, target mold thickness and target size parameters of the precast concrete wall mold in the customer order based on the first semantic analysis model, otherwise executing step 2; Step 2, performing a second semantic analysis model analysis: obtaining the environmental requirements, material requirements, size requirements and customer required functions of the precast concrete wall mold in the customer order based on the second semantic analysis model; Step 3, first knowledge graph matching and retrieval: Use the first knowledge graph containing historical data, standardization requirements, and industry historical practice data related to the mold design to match and screen environmental requirements, material requirements, and size requirements to generate pre-selected material types, mold thickness ranges, and size parameter ranges; Step 4, second knowledge graph matching and retrieval: using the second knowledge graph containing actual construction data, production feedback and cost-benefit analysis information in the industry, match and screen environmental requirements, material requirements, size requirements and customer-required functions to generate a production cost requirement range and a production difficulty requirement range; Step 5, third knowledge graph matching and retrieval: Using the third knowledge graph containing technical specifications, material application guidelines, construction processes and international building standards, the pre-selected material types, mold thickness ranges and size parameter ranges are searched and matched under the constraints of the production cost requirement range and the production difficulty requirement range to generate the target material type, target mold thickness and target size parameters.

5. The construction mold optimization system based on NSGA2 according to claim 4 is characterized in that: In the multi-objective optimization module, the production cost objective function is the sum of material cost, processing cost, surface treatment cost and energy consumption processing cost; the mold strength objective function is the ratio of the yield strength of the material to the maximum working stress of the mold during use; the durability objective function is the tested service life of the material plus the extended service life of the surface treatment minus the duration of durability weakening due to environmental factors; the material utilization objective function is the ratio of the volume of effectively used materials to the total material volume; and the construction convenience objective function is the reciprocal of the sum of the standardized values ​​of the total weight of the mold, the mold installation time and the mold disassembly time.

6. The construction mold optimization system based on NSGA2 according to claim 5 is characterized in that: In the multi-objective optimization module, when the non-dominated sorting layers are the same, the solution with the larger crowding distance between the two solutions in the same non-dominated level is preferred. The formula is: exist hour, ; Where: , are two solutions in the same non-dominated hierarchy obtained by NSGA2, , The solution is , The corresponding crowding distance is ,in is the target function index, , The solution is The neighboring solutions on this objective function are: , The solution is In the The objective function values ​​corresponding to adjacent solutions on the objective function are: , Respectively The maximum and minimum values ​​of the objective function, The same method , has precedence over the operator.

7. The NSGA2-based building mold optimization system according to claim 6 is characterized in that: The mold design module includes a CAD geometry design and modeling module and a finite element analysis and structural optimization module. The CAD geometry design and modeling module automatically generates the geometric shape of the mold according to the Pareto optimal solution, adds preset holes, grooves and connectors according to the mold design requirements, uses Boolean operations to add and subtract different geometric shapes, obtains the hollowing and depressions of the mold, automatically sets the mold thickness and material properties, and transmits the generated model to the finite element analysis and structural optimization module. The finite element analysis and structural optimization module performs stress analysis and deformation analysis based on the finite element analysis mold, adjusts the support structure and thickness of the model according to the analysis results, generates an optimized precast concrete mold model, and outputs the model as a CAD drawing and a 3D model.

8. The construction mold optimization system based on NSGA2 according to claim 1 is characterized in that: In the multi-objective optimization module, the crossover operation of NSGA2 is a single-point crossover operation, and the mutation operation is to generate diversified solutions by changing the individual decision variables by a preset amplitude.

9. The construction mold optimization system based on NSGA2 according to claim 1 is characterized in that: In the multi-objective optimization module, the specific implementation process of generating diversified solutions by changing individual decision variables by preset amplitudes is as follows: Step 11, define decision variables: select decision variables from material type, mold thickness, size parameters, surface treatment type and production process parameters; Step 22, select the mutation method according to the selected decision variables: for discrete variables, random reselection is used to achieve mutation. If the mutated value is not in the specified set, a value that conforms to the specified set is reselected to replace it. For continuous variables, Gaussian perturbation is used to perform Gaussian random perturbation on the value of the current solution. During the mutation process, if the mutated value exceeds the domain of definition, the mutated value is corrected to the selected boundary value.

Images