Intelligent design system and method for plasticizing screw based on material characteristics

Abstract

The present application relates to a material property-based plasticizing screw intelligent design system and method, belonging to the field of material processing and forming. The material property-based plasticizing screw intelligent design system comprises a material property database, a screw structure parameter database, an experimental design module, a computer-aided design module, a computer-aided numerical simulation module, a screw plasticizing performance evaluation module, a screw structure optimization module, an optimal screw structure machine learning module, and an optimal screw structure generation module. The system and method can output optimal plasticizing screw structure geometric parameter data based on the plasticizing performance evaluation objective standard by only inputting material property parameters, and a screw model can be established according to the output screw structure geometric parameter data, which is used for the next step of screw processing and manufacturing and applied to the plasticizing production of the corresponding material.

Description

Technical Field

[0001] The present invention relates to a material property-based plasticizing screw intelligent design system and method, and belongs to the field of material processing and molding. Background Art

[0002] In the modern plastics industry, injection molding and extrusion are the primary methods for mass-producing plastic products. The specifications, screw configuration, mold structure, material properties, and various configurable process parameters of the injection molding machine and extruder significantly impact the quality of the final product. Plasticization is essential to the plastic molding process. Plastic is fed into the barrel of the injection molding machine or extruder through a hopper. Plasticization, the transformation of the plastic from a solid to a liquid state, is achieved primarily through heat generated and transferred by external barrel heaters and the rotational shear of the plasticizing screw within the barrel. As one of the most critical core components, the plasticizing screw significantly impacts the performance of plastic molding equipment and products.

[0003] The plasticizing capacity of the screw directly affects the plastic processing cycle and production efficiency. Based on the rheological properties of the material, a flow model is established under isothermal conditions in the cavity flow channel established by the screw and barrel, namely the melt conveying theory, which can be used to describe the plasticizing capacity of the screw. The screw melt conveying theory model shows that under certain process parameters (such as barrel temperature and screw speed), the plasticizing capacity of the screw is related to the machinability of the material and the structure and geometric dimensions of the screw. According to the different thermophysical properties, crystallization morphology and rheological properties of the processed polymer materials and their composite materials, the plasticizing screw should be designed with different structures and geometric dimensions to achieve effective plasticization and improve production efficiency. The plasticizing screw is usually divided into three sections in the axial length, namely the conveying section, the compression section and the metering section, to match the three stages of the plasticization process of the material from solid to liquid. The plasticization process of the material in the screw is: 1) Plastic particles are added to the barrel of the injection molding machine from the hopper, and the solid plastic particles are initially transported and compacted in the conveying section of the screw; 2) With the rotation of the screw and the action of the axial force of the screw ribs, the plastic gradually enters the compression section of the screw and is compacted more tightly; at the same time, under the action of the heat transmitted from the heater outside the barrel, part of the plastic above the screw groove is transformed from solid to liquid to form a molten film; under the action of the shear force generated by the rotation of the screw, the molten film gradually expands and a molten pool is generated at the screw ribs. The closer to the front end of the screw push, the larger the molten pool is, and the smaller the solid bed formed by the solid plastic is; 3) Under the action of the axial force of the screw, the plastic enters the metering section of the screw, and the solid plastic particles are completely converted into liquid. Under the precise size control of the metering section, accurate metering and flow control of the liquid plastic can be achieved. The screw groove of the conveying section is the deepest, which mainly plays the role of transporting and initially compacting the plastic particles; the screw groove depth of the compression section usually changes gradually from deep to shallow, and the extrusion effect on the material is more obvious. As the screw groove depth decreases, the shearing effect of the screw becomes more obvious. Therefore, the compression section is an important transition stage for the polymer material to transform from a glassy state to a highly elastic state; the screw groove depth of the metering section is the shallowest and the shearing effect is the strongest. Under the action of high shear and continuous heating of the barrel temperature, the polymer material has now transformed from a highly elastic state to a viscous flow state.

[0004] In addition to the common three-stage screw, there are many new types of screws, including gradual-change screws, sudden-change screws, screws without metering sections, corrugated screws, split-flow screws, and separating screws. These three screw structures are the most common. The gradual-change screw has a longer compression section, generally accounting for more than 40% of the total thread length, and a shorter metering section, generally no more than 15% of the total thread length. The longer compression section allows the screw channel depth to change slowly, resulting in gentle pressure and temperature changes in the material, making it suitable for processing materials with poor thermal stability. The sudden-change screw is characterized by a relatively short compression section, accounting for approximately 15% of the effective screw length, and a longer feed section, generally accounting for 70% of the effective screw length. Its structural characteristic is that the screw channel depth changes dramatically in the short compression section, resulting in a rapid increase in shear rate and temperature during processing, making it suitable for processing materials with inherently sudden changes in properties. The separation screw is characterized by the presence of secondary threads, which can separate the solid and liquid phases of the polymer during the plasticizing process, allowing the gas in the compressed plastic particles to be discharged, which is conducive to the reduction of bubbles.

[0005] The plasticized state of the material before entering the mold directly impacts the quality of the finished product, and this effect cannot usually be balanced by adjusting machine parameters. Therefore, precise design of the screw structure and dimensions is extremely important. The influence of the screw structure on the material's phase behavior exhibits multivariable, nonlinear coupling. However, traditional screw design is typically empirically based on melt delivery theory and plastics machinery design manuals. With the continuous development of the plastics industry and the emergence of new materials, customized screw design based on the characteristics of different materials has become a pressing need in the plastics processing and molding field.

[0006] With the advent of Industry 4.0, smart manufacturing has entered various industries, ushering in a new chapter for the plastics industry. Precision control for plastic products differs from machining metal products. Due to the material properties, universal injection molding screws cannot be compatible with all types of materials. Traditionally, injection molding screw design relied on experience and extensive trial and error to create a specialized screw. However, with the maturity of computer-aided technology, screw design has gradually shifted from traditional trial-and-error methods to computer-aided production. In recent years, numerous studies have utilized computer-aided technology to guide screw design, but research using machine learning methods to aid screw design has been rare. Machine learning methods offer advantages such as safety, reliability, consistency, high efficiency, and high precision, making them a popular auxiliary tool for many engineers in engineering. By leveraging the digitally driven nature of machine learning methods, computer-aided design can digitize the complexities of screw design, enabling the design of specialized plasticizing screws tailored to the characteristics of different materials.

[0007] Patent CN102194033A discloses a variant design method, which first establishes a product model and then performs product structure variant design, including a description of the relationship between object characteristics and a variant design process. During the variant design process, it is necessary to determine the basic characteristics, and based on the constraint relationship set between the basic characteristics and the derived characteristics, determine or calculate the value of the derived characteristics. The basic characteristics and derived characteristics of the object characteristic table are used to drive the three-dimensional model in the CAD system to perform variant design, and finally perform process variant design. Patent CN109658499A invents a model establishment method, device, and storage medium, which obtains the target development activity requirements of an instance object; obtains the target category of the instance object; obtains the specific attribute items of the instance object from a technology spectrum model based on the target development activity requirements and the target category; and establishes a full information attribute model of the instance object based on the target development activity requirements, the target category, and the specific attribute items. Patent US9959684B2 proposes a method for changing the shape of geometric solid models using deformation using a computer graphics system. The method is characterized by being used to generate mutually mating geometric solids, where the geometric solids are relative to each other and have mating surfaces that need to be manufactured with precise accuracy (up to 10 microns). Examples include toothed screws (cylindrical, conical) for screw compressors, gears for gear transmissions, gear pumps, and similar mating pairs of geometric solids. The proposed method involves deforming a semi-finished model of a first mating solid using a model of a first tool, the surface of which is automatically calculated using several mathematical laws, each of which is a polynomial with at least one coefficient representing one of several features defining the surface shape of a second geometric solid. The semi-finished model of the second mating solid is deformed using a model of a second tool, which represents the deformed target model of the first solid. Furthermore, the deformation is performed according to a motion law that mimics the geometric solids and that cooperates with each other during use. Patent KR20150031652A relates to a design method for a plasticizing screw for an injection molding machine or an extruder, and more particularly to a three-dimensional modeling method for a plasticizing screw, characterized in that the diameter and pitch values ​​of the screw can be variably set, including forming a two-dimensional spiral line, then applying a sweep to form a three-dimensional solid, and then forming the screw axis. It is characterized in that the screw diameter value and the pitch value can be variably set within the said interval, thereby providing a three-dimensional modeling method for a plasticizing screw. In actual operation, the dimensional accuracy of the plasticizing screw can be improved, the design freedom of the plasticizing screw to be manufactured can be ensured, and various shapes can be realized accordingly. However, although the invention patents listed above have certain applicability for the design of certain mechanical parts, they lack a digital process for constraint parameters and conditions. In particular, for the design of plasticizing screw parts that are closely related to material properties, it is impossible to establish a relationship between the part shape and the material properties.

[0008] Patent CN110809510A discloses a screw shape inference device, a screw shape inference method, and a screw shape inference program. The acquisition module obtains input information including resin physical properties, and obtains the required value of the physical quantity related to the mixed resin or the mixing device as output information; the storage module is used to store a knowledge file containing the input information, physical quantities, and the correlation between multiple screw shapes; the search module and the shape making module infer the screw shape that meets the required value based on the input information and the knowledge file. Patent KR20230026290A discloses a screw recommendation function and a screw recommendation method in an artificial intelligence-based injection molding system. The artificial intelligence-based injection molding system with a screw recommendation function includes: a resin information input part, which receives resin information including the type of resin used for injection molding; a screw diameter recommendation part, which determines the candidate screw diameter of the resin and the candidate screw model supporting the candidate screw diameter based on the physical property data of the resin corresponding to the resin and the injection molding data mapped to the resin; and a screw diameter recommendation part, which determines the candidate screw diameter of the resin and the candidate screw model supporting the candidate screw diameter for each candidate screw diameter. a first selection part, in which a target screw diameter among the candidate screw diameters is input from a user, and a target screw model among the candidate screw models is input; and a screw material and design recommendation part, in which a candidate screw material and design is determined based on the physical property data, the injection molding data, the target screw diameter, the target screw model and the screw design data; a second selection part, for receiving the target screw material and design among the input candidate screw materials and designs; and a selection result output part, for outputting a selection result of the target screw diameter, the target screw model and the target screw material and design. The above two patents are the prior art that are closest to the purpose of the present invention. However, although the above two patents and prior art mention the procedure of learning and inferring or recommending screw shape through resin physical property information, there is no clear relevant information data, especially on what information data to infer or recommend screw shape, how to evaluate screw shape, on what indicators to infer and evaluate screw shape, what algorithm or model to obtain the optimal screw shape, whether the inferred screw shape meets the plasticization conditions required by the resin, and many other issues. No supporting measures and methods are given. The physical quantity parameter data established by the relevant prior art is relatively limited. Patent CN110809510A only mentions the inference of screw shape without the specific design function of the screw shape related dimensions matching the material properties. Patent KR20230026290A only mentions the matching of screw diameter, model and screw material by information on resin type and additive type. Therefore, the screw models that can be output by these two patents are relatively rough and do not have the specific design function corresponding to the screw shape and size.

[0009] Patent P2002-214107A discloses a twin-screw extruder screw segment configuration method, which creates a database by flow analysis and kneading extrusion test on various resin materials in the segmented configuration of various size twin-screw kneading extruders, and provides data corresponding to the user's required conditions, while proposing the best solution to the screw segment configuration configuration under the relevant conditions; when used by the user, there is no need to separately perform flow analysis and mixing extrusion test of the twin-screw mixing extruder, only the required conditions need to be sent to obtain the corresponding data, so as to obtain the screw segment configuration suggestion most suitable for the conditions. Patent US2022 / 0080646Al discloses an apparatus, method and computer program for screw configuration reasoning, an acquisition unit is used to acquire input information including resin characteristics, and to acquire specified values of physical quantities related to mixed resin or mixing machine as output information; a storage unit is used to store knowledge information containing the correlation between input information, physical quantities and multiple screw configurations; a search unit and a configuration generation unit are used to deduce a screw configuration capable of meeting the specified values according to the input information and the knowledge information. The above-mentioned patent technology clearly defines the application target of screw configuration based on resin material information, and clearly defines the device application of twin-screw extruder and resin mixing machine, but there are still many problems in specific implementation. Because the twin-screw extruder used for mixing involves a particularly complex and diverse material system, the screw configuration of the twin-screw is different from the conventional single screw, and is particularly complex, and only the conventional screw elements include various types of conveying elements, meshing elements, mixing elements, etc. of different shapes and sizes, which can be combined in different orders to form different overall screw configurations. Patent P2002-214107A clearly defines that the creation of the database is completed through experimental test, and how to complete the flow analysis and kneading extrusion test under different screw configurations of different material systems to create the database becomes the main problem of the implementation of this patent. The cost, especially the time cost, of collecting data and creating a database is too high, and the data reliability is low due to experimental errors; Patent US2022 / 0080646Al does not clearly define the specific measures and methods of database creation, and the basis of its optimal screw configuration is the mixing effect of the resin material, which clearly defines resin temperature, machine power, solid phase ratio, residence time, maximum pressure and maximum torque as indicators, but does not clearly define the acquisition approach of related parameter indicators. Through experiments, the time cost and experimental errors also affect the reliability of the data. Furthermore, Patent P2002-214107A does not give the actual algorithm, indicators or standards for obtaining the optimal screw configuration, and Patent US2022 / 0080646Al mentions a neural network algorithm model, but does not clearly define the applicability of the related model. In addition, the above-mentioned patents only perform selective combination configuration from the screw configuration database, and do not have the function of designing screw shapes and specific sizes.

[0010] Patent WO 2011 / 033522 Al discloses a method and system for extruder configuration and visualization, the system comprising an extruder information database including detailed information about extruders, extruder screw elements, and extruder barrels; a user interface receiving user input, including extruders, screw elements, and barrels selected by the user from the extruder information database for addition to the extruder information database; a component positioning module including a database of incompatible components; the component positioning module including an incompatible component database for reviewing user selections and preventing incompatible extruder components from being positioned adjacent to each other; a processor for creating an extruder configuration based on the user input; and a display module for creating an extruder configuration based on the user input; a display module for creating an extruder configuration based on the user input; a processor for creating an extruder configuration based on the user input; and a display module for displaying the extruder configuration. This patent clarifies that it is only applicable to the field of extruders and only provides the function of manually combining screw element configurations, and does not have the function of designing the structure, shape, or size of the screw elements.

[0011] US Patent No. 10549453B2 discloses a simulation device, simulation method, and non-transitory computer-readable medium containing a simulation program for analyzing the fluid flow of materials in a twin-screw kneading and mixing apparatus. The simulation apparatus includes a one-dimensional analysis unit that performs a low-dimensional fluid flow analysis of the material within an arithmetic object domain of the apparatus based on configuration information including material physical properties, as well as configuration data and operating conditions of a kneading apparatus used to mix the material; a region selection unit that receives a selection of a region within the arithmetic object domain as an object for high-dimensional fluid flow analysis; and a three-dimensional analysis unit that extracts material physical quantities related to the region based on the results of the low-dimensional fluid flow analysis and performs a high-dimensional fluid flow analysis of the material within the region based on the extracted physical quantities and the configuration information. This patent outputs information on the flow parameters of the mixing process by inputting information such as material density, thermal conductivity, specific heat capacity, and rheological data, as well as screw element configuration information. This patent does not provide functionality for designing the structure, shape, or size of the screw elements.

[0012] It can be seen that based on the specific parameter data of material characteristics, combined with computer-aided design and numerical simulation to obtain plasticizing performance parameter data, establish plasticizing evaluation methods and indicators, and further establish an artificial intelligence machine learning model to obtain the optimal plasticizing screw shape, structure, and size, it is possible to produce an intelligent plasticizing screw design system and method based on material characteristics. This system has many advantages: matching material properties, conforming to theoretical laws without human experience interference, clear and reliable parameter information data, and intelligent, automated, and rapid generative design. Existing technologies do not have these advantages. Summary of the Invention

[0013] The purpose of the present invention is to provide an intelligent design system and method for a plasticizing screw based on material properties, which can replace the conventional experience-based manual design method. By combining material characteristic parameters, computer-aided design, computer-aided numerical simulation, and experimental design method, a database of material characteristic parameters and plasticizing screw structural dimensions is established to provide effective data for artificial intelligence machine learning. An artificial intelligence model is obtained through artificial intelligence machine learning, and ultimately the purpose of outputting plasticizing screw structural geometric parameter data by simply inputting material characteristic parameters is achieved. A screw model can be established based on the output screw structural geometric parameter data for the next step of screw processing and manufacturing and applied to the plasticizing production of the corresponding material.

[0014] The objectives of the present invention are achieved through the following technical solutions.

[0015] An intelligent design system for plasticizing screws based on material properties, including a material property database, a screw structure parameter database, an experimental design module, a computer-aided design module, a computer-aided numerical simulation module, a screw plasticizing performance evaluation module, a screw structure optimization module, an optimal screw structure machine learning module, and an optimal screw structure generation module, characterized in that: the material property database stores physical property parameter data of n kinds of materials, and the material physical property parameters mainly include material density, specific heat capacity, thermal conductivity, and rheological parameters, forming n groups of material physical property parameter data; the screw structure parameter database includes data and variation ranges of geometric parameters related to the screw structure type, and the screw structure parameters mainly include screw diameter, screw aspect ratio, screw Groove depth, screw ridge width, helix angle, pitch, compression ratio; the experimental design module includes a parameter design part based on the experimental design method, which establishes m types of screw structures according to the screw structure parameters and data range and the experimental design algorithm. The experimental design algorithm mainly includes the orthogonal design method, the Taguchi design method, and the central design method to form m groups of screw structure parameter data; the computer-aided design module establishes m corresponding screw models and barrel flow channel models based on the m groups of screw structure parameter data established by the experimental design module; the computer-aided numerical simulation module carries out numerical simulation of the screw plasticizing process in turn according to the n groups of material physical property parameter data and the m screw models established by the computer-aided design module, calculates the flow field parameters in the model flow channel, and the flow field The parameters mainly include temperature and viscosity. Through numerical simulation calculation, n×m groups of flow field parameter data can be obtained, that is, each material corresponds to m groups of flow field parameter data; the screw plasticizing performance evaluation module determines the plasticizing parameters and weight ratios. The plasticizing parameters mainly include the temperature difference, viscosity difference and average viscosity at the front end of the screw, i.e., the outlet end of the flow channel in the barrel, obtained by numerical simulation calculation. The plasticizing performance index value is obtained according to the plasticizing parameters and weight calculation, which is the sum of the product of each plasticizing parameter and its weight. Based on the numerical simulation calculation results, n×m groups of plasticizing performance index data are calculated, and each material corresponds to m plasticizing performance index values; the screw structure optimization module performs the calculation on the m groups of screw structure parameter data established by the experimental design module and the screw plasticizing performance evaluation module. The m plasticizing performance index value results are processed, and the screw structure parameter data that can make the plasticizing performance index value closest to the expected value is obtained by using the experimental design algorithm and the expected value size of the plasticizing performance index data, that is, the optimal screw structure parameter data is obtained, and each material corresponds to a set of optimal screw structure parameter data, and n materials correspond to n sets of optimal screw structure parameter data; the optimal screw structure machine learning module mainly includes an input data interface, an output data interface, a machine learning model, and a hyperparameter optimization module, the input data interface inputs the physical property parameter data of n materials, and the output data interface inputs the n sets of optimal screw structure parameter data corresponding to the n materials, and the machine learning model is used to perform machine learning training to obtain an artificial intelligence model;The optimal screw structure generation module mainly includes a data input module, a data output module, and a model input module. The data input module inputs new material physical property parameter data, the model input module inputs an artificial intelligence model, and the data output module is used by the artificial intelligence model to calculate the corresponding new optimal screw structure parameter data based on the input new material physical property parameter data.

[0016] Furthermore, the rheological parameters include melt flow index.

[0017] Furthermore, the rheological parameters include equation parameters obtained by regressing rheological test experimental data using conventional rheological equations.

[0018] Furthermore, the plasticizing performance index of the screw plasticizing performance evaluation module determines the weight of each plasticizing parameter through the plasticizing criterion. The screw structure optimization module takes the plasticizing criterion, i.e., the optimal plasticizing performance, as the optimization target during optimization, and the corresponding plasticizing performance index is minimized with the expected value of the plasticizing parameter temperature difference, viscosity difference and average viscosity.

[0019] Furthermore, the flow field parameters calculated by the computer-aided numerical simulation module also include screw torque, and the plasticizing parameters of the screw plasticizing performance evaluation module also include screw torque.

[0020] Furthermore, the screw plasticizing performance evaluation module also includes a plasticizing criterion and a balance criterion and an energy-saving criterion that take into account the energy consumption generated by the screw torque. The plasticizing criterion corresponds to the plasticizing parameter weight assigned with the minimum expected value of the temperature difference, viscosity difference and average viscosity to obtain the plasticizing performance index value; the balance criterion corresponds to the plasticizing parameter weight assigned with the minimum expected value of the temperature difference, viscosity difference, average viscosity and screw torque to obtain the plasticizing performance index value; under the premise of a total weight of 100%, the balance criterion preferably prefers that the temperature difference, viscosity difference and average viscosity together account for 50% of the weight, and the screw torque accounts for 50% of the weight; the energy-saving criterion preferably prefers that the temperature difference, viscosity difference and average viscosity together account for less than 50% of the weight, and the screw torque accounts for the remaining weight.

[0021] Furthermore, the screw plasticizing performance evaluation module performs dimensionless processing on the plasticizing parameter data before determining the plasticizing performance index value.

[0022] Furthermore, the dimensionless processing method is preferably a maximum-minimum normalization method.

[0023] Furthermore, the plasticizing parameter weights of the screw plasticizing performance evaluation module have the function of assigning weights based on interventional artificial experience, that is, the plasticizing parameter weight values ​​can be modified artificially based on experience.

[0024] Furthermore, the material property-based plasticizing screw intelligent design system also includes a screw plasticizing performance machine learning module and a screw plasticizing performance prediction module. The screw plasticizing performance machine learning module mainly includes a material physical property parameter and screw structure parameter input interface, a plasticizing parameter output interface, a machine learning model, and a hyper-parameter optimization module; the screw plasticizing performance prediction module mainly includes a new material physical property parameter and screw structure parameter input interface, a model input interface after training of the screw plasticizing performance machine learning module, and a plasticizing parameter output interface.

[0025] Furthermore, the material property-based plasticizing screw intelligent design system also includes a computer-aided design model generation system, which can directly generate a screw model file in a format readable by computer-aided design software based on the optimal screw structure parameter data output by the optimal screw structure generation module.

[0026] Furthermore, the machine learning model is a supervised learning model, which mainly includes a multiple linear regression model, a random forest model, an XGB model, a GBDT model and a neural network model. The prediction accuracy of the machine learning model is calculated by comparing the predicted results with the actual results to obtain the mean square error (MSE) and the coefficient of determination (R 2 ), the coefficient of determination ranges from 0 to 1. The closer it is to 1, the higher the model accuracy; the closer the mean square error is to 0, the higher the model accuracy.

[0027] Furthermore, the machine learning model is preferably a random forest model.

[0028] Furthermore, the material property-based plasticizing screw intelligent design system is also provided with a transfer learning module.

[0029] Furthermore, the material property database contains at least 50 materials.

[0030] The intelligent design method for a plasticizing screw based on material properties is characterized in that: the method uses the above-mentioned intelligent design system for a plasticizing screw based on material properties, and the specific steps of the method are as follows:

[0031] Step 1: Store the physical property parameters of n kinds of materials of different types, manufacturers and brands in a material property database to establish a database of n sets of material physical property parameter data;

[0032] Step 2: storing screw structural parameter data of different screw types in a screw structural parameter database, and establishing a database containing different screw structural parameters and value variation ranges;

[0033] Step 3: Using the experimental design module, select an experimental design algorithm, and establish m types of screw structures corresponding to m groups of screw structure parameter data according to the screw structure parameters and data ranges in the screw structure parameter database;

[0034] Step 4: Using a computer-aided design module, establish m corresponding screw models and barrel flow channel models based on the m sets of screw structural parameter data established by the experimental design module;

[0035] Step 5: Using a computer-aided numerical simulation module, based on the n sets of material physical property parameter data and the m screw models established by the computer-aided design module, numerical simulations of the screw plasticizing process are sequentially performed to calculate and obtain n×m sets of flow field parameter data, i.e., each material corresponds to m sets of flow field parameter data;

[0036] Step 6: Determine the plasticizing parameters and weight ratios using the screw plasticizing performance evaluation module, and calculate the plasticizing performance index values ​​based on the plasticizing parameters and weights. This is the sum of the products of the plasticizing parameters and their respective weights. Based on the numerical simulation results, n×m groups of plasticizing performance index data are calculated, with each material corresponding to m plasticizing performance index values.

[0037] Step 7: Processing the m groups of screw structure parameter data established by the experimental design module and the m plasticizing performance index values ​​obtained by the screw plasticizing performance evaluation module through the screw structure optimization module, and obtaining the screw structure parameter data that can make the plasticizing performance index value closest to the expected value by using the experimental design algorithm and the expected value analysis of the plasticizing performance index data, that is, obtaining the optimal screw structure parameter data. Each material corresponds to one group of optimal screw structure parameter data, and n materials correspond to n groups of optimal screw structure parameter data.

[0038] Step 8: Input n sets of material physical property parameter data through the input data interface of the optimal screw structure machine learning module, input n sets of optimal screw structure parameter data corresponding to n materials through the output data interface, and perform machine learning training through the machine learning model to obtain an artificial intelligence model;

[0039] Step 9: Input the physical property parameter data of the new material through the data input module of the optimal screw structure generation module, and input the artificial intelligence model trained by the optimal screw structure machine learning module through the model input module. The artificial intelligence model calculates the corresponding new optimal screw structure parameter data based on the input new material physical property parameter data, and outputs the new optimal screw structure parameter data through the data output module to complete the intelligent design task of the plasticizing screw based on material properties.

[0040] Beneficial effects

[0041] 1. The material property-based plasticizing screw intelligent design system and method provided by the present application, through the material property parameters provided by the material database, the flow field parameters are obtained by using the computer-aided numerical simulation module under the condition of the plasticizing screw model established by the experimental design module and the computer-aided design module, and then the optimal screw structure parameters of all materials in the material database are obtained through the screw plasticizing performance evaluation module and the screw structure optimization module, and further the model of the optimal screw structure generation module is obtained through the machine learning and training of the optimal screw structure machine learning module, and under the support of the model, the optimal screw structure parameters corresponding to any input new material property parameters can be matched;

[0042] 2. The material property-based plasticizing screw intelligent design system and method provided by the present application can realize the function of continuous learning, continuously increase the data set of the material property parameters and the optimal screw structure parameters of new materials on the basis of the existing material database, and continuously improve the accuracy of intelligent design;

[0043] 3. The material property-based plasticizing screw intelligent design system and method provided by the present application can provide the material parameter selection function, and the corresponding optimal screw structure has more applicability according to the need to select the material property parameters to be considered.

[0044] 4. The material property-based plasticizing screw intelligent design system and method provided by the present application, the screw plasticizing performance evaluation module can be customized, the evaluation parameters are selected as needed, and the weights of the evaluation parameters are artificially assigned; the corresponding optimal screw can be intelligently designed according to the evaluation index selected by the artificial selection;

[0045] 5. The material property-based plasticizing screw intelligent design system and method provided by the present application provides a plurality of machine learning models, and the corresponding model can be manually specified for machine learning and training;

[0046] 6. The material property-based plasticizing screw intelligent design system and method provided by the present application further comprises a screw plasticizing performance machine learning module and a screw plasticizing performance prediction module, and has the prediction function of outputting the flow field parameters and the plasticizing parameter data by inputting the material property parameters and the screw structure parameters.

[0047] 7. The more the types of materials contained in the material database of the material property-based plasticizing screw intelligent design system and method provided by the present application, the higher the machine learning accuracy, and the more accurate the optimal screw structure obtained by the intelligent design. BRIEF DESCRIPTION OF DRAWINGS

[0048] Figure 1 is the architecture relationship and method sequence diagram of the material property-based plasticizing screw intelligent design system and method of example 1;

[0049] Figure 2This is a diagram showing the architecture and method sequence of a plasticizing screw intelligent design system and method based on material properties in Example 2;

[0050] Figure 3 is the orthogonal table of screw structure parameters established by the experimental design module of Example 2;

[0051] Figure 4 The screw model and barrel flow channel model are established by the computer-aided design module of Example 2;

[0052] Figure 5 The temperature and viscosity flow field distribution at the front end of the screw, i.e., the outlet end of the flow channel in the barrel, is obtained by numerical simulation of the PBT material with the brand VALOX 215HPR under a certain screw structure calculated by the computer-aided simulation module of Example 2;

[0053] Figure 6 This is a mean main effect diagram and a mean response table of the plasticizing performance index value data of the plasticizing criterion obtained by the orthogonal experimental design algorithm for the PBT material with the brand VALOX 215HPR in Example 2;

[0054] Figure 7 These are the three optimal screw structures of the new PP material with the brand name Braskem PP CP 442XP Bras obtained by the artificial intelligence model in Example 2;

[0055] Figure 8 This is a data table and correlation diagram comparing the three optimal screw structural parameters of the new PP material with the brand Braskem PP CP 442XP Bras obtained by the artificial intelligence model in Example 2 and the three optimal screw structural parameters obtained by the orthogonal experimental design method.

[0056] In the figure: 1—material property database, 2—screw structure parameter database, 3—experimental design module, 4—computer-aided design module, 5—computer-aided numerical simulation module, 6—screw plasticizing performance evaluation module; 7—screw structure optimization module, 8—optimal screw structure machine learning module, 9—optimal screw structure generation module; A—n groups of material physical property parameter data, B—m groups of screw structure parameter data, C—m screw models, D—n×m groups of flow field parameter data, E—n×m groups of plasticizing performance index data, F—n groups of optimal screw structure parameter data, G—artificial intelligence model; H—new material physical property parameter data, I—new optimal screw structure parameter data. DETAILED DESCRIPTION

[0057] The preferred embodiments of the present invention are further described in detail below with reference to the accompanying drawings.

[0058] Example 1

[0059] Reference Attachment Figure 1As shown, a plasticizing screw intelligent design system based on material properties includes a material property database (1), a screw structure parameter database (2), an experimental design module (3), a computer-aided design module (4), a computer-aided numerical simulation module (5), a screw plasticizing performance evaluation module (6), a screw structure optimization module (7), an optimal screw structure machine learning module (8), and an optimal screw structure generation module (9), characterized in that: the material property database (1) stores physical property parameter data of n kinds of materials, and the material physical property parameters mainly include material density, specific heat capacity, thermal conductivity, and rheological parameters, forming n groups of material physical property parameter data (A); the screw structure The parameter database (2) includes data and variation ranges of geometric parameters related to the screw structure type, wherein the screw structure parameters mainly include screw diameter, screw aspect ratio, screw groove depth, screw ridge width, helix angle, screw pitch, and compression ratio; the experimental design module (3) includes a parameter design part based on the experimental design method, and establishes m types of screw structures according to the screw structure parameters and data range and the experimental design algorithm, wherein the experimental design algorithm mainly includes the orthogonal design method, the Taguchi design method, and the central design method, thereby forming m groups of screw structure parameter data (B); the computer-aided design module (4) establishes corresponding m screw models ( C) and a barrel flow channel model; the computer-aided numerical simulation module (5) sequentially carries out numerical simulation of the screw plasticizing process based on n groups of material physical parameter data (A) and m screw models (C) established by the computer-aided design module (4), and calculates and obtains n×m groups of flow field parameter data (D), that is, each material corresponds to m groups of flow field parameter data; the screw plasticizing performance evaluation module (6) determines the plasticizing parameters and weight ratios, the plasticizing parameters mainly including the temperature difference, viscosity difference and average viscosity at the front end of the screw, i.e., the outlet end of the flow channel in the barrel, obtained by numerical simulation calculation, and the plasticizing performance index value is obtained by calculation based on the plasticizing parameters and weights, that is, the sum of the products of each plasticizing parameter and its weight, based on the plasticizing performance evaluation module (6). Based on the numerical simulation calculation results, n×m groups of plasticizing performance index data (E) are calculated, and each material corresponds to m plasticizing performance index values; the screw structure optimization module (7) processes the m groups of screw structure parameter data (B) established by the experimental design module (3) and the m plasticizing performance index value results obtained by the screw plasticizing performance evaluation module (6), and obtains the screw structure parameter data that can make the plasticizing performance index value closest to the expected value by using the experimental design algorithm and the expected value size analysis of the plasticizing performance index data, that is, obtains the optimal screw structure parameter data, each material corresponds to one group of optimal screw structure parameter data, and n materials correspond to n groups of optimal screw structure parameter data (F);The optimal screw structure machine learning module (8) mainly includes an input data interface, an output data interface, a machine learning model, and a hyperparameter optimization module. The input data interface inputs n types of material physical property parameter data (A), and the output data interface inputs n groups of optimal screw structure parameter data (F) corresponding to the n types of materials. The machine learning model is used for machine learning training to obtain an artificial intelligence model (G). The optimal screw structure generation module (9) mainly includes a data input module, a data output module, and a model input module. The data input module inputs new material physical property parameter data (H), the model input module inputs the artificial intelligence model (G), and the data output module is used for the artificial intelligence model (G) to calculate the corresponding new optimal screw structure parameter data (I) based on the input new material physical property parameter data (H).

[0060] The steps for using the above-mentioned intelligent design system for plasticizing screws based on material properties are:

[0061] Step 1: storing the physical property parameters of n kinds of materials of different types, different manufacturers and different brands in a material property database (1), and establishing a database of n groups of material physical property parameter data (A);

[0062] Step 2, storing screw structure parameter data of different screw types in a screw structure parameter database (2), and establishing a database containing different screw structure parameters and value variation ranges;

[0063] Step 3, through the experimental design module (3), select the experimental design algorithm, and establish m types of screw structures according to the screw structure parameters and data ranges in the screw structure parameter database, corresponding to m groups of screw structure parameter data (B);

[0064] Step 4, using the computer-aided design module (4), according to the m groups of screw structure parameter data (B) established by the experimental design module (3), establish corresponding m screw models and barrel flow channel models;

[0065] Step 5, using the computer-aided numerical simulation module (5) to sequentially carry out numerical simulation of the screw plasticizing process based on the n sets of material physical property parameter data (A) and the m screw models (C) established by the computer-aided design module (4), and calculate and obtain n×m sets of flow field parameter data (D), that is, each material corresponds to m sets of flow field parameter data;

[0066] Step 6, determining the plasticizing parameters and weight ratios through the screw plasticizing performance evaluation module (6), and obtaining the plasticizing performance index values ​​according to the plasticizing parameters and weights, which are the sum of the products of each plasticizing parameter and its weight, and obtaining n×m groups of plasticizing performance index data (E) based on the numerical simulation calculation results, where each material corresponds to m plasticizing performance index values;

[0067] Step 7, performing data processing on the m groups of screw structure parameter data (B) established by the experimental design module (3) and the m plasticizing performance index value results obtained by the screw plasticizing performance evaluation module (6) through the screw structure optimization module (7), and obtaining the screw structure parameter data that can make the plasticizing performance index value closest to the expected value by analyzing the experimental design algorithm and the expected value of the plasticizing performance index data, that is, obtaining the optimal screw structure parameter data, each material corresponds to one group of optimal screw structure parameter data, and n materials correspond to n groups of optimal screw structure parameter data (F);

[0068] Step 8: Input n sets of material property parameter data (A) through the input data interface of the optimal screw structure machine learning module (8), input n sets of optimal screw structure parameter data (F) corresponding to the n materials through the output data interface, and perform machine learning training through the machine learning model to obtain an artificial intelligence model (G);

[0069] Step 9: Input the new material physical property parameter data (H) through the data input module of the optimal screw structure generation module (9), input the artificial intelligence model (G) through the model input module, the artificial intelligence model (G) calculates the corresponding new optimal screw structure parameter data (I) based on the input new material physical property parameter data (H), and outputs the new optimal screw structure parameter data (I) through the data output module, thereby completing the intelligent design task of the plasticizing screw based on material properties.

[0070] Example 2

[0071] Reference Attachment Figure 2 As shown, this embodiment is similar to the above embodiment, except that the material property database (1) of this embodiment stores the physical property parameter data of 62 different brands of materials, including density (unit kg / m 3 ), specific heat capacity (unit J / (kg·K)), thermal conductivity (unit W / (m·K)), zero shear viscosity (unit Pa·s), power law exponent, time constant (unit s), reference temperature (unit K), and activation energy parameter (unit K), forming 62 sets of material property parameter data (A). The zero shear viscosity, power law exponent, time constant, reference temperature, and activation energy parameter of the above material property parameters are rheological parameters, which are obtained by regressing rheological test experimental data through conventional rheological equations. The cross rheological equation adopted in this utility model is as follows:

[0072] η=η0 / (1+(λ·γ·) (1-n) )

[0073] H(T)=exp[α(1 / (T-T0)-1 / (T α -T0))]

[0074] where η is the melt viscosity, η0is the zero shear viscosity of the melt, λ is the time constant (i.e. the inverse of the shear rate at which the fluid changes from Newtonian to power law behavior), n is the power law index, a is the activation energy parameter (i.e. the ratio of the activation energy to the thermal constant), T is the melt temperature, T α is the reference temperature, and To is the temperature offset, which is taken as 0 in this case.

[0075] Take the PBT material produced by SABIC with the brand VALOX 215HPR as an example, the corresponding material physical property parameters are: density 1098 kg / m 3 , specific heat capacity 1935 J / (kg·K), thermal conductivity 0.2 W / (m·K), no shear viscosity 108.6108 Pa·s, power law index 0.38576, time constant 0.007555 s, reference temperature 529 K, activation energy parameter 6435.535 K.

[0076] The screw type of the screw structure parameter database (2) of the embodiment belongs to a conventional three-section injection molding screw, mainly including a conveying section, a compression section and a metering section. The fixed screw diameter of the embodiment is 30 mm, the screw pitch is 30 mm and the length-diameter ratio is 18. The screw structure parameters to be optimized correspond to four parameters including the conveying / compression / metering three-section length ratio, the compression ratio, the metering section groove depth and the screw flight width.

[0077] The experimental design algorithm of the experimental design module (3) of this embodiment adopts the orthogonal design method. The method of dividing the length ratio of the three-stage screw is to divide the total length of the screw into 20 parts in equal proportion. The ratio of 1 represents 5% of the total length of the screw. For example, the length ratio of the three sections of conveying / compression / metering is the length of the conveying section: the length of the compression section: the length of the metering section = 9:7:4, which means that the length of the conveying section of the screw accounts for 45% of the total length, the length of the compression section accounts for 35% of the total length, and the length of the metering section accounts for 20% of the total length. The selection of the level value of the three-section length ratio is based on the fact that each section length is within the range of the conventional three-section injection molding screw design criteria, and based on this, four levels are divided, namely 9:7:4, 8:6:6, 10:5:5, and 10:4:6; the division of the compression ratio takes into account the different compression ratios. The scaling ratio is applicable to materials with different viscosities, so the coverage range of its four levels includes high-viscosity materials and low-viscosity materials, with values ​​of 2, 2.5, 3, and 3.5; the design range of the groove depth of the metering section of the three-stage injection molding screw is specified as 0.04 to 0.07 times the screw diameter according to conventional screw design criteria, so as to be applicable to materials with different sensitivities to shear rates. This embodiment is evenly divided into four levels, namely 1.2mm, 1.5mm, 1.8mm, and 2.1mm; the design range of the screw fin width of the three-stage injection molding screw is specified as 0.08 to 0.12 times the screw diameter according to conventional screw design criteria. This embodiment is evenly divided into four levels within its range, namely 2.4mm, 2.85mm, 3.15mm, and 3.6mm. According to the above-mentioned factor and level design of the orthogonal experiment, the L16 (4 factors and 4 levels) orthogonal table is obtained, as shown in the attached figure. Figure 3 As shown, it includes 16 sets of screw structure parameter data (B).

[0078] The computer-aided design module (4) establishes 16 corresponding screw models (C) and barrel flow channel models based on the 16 sets of screw structure parameter data established by the experimental design module (3). Figure 4 The screw model and barrel flow channel model are established. The computer-aided numerical simulation module (5) carries out numerical simulation of the screw plasticizing process in sequence based on the 62 sets of material physical parameter data (A) and the 16 screw models (C) established by the computer-aided design module (4), and calculates and obtains 62×16=992 sets of flow field parameter data (D), that is, each material corresponds to 16 sets of flow field parameter data. Figure 5 The temperature and viscosity flow field distribution at the screw tip (i.e., the barrel outlet) is calculated using numerical simulations for PBT material with the grade VALOX 215HPR under a specific screw configuration. The simulation results provide the temperature difference, viscosity difference, and average viscosity at the screw tip (i.e., the barrel outlet), as well as the calculated screw torque, as plasticizing parameter data.

[0079] The screw plasticizing performance evaluation module (6) determines the plasticizing parameters and weight ratios. The plasticizing parameters mainly include the temperature difference, viscosity difference, average viscosity, and screw torque at the front end of the screw, i.e., the outlet end of the flow channel in the barrel, obtained by numerical simulation calculation. The plasticizing performance index value is obtained based on the plasticizing parameters and weight calculation, which is the sum of the product of each plasticizing parameter and its weight. Based on the numerical simulation calculation results, 992 sets of plasticizing performance index data (E) are calculated, and each material corresponds to 16 plasticizing performance index values. In addition, considering that there is an order of magnitude difference between the dimensions and values ​​of the rheological parameters, in order to avoid large errors, this embodiment performs dimensionless processing on the plasticizing parameters before calculating the plasticizing performance index value, using the maximum-minimum normalization method, and the corresponding formula is as follows:

[0080] x'=[x-min(x)] / [max(x)-min(x)]

[0081] Where x is the data value of the corresponding parameter, max(x) is the maximum value of all data for that parameter, min(x) is the minimum value of all data for that parameter, and x' is the normalized value of the data. The plasticizing performance index value is the sum of the normalized data of each plasticizing parameter multiplied by its weight. This embodiment adopts the plasticizing criterion, the balance criterion, and the energy-saving criterion to set different plasticizing performance index values. The plasticizing criterion emphasizes the plasticizing performance, and the weights of the four plasticizing parameters of temperature difference, viscosity difference, average viscosity, and screw torque are each set to 1; the plasticizing performance index value = normalized temperature difference + normalized viscosity difference + normalized average viscosity + normalized screw torque; the balance criterion sets the weights of the four plasticizing parameters of temperature difference, viscosity difference, average viscosity, and screw torque to 1, 1, 1, and 3, respectively, and the plasticizing performance index value = normalized temperature difference + normalized viscosity difference + normalized average viscosity + 3×normalized screw torque; the energy-saving criterion sets the weights of the four plasticizing parameters of temperature difference, viscosity difference, average viscosity, and screw torque to 1, 1, 1, and 9, respectively, and the plasticizing performance index value = normalized temperature difference + normalized viscosity difference + normalized average viscosity + 9×normalized screw torque.

[0082] Taking the plasticizing criterion as an example, the screw structure optimization module (7) processes the 16 sets of screw structure parameter data (B) established by the experimental design module (3) and the 16 plasticizing performance index value results obtained by the screw plasticizing performance evaluation module (6), and uses the experimental design algorithm and the expected value size analysis of the plasticizing performance index data to obtain the screw structure parameter data that can make the plasticizing performance index value closest to the expected value, that is, to obtain the optimal screw structure parameter data; each material corresponds to one set of optimal screw structure parameter data, and 62 materials correspond to 62 sets of optimal screw structure parameter data (F). Taking the PBT material of brand VALOX 215HPR as an example, the attached Figure 6The plasticizing criterion mean main effect diagram and mean response table of the plasticizing performance index value data obtained by the orthogonal experimental design algorithm are displayed. The plasticizing performance index value determined by the plasticizing criterion = temperature difference + viscosity difference + average viscosity + screw torque. According to the plasticizing performance index value, which is small, the optimal screw structure parameters are: three-section length ratio 8:6:6, compression ratio 3, metering section groove depth 1.2mm, and screw flight width 3.6mm.

[0083] Corresponding to the balance criterion and energy-saving criterion, taking the PBT material of brand VALOX 215HPR as an example, according to the small value of plasticizing performance index, the optimal screw structure parameters corresponding to the balance criterion are: three-section length ratio 8:6:6, compression ratio 3, metering section groove depth 1.2mm, screw flight width 3.15mm; the optimal screw structure parameters corresponding to the energy-saving criterion are: three-section length ratio = 8:6:6, compression ratio 3.5, metering section groove depth 1.2mm, screw flight width 2.4mm.

[0084] The optimal screw structure machine learning module (8) of this embodiment mainly includes an input data interface, an output data interface, a machine learning model, and a super parameter optimization module. The input data interface inputs 62 kinds of material property parameter data (A), and the output data interface inputs 62 groups of optimal screw structure parameter data (F) corresponding to the 62 kinds of materials. The machine learning model is used to perform machine learning training to obtain an artificial intelligence model (G); the machine learning model of this embodiment selects a random forest model. After grid search and cross-validation through the super parameter optimization module, the number of trees is determined to be 50, the maximum depth of each tree is 10, the minimum number of samples required before node splitting is 5, the minimum number of samples required for leaf nodes is 3, the maximum number of features considered for each split is 1, and the bootstrap method is used to train each tree. The criterion for measuring the quality of the split is "gini", and the random seed is 5. The machine learning setting sets all 62 groups of data for each model, and the test set is random 25% data to ensure the adequacy of the training data and the accuracy of the test set. The prediction accuracy of the machine learning model is calculated by comparing the predicted results with the actual results to obtain the mean square error (MSE) and the coefficient of determination (R 2 ) is determined, and the coefficient of determination ranges from 0 to 1. The closer it is to 1, the higher the model accuracy; the closer the mean square error is to 0, the higher the model accuracy. In this embodiment, a random forest model is used to perform machine learning on 62 sets of optimal screw structure parameters obtained according to the plasticization criteria for 62 materials. The screw structure parameter data results predicted by the artificial intelligence model (G) obtained after machine learning training are compared with the actual results to calculate the coefficient of determination R 2 =97.98%, indicating that the obtained artificial intelligence model (G) has high accuracy. Three corresponding artificial intelligence models (G) can be obtained according to the plasticization criterion, balance criterion and energy saving criterion respectively.

[0085] The optimal screw structure generation module (9) of this embodiment mainly includes a data input module, a data output module and a model input module. The data input module is used to input the new material physical property parameter data (H), the model input module inputs the artificial intelligence model (G), and the data output module is used for the artificial intelligence model (G) to calculate the corresponding new optimal screw structure parameter data (I) based on the input new material physical property parameter data (H). This embodiment uses a PP material with the brand name Braskem PP CP 442XPBras for applicability verification. This material is a new material not included in the 62 materials used for machine learning. The new material physical property parameter data (H) include: density 709.22kg / m 3 , specific heat capacity 2870 J / (kg·K), thermal conductivity 0.16 W / (m·K), no-shear viscosity 572 Pa·s, power law index 0.299, time constant 0.014250125s, reference temperature 597.6845 K, activation energy parameter 513 K. The new material physical property parameter data (H) are respectively input into the optimal screw structure generation module (9), and three sets of new optimal screw structure parameter data (I) can be obtained according to the three artificial intelligence models (G) corresponding to the plasticization criterion, balance criterion and energy saving criterion. The new optimal screw structure parameter data (I) obtained according to the plasticizing criterion are: conveying section length 228mm, compression section length 151mm, metering section length 161mm, compression ratio 3.393, metering section groove depth 1.21mm, screw flight width 3.45mm; the new optimal screw structure parameter data (I) obtained according to the balance criterion are: conveying section length 222mm, compression section length 152mm, metering section length 166mm, compression ratio 3.499, metering section groove depth 1.29mm, screw flight width 3.13mm; the new optimal screw structure parameter data (I) obtained according to the energy-saving criterion are: conveying section length 244mm, compression section length 164mm, metering section length 132mm, compression ratio 3.5, metering section groove depth 1.54mm, screw flight width 2.64mm. Figure 7 The three optimal screw structures of the new material obtained by the artificial intelligence model are shown. From top to bottom, the screw structures are obtained based on the plasticization criterion, balance criterion and energy-saving criterion respectively.

[0086] In order to verify the accuracy of the method, the new material physical parameter data (H) of the new PP material with the brand name Braskem PP CP 442XP Bras was used; then the computer-aided design module (4) established 16 corresponding screw models (C) and barrel flow channel models based on the 16 sets of screw structure parameter data established by the experimental design module (3); the computer-aided numerical simulation module (5) was used to carry out numerical simulation calculations to obtain flow field parameters, and then the temperature difference, viscosity difference and average viscosity at the front end of the screw, i.e., the outlet end of the flow channel in the barrel, as well as the screw torque obtained by overall calculation were extracted based on the flow field parameters as plasticizing parameter data, and then the plasticizing parameters and weight ratios were determined according to the screw plasticizing performance evaluation module (6), and the corresponding plasticizing performance index values ​​were calculated according to the plasticizing criterion, balance criterion and energy-saving criterion respectively. ; 16 screw models correspond to 16 plasticizing performance index values, each criterion corresponds to 16 plasticizing performance index values, and there are 3 groups of 16 plasticizing performance index values ​​in total; the screw structure optimization module (7) is used again to process the 16 groups of screw structure parameter data (B) established by the experimental design module (3) and the 3 groups of 16 plasticizing performance index value results obtained by the screw plasticizing performance evaluation module (6), and the screw structure parameter data that can make the plasticizing performance index value closest to the expected value is obtained by analyzing the experimental design algorithm and the expected value of the plasticizing performance index data, that is, 3 groups of optimal screw structure parameter data corresponding to the new material are obtained as data for verifying the accuracy of the artificial intelligence model of this embodiment. Appendix Figure 8 The table and correlation diagram show the comparison of the three optimal screw structure parameters of the new material obtained by the artificial intelligence model and the three optimal screw structure parameters obtained by the orthogonal experimental design method. The correlation determination coefficient R of the optimal screw structure parameter data corresponding to the plasticization criterion, balance criterion and energy saving criterion 2 The prediction accuracy of the AI ​​model is 98.59%, 99.3%, and 93.33%, respectively, demonstrating the high accuracy of the prediction. While the orthogonal experimental design method only allows for the selection of 16 screw parameter combinations in the orthogonal experimental design table, the AI ​​model's results are not limited to these 16 combinations. Any value within the screw parameter variation range can be calculated using the AI ​​model as the target plasticizing performance index. Therefore, the screw parameters obtained using the AI ​​model provide superior plasticizing performance.

[0087] The present invention includes but is not limited to the above embodiments. Any equivalent replacement or partial improvement made under the spirit and principle of the present invention shall be deemed to be within the protection scope of the present invention.

Claims

1. A material property-based plasticizing screw intelligent design system, comprising a material property database, a screw structure parameter database, an experimental design module, a computer-aided design module, a computer-aided numerical simulation module, a screw plasticizing performance evaluation module, a screw structure optimization module, an optimal screw structure machine learning module, and an optimal screw structure generation module, characterized in that: The material property database stores the physical property parameter data of n kinds of materials, and the computer-aided design module establishes m groups of screw structure parameter data and corresponding m screw models and barrel flow channel models according to the experimental design module; the computer-aided numerical simulation module sequentially carries out numerical simulation of the screw plasticizing process according to the n groups of material physical property parameter data and the m screw models established by the computer-aided design module, calculates and obtains the flow field parameters in the model flow channel, and obtains n×m groups of flow field parameter data through numerical simulation calculation, wherein each material corresponds to m groups of flow field parameter data; the screw plasticizing performance evaluation module determines the plasticizing parameters and weight ratio, obtains the plasticizing performance index value according to the plasticizing parameters and weight ratio, and obtains n×m groups of plasticizing performance index data based on the numerical simulation calculation results, wherein each material corresponds to m plasticizing performance index values; the screw structure optimization module determines the plasticizing parameters and weight ratio, and obtains the plasticizing performance index value according to the plasticizing parameters and weight ratio, and obtains n×m groups of plasticizing performance index data based on the numerical simulation calculation results, wherein each material corresponds to m plasticizing performance index values; the screw structure optimization module determines the plasticizing parameters and weight ratio, and obtains the plasticizing performance index value according to the plasticizing parameters and weight ratio, and obtains n×m groups of plasticizing performance index data based on the numerical simulation calculation results, and .... The module processes the m groups of screw structure parameter data established by the experimental design module and the m plasticizing performance index value results obtained by the screw plasticizing performance evaluation module, and obtains the screw structure parameter data that can make the plasticizing performance index value closest to the expected value by using the experimental design algorithm and the expected value size analysis of the plasticizing performance index data. Each material corresponds to a group of optimal screw structure parameter data, and n materials correspond to n groups of optimal screw structure parameter data; the optimal screw structure machine learning module performs machine learning training on the physical property parameter data of n materials and the corresponding n groups of optimal screw structure parameter data through a machine learning model to obtain an artificial intelligence model; the optimal screw structure generation module uses the artificial intelligence model obtained by the optimal screw structure machine learning module to calculate the corresponding new optimal screw structure parameter data according to the physical property parameter data of the input new material.

2. The intelligent design system for plasticizing screws based on material properties according to claim 1, characterized in that: The material properties of the material property database include material density, specific heat capacity, thermal conductivity, and rheological parameters, and each of n types of materials has n sets of material property parameter data. The screw structure parameter database includes data and variation ranges of geometric parameters related to the screw structure type. The screw structure parameters include screw diameter, screw aspect ratio, screw groove depth, screw flight width, helix angle, pitch, and compression ratio.

3. The intelligent design system for plasticizing screws based on material properties according to claim 1, characterized in that: The experimental design module includes a parameter design part based on the experimental design method, which establishes m types of screw structures according to the screw structure parameters and data range and the experimental design algorithm. The experimental design algorithm includes the orthogonal design method, the Taguchi design method, and the central design method to form m groups of screw structure parameter data.

4. The intelligent design system for plasticizing screws based on material properties according to claim 1, characterized in that: The flow field parameters calculated by the computer-aided numerical simulation module include temperature and viscosity; the plasticizing parameters of the screw plasticizing performance evaluation module include the temperature difference, viscosity difference and average viscosity at the front end of the screw obtained by numerical simulation calculation.

5. The intelligent design system for plasticizing screws based on material properties according to claim 1, characterized in that: The optimal screw structure machine learning module includes an input data interface, an output data interface, a machine learning model, and a hyperparameter optimization module. The input data interface inputs n types of material physical property parameter data, and the output data interface inputs n groups of optimal screw structure parameter data corresponding to the n types of materials. Machine learning training is performed through the machine learning model to obtain an artificial intelligence model; the optimal screw structure generation module includes a data input module, a data output module, and a model input module. The data input module is used to input new material physical property parameter data, the model input module is used to input the artificial intelligence model, and the data output module is used for the artificial intelligence model to calculate the corresponding new optimal screw structure parameter data based on the input new material physical property parameter data.

6. The intelligent design system for plasticizing screws based on material properties according to claim 1, characterized in that: The plasticizing performance index of the screw plasticizing performance evaluation module determines the weight of each plasticizing parameter through the plasticizing criterion. The screw structure optimization module takes the optimal plasticizing performance as the optimization goal during optimization, and the corresponding plasticizing performance index is minimized by the temperature difference, viscosity difference and average viscosity expected value of the plasticizing parameters.

7. The intelligent design system for plasticizing screws based on material properties according to any one of claims 1 to 6, characterized in that: The flow field parameters calculated by the computer-aided numerical simulation module also include screw torque, and the plasticizing parameters of the screw plasticizing performance evaluation module also include screw torque; the screw plasticizing performance evaluation module also includes a plasticizing criterion and a balance criterion and an energy-saving criterion that take into account the energy consumption generated by the screw torque, and the plasticizing criterion corresponds to the plasticizing parameter weights assigned by minimizing the expected values ​​of the temperature difference, viscosity difference, and average viscosity to obtain a plasticizing performance index value; The balance criterion corresponds to the plasticizing parameter weights assigned with the minimum expected values ​​of temperature difference, viscosity difference, average viscosity and screw torque to obtain the plasticizing performance index value; under the premise of a total weight of 100%, the balance criterion selects a weight of 50% for the temperature difference, viscosity difference and average viscosity, and a weight of 50% for the screw torque; the energy-saving criterion selects a weight of less than 50% for the temperature difference, viscosity difference and average viscosity, and the screw torque accounts for the remaining weight.

8. The intelligent design system for plasticizing screws based on material properties according to any one of claims 1 to 7, characterized in that: The screw plasticizing performance evaluation module performs dimensionless processing on the plasticizing parameter data before determining the plasticizing performance index value.

9. The intelligent design system for plasticizing screws based on material properties according to claim 1, characterized in that: The plasticizing parameter weights of the screw plasticizing performance evaluation module have the function of assigning weights by interventional artificial experience.

10. The intelligent design system for plasticizing screws based on material properties according to claim 1, characterized in that: The material property-based plasticizing screw intelligent design system also includes a screw plasticizing performance machine learning module and a screw plasticizing performance prediction module. The screw plasticizing performance machine learning module includes an input interface for material physical properties and screw structure parameters, a plasticizing parameter output interface, a machine learning model, and a hyper-parameter optimization module; the screw plasticizing performance prediction module includes an input interface for new material physical properties and screw structure parameters, an input interface for the model trained by the screw plasticizing performance machine learning module, and a plasticizing parameter output interface.

11. The intelligent design system for plasticizing screws based on material properties according to claim 1, characterized in that: The material property-based plasticizing screw intelligent design system also includes a computer-aided design model generation system, which can directly generate a screw model file in a format readable by computer-aided design software based on the optimal screw structure parameter data output by the optimal screw structure generation module.

12. The intelligent design system for plasticizing screws based on material properties according to claim 1, characterized in that: The machine learning model is a supervised learning model, including a multiple linear regression model, a random forest model, an XGB model, a GBDT model, and a neural network model. The prediction accuracy of the machine learning model is calculated by comparing the predicted results with the actual results to obtain the mean square error (MSE) and the coefficient of determination (R 2 ), the coefficient of determination ranges from 0 to 1. The closer it is to 1, the higher the model accuracy; the closer the mean square error is to 0, the higher the model accuracy.

13. The intelligent design system for plasticizing screws based on material properties according to claim 1, characterized in that: The material property-based plasticizing screw intelligent design system is also provided with a transfer learning module.

14. An intelligent design method for a plasticizing screw based on material properties, characterized by: The method uses the intelligent design system for plasticizing screws based on material properties according to any one of claims 1 to 13, and the specific steps of the method are as follows: Step 1: Store the physical property parameters of n kinds of materials of different types, manufacturers and brands in a material property database to establish a database of n sets of material physical property parameter data; Step 2: storing screw structural parameter data of different screw types in a screw structural parameter database, and establishing a database containing different screw structural parameters and value variation ranges; Step 3: Using the experimental design module, select an experimental design algorithm, and establish m types of screw structures corresponding to m groups of screw structure parameter data according to the screw structure parameters and data ranges in the screw structure parameter database; Step 4: Using a computer-aided design module, establish m corresponding screw models and barrel flow channel models based on the m sets of screw structural parameter data established by the experimental design module; Step 5: Using a computer-aided numerical simulation module, based on the n sets of material physical property parameter data and the m screw models established by the computer-aided design module, numerical simulations of the screw plasticizing process are sequentially performed to calculate and obtain n×m sets of flow field parameter data, where each material corresponds to m sets of flow field parameter data; Step 6: Determine the plasticizing parameters and weight ratios through the screw plasticizing performance evaluation module, calculate the plasticizing performance index values ​​based on the plasticizing parameters and weights, and calculate n×m groups of plasticizing performance index data based on the numerical simulation results, with each material corresponding to m plasticizing performance index values; Step 7: Processing the m groups of screw structure parameter data established by the experimental design module and the m plasticizing performance index values ​​obtained by the screw plasticizing performance evaluation module through the screw structure optimization module, and obtaining the screw structure parameter data that can make the plasticizing performance index value closest to the expected value by using the experimental design algorithm and the expected value analysis of the plasticizing performance index data. Each material corresponds to one group of optimal screw structure parameter data, and n materials correspond to n groups of optimal screw structure parameter data. Step 8: Input n sets of material physical property parameter data through the input data interface of the optimal screw structure machine learning module, input n sets of optimal screw structure parameter data corresponding to n materials through the output data interface, and perform machine learning training through the machine learning model to obtain an artificial intelligence model; Step 9: Input the physical property parameter data of the new material through the data input module of the optimal screw structure generation module, and input the artificial intelligence model trained by the optimal screw structure machine learning module through the model input module. The artificial intelligence model calculates the corresponding new optimal screw structure parameter data based on the input new material physical property parameter data, and outputs the new optimal screw structure parameter data through the data output module to complete the intelligent design task of the plasticizing screw based on material properties.

Images