Process parameter determination method and system, terminal equipment and storage medium

Abstract

The invention relates to the technical field of mold machining, and discloses a process parameter determination method and system, terminal equipment and a storage medium, and the process parameter determination method comprises the steps: obtaining model training data, carrying out the labeling processing of the model training data, and obtaining the processed model training data; performing model training based on the processed model training data to obtain a process parameter model, the process parameter model comprising a meta learner and a plurality of base learners; inputting the process processing data into each base learner to obtain initial process parameters; and inputting each initial process parameter into the meta-learner to obtain a target process parameter. According to the method, the target technological parameters are automatically determined through the technological parameter model, the problem that determination of the mold technological parameters depends on artificial experience is solved, the technological parameter model predicts the final target technological parameters based on the multiple base learners and the meta learners, single model deviation is avoided, and the accuracy and stability of technological parameter prediction are remarkably improved.

Description

Technical Field

[0001] This application relates to the field of mold processing technology, and in particular to a method, system, terminal equipment and storage medium for determining process parameters. Background Technology

[0002] In the extrusion molding process of profiles, the proper setting of mold process parameters directly affects the molding quality, production efficiency, and mold life. These mold process parameters need to be determined comprehensively based on factors such as the characteristics of the original material, the complexity of the product structure, and the mold design features. Currently, mold process parameters mainly rely on production technicians manually setting them according to process instruction manuals and their long-term accumulated operational experience. However, the effectiveness of setting mold process parameters is highly dependent on the experience level of technicians, requiring a long training period; the process instruction manuals provide range values ​​for process parameters, and these range values ​​are relatively broad and lack specificity, necessitating reliance on manual experience. Summary of the Invention

[0003] In view of this, embodiments of this application provide a method, system, terminal device, and storage medium for determining process parameters, which can effectively solve the problem that the determination of mold process parameters depends on human experience.

[0004] In a first aspect, embodiments of this application provide a method for determining process parameters, including: Obtain model training data, and label the model training data to obtain processed model training data. Based on the processed model training data, the model is trained to obtain a process parameter model, which includes a meta-learner and multiple base learners. The process data is input into each of the base learners to obtain the initial process parameters determined by each base learner based on the process data; Each of the initial process parameters is input into the meta-learner to obtain the target process parameters determined by the meta-learner based on each of the initial process parameters.

[0005] In a first possible embodiment of the first aspect, the model training data includes model input data and model output data, the model input data including adjustable parameters and non-adjustable parameters, and the method further includes: Adjust the adjustable parameters in the original model input data to obtain new model input data; The new model input data is input into the process parameter model to obtain the new model output data; Under the condition that the new model output data meets the preset reward standard, the new model input data and the corresponding new model output data are used as the optimal model training data; The process parameter model is trained based on the optimal model training data. In a second possible embodiment of the first aspect, the model training data includes categorical data and continuous data, and the step of labeling the model training data to obtain processed model training data includes: Encode each category of data into a continuous numerical label; Calculate the standardized deviation of each value of each of the continuous data, and encode each value as a corresponding target value label based on the standardized deviation, wherein the standardized deviation is the standardized distance of each value of the continuous data relative to the mean of the continuous data.

[0006] In a third possible embodiment of the first aspect, the step of training the model based on the processed model training data to obtain the process parameter model includes: Multiple machine learning models are used as the base learners, and a linear model is used as the meta-learner; Each base learner and the meta learner are trained based on the model training data. The meta-learner is optimized by using a regularization penalty term so that it determines the optimal weights for each of the base learners.

[0007] In a fourth possible embodiment of the first aspect, the step of inputting each of the initial process parameters into the meta-learner to obtain the target process parameters determined by the meta-learner based on each of the initial process parameters includes: The initial process parameters are weighted and summed based on the optimal weights to obtain the target process parameters.

[0008] In a fifth possible embodiment of the first aspect, the model output data includes a recommended die extrusion speed and other target process parameters, the new model output data includes a new recommended die extrusion speed and new other target process parameters, and the step of using the new model input data and the corresponding new model output data as optimal model training data under the condition that the new model output data meets a preset reward standard includes: Under the condition that the new recommended die extrusion speed is greater than the original recommended die extrusion speed, and that none of the other new target process parameters exceed a preset threshold, the new model input data and the corresponding new model output data are used as the optimal model training data. In a sixth possible embodiment of the first aspect, the target numerical label includes a first target numerical value, a second target numerical value, and a third target numerical value, wherein encoding each numerical value as a corresponding target numerical label according to the standardization deviation includes: Under the condition that the standardization deviation is less than zero, the numerical value is encoded as the first target numerical value; Under the condition that the standardization deviation is equal to zero, the numerical value is encoded as the second target numerical value; Under the condition that the standardization deviation is greater than zero, the numerical value is encoded as the third target numerical value.

[0009] Secondly, embodiments of this application provide a process parameter determination system, comprising: The training data acquisition module is used to acquire model training data, and to label the model training data to obtain processed model training data. The training module is used to train the model based on the processed model training data to obtain a process parameter model, which includes multiple base learners and meta learners. An initial parameter prediction module is used to input process data into each of the base learners to obtain initial process parameters determined by each base learner based on the process data. The target parameter prediction module is used to input each of the initial process parameters into the meta-learner to obtain the target process parameters determined by the meta-learner based on each of the initial process parameters.

[0010] Thirdly, embodiments of this application provide a terminal device, including a memory and a processor. The memory stores a computer program, and the computer program executes the above-described process parameter determination method when the processor is running.

[0011] Fourthly, embodiments of this application provide a computer-readable storage medium storing a computer program that executes the above-described process parameter determination method when run on a processor.

[0012] The embodiments of this application have the following beneficial effects: This embodiment of a method for determining process parameters includes: acquiring model training data; labeling the model training data to obtain processed model training data; training the model based on the processed model training data to obtain a process parameter model, the process parameter model including a meta-learner and multiple base learners; inputting process processing data into each base learner to obtain initial process parameters determined by each base learner based on the process processing data; and inputting each initial process parameter into the meta-learner to obtain target process parameters determined by the meta-learner based on the fusion of each initial process parameter. This application solves the problem of mold process parameter determination relying on human experience by building a process parameter model and automatically calculating the target process parameters through the process parameter model. The process parameter model uses multiple base learners, and the predicted initial process parameters are used as new columns and input into the meta-learner to predict the final target process parameters, avoiding single-model bias and significantly improving the accuracy and stability of process parameter prediction. Attached Figure Description

[0013] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0014] Figure 1 A schematic flowchart of a first method for determining process parameters according to an embodiment of this application is shown; Figure 2 A second flowchart illustrating the method for determining process parameters according to an embodiment of this application is shown. Figure 3 A third flowchart illustrating the process parameter determination method according to an embodiment of this application is shown. Figure 4 A schematic diagram of the process parameter determination system according to an embodiment of this application is shown.

[0015] Explanation of key component symbols: 200 - Process parameter determination system; 210 - Training data acquisition module; 220 - Training module; 230 - Initial parameter prediction module; 240 - Target parameter prediction module. Detailed Implementation

[0016] The technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments.

[0017] The components of the embodiments of this application described and illustrated in the accompanying drawings can be arranged and designed in a variety of different configurations. Therefore, the following detailed description of the embodiments of this application provided in the drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of the application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0018] In the following text, the terms "comprising," "having," and their cognates, which may be used in various embodiments of this application, are intended only to indicate a particular feature, number, step, operation, element, component, or combination thereof, and should not be construed as primarily excluding the presence of one or more other features, numbers, steps, operations, elements, components, or combinations thereof, or adding the possibility of one or more combinations thereof. Furthermore, the terms "first," "second," "third," etc., are used only for distinguishing descriptions and should not be construed as indicating or implying relative importance.

[0019] Unless otherwise specified, all terms used herein (including technical and scientific terms) shall have the same meaning as commonly understood by one of ordinary skill in the art to which the various embodiments of this application pertain. Terms (such as those defined in commonly used dictionaries) shall be interpreted as having the same meaning as in their contextual meaning in the relevant technical field and shall not be construed as having an idealized or overly formal meaning, unless clearly defined in the various embodiments of this application.

[0020] The following detailed description of some embodiments of this application is provided in conjunction with the accompanying drawings. Unless otherwise specified, the following embodiments and features can be combined with each other.

[0021] Figure 1 A flowchart illustrating a method for determining process parameters according to an embodiment of this application is shown. Exemplarily, the method for determining process parameters includes the following steps: S110: Obtain model training data, label the model training data, and obtain the processed model training data.

[0022] In this embodiment, the model training data consists of manual experience-based process data from the extrusion production of profiles using a die. The model training data includes model input data and model output data. The model input data includes, but is not limited to, profile weight per meter, profile processing difficulty coefficient, production batch output, profile surface treatment method, processing length, alloy grade, extrusion quantity, profile raw material temperature, and profile raw material length. Profile weight per meter refers to the mass of the profile per unit length. A higher profile weight per meter indicates a more complex cross-section or thicker wall, requiring greater extrusion pressure and a correspondingly lower speed. The profile processing difficulty coefficient quantifies the ease or difficulty of producing a particular profile. Production batch output refers to the total quantity required for the current order or task batch. Profile surface treatment method refers to the surface treatment process performed on the finished profile. Processing length refers to the length obtained from the extrusion of a single profile. Alloy grade indicates the chemical composition of the profile. Extrusion quantity represents the number of profiles extruded simultaneously from the die during a single extrusion process. Profile raw material temperature is the temperature to which the profile raw material is heated before entering the extruder. Profile raw material length is the length of the profile before extrusion.

[0023] The model output data includes, but is not limited to, recommended die extrusion speed, recommended bar feed zone temperature setting, recommended bar output zone temperature setting, recommended extrusion exit temperature, recommended die mounting temperature, recommended quenching exit temperature, recommended residual thickness, recommended head and tail length, and feed stretching amount. The recommended bar feed zone temperature setting is the temperature at which the profile raw material enters the die; the recommended bar output zone temperature setting is the temperature of the profile extrusion die; the recommended extrusion exit temperature is the die extrusion exit temperature; the recommended die mounting temperature is the temperature at which the die operates; the recommended quenching exit temperature is the suggested temperature the profile should reach after cooling; the recommended residual thickness is the thickness of the residual profile raw material that should be retained after each extrusion; the recommended head and tail length is the suggested length of the unstable section of profile that needs to be removed; and the feed stretching amount is the amount of deformation to be applied during the stretching and straightening process.

[0024] For example, the model training data includes categorical data and continuous data. Continuous data is data that can take any real value within a certain range, while categorical data represents the category or state to which an object belongs, and its values ​​are a finite number of discrete labels.

[0025] In one implementation, each category of data is categorized into continuous numerical labels. For continuous data with significant fluctuations, the mean and square root of variance are calculated using statistical methods, followed by the standardized deviation, to determine new target numerical labels for model learning. For example, three alloy grades are categorized into two labels, 0 and 1, and the four categories of profile surface treatment methods are coded into label data 0, 1, 2, and 3.

[0026] In one embodiment, the standardized deviation of each value in each consecutive data point is calculated. Each value is then encoded into a corresponding target value label based on the standardized deviation. The standardized deviation is the standardized distance of each value in the consecutive data points relative to the mean of the consecutive data points. In this embodiment, the formula for the standardized deviation is:

[0027] This represents the standardization deviation of a specific value in continuous data. Represents the first continuous data Numerical value This represents the average value of continuous data. It represents the square root of the variance of continuous data.

[0028] In one embodiment, the target numerical label includes a first target numerical value, a second target numerical value, and a third target numerical value. The numerical value is encoded as the first target numerical value when the standardized deviation is less than zero; the numerical value is encoded as the second target numerical value when the standardized deviation is equal to zero; and the numerical value is encoded as the third target numerical value when the standardized deviation is greater than zero. For example, the first, second, and third target numerical values ​​can be set as consecutive values. For instance, the first target numerical value can be set to 1, and the second and third target numerical values ​​can both be set to 0. If the standardized deviation of the length of a certain profile raw material is less than zero, the length of the profile raw material is encoded as 1; if the standardized deviation of the length of a certain profile raw material is greater than or equal to zero, the length of the profile raw material is encoded as 0.

[0029] S120, based on the processed model training data, performs model training to obtain the process parameter model, which includes a meta-learner and multiple base learners.

[0030] As an example, the process parameter model is a stacked model. The prediction results of multiple base learners are used as new features through learning, and a meta learner is trained to make the final prediction.

[0031] In one embodiment, such as Figure 2 As shown, the training process for the process parameter model includes the following steps: S121 employs multiple machine learning models as base learners and a linear model as a meta-learner.

[0032] S122, each base learner and meta learner are trained based on the model training data.

[0033] In one implementation, the machine learning model includes, but is not limited to, random forests, catboost, and xgboost. Random forests are a powerful algorithm based on the Bagging ensemble learning concept, which makes predictions by constructing multiple decision trees and combining their results. During training, its core lies in introducing double randomness: first, by randomly selecting a subset of model training data samples through bootstrap sampling to train each tree; second, when splitting a node in each tree, finding the optimal split point only from the randomly selected subset of model training data. This method effectively reduces the correlation between models, thereby significantly improving the model's generalization ability and making it highly robust to overfitting. Random forests can naturally evaluate feature importance and are insensitive to noise and missing values ​​in the data, making them suitable as base learners to capture stable, generalized patterns in the data.

[0034] CatBoost, a member of the gradient boosting algorithm family developed by Yandex, excels in handling categorical features. Its core advantage lies in its use of techniques such as ordered target encoding, which automatically and efficiently processes categorical features without cumbersome preprocessing, while avoiding target leakage issues. During training, CatBoost uses a symmetric tree structure and an ordered boosting strategy to mitigate gradient bias, resulting in a more stable training process and stronger generalization ability. For training data containing a large number of categorical variables or complex nonlinear relationships between features, CatBoost, as a base learner, can provide very accurate predictions.

[0035] XGBoost is a highly optimized gradient boosting decision tree algorithm renowned for its exceptional efficiency and performance. During training, XGBoost makes several key improvements to the traditional gradient boosting framework, including introducing a regularization term into the objective function to control model complexity and utilizing the second derivative of the loss function to more accurately fit the residuals, thereby accelerating convergence. XGBoost also supports parallel processing, optimizing computational speed. Due to its powerful fitting ability and fine-grained control over model complexity, XGBoost produces very accurate predictions and, as a base learner, excels at capturing more subtle and complex patterns in data.

[0036] S123, the meta-learner is optimized and trained by a regularization penalty term so that the meta-learner can determine the optimal weights of each base learner.

[0037] In one embodiment, the linear model employs a ridge regression model, which is selected as the meta-learner to integrate the predictions of three powerful base learners. During training, a regularization penalty term is introduced to address the overfitting problem that may exist in ordinary linear regression, especially in scenarios with highly correlated features (multicollinearity). The regularization penalty term is L2 regularization. This application uses the ridge regression model as the meta-learner and performs L2-regularized linear weighted fusion of the initial process parameters output by multiple base learners, which can effectively suppress the risk of overfitting while ensuring model interpretability. Especially when there is a strong correlation between the predictions of multiple base learners, L2 regularization can automatically adjust the weight distribution of each model to obtain the optimal weights of each base learner, avoiding abnormal amplification of the weights of individual models, thereby improving the overall stability and generalization ability of the ensemble model. Compared with the unregularized ordinary linear regression fusion method, this application significantly improves the reliability of parameter prediction under new operating conditions.

[0038] In one embodiment, the meta-learner learns a set of optimal weights, corresponding to the influence of the three base learners on the target process parameters. If a certain weight is particularly large, it indicates that the meta-learner is overly dependent on a single base learner, and the prediction of the target process parameters is unstable under the new operating conditions. Ridge regression introduces the square of the L2 norm to calculate the regularization term. The formula for calculating the regularization term is:

[0039] Represents the regularization term. This represents the regularization coefficient, used to control the strength of the penalty applied to model complexity. The regularization term represents the weights of the z-th base learner. The larger the regularization term, the greater the difference in weights and the uneven distribution. The smaller the regularization term, the smaller the difference in weights and the uneven distribution. During the training process of the meta-learner, it will automatically tend to select the optimal set of weights that are evenly distributed and not overly concentrated.

[0040] S130, input the process data into each base learner to obtain the initial process parameters determined by each base learner based on the process data.

[0041] Exemplary, the processing data serves as the model input, and the initial process parameters are the model output. After training, the base learners can independently predict the corresponding initial process parameters based on the processing data from the die-extruded profile. These initial process parameters reflect the learning outcomes of different machine learning models from their respective modeling perspectives, focusing on the implicit patterns in the training data. The predictions output in parallel by multiple base learners collectively constitute the input feature space of the meta-learner, providing high-quality and diverse decision-making support for subsequently generating the final optimized target process parameters through ridge regression.

[0042] In this embodiment, by integrating the prediction results of multiple base learners, the process parameter model can effectively reduce the bias and variance of a single model and improve the overall prediction accuracy. The process parameter model supports the use of base learners with different algorithms and architectures, and this diversity can reduce the risk of overfitting.

[0043] S140, input each initial process parameter into the meta-learner to obtain the target process parameters determined by the meta-learner based on each initial process parameter.

[0044] In one embodiment, the initial process parameters are weighted and summed based on optimal weights to obtain the target process parameters. In this embodiment, the formula for calculating the weighted sum of the initial process parameters is as follows:

[0045] For a specific target process parameter, such as the target extrusion speed, The initial process parameters predicted by the first base learner. The optimal weights for the second base learner. The initial process parameters predicted by the second base learner. The optimal weights for the third base learner The initial process parameters predicted by the third base learner.

[0046] In another embodiment, the model input data includes adjustable parameters and non-adjustable parameters. The adjustable parameters are continuous data from the model input data, and the non-adjustable parameters are categorical data from the model input data, such as... Figure 3 As shown, the method for determining process parameters also includes: S150: Adjust the adjustable parameters in the original model input data to obtain new model input data.

[0047] By way of example, this application arbitrarily adjusts the adjustable parameters in the model input data based on the adjustment range. For example, the adjustment range of the processing length is 100. Each time the adjustment is made, the processing length is increased or decreased by 100. By adjusting the adjustable parameters, new model input data is obtained.

[0048] S160: Input the new model input data into the process parameter model to obtain the new model output data.

[0049] S170: Under the condition that the new model output data meets the preset reward standard, the new model input data and the corresponding new model output data are used as the optimal model training data.

[0050] In one embodiment, upon receiving new model input data, the corresponding model output data also changes from the old model output data to the new model output data. The model output data includes the recommended die extrusion speed and other target process parameters. The new model output data includes the new recommended die extrusion speed and the new target process parameters, and the corresponding recommended die extrusion speed also changes to the new recommended die extrusion speed. In this application, under the condition that the new recommended die extrusion speed is greater than the original recommended die extrusion speed, and each new target process parameter does not exceed a preset threshold, the new model input data and the corresponding new model output data are used as the optimal model training data. The preset threshold is the effective range within which the target process parameters satisfy the die processing performance indicators.

[0051] S180 trains the process parameter model based on the optimal model training data.

[0052] In this embodiment, the present application utilizes the accumulated optimal model training data to further incrementally train the process parameter model, thereby continuously improving the model's ability to recommend high-efficiency and high-quality process solutions. This training method is consistent with the model training method described above and will not be elaborated upon here. Compared to traditional static models that can only rely on historical data for training, this application can proactively explore potential better operating conditions and select new samples that are both efficient and safe for model iteration and upgrades.

[0053] Figure 4 A schematic diagram of a process parameter determination system 200 according to an embodiment of this application is shown. Exemplarily, the process parameter determination system 200 includes: The training data acquisition module 210 is used to acquire model training data, label the model training data, and obtain processed model training data.

[0054] The training module 220 is used to train the model based on the processed model training data to obtain the process parameter model, which includes multiple base learners and meta learners.

[0055] The initial parameter prediction module 230 is used to input the process data into each base learner to obtain the initial process parameters determined by each base learner based on the process data.

[0056] The target parameter prediction module 240 is used to input each initial process parameter into the meta-learner to obtain the target process parameter determined by the meta-learner based on the fusion of each initial process parameter.

[0057] It is understood that the system in this embodiment corresponds to the process parameter determination method in the above embodiment, and the options in the above embodiment are also applicable to this embodiment, so they will not be described again here.

[0058] This application also provides a terminal device, exemplary of which includes a processor and a memory, wherein the memory stores a computer program, and the processor executes the computer program to enable the terminal device to perform the functions of the above-described process parameter determination method or the various modules in the above-described process parameter determination system.

[0059] The processor can be an integrated circuit chip with signal processing capabilities. The processor can be a general-purpose processor, including at least one of a Central Processing Unit (CPU), Graphics Processing Unit (GPU), Network Processor (NP), Digital Signal Processor (DSP), Application-Specific Integrated Circuit (ASIC), Field-Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. The general-purpose processor can be a microprocessor or any conventional processor, capable of implementing or executing the methods, steps, and logic block diagrams disclosed in the embodiments of this application.

[0060] The memory can be, but is not limited to, Random Access Memory (RAM), Read Only Memory (ROM), Programmable Read-Only Memory (PROM), Erasable Programmable Read-Only Memory (EPROM), Electrically Erasable Programmable Read-Only Memory (EEPROM), etc. The memory is used to store computer programs, and the processor can execute the computer programs accordingly after receiving execution instructions.

[0061] This application also provides a computer-readable storage medium for storing the computer program used in the aforementioned terminal device. For example, the computer-readable storage medium may include, but is not limited to, various media capable of storing program code, such as a USB flash drive, a portable hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0062] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can also be implemented in other ways. The apparatus embodiments described above are merely illustrative. For example, the flowcharts and block diagrams in the accompanying drawings show the architecture, functionality, and operation of possible implementations of apparatus, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that, in alternative implementations, the functions marked in the blocks may occur in a different order than those marked in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagram and / or flowchart, and combinations of blocks in the block diagram and / or flowchart, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

[0063] In addition, the functional modules or units in the various embodiments of this application can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0064] If the aforementioned functions are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a smartphone, personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application.

[0065] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A method for determining process parameters, characterized in that, include: Obtain model training data, and label the model training data to obtain processed model training data. Based on the processed model training data, the model is trained to obtain a process parameter model, which includes a meta-learner and multiple base learners. The process data is input into each of the base learners to obtain the initial process parameters determined by each base learner based on the process data; Each of the initial process parameters is input into the meta-learner to obtain the target process parameters determined by the meta-learner based on each of the initial process parameters.

2. The method for determining process parameters according to claim 1, characterized in that, The model training data includes model input data and model output data, the model input data includes adjustable parameters and non-adjustable parameters, and the method further includes: Adjust the adjustable parameters in the original model input data to obtain new model input data; The new model input data is input into the process parameter model to obtain the new model output data; Under the condition that the new model output data meets the preset reward standard, the new model input data and the corresponding new model output data are used as the optimal model training data; The process parameter model is trained based on the optimal model training data.

3. The method for determining process parameters according to claim 1, characterized in that, The model training data includes categorical data and continuous data. The step of labeling the model training data to obtain processed model training data includes: Encode each category of data into a continuous numerical label; Calculate the standardized deviation of each value of each of the continuous data, and encode each value as a corresponding target value label based on the standardized deviation, wherein the standardized deviation is the standardized distance of each value of the continuous data relative to the mean of the continuous data.

4. The method for determining process parameters according to claim 1, characterized in that, The process parameter model is obtained by training the model based on the processed model training data, including: Multiple machine learning models are used as the base learners, and a linear model is used as the meta-learner; Each base learner and the meta learner are trained based on the model training data. The meta-learner is optimized by using a regularization penalty term so that it determines the optimal weights for each of the base learners.

5. The method for determining process parameters according to claim 4, characterized in that, The step of inputting each of the initial process parameters into the meta-learner to obtain the target process parameters determined by the meta-learner based on each of the initial process parameters includes: The initial process parameters are weighted and summed based on the optimal weights to obtain the target process parameters.

6. The method for determining process parameters according to claim 2, characterized in that, The model output data includes recommended die extrusion speed and other target process parameters. The new model output data includes new recommended die extrusion speed and new other target process parameters. The step of using the new model input data and the corresponding new model output data as optimal model training data, under the condition that the new model output data meets a preset reward standard, includes: Under the condition that the new recommended die extrusion speed is greater than the original recommended die extrusion speed, and that none of the other new target process parameters exceed a preset threshold, the new model input data and the corresponding new model output data are used as the optimal model training data.

7. The method for determining process parameters according to claim 3, characterized in that, The target numerical labels include a first target numerical value, a second target numerical value, and a third target numerical value. Encoding each numerical value into a corresponding target numerical label based on the standardized deviation includes: Under the condition that the standardization deviation is less than zero, the numerical value is encoded as the first target numerical value; Under the condition that the standardization deviation is equal to zero, the numerical value is encoded as the second target numerical value; Under the condition that the standardization deviation is greater than zero, the numerical value is encoded as the third target numerical value.

8. A process parameter determination system, characterized in that, include: The training data acquisition module is used to acquire model training data, and to label the model training data to obtain processed model training data. The training module is used to train the model based on the processed model training data to obtain a process parameter model, which includes multiple base learners and meta learners. An initial parameter prediction module is used to input process data into each of the base learners to obtain initial process parameters determined by each base learner based on the process data. The target parameter prediction module is used to input each of the initial process parameters into the meta-learner to obtain the target process parameters determined by the meta-learner based on each of the initial process parameters.

9. A terminal device, characterized in that, It includes a memory and a processor, the memory storing a computer program that executes the process parameter determination method according to any one of claims 1 to 7 when the processor is running.

10. A computer-readable storage medium, characterized in that, It stores a computer program that, when run on a processor, executes the process parameter determination method according to any one of claims 1 to 7.

Images