Intelligent generation method, device and equipment of injection molding machine molding process parameters and medium

Abstract

The application discloses an injection molding machine forming process parameter intelligent generation method and device, equipment and medium, relates to injection molding intelligent manufacturing and industrial artificial intelligence technical field. The method comprises the following steps: acquiring multi-source data corresponding to a current injection molding task, and encoding the multi-source data into a current working condition feature vector; inputting the current working condition feature vector and a current working condition defect risk prior vector into a pre-trained conditional diffusion generation model to obtain multiple groups of candidate forming process parameter packages; performing constraint projection correction on the multiple groups of candidate forming process parameter packages, screening a target forming process parameter package from the multiple groups of candidate forming process parameter packages through a constraint Bayesian optimization algorithm and a pre-trained quality prediction model, and converting the target forming process parameter package into an executable forming process parameter package of a target injection molding machine based on a controller parameter dictionary of the target injection molding machine. The method can automatically generate the injection molding machine forming process parameters, and greatly improves the production efficiency.

Description

Technical Field

[0001] This application relates to the fields of intelligent manufacturing and industrial artificial intelligence in injection molding, and in particular to a method, apparatus, equipment and medium for intelligent generation of injection molding process parameters. Background Technology

[0002] As an important process in the production of plastic products, injection molding is widely used in home appliances, automobiles, packaging and industrial parts. Its process parameters, such as barrel temperature, mold temperature, injection speed and injection pressure, can directly affect the dimensional accuracy of the product, molding cycle, energy consumption and defect rate.

[0003] Currently, the molding process parameters of injection molding machines in related technologies mainly rely on the experience of process engineers to set, resulting in a large number of trial moldings. This is especially true for new products, new materials, new molds, or machine replacement needs, as the development process is time-consuming and labor-intensive, which seriously restricts production efficiency. Summary of the Invention

[0004] In view of the above-mentioned defects or deficiencies in the related technologies, it is desirable to provide a method, device, equipment and medium for intelligent generation of injection molding process parameters, which can automatically generate injection molding process parameters that meet process constraints, quality requirements and equipment execution conditions, thereby improving production efficiency.

[0005] Firstly, this application provides a method for intelligently generating injection molding process parameters, the method comprising: Acquire multi-source data corresponding to the current injection molding task, and encode the multi-source data into a feature vector of the current working condition. The multi-source data includes product structure data, material performance data, mold structure data, equipment capacity data, and production target data. The current working condition feature vector is mapped to a pre-constructed injection molding process knowledge graph, and the current working condition defect risk prior vector is output through a pre-trained graph neural network. The current working condition feature vector and the current working condition defect risk prior vector are input into a pre-trained conditional diffusion generation model to obtain multiple sets of candidate molding process parameter packages. Constrained projection correction is performed on the multiple sets of candidate molding process parameter packages, and the target molding process parameter package is selected from the multiple sets of candidate molding process parameter packages through constrained Bayesian optimization algorithm and pre-trained quality prediction model. Based on the controller parameter dictionary of the target injection molding machine, the target molding process parameter package is converted into the executable molding process parameter package of the target injection molding machine.

[0006] Optionally, in some embodiments of this application, the conditional diffusion generative model is trained during the pre-training phase by... Noise is gradually added. For the first The parameter packet after adding noise. For noise scheduling parameters, This is a package of successful process parameters based on historical operating conditions. Use Gaussian noise and train a denoising network. The training loss function is , This is a feature vector representing historical operating conditions. This is the prior vector for defects and risks under historical operating conditions.

[0007] Optionally, the constraint projection correction of the multiple sets of candidate molding process parameter packages described in some embodiments of this application includes: Each candidate molding process parameter package is verified sequentially, and the hard constraint violation loss, soft constraint deviation loss, and conflict constraint loss of each candidate molding process parameter package are calculated respectively. If the hard constraint violation loss and conflict constraint loss of the candidate molding process parameter package are not zero, or the soft constraint deviation loss of the candidate molding process parameter package exceeds a preset allowable threshold, then the primary goal is to keep the original values ​​of the candidate molding process parameter package unchanged, and the correction goal is to reduce the constraint violation loss, so as to obtain a corrected parameter package.

[0008] Optionally, the modified parameter package described in some embodiments of this application , For the parameter package during the correction process, This is a candidate molding process parameter package. For hard constraints, the penalty for violation The corresponding weights For soft constraint deviation loss The corresponding weights To constrain losses in conflict The corresponding weights.

[0009] Optionally, the hard constraint violation loss described in some embodiments of this application , For the first hard constraint The corresponding weights It is a linear rectified function. For the first hard constraint The corresponding maximum constraint upper limit value; The soft constraint deviation loss , For the first Soft constraints The corresponding weights For the first Soft constraints The corresponding target value; and the conflict constraint loss. , For the first Individual risks The corresponding weights For the first Individual risks The corresponding maximum risk value.

[0010] Optionally, the hard constraints described in some embodiments of this application This includes material processing window constraints, equipment capacity boundary constraints, mold structure limitation constraints, and safety boundary constraints; the soft constraints... This includes molding cycle targets, energy consumption targets, quality targets, and stability targets, as well as the aforementioned risks. This includes the risks of burning, flash, and shrinkage.

[0011] Optionally, in some embodiments of this application, the method of selecting the target molding process parameter package from the multiple sets of candidate molding process parameter packages using a constrained Bayesian optimization algorithm and a pre-trained quality prediction model includes: The candidate molding process parameter packages are processed from different aspects using the tabular feature sub-prediction model, the time-series process sub-prediction model, and the graphic risk sub-prediction model of the quality prediction model to obtain a comprehensive prediction result. Specifically, the tabular feature sub-prediction model processes structured features and outputs a basic quality prediction result; the time-series process sub-prediction model processes curve process data and outputs processing parameter prediction results; the graphic risk sub-prediction model processes prior defect risks and outputs defect risk prediction results; and the comprehensive prediction result includes the basic quality prediction result, the processing parameter prediction result, and the defect risk prediction result. With the optimization objectives of maximizing the basic quality prediction result, minimizing the processing parameter prediction result, and minimizing the defect risk prediction result, the constrained Bayesian optimization algorithm is used to calculate the acquisition function of each candidate molding process parameter package, and the candidate molding process parameter package with the highest acquisition function value is selected as the target molding process parameter package.

[0012] Secondly, this application provides an intelligent generation device for injection molding process parameters, the intelligent generation device for injection molding process parameters comprising: The acquisition module is used to acquire multi-source data corresponding to the current injection molding task and encode the multi-source data into a current working condition feature vector. The multi-source data includes product structure data, material performance data, mold structure data, equipment capacity data, and production target data. The candidate module is used to map the current working condition feature vector to a pre-constructed injection molding process knowledge graph, output the current working condition defect risk prior vector through a pre-trained graph neural network, and input the current working condition feature vector and the current working condition defect risk prior vector into a pre-trained conditional diffusion generation model to obtain multiple sets of candidate molding process parameter packages. The conversion module is used to perform constraint projection correction on the multiple sets of candidate molding process parameter packages, and to select the target molding process parameter package from the multiple sets of candidate molding process parameter packages through a constraint Bayesian optimization algorithm and a pre-trained quality prediction model. Based on the controller parameter dictionary of the target injection molding machine, the target molding process parameter package is converted into an executable molding process parameter package of the target injection molding machine.

[0013] Thirdly, this application provides an electronic device, which includes a processor and a memory. The memory stores at least one instruction, at least one program, a code set, or an instruction set. The instruction, the program, the code set, or the instruction set is loaded and executed by the processor to implement the steps of the intelligent generation method for injection molding process parameters of any one of the first aspects.

[0014] Fourthly, this application provides a computer-readable storage medium storing one or more programs that can be executed by one or more processors to implement the steps of the intelligent generation method for injection molding process parameters of any one of the first aspects.

[0015] As can be seen from the above technical solutions, the embodiments of this application have the following advantages: This application provides an intelligent generation method, apparatus, equipment, and medium for injection molding machine molding process parameters. It acquires multi-source data corresponding to the current injection molding task, including product structure data, material performance data, mold structure data, equipment capacity data, and production target data, taking all factors into account. The multi-source data is encoded into a current operating condition feature vector, which is then automatically mapped to a pre-constructed injection molding process knowledge graph. A pre-trained graph neural network outputs a prior vector for the current operating condition's defect risk. The current operating condition feature vector and the prior vector for the current operating condition's defect risk are input into a pre-trained conditional diffusion generation model to obtain multiple sets of candidate molding process parameter packages. Constrained projection correction is then applied to these multiple sets of candidate molding process parameter packages. A constrained Bayesian optimization algorithm and a pre-trained quality prediction model are used to select the target molding process parameter package from the multiple sets of candidate molding process parameter packages. Based on the controller parameter dictionary of the target injection molding machine, the target molding process parameter package is converted into an executable molding process parameter package for the target injection molding machine. This method is objective, convenient, and significantly improves production efficiency. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 A flowchart illustrating an intelligent generation method for injection molding process parameters provided in this application embodiment; Figure 2 A structural block diagram of an intelligent generation device for injection molding process parameters provided in this application embodiment; Figure 3 This is a structural block diagram of an electronic device provided in an embodiment of this application. Detailed Implementation

[0018] The technical solutions of 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. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0019] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0020] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. The following examples illustrate this. Figures 1 to 3 This application provides a detailed description of the intelligent generation method, apparatus, equipment, and medium for injection molding process parameters provided in the embodiments of this application.

[0021] Please refer to Figure 1 This is a flowchart illustrating an intelligent method for generating injection molding process parameters according to an embodiment of this application. The method specifically includes the following steps: S101: Obtain the multi-source data corresponding to the current injection molding task and encode the multi-source data into the current working condition feature vector. The multi-source data includes product structure data, material performance data, mold structure data, equipment capacity data, and production target data.

[0022] In some embodiments of this application, product structure data includes, but is not limited to, product weight, maximum wall thickness, minimum wall thickness, wall thickness change rate, flow length ratio, projected area, appearance grade, and dimensional tolerance grade, the corresponding codes of which are: , For product weight, For the maximum wall thickness, This is the minimum wall thickness, and so on. Material property data includes, but is not limited to, material type, melting temperature, melt flow index, shrinkage rate, drying temperature, drying time, heat sensitivity rating, water absorption rating, and shear sensitivity rating, the corresponding codes of which are... , Material type The melting temperature is... This refers to the melt flow index, and so on. Mold structure data includes, but is not limited to, the number of cavities, gate type, number of gates, runner type, whether it is a hot runner, venting capacity, number of cooling water channels, and cooling uniformity index, etc., and their corresponding codes are: , For the number of cavities, For gate type, This refers to the number of gates, and so on. Equipment capacity data includes, but is not limited to, the maximum injection pressure, maximum injection speed, screw diameter, maximum injection volume, clamping force, number of heating zones, maximum mold opening stroke, and controller parameter dictionary, etc., with corresponding codes as follows: , Maximum injection pressure, For maximum injection speed, For the screw diameter, and so on. For each parameter item in the controller parameter dictionary, extract the parameter name, parameter type, unit, upper and lower limits, precision, number of segments, writability, communication protocol field, and writing order, and then encode them. Additionally, production target data includes, but is not limited to, molding cycle targets, yield targets, energy consumption targets, dimensional stability targets, appearance quality targets, and defect control targets, and their corresponding codes are as follows: , To achieve the molding cycle target, To achieve the yield target, For energy consumption targets, the same principle applies. For continuous data, normalization is used for encoding; for discrete data, one-hot encoding, tag encoding, or embedding encoding are used. Then, the current operating condition feature vector is obtained through encoder fusion. .

[0023] S102 maps the current working condition feature vector to the pre-constructed injection molding process knowledge graph, outputs the current working condition defect risk prior vector through a pre-trained graph neural network, and inputs the current working condition feature vector and the current working condition defect risk prior vector into a pre-trained conditional diffusion generation model to obtain multiple sets of candidate molding process parameter packages.

[0024] In some embodiments of this application, an injection molding process knowledge graph is included. , For a set of nodes, Let be the set of edges. This is a set of process constraint rules. The node set includes, but is not limited to, product nodes, material nodes, mold nodes, equipment nodes, process parameter nodes, and defect type nodes. The edge set includes, but is not limited to, the influence relationships between materials and process parameters, the influence relationships between product structure and process parameters, the restrictive relationships between mold structure and process parameters, the boundary constraint relationships between equipment capacity and process parameters, the inducing relationships between process parameters and defect risk, the suppressing relationships between process parameters and defect risk, the synergistic relationships between multiple process parameters, and the conflicting relationships between multiple process parameters. In the current operating condition feature vector... When mapping the injection molding process knowledge graph, the corresponding product nodes, material nodes, mold nodes, and equipment nodes are activated, and the prior vector of defect risk is calculated through a graph neural network. Defect risks include, but are not limited to, scorching, flash, shrinkage, short shots, silver streaks, and warpage. The graph neural network uses an existing structure, with training samples consisting of historical trial molding records and stable production records. Each sample includes product structure data, material performance data, mold structure data, equipment capacity data, actual executed process parameters, process curves, quality inspection results, and defect labels. During training, corresponding nodes in the knowledge graph are activated based on the working condition characteristics of each sample, using defect labels as supervisory signals to output the prior risks of various defects under the current working condition. Besides the injection molding process knowledge graph and graph neural network implementation, process rule bases, expert systems, constraint matrices, defect mechanism rule tables, and causal graph models can also be used to express process knowledge constraints.

[0025] In some embodiments of this application, the conditional diffusion generation model adopts an existing structure, and its training samples also consist of historical trial molding records and stable production records. Each sample includes product structure data, material performance data, mold structure data, equipment capacity data, production target data, prior defect risk, and a package of molding process parameters that have been verified as qualified through trial molding. In the pre-training stage, through... Noise is gradually added. For the first The parameter packet after adding noise. For noise scheduling parameters, This is a package of successful process parameters based on historical operating conditions. Use Gaussian noise and train a denoising network. The training loss function is , This is a feature vector representing historical operating conditions. The model uses a prior vector representing historical defects and risks to learn and recover a reasonable set of process parameters given the characteristics of the operating conditions and the prior risk of defects. During the actual parameter generation stage, the model progressively denoises from random noise, ultimately obtaining multiple sets of candidate molding process parameter packages. These candidate packages include, but are not limited to, barrel temperatures, nozzle temperatures, mold temperatures, segmented injection speeds, upper limit of injection pressure, segmented holding pressure, holding time, cooling time, back pressure, screw speed, unwinding amount, clamping force, mold opening and closing parameters, and ejection parameters. Besides the conditional diffusion generation model, parameters can also be generated using conditional generative adversarial networks, variational autoencoders, large language models, and reinforcement learning policy networks.

[0026] S103 performs constraint projection correction on multiple candidate molding process parameter packages, and selects the target molding process parameter package from multiple candidate molding process parameter packages through constraint Bayesian optimization algorithm and pre-trained quality prediction model. Based on the controller parameter dictionary of the target injection molding machine, the target molding process parameter package is converted into an executable molding process parameter package of the target injection molding machine.

[0027] In some embodiments of this application, during the correction process, each candidate molding process parameter package can be sequentially verified, and the hard constraint violation loss, soft constraint deviation loss, and conflict constraint loss of the candidate molding process parameter package can be calculated respectively. If the hard constraint violation loss and conflict constraint loss of the candidate molding process parameter package are not zero, or the soft constraint deviation loss of the candidate molding process parameter package exceeds a preset allowable threshold, then maintaining the original values ​​of the candidate molding process parameter package remains unchanged is the primary objective, and reducing the constraint violation loss is the correction objective, thus obtaining the corrected parameter package. For example, the corrected parameter package... , For the parameter package during the correction process, This is a candidate molding process parameter package. For hard constraints, the penalty for violation The corresponding weights For soft constraint deviation loss The corresponding weights To constrain losses in conflict The corresponding weights can be preset based on historical trial molding records and stable production records. Another example is the loss due to hard constraint violation. , For the first hard constraint The corresponding weights It is a linear rectified function. For the first hard constraint The corresponding maximum constraint upper limit, soft constraint deviation loss , For the first Soft constraints The corresponding weights For the first Soft constraints The corresponding target value, and the conflict constraint loss. , For the first Individual risks The corresponding weights For the first Individual risks The corresponding maximum risk values, and these weights can also be preset based on historical trial molding records and stable production records.

[0028] hard constraints These include, but are not limited to, material processing window constraints, equipment capacity boundary constraints, mold structure limitation constraints, and safety boundary constraints, for example... , For barrel temperature, This is the lower limit of the material's melt processing temperature. This is the upper limit of the material's melting and processing temperature. , For injection pressure, The equipment safety factor is typically a value between 0 and 1. This refers to the maximum allowable injection pressure of the injection molding machine. (Soft constraint) This includes targets for molding cycle time, energy consumption, quality, and stability, as well as risks. This includes risks of scorching, flash, and shrinkage, such as the scorching risk value. , Typically, a value between 0 and 1 is used; a higher value indicates a higher risk of burning. For injection speed, The melting temperature is... The mold venting capability level indicates whether the mold venting is sufficient and the risk value of flash. , Typically, a value between 0 and 1 is used; a higher value indicates a higher risk of edge defects. For injection pressure, To maintain pressure, For the clamping force of the injection molding machine, and the shrinkage risk value. , Typically, a value between 0 and 1 is used; a larger value indicates a higher risk of shrinkage. To maintain pressure, For the pressure holding time, Cooling time, For product wall thickness, , and The risk calculation function can be implemented by a rule function, a machine learning model, or a neural network model. The model is obtained through supervised training using labeled historical trial records and stable production records.

[0029] In some embodiments of this application, during the screening process, the tabular feature sub-prediction model, the temporal process sub-prediction model, and the graphical risk sub-prediction model of the quality prediction model can be used to process each candidate molding process parameter package from different aspects to obtain a comprehensive prediction result. The tabular feature sub-prediction model is used to process structured features such as working condition feature vectors and working condition defect risk prior vectors and output basic quality prediction results. The temporal process sub-prediction model is used to process curve process data such as injection pressure curve, screw position curve, mold cavity pressure curve, and temperature curve and output processing parameter prediction results. The graphical risk sub-prediction model is used to process defect risk priors such as scorching risk, flash risk, and shrinkage risk and output defect risk prediction results. The basic quality prediction results, processing parameter prediction results, and defect risk prediction results are fused through an attention fusion network to obtain a comprehensive prediction result. For example, the table feature sub-prediction model uses a combination of the existing TabTransformer and a multilayer perceptron; the temporal process sub-prediction model uses a combination of the existing TCN temporal convolutional network; and the graph risk sub-prediction model uses a combination of the existing heterogeneous relational graph attention network R-GAT and a multilayer perceptron. The training samples for these sub-prediction models consist of historical trial molding records and stable production records, and they are trained through supervised learning. Therefore, with the optimization objectives of maximizing the basic quality prediction result, minimizing the processing parameter prediction result, and minimizing the defect risk prediction result, the constrained Bayesian optimization algorithm is used to calculate the acquisition function for each candidate molding process parameter package. , For the desired supervolume improvement value, The probability that the parameter package satisfies the constraints is calculated, and the candidate molding process parameter package with the highest acquisition function value is selected as the target molding process parameter package.

[0030] Furthermore, during conversion, mapping can be performed based on parameter name, unit, precision, upper and lower limits, number of segments, writing order, and communication protocol fields. For example, "segmented injection speed" in the target molding process parameter package can be mapped to "first-stage injection speed, second-stage injection speed, third-stage injection speed," etc., and "holding pressure curve" in the target molding process parameter package can be mapped to the number of holding stages, pressure value, and time value. In addition to the above methods, parameter template mapping, protocol field mapping, manual confirmation mapping, equipment model matching tables, and self-learning parameter mapping models can also be used to facilitate adaptation to different brands and models of injection molding machines.

[0031] In some embodiments of this application, after the trial molding is completed, actual production data such as actual issued parameters, injection pressure curve, screw position curve, mold cavity pressure curve, temperature curve, product weight, dimensional inspection results, appearance defect results, yield rate, and manual adjustment records by process engineers can be collected. Deviations are calculated, and samples with smaller deviations are prioritized for back-updating the edge weights, parameter feasible domain boundaries, defect risk thresholds, conditional diffusion generation model parameters, and quality prediction model parameters of the injection molding process knowledge graph. For example, when the predicted flash risk is low, but flash appears in the actual trial molding result, the edge weights of graphs such as "holding pressure - flash risk," "injection pressure - flash risk," and "insufficient clamping force - flash risk" are increased. , For the new graph edge weights, For the old graph edge weights, For learning rate, To predict the deviation between risks and actual defects, this setting has the advantage of continuously improving the accuracy of parameter generation and process adaptability as production data accumulates, resulting in greater stability. In addition to the above methods, closed-loop optimization can also be achieved through incremental learning, online learning, transfer learning, federated learning, reinforcement learning, manual rule updates, and parameter boundary adjustments based on defect labels.

[0032] The intelligent generation method for injection molding process parameters provided in this application acquires multi-source data corresponding to the current injection molding task, including product structure data, material performance data, mold structure data, equipment capacity data, and production target data. This comprehensive approach considers all factors and encodes the multi-source data into a current operating condition feature vector. The method then automatically maps this feature vector to a pre-constructed injection molding process knowledge graph. A pre-trained graph neural network outputs a priori vector of current operating condition defect risk. The current operating condition feature vector and the priori vector of current operating condition defect risk are input into a pre-trained conditional diffusion generation model to obtain multiple sets of candidate molding process parameter packages. These packages are then constrained by projection correction. A constrained Bayesian optimization algorithm and a pre-trained quality prediction model are used to select the target molding process parameter package from the multiple candidate packages. Finally, based on the controller parameter dictionary of the target injection molding machine, the target molding process parameter package is converted into an executable molding process parameter package for the target injection molding machine. This method is objective, convenient, and significantly improves production efficiency.

[0033] Based on the foregoing embodiments, this application provides an intelligent generation device for injection molding process parameters. This intelligent generation device 100 for injection molding process parameters can be applied to… Figure 1 In the corresponding embodiment of the intelligent generation method for injection molding process parameters, please refer to... Figure 2This is a structural block diagram of an intelligent injection molding process parameter generation device provided in an embodiment of this application. The intelligent injection molding process parameter generation device 100 includes: The acquisition module 101 is used to acquire multi-source data corresponding to the current injection molding task and encode the multi-source data into a feature vector of the current working condition. The multi-source data includes product structure data, material performance data, mold structure data, equipment capacity data and production target data. Candidate module 102 is used to map the current working condition feature vector to the pre-constructed injection molding process knowledge graph, output the current working condition defect risk prior vector through a pre-trained graph neural network, and input the current working condition feature vector and the current working condition defect risk prior vector into a pre-trained conditional diffusion generation model to obtain multiple sets of candidate molding process parameter packages. The conversion module 103 is used to perform constraint projection correction on multiple sets of candidate molding process parameter packages, and to select the target molding process parameter package from multiple sets of candidate molding process parameter packages through constraint Bayesian optimization algorithm and pre-trained quality prediction model, and to convert the target molding process parameter package into an executable molding process parameter package for the target injection molding machine based on the controller parameter dictionary of the target injection molding machine.

[0034] Optionally, in some embodiments of this application, the conditional diffusion generative model, during the pre-training stage, through... Noise is gradually added. For the first The parameter packet after adding noise. For noise scheduling parameters, This is a package of successful process parameters based on historical operating conditions. Use Gaussian noise and train a denoising network. The training loss function is , This is a feature vector representing historical operating conditions. This is the prior vector for defects and risks under historical operating conditions.

[0035] Optionally, in some embodiments of this application, the conversion module 103 is specifically used to sequentially verify each candidate molding process parameter package and calculate the hard constraint violation loss, soft constraint deviation loss and conflict constraint loss of the candidate molding process parameter package respectively; If the hard constraint violation loss and conflict constraint loss of the candidate molding process parameter package are not zero, or the soft constraint deviation loss of the candidate molding process parameter package exceeds the preset allowable threshold, then the primary goal is to keep the original values ​​of the candidate molding process parameter package unchanged, and the correction goal is to reduce the constraint violation loss, so as to obtain the corrected parameter package.

[0036] Optionally, in some embodiments of this application, the modified parameter package , For the parameter package during the correction process, This is a candidate molding process parameter package. For hard constraints, the penalty for violation The corresponding weights For soft constraint deviation loss The corresponding weights To constrain losses in conflict The corresponding weights.

[0037] Optionally, in some embodiments of this application, hard constraint violation loss , For the first hard constraint The corresponding weights It is a linear rectified function. For the first hard constraint The corresponding maximum constraint upper limit value; Soft constraint deviation loss , For the first Soft constraints The corresponding weights For the first Soft constraints The corresponding target value; and the conflict constraint loss. , For the first Individual risks The corresponding weights For the first Individual risks The corresponding maximum risk value.

[0038] Optionally, hard constraints are included in some embodiments of this application. These include material processing window constraints, equipment capacity boundary constraints, mold structure limitation constraints, and safety boundary constraints, as well as soft constraints. This includes targets for molding cycle time, energy consumption, quality, and stability, as well as risks. This includes the risks of burning, flash, and shrinkage.

[0039] Optionally, in some embodiments of this application, the conversion module 103 is specifically used to process each candidate molding process parameter package in different aspects using the table feature sub-prediction model, the time-series process sub-prediction model, and the graphic risk sub-prediction model of the quality prediction model, respectively, to obtain a comprehensive prediction result; wherein, the table feature sub-prediction model is used to process structured features and output basic quality prediction results, the time-series process sub-prediction model is used to process curve process data and output processing parameter prediction results, the graphic risk sub-prediction model is used to process defect risk priors and output defect risk prediction results, and the comprehensive prediction result includes basic quality prediction results, processing parameter prediction results, and defect risk prediction results; With the optimization objectives of maximizing the basic quality prediction results, minimizing the processing parameter prediction results, and minimizing the defect risk prediction results, the constrained Bayesian optimization algorithm is used to calculate the acquisition function of each candidate molding process parameter package, and the candidate molding process parameter package with the highest acquisition function value is selected as the target molding process parameter package.

[0040] It should be noted that the descriptions of the same steps and contents as in other embodiments in this embodiment can be found in the descriptions in other embodiments, and will not be repeated here.

[0041] The intelligent injection molding process parameter generation device provided in this application acquires multi-source data corresponding to the current injection molding task, including product structure data, material performance data, mold structure data, equipment capacity data, and production target data. It comprehensively considers various factors and encodes the multi-source data into a current working condition feature vector. Then, it automatically maps the current working condition feature vector to a pre-constructed injection molding process knowledge graph. A pre-trained graph neural network outputs a prior vector of current working condition defect risk. The current working condition feature vector and the prior vector of current working condition defect risk are input into a pre-trained conditional diffusion generation model to obtain multiple sets of candidate molding process parameter packages. Constrained projection correction is then applied to these multiple sets of candidate molding process parameter packages. A constrained Bayesian optimization algorithm and a pre-trained quality prediction model are used to select the target molding process parameter package from the multiple sets of candidate molding process parameter packages. Based on the controller parameter dictionary of the target injection molding machine, the target molding process parameter package is converted into an executable molding process parameter package for the target injection molding machine. This method is objective, convenient, and significantly improves production efficiency.

[0042] Based on the foregoing embodiments, this application provides an electronic device. Please refer to... Figure 3 The electronic device 200 may include a processor 201 and a memory 202. The memory 202 stores at least one instruction, at least one program, code set, or instruction set, which is loaded and executed by the processor 201 to implement [the desired functionality]. Figure 1 The steps of the intelligent generation method for injection molding process parameters in the corresponding embodiment.

[0043] In another aspect, embodiments of this application provide a computer-readable storage medium for storing program code for executing the aforementioned... Figure 1 Any implementation of the intelligent generation method for injection molding process parameters in the corresponding embodiment.

[0044] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and modules described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0045] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative. For instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed between each other can be through some interfaces, indirect coupling or communication connection between devices or modules, and can be electrical, mechanical, or other forms. Modules described as separate components may or may not be physically separate, and components shown as modules may or may not be physical units, i.e., they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0046] Furthermore, the functional modules in the various embodiments of this application can be integrated into one processing unit, or each module can exist physically separately, or two or more units can be integrated into one module. The integrated unit can be implemented in hardware or as a software functional unit. If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium.

[0047] Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or 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 personal computer, server, or network device, etc.) to execute all or part of the steps of the intelligent generation method for injection molding process parameters of various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0048] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0049] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for intelligently generating injection molding process parameters, characterized in that, The intelligent generation method for injection molding process parameters includes: Acquire multi-source data corresponding to the current injection molding task, and encode the multi-source data into a feature vector of the current working condition. The multi-source data includes product structure data, material performance data, mold structure data, equipment capacity data, and production target data. The current working condition feature vector is mapped to a pre-constructed injection molding process knowledge graph, and the current working condition defect risk prior vector is output through a pre-trained graph neural network. The current working condition feature vector and the current working condition defect risk prior vector are input into a pre-trained conditional diffusion generation model to obtain multiple sets of candidate molding process parameter packages. Constrained projection correction is performed on the multiple sets of candidate molding process parameter packages, and the target molding process parameter package is selected from the multiple sets of candidate molding process parameter packages through constrained Bayesian optimization algorithm and pre-trained quality prediction model. Based on the controller parameter dictionary of the target injection molding machine, the target molding process parameter package is converted into the executable molding process parameter package of the target injection molding machine.

2. The intelligent generation method for injection molding process parameters according to claim 1, characterized in that, The conditional diffusion generative model, during the pre-training phase, through... Noise is gradually added. For the first The parameter packet after adding noise. For noise scheduling parameters, This is a package of successful process parameters based on historical operating conditions. Use Gaussian noise and train a denoising network. The training loss function is , This is a feature vector representing historical operating conditions. This is the prior vector for defects and risks under historical operating conditions.

3. The intelligent generation method for injection molding process parameters according to claim 1, characterized in that, The constraint projection correction of the multiple sets of candidate molding process parameter packages includes: Each candidate molding process parameter package is verified sequentially, and the hard constraint violation loss, soft constraint deviation loss, and conflict constraint loss of each candidate molding process parameter package are calculated respectively. If the hard constraint violation loss and conflict constraint loss of the candidate molding process parameter package are not zero, or the soft constraint deviation loss of the candidate molding process parameter package exceeds a preset allowable threshold, then the primary goal is to keep the original values ​​of the candidate molding process parameter package unchanged, and the correction goal is to reduce the constraint violation loss, so as to obtain a corrected parameter package.

4. The intelligent generation method for injection molding process parameters according to claim 3, characterized in that, The corrected parameter package , For the parameter package during the correction process, This is a candidate molding process parameter package. For hard constraints, the penalty for violation The corresponding weights For soft constraint deviation loss The corresponding weights To constrain losses in conflict The corresponding weights.

5. The intelligent generation method for injection molding process parameters according to claim 4, characterized in that, The hard constraint violation loss , For the first hard constraint The corresponding weights It is a linear rectified function. For the first hard constraint The corresponding maximum constraint upper limit value; The soft constraint deviation loss , For the first Soft constraints The corresponding weights For the first Soft constraints The corresponding target value; and the conflict constraint loss. , For the first Individual risks The corresponding weights For the first Individual risks The corresponding maximum risk value.

6. The intelligent generation method for injection molding process parameters according to claim 5, characterized in that, The hard constraint This includes material processing window constraints, equipment capacity boundary constraints, mold structure limitation constraints, and safety boundary constraints; the soft constraints... This includes molding cycle targets, energy consumption targets, quality targets, and stability targets, as well as the aforementioned risks. This includes the risks of burning, flash, and shrinkage.

7. The intelligent generation method for injection molding process parameters according to any one of claims 1 to 6, characterized in that, The process of selecting the target molding process parameter package from the multiple candidate molding process parameter packages using a constrained Bayesian optimization algorithm and a pre-trained quality prediction model includes: The candidate molding process parameter packages are processed from different aspects using the tabular feature sub-prediction model, the time-series process sub-prediction model, and the graphic risk sub-prediction model of the quality prediction model to obtain a comprehensive prediction result. Specifically, the tabular feature sub-prediction model processes structured features and outputs a basic quality prediction result; the time-series process sub-prediction model processes curve process data and outputs processing parameter prediction results; the graphic risk sub-prediction model processes prior defect risks and outputs defect risk prediction results; and the comprehensive prediction result includes the basic quality prediction result, the processing parameter prediction result, and the defect risk prediction result. With the optimization objectives of maximizing the basic quality prediction result, minimizing the processing parameter prediction result, and minimizing the defect risk prediction result, the constrained Bayesian optimization algorithm is used to calculate the acquisition function of each candidate molding process parameter package, and the candidate molding process parameter package with the highest acquisition function value is selected as the target molding process parameter package.

8. An intelligent device for generating molding process parameters for an injection molding machine, characterized in that, The intelligent generation device for injection molding process parameters includes: The acquisition module is used to acquire multi-source data corresponding to the current injection molding task and encode the multi-source data into a current working condition feature vector. The multi-source data includes product structure data, material performance data, mold structure data, equipment capacity data, and production target data. The candidate module is used to map the current working condition feature vector to a pre-constructed injection molding process knowledge graph, output the current working condition defect risk prior vector through a pre-trained graph neural network, and input the current working condition feature vector and the current working condition defect risk prior vector into a pre-trained conditional diffusion generation model to obtain multiple sets of candidate molding process parameter packages. The conversion module is used to perform constraint projection correction on the multiple sets of candidate molding process parameter packages, and to select the target molding process parameter package from the multiple sets of candidate molding process parameter packages through a constraint Bayesian optimization algorithm and a pre-trained quality prediction model. Based on the controller parameter dictionary of the target injection molding machine, the target molding process parameter package is converted into an executable molding process parameter package of the target injection molding machine.

9. An electronic device, characterized in that, The electronic device includes a processor and a memory, wherein the memory stores at least one instruction, at least one program, code set, or instruction set, and the instruction, program, code set, or instruction set is loaded and executed by the processor to implement the steps of the intelligent generation method for injection molding process parameters of any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores one or more programs that can be executed by one or more processors to implement the steps of the intelligent generation method for injection molding process parameters of any one of claims 1 to 7.

Images