Injection molding condition generation system and method

Abstract

The present application provides an injection molding condition generation system capable of improving the quality of injection molding. The injection molding condition generation system (1) generates an injection molding condition (417) for injection molding using a first resin material, based on a target value of a quality parameter related to the quality of a molded product, a first material characteristic value of the first resin material, and a predetermined relational expression, when the first material characteristic value is acquired, the predetermined relational expression being an expression (416) representing the relationship among the material characteristic value of the resin material, a plurality of injection molding conditions input to an injection molding machine, and the quality parameter related to the quality of a molded product molded by the injection molding machine based on the material characteristic value and each injection molding condition, the predetermined relational expression being generated based on data obtained by accumulating the material characteristic value of the resin material, the injection molding condition, and the quality parameter in association with each other.

Description

TECHNICAL FIELD

[0001] The present application relates to an injection molding condition generation system and method. BACKGROUND

[0002] In an injection molding process, various factors such as a plurality of injection molding conditions input to a molding machine, a material characteristic variation of a resin, and the like have an influence on a molded product quality. Therefore, it is not an easy job to adjust the molding conditions in order to stabilize the molded product quality, and requires a skilled operator's technique and time. In view of this problem, a method for optimizing the injection molding conditions independently of the operator has been studied (Patent Literature 1).

[0003] In the method described in Patent Literature 1, an injection molding controller for providing a control signal for partially determining an injection molding pressure in an injection molding process is connected to an injection molding machine, a first control signal from a pressure control output is measured in a first time in an injection molding cycle, a second control signal from the pressure control output is measured in a second time in a subsequent injection molding cycle, and the control signal of the pressure in a third time in the injection molding cycle is adjusted in accordance with the result of comparing the first control signal with the second control signal. That is, in Patent Literature 1, a variation in the material characteristic in the injection molding process is measured as a variation in the controller control signal, and pressure adjustment based on the variation is performed, whereby the injection molding condition (pressure) considering the variation in the material characteristic can be controlled independently of the operator.

[0004] PRIOR ART DOCUMENTS

[0005] PATENT LITERATURE

[0006] Patent Literature 1: Japanese Patent Application Publication No. 2016-527109 SUMMARY

[0007] PROBLEMS TO BE SOLVED BY THE INVENTION

[0008] In the method described in Patent Literature 1, it is premised that the pressure is adjusted in each cycle of the injection molding operation. Therefore, in the method described in Patent Literature 1, a special controller for measuring the control signal from the pressure control output and performing pressure control corresponding to the measured value in the injection molding operation, and development for connecting the controller to the injection molding machine at all times are required, and thus additional development and man-hours are required.

[0009] The relationship between molding conditions and quality in injection molding is influenced by the material properties of the resin. For example, even with the same kind of material (e.g., polypropylene), the flowability differs in grade units, supplier units, so even if the same molding conditions are input, the behavior of the resin within the metal mold differs greatly, and the quality of the molded product also has deviations.

[0010] Here, in recent years, due to the problem of ocean pollution caused by plastic waste, measures to ban the import of plastic waste in China and Southeast Asia, there has been greater attention than ever before to the effective use of plastic recycling materials. In Europe, a part of the region also studies a tax or regulation against the use of raw materials, and for manufacturers who manufacture products using plastic, the effective use of recycled materials becomes an urgent issue. However, compared to raw materials, recycled materials have large deviations in material properties due to the heat history at the time of molding, degradation caused by the use environment, and the mixing of foreign matter or heat history at the time of recycling, so the deviation in the quality of the molded product also increases above the raw material.

[0011] Figure 17 is a graph that shows a comparative example for explicitly showing the advantages of the embodiments described later, and is not background art. Figure 17 In, the distribution of the molded product weight when molding is performed using three recycled materials, Batch A, Batch B, and Batch C, is shown. Figure 17 The horizontal axis of indicates the molded product weight, Figure 17 The vertical axis of indicates the probability density. Figure 17 In, the distribution of the molded product weight when molding is performed using three recycled materials, Batch A, Batch B, and Batch C, is shown.

[0012] Hereinafter, the difference based on the delivery period will be labeled as a batch. In Figure 17 In, the number of molded products per 1 batch is 40. Figure 17 The distribution of the molded product weight of is all normalized by the number of molded products.

[0013] According to Figure 17 It can be seen that even when molding is performed under the same molding conditions, the weight deviation between batches is sufficiently larger than the deviation within a batch. Therefore, it can be confirmed that in recycled materials, the difference in material properties of each delivery batch affects the relationship between molding conditions and quality.

[0014] Regarding the influence on the material properties of the resin, in Patent Literature 1, the change in the material properties during the injection molding operation is measured as a change in the controller control signal. In Patent Literature 1, the pressure of the injection molding condition is controlled based on the change in the controller control signal, and thus adjustment corresponding to the change in the material properties is performed. However, in the control of only the pressure, in the case where the temperature properties and the flowability of the resin greatly change, the amount of adjustment of the pressure becomes large. Therefore, in the method described in Patent Literature 1, in the case where the pressure becomes too low, the possibility of underfilling increases, and in the case where the pressure becomes too large, the possibility of appearance defects such as burrs increases.

[0015] The present application has been made in view of the above problems, and an object thereof is to provide an injection molding condition generation system and method capable of improving the quality of an injection molded product.

[0016] Means for solving the problems

[0017] To solve the above problems, the injection molding condition generation system of the present application is an injection molding condition generation system that generates an injection molding condition for injection molding using a computer, the computer including a processor and a memory used by the processor, the processor generating an injection molding condition for injection molding using a first resin material based on a target value of a quality parameter related to the quality of a molded product, a first material property value of the first resin material, and a predetermined relational expression representing the relationship of a material property value of a resin material, a plurality of injection molding conditions input to an injection molding machine, and a quality parameter related to the quality of a molded product molded by the injection molding machine based on the material property value and each injection molding condition, the predetermined relational expression being generated based on data obtained by accumulating the material property value of the resin material, the injection molding condition, and the quality parameter in the memory in association with each other.

[0018] Effects of the Invention

[0019] According to the present application, it is possible to obtain an injection molding condition suitable for the material property value of a resin material used in an injection molding machine and the quality of a molded product. BRIEF DESCRIPTION OF DRAWINGS

[0020] Figure 1 is a functional block diagram of a molding condition optimization system of an injection molding machine.

[0021] Figure 2 is a structural diagram of a computer that can be used to implement an injection molding condition generation system.

[0022] Figure 3 is a conceptual diagram of an injection molding machine.

[0023] Figure 4is a conceptual diagram of a metal mold used in experiments for verifying optimization of molding conditions.

[0024] Figure 5 is a block diagram showing extraction of a feature quantity from a physical quantity obtained by a sensor and recording to a feature quantity database.

[0025] Figure 6 is a graph showing time-series changes in acquired data of a pressure sensor and a resin sensor.

[0026] Figure 7 is an explanatory diagram of a feature quantity extraction result example.

[0027] Figure 8 is an explanatory diagram of a process data set.

[0028] Figure 9 is a flowchart of a learning mode.

[0029] Figure 10 is a graph showing an example of a result of conversion of a feature quantity data set by UMAP.

[0030] Figure 11 is an explanatory diagram of an example of a feature quantity data set after dimension reduction.

[0031] Figure 12 is an explanatory diagram of a learning data set example.

[0032] Figure 13 is a graph for explaining performance evaluation of a learned regression model.

[0033] Figure 14 is a flowchart of an optimization mode.

[0034] Figure 15 is an explanatory diagram showing an example of a combination of explanatory variables in a grid search method.

[0035] Figure 16 is a weight distribution diagram of a molded product for confirming appropriateness of a molding condition after optimization.

[0036] Figure 17 is a weight distribution diagram of a molded product when a recycled material for different delivery periods is molded under the same molding condition as a comparative example. DETAILED DESCRIPTION

[0037] Hereinafter, an embodiment of the present application will be described based on the drawings. In the present embodiment, a technology for suppressing a variation in quality of a molded product caused by a variation in material characteristics is provided. The injection molding condition generation system of the present embodiment does not need to be connected to an injection molding machine at all times, and can provide an appropriate injection molding condition to the injection molding machine.

[0038] The embodiments described below do not limit the scope of the invention pertaining to the scope of protection, and in addition, not all of the elements described in the embodiments are essential to the solution means of the invention.

[0039] The system of the present embodiment optimizes the injection molding conditions in accordance with the material property values of the resin material and the required quality. The system, for example, generates in advance a relational expression indicating the correlation among the material property values of the resin material used in injection molding, the plurality of injection molding conditions input to the injection molding machine, and the quality parameter relating to the quality of the molded product molded by the injection molding machine based on the resin material and the injection molding conditions, acquires the first material property values of a first material, and generates the injection molding conditions suitable for the first material in accordance with the first material property values, a target value of the quality parameter, and the predetermined relational expression.

[0040] According to the present embodiment, even in the case where the material properties of the resin fluctuate, optimal molding conditions for stabilizing the quality of the molded product can be obtained without depending on the skills of the operators and without requiring real-time control in the injection molding process.

[0041] Example 1

[0042] Use Figures 1 to 16 Example 1 is described. Figure 1 A functional block diagram showing the system 1 that generates the molding conditions of the injection molding machine.

[0043] The injection molding condition generation system 1 includes, for example, a production management system 2, a manufacturing execution system 3, a learning and optimization system 4, and a manufacturing plant 5. Part or all of the functions of each of the functions of the injection molding condition generation system 1 described below can be configured as software, can be realized as a cooperation of software and hardware, and can also be realized using hardware having a fixed circuit. At least a part of the functions possessed by the production management system 2, the manufacturing execution system 3, and the manufacturing plant 5 can also be executed by an operator.

[0044] The production management system 2 is a system that manages production plans, and includes at least a production plan management section 21. The production plan management section 21 is a function that generates production plans including production specifications, quantities, and time periods, and the like in accordance with order situations and inventory situations.

[0045] The manufacturing execution system 3 is a system that instructs the execution of production to the manufacturing plant 5. The manufacturing execution system 3 decides manufacturing conditions based on the production plan generated by the production management system 2, and transmits a production instruction including the manufacturing conditions to the manufacturing plant 5. In the manufacturing conditions, for example, information for determining an injection molding machine used for production (injection molding), information for determining a metal mold used for production, information for determining a material used for production, the number of molded products of production, a production period, a required quality, and the like can be included.

[0046] The manufacturing execution system 3 will be described. The manufacturing execution system 3 has, for example, a manufacturing condition deciding section 31, a production performance storage section 32, a production performance acquiring section 33, a manufacturing execution instruction section 34, and a production performance recording section 35.

[0047] The manufacturing condition deciding section 31 is a function that decides the above-described manufacturing conditions based on the production plan generated by the production plan management section 21 of the production management system 2. The manufacturing condition deciding section 31 can transmit information related to the manufacturing conditions to the learning and optimization system 4. The information related to the manufacturing conditions can include predetermined information related to the metal mold, the injection molding machine, and the material.

[0048] The predetermined information includes, for example, the capacity of the metal mold, the runner structure of the metal mold. As the predetermined information, for example, the required quality of the molded product of production, the manufacturing number of the supplier unit of the resin material used, and the delivery number of each supplier, and the like can be included. The learning and optimization system 4 inputs the required quality information of the molded product received from the manufacturing condition deciding section 31 to the optimum condition generating section 417 in order to optimize the molding conditions, and generates the optimized molding conditions.

[0049] The production performance storage section 32 is a function that stores production performance. In the present embodiment, the production performance indicates a molding condition in which it is confirmed that a good molded product quality can be obtained for a combination of the injection molding machine, the metal mold, and the material. The good molded product quality means that the molding quality satisfies the required quality decided in the manufacturing execution system 3.

[0050] The production performance acquiring section 33 is a function that acquires production performance from the production performance storage section 32. The production performance acquiring section 33 reads out and acquires production performance based on the metal mold (hereinafter, referred to as a first metal mold) decided by the manufacturing condition deciding section 31 and the material (hereinafter, referred to as a first material) decided by the manufacturing condition deciding section 31 from the production performance storage section 32.

[0051] When there is no production performance based on the combination of the first metal mold and the first material, the production performance acquiring section 33 requests the manufacturing execution instruction section 34 to optimize the molding conditions based on the combination of the first metal mold and the first material, or to perform molding based on the optimization result.

[0052] The request for molding condition optimization means that, in the manufacturing plant 5, trial molding is performed with a preset injection molding condition that becomes a reference (hereinafter, referred to as a reference condition), information related to the trial molding is input to the learning and optimization system 4, and molding conditions that satisfy the requested quality are generated. Examples of the information input to the learning and optimization system 4 are described later. In addition, molding based on the optimization result means that molding is performed with the good molding condition output from the learning and optimization system 4 and the expected molding condition.

[0053] On the other hand, in a case where there is a production history based on the combination of the first metal mold and the first material, the production history acquisition unit 33 requests the manufacturing execution instruction unit 34 to perform molding under the manufacturing history or to perform trial molding for learning (hereinafter, referred to as trial molding for learning).

[0054] Here, the trial molding for learning means that injection molding is performed while changing the molding condition, the obtained information is input to the learning and optimization system 4, and is saved as data for extracting the best molding condition. Details of the trial molding for learning are described later. Figure 8

[0055] The manufacturing execution instruction unit 34 is a function of instructing execution of manufacturing in the manufacturing plant 5. In addition, the execution of manufacturing can be referred to as production. In the manufacturing execution instruction, for example, there are included a request for molding condition optimization input from the production history acquisition unit 33, a request for molding based on the optimization result, a request for molding under the manufacturing history, and a request for trial molding for learning.

[0056] The production history recording unit 35 is a function of recording the molding condition in which the good molded product quality is confirmed in the manufacturing plant 5 to the production history storage unit 32. The production history recording unit 35 registers the molding condition in which the good molded product quality is obtained in the production history storage unit 32, based on the information indicating the quality result of the molded product obtained from the quality inspection unit 57 of the manufacturing plant 5.

[0057] The manufacturing plant 5 is described. The manufacturing plant 5 receives the manufacturing execution instruction from the manufacturing execution system 3, and executes any one or a plurality of the injection molding processes 53 to 56. Hereinafter, the injection molding is sometimes abbreviated as "IM".

[0058] The manufacturing plant 5 has, for example, a manufacturing execution unit 51, a plurality of injection molding machines 50 (described later in Figure 3 ), a plurality of metal molds (described later in Figure 3 ), a molding condition generation unit 52, and a molded product quality inspection unit 57. Hereinafter, the molded product quality inspection unit 57 is sometimes abbreviated as the quality inspection unit 57.

[0059] ​The manufacturing execution section 51 executes the injection molding process based on the manufacturing conditions input from the manufacturing execution instruction section 34 of the manufacturing execution system 3. When the molding with production performance is requested, the manufacturing execution section 51 inputs the production performance to the injection molding machine for the indicated combination of the metal mold and the material, thereby executing the injection molding process 53. That is, the injection molding process 53 is the injection molding process performed under the molding conditions with the production performance of producing a qualified product using the specified combination of the metal mold and the material.

[0060] When the trial molding for learning is requested, the manufacturing execution section 51 instructs the molding condition generation section 52 of the trial molding for learning. When the molding condition generation section 52 receives the instruction of the trial molding for learning from the manufacturing execution section 51, it generates a plurality of molding conditions for saving in the learning database 412 of the learning and optimization system 4 for the indicated combination of the metal mold and the material. The manufacturing execution section 51 changes the generated plurality of molding conditions for each predetermined injection unit and inputs them to the injection molding machine, thereby executing the injection molding process 54. That is, the injection molding process 54 is the process of changing the molding conditions for each predetermined injection unit and performing the trial molding according to the molding conditions.

[0061] When the molding condition optimization is requested, the manufacturing execution section 51 instructs the molding condition generation section 52 of the molding under the reference conditions. When the molding condition generation section 52 receives the instruction of the molding under the reference conditions from the manufacturing execution section 51, it inputs the reference conditions specified in advance to the injection molding machine for the indicated combination of the metal mold and the material, thereby executing the injection molding process 55. That is, the injection molding process 55 is the process of performing the injection molding according to the reference conditions.

[0062] When the molding based on the optimization result is requested, the manufacturing execution section 51 instructs the molding condition generation section 52 of the molding based on the optimization result. When the molding condition generation section 52 receives the instruction of the molding based on the optimization result from the manufacturing execution section 51, it receives the optimized molding conditions generated by the optimum condition generation section 417 for the indicated combination of the metal mold and the material from the learning and optimization system 4, and inputs the received molding conditions to the injection molding machine, thereby executing the injection molding process 56. That is, the injection molding process 56 is the process of performing the injection molding according to the optimized molding conditions.

[0063] The quality inspection section 57 is a function of determining the quality of the molded product obtained by the injection molding process. For example, the molded product quality is evaluated based on the size, the warping amount, the burr, the scratch, the gloss, the color, and the like. The quality inspection of the molded product can be performed automatically, manually by an inspector, or semi-automatically.

[0064] The quality inspection section 57 outputs the manufacturing conditions, the combination of the injection molding machine and the metal mold, the molding conditions, and the inspection results of the molded product quality to the production performance recording section 35 of the manufacturing execution system 3 in the case where the molded product is of good quality. In addition, the manufacturing conditions, the combination of the injection molding machine and the metal mold, the molding conditions, and the inspection results of the molded product quality are output to the process data recording section 407 of the learning and optimization system 4 in the case where the injection molding process 54 to 56 is performed.

[0065] In addition, in the present embodiment, as for the information related to the material properties of each material used when using the first metal mold, the physical quantity at a predetermined position in the metal mold is measured in advance using the sensor 58 mounted on each injection molding machine and the metal mold possessed by the manufacturing plant 5, and the physical quantity is output to the sensor information recording section 401 of the learning and optimization system 4. As for the details of the sensor mounted on the metal mold, the use Figure 4 This will be described later. Here, the information related to the material properties refers to, for example, the flowability of the material, the material properties, and the physical quantities related thereto.

[0066] The predetermined position of the injection molding machine is, for example, the tip of the nozzle or the like. The predetermined position in the metal mold is, for example, the resin inflow port of the metal mold or the like. The physical quantity includes, for example, the pressure of the resin, the temperature of the resin, the speed of the resin, the material properties of the resin, and the opening amount of the metal mold (mold opening amount). The material properties are, for example, the density of the resin, the viscosity of the resin, the distribution of the fiber length of the resin (when it is a material containing reinforcing fibers), and the like. Among them, the physical quantity most related to the flowability of the material is the viscosity of the resin, but other characteristic quantities related to the flowability calculated from the pressure, the temperature, and the speed can also be used.

[0067] The learning and optimization system 4 will be described. The learning and optimization system 4 is a function of generating appropriate injection molding conditions without the production performance based on the specified combination of the metal mold and the resin material. In addition, the optimal injection molding conditions referred to in the present embodiment mean appropriate injection molding conditions.

[0068] The learning and optimization system 4 is a function of, for the material (hereinafter, referred to as the second material) for which the condition optimization is performed without the production performance based on the combination of the first metal mold and the first material in the production performance storage section 32 of the manufacturing execution system 3, generating the optimal molding conditions for the second material to satisfy the required quality, using the information based on the past executed IM process 54 and the IM process 55 using the second material and the required quality as inputs.

[0069] The learning and optimization system 4 includes, for example, a sensor information recording unit 401, a feature quantity extraction unit 402, a feature quantity database 403, a dimension reduction model learning unit 404, a learned dimension reduction model storage unit 405, a learned dimension reduction model storage unit 406, a process data recording unit 407, a process database 408, a dimension reduction model reading unit 409, a dimension reduction execution unit 410, a linking processing unit 411, a learning database 412, a regression model learning unit 413, a regression model storage unit 414, a learned model storage unit 415, a regression model reading unit 416, and an optimum condition generation unit 417.

[0070] The sensor information recording unit 401 is a function that records physical quantities at predetermined positions in a metal mold obtained by the sensor 58 in the injection molding process 55 of the manufacturing plant 5. The feature quantity extraction unit 402 extracts feature quantities from the physical quantities temporarily recorded by the sensor information recording unit and records the predetermined material unit, the combination of the injection molding machine and the metal mold, and the extracted feature quantities in the feature quantity database 403. Here, the predetermined material unit refers to, for example, a model or a lot number of each material supplier, or the like, which is a unit for distinguishing materials.

[0071] In the present embodiment, the data group recorded in the feature quantity database 403 is referred to as a feature quantity data set. Regarding the processing in the feature quantity extraction unit 402, the following is used. Figure 6 Figure 7 Hereinafter. In a state where the combination of the injection molding machine and the metal mold is fixed and the molding conditions in the injection molding process 55 are fixed, in a case where only the predetermined material unit is changed, the feature quantities extracted from the physical quantities obtained by the sensor 58 are strongly affected by the change in the material information (for example, flowability, physical property values) between the material units.

[0072] Therefore, it is possible to record the deviation in the material information between the material units as the deviation between the feature quantities. Here, the feature quantities extracted from the physical quantities obtained by the sensor 58 are affected by the mechanical errors of the injection molding machine and the metal mold, and therefore the deviation in the material information between the material units is stored as the deviation between the feature quantities. That is, the feature quantity data set is recorded for each combination of the injection molding machine and the metal mold. The feature quantity data sets described below all fix the combination of the injection molding machine and the metal mold.

[0073] ​The dimension reduction model learning unit 404 is a function of converting the dimension of the feature quantity data set recorded in the feature quantity database into a lower dimension vector using a dimension reduction model. As for the dimension after the conversion, the user of the learning and optimization system can set an arbitrary dimension. Dimension reduction refers to an unsupervised learning method of extracting a vector of a lower dimension than the input vector in a manner that loses as little information of the input vector as possible, and in the dimension reduction model, for example, there are principal component analysis, auto-encoding, UMAP, and the like.

[0074] In general, by converting a multi-dimensional vector into a lower-dimensional vector, it is possible to improve the interpretability of the data or to improve the general performance when generating a regression model. The dimension reduction model is learned using the feature quantity data set of the feature quantity database 403, and thus a learned dimension reduction model is generated. When the feature quantity data recorded in the feature quantity database is input to the learned dimension reduction model, a vector converted into the dimension number specified by the user is output.

[0075] The learned dimension reduction model holding unit 405 is a function of recording the dimension reduction model generated by the dimension reduction model learning unit 404 in the learned dimension reduction model storage unit 406.

[0076] The process data recording unit 407 will be described. The process data recording unit 407 records the molded product quality molded by either of the injection molding processes 54 or 55, the molding conditions input to the injection molding machine, the predetermined material unit, and the combination of the injection molding machine and the metal mold in association with each other in the process database 408.

[0077] In the present embodiment, a group of data recorded in the process database is referred to as a process data set. As with the feature quantity data set, the process data set is recorded for each combination of the injection molding machine and the metal mold. In the process data set described below, the combination of the injection molding machine and the metal mold is set to be fixed.

[0078] The dimension reduction model reading unit 409 is a function of reading out the learned dimension reduction model recorded in the learned dimension reduction model storage unit 406. The read-out dimension reduction model is output to the dimension reduction execution unit 410.

[0079] The dimension reduction execution unit 410 is a function of reducing the dimension of the feature quantity database using the learned dimension reduction model read out from the dimension reduction model reading unit 409 with respect to the feature quantity data set read from the feature quantity database 403. As for the details of the dimension reduction result, the dimension reduction model used is used Figures 9 to 11 This will be described later. The data set after the dimension reduction is output to the linking processing unit 411.

[0080] The linking processing section 411 generates a data set in which the molding conditions, the molding product quality, and the dimensionally-reduced vectors are associated with each other using the material information as a combination key, from the dimensionally-reduced feature amount data set associated with the material information output from the dimension reduction execution section 410 and the process data set associating the material information, the molding conditions, and the molding product quality, which is acquired from the process database 408, and records the data set in a learning database 412 (hereinafter, referred to as a learning data set).

[0081] Here, the linking processing section 411 does not necessarily have to use the dimensionally-reduced feature amount data set, and can generate the learning data set by combining the feature amount data set associated with the material information from the feature amount database 403 with the process data set, instead of the dimensionally-reduced feature amount data set, and record the learning data set in the learning database 412. In the following description of the present embodiment, a case in which the learning data set is generated using the dimensionally-reduced vectors is described.

[0082] The regression model learning section 413 is a function of acquiring the learning data set from the learning database 412, setting the explanatory variables to the dimensionally-reduced vectors and the molding conditions, setting the target variable to the molding product quality, learning a regression model that predicts the target variable from the explanatory variables using the regression model, and generating a learned regression model.

[0083] In general, the regression model refers to a model that predicts a target variable (y) from an explanatory variable (X) (y = f(X)), and the parameters within the model are determined based on the learning data. In the description of the present embodiment, when described as "regression model", it refers to the entire regression model, and when described as "learned regression model", it refers to the regression model in which the model parameters are determined based on the learning data.

[0084] In the present embodiment, the regression model can use, for example, a linear regression, a ridge regression, a support vector machine, a neural network, a random forest regression, or a regression model in which these are combined. In addition, in a case in which the regression model used can have one or more target variables, such as a neural network, one or more molding product qualities can be selected as the target variable. Regarding the learning of the specific regression model, a neural network is used. Figures 9 to 13 This is described later.

[0085] The regression model storage section 414 is a function of recording the learned regression model generated in the regression model learning section 413 in a learned regression model storage section 415. The regression model readout section 416 acquires the learned regression model from the learned regression model storage section 415 and inputs it to the optimum condition generation section 417.

[0086] The optimum condition generation section 417 is a function that generates a molding condition (injection molding condition) for achieving a required quality. The optimum condition generation section 417 acquires the required quality of the molded product from the manufacturing condition decision section 31, acquires the learned regression model from the regression model reading section 416, and acquires the dimension-reduced feature amount data for the second material from the concatenation processing section 411, thereby generating an optimum molding condition for achieving the required quality. Regarding the generation of the optimum molding condition, the following equation is used Figures 14 to 16 This will be described later.

[0087] Figure 2 An example of a structure of the computer 10 that can be used to implement the injection molding condition generation system 1 is shown. Here, a case where the injection molding condition generation system 1 is implemented by one computer 10 is described, but the present application is not limited thereto, and one or more injection molding condition generation systems 1 can be constructed by causing a plurality of computers to cooperate. In addition, as described above, the production management system 2, the manufacturing execution system 3, and the manufacturing plant 5 can also not use dedicated software or hardware, and a part or all of each function can be performed by an operator, thereby implementing the injection molding condition generation system 1.

[0088] The computer 10 has, for example, an arithmetic device 11, a memory 12, a storage device 13, an input device 14, an output device 15, a communication device 16, and a medium interface section 17, which are connected by a communication path CN1. The communication path CN1 is, for example, an internal bus, a LAN (Local Area Network), or the like.

[0089] The arithmetic device 11 is constituted by, for example, a microprocessor or the like. The arithmetic device 11 is not limited to a microprocessor, and can include, for example, a DSP (Digital Signal Processor), a GPU (Graphics Processing Unit), or the like. The arithmetic device 11 reads out a computer program recorded in the storage device 13 to the memory 12 and executes it, thereby implementing each functional module of the injection molding condition generation system 1, such as 21, 31 to 35, 401 to 417, 51, 52, and 60.

[0090] The storage device 13 is a device that stores computer programs and data, and has, for example, a rewritable storage medium such as a flash memory or a hard disk. In the storage device 13, a computer program for implementing the GUI section 60 that provides a GUI (Graphical User Interface) to an operator, and computer programs for implementing each functional module described above, such as 21, 31 to 35, 401 to 417, and 51 to 52, are stored.

[0091] The input device 14 is a device by which an operator inputs information to the computer 10. As the input device 14, for example, there are a keyboard, a touch panel, a pointing device such as a mouse, a voice instruction device (none of which is shown), and the like. The output device 15 is a device by which the computer 10 outputs information. As the output device 15, for example, there are a display, a printer, a voice synthesizing device (none of which is shown), and the like.

[0092] The communication device 16 is a device by which an external information processing device communicates with the computer 10 via a communication path CN2. As the external information processing device, in addition to a computer (none of which is shown), there is an external storage device 19. The computer 10 is capable of reading in data (computer-specific information, production performance, and the like) and a computer program recorded in the external storage device 19. The computer 10 is also capable of transmitting all or a part of the computer program and data stored in the storage device 13 to the external storage device 19 to store.

[0093] The medium interface section 17 is a device that reads and writes the external recording medium 18. As the external recording medium 18, for example, there are a USB (Universal Serial Bus) memory, a memory card, a hard disk, and the like. It is also possible to transfer all or a part of the computer program and data stored in the storage device 13 to the external recording medium 18 to store from the external recording medium 18.

[0094] Figure 3 is a conceptual view of an injection molding machine 50. The injection molding machine 50 is used for the production of a molded article 100. Figure 3 The processes of the injection molding process will be described. In the present embodiment, the molding phenomenon indicates a series of phenomena that occur in the injection molding process. In the present embodiment, the injection molding process is roughly divided into a metering and plasticizing process, an injection and packing process, a cooling process, and a take-out process.

[0095] In the metering and plasticizing process, the screw 502 is retracted with the plasticizing motor 501 as a driving force, and the resin pellets 504 are supplied from the hopper 503 into the cylinder 505. Then, the resin is plasticized to a uniform molten state by heating by the heater 506 and rotation of the screw 502. Depending on the settings of the back pressure and the rotation speed of the screw 502, the density of the molten resin and the degree of breakage of the reinforcing fibers change, and these changes affect the quality of the molded article.

[0096] During the injection and the pressure maintaining, the screw 502 is advanced by the driving force of the injection motor 507, and the molten resin is injected into the metal mold 509 via the nozzle 508. In the molten resin injected into the metal mold 509, the cooling from the wall surface of the metal mold 509 and the shearing heat caused by the flow act in parallel. That is, the molten resin flows in the metal mold 509 while being subjected to the cooling and the heating. In the case where the closing force of the metal mold 509, that is, the clamping force is small, a slight metal mold opening occurs after the molten resin is solidified, and the quality of the molded product is affected due to the slight gap.

[0097] During the cooling, the molten resin is cooled to below the solidification temperature by the metal mold 509 maintained at a constant temperature. The residual stress generated during the cooling process affects the quality of the molded product. The residual stress is generated in association with the anisotropy of the material properties due to the flow in the metal mold, the density distribution based on the pressure maintaining, and the unevenness of the molding shrinkage.

[0098] During the ejection, the metal mold 509 is opened by driving the clamping mechanism 512 using the motor 511 for opening and closing the metal mold 509 as the driving force. Then, in the case where the solidified molded product is ejected from the metal mold 509 by driving the ejection mechanism 514 using the ejection motor 513 as the driving force, when sufficient ejection force is not uniformly applied to the molded product, the residual stress remains in the molded product, and the quality of the molded product is affected.

[0099] In the injection molding machine 50, the pressure control is performed so that the pressure value of the load cell 510 approaches the pressure value in the molding condition input. The temperature of the cylinder 505 is controlled by the plurality of heaters 506. Depending on the shape of the screw 502, the shape of the cylinder 505, and the shape of the nozzle 508, different pressure losses are generated in each injection molding machine. Due to this, the pressure at the resin flow inlet of the metal mold 509 becomes a value lower than the pressure indicated by the molding condition input to the injection molding machine. Also, due to the arrangement of the heaters 506 and the shearing heat of the resin in the nozzle portion, the resin temperature at the resin flow inlet of the metal mold 509 is sometimes different from the resin temperature indicated by the molding condition input to the injection molding machine.

[0100] The structure of the injection mechanism (the shape of the screw 502, the shape of the cylinder 505, the shape of the nozzle 508, the arrangement of the heaters 506, and the like) differs depending on the injection molding machine, and sometimes becomes a mechanical error to affect the quality of the molded product.

[0101] The quality of the molded product is evaluated in terms of shape characteristics (weight, length, thickness, sink marks, burrs, warpage, etc.), surface characteristics (seams, crazing, burns, whitening, scuff marks, blisters, peeling, flow marks, jet marks, color / gloss, etc.), and mechanical / optical characteristics (tensile strength, impact resistance, transmittance, etc.).

[0102] The shape characteristics have a strong correlation with the pressure and temperature history and the clamping force during the injection and holding processes and the cooling process. As for the surface characteristics, the causes of the phenomena that occur are different for each, but, for example, flow marks and jet marks have a strong correlation with the temperature and speed of the resin during the injection process. As for the mechanical and optical characteristics, for example, in the case of the tensile strength, evaluation is required using a destruction test, so other quality indicators such as the weight are used for evaluation.

[0103] The parameters corresponding to each process of the injection molding process are set in the molding conditions. As for the metering and plasticizing process, the metering position, back suction, back pressure, back pressure speed, and rotational speed, etc. are set. As for the injection and holding processes, the pressure, temperature, time, and speed, etc. are set for each. As for the injection and holding processes, the screw position at which the injection and pressure are switched (VP switching position) and the clamping force of the metal mold 509 are also set. As for the cooling process, the cooling time after the holding is set. As parameters related to the temperature, the temperature of the plurality of heaters 506, and the temperature and flow rate of the refrigerant for cooling the metal mold 509, etc. are set.

[0104] Figure 4 An outline of the metal mold used in the experiment for verifying the optimization of the molding conditions in the present embodiment is shown. Figure 4 A plan view 70 of the product portion, a side view 71 of the product portion, and a plan view 72 of the runner portion are shown. The metal mold is a configuration in which the resin flows into the product portion from the runner portion in a 5-point gate manner. In the actual molding experiment, a pressure sensor and a resin sensor (both not shown) are arranged at a sensor arrangement portion 73 of the runner, and the time variation thereof is obtained. As for the material used in the molding, polypropylene (PP) is used. As for the injection molding machine, an electric injection molding machine having a maximum clamping force of 150 t and a screw diameter of 44 mm is used.

[0105] Figure 5 is a block diagram showing an example of a method of obtaining material information by performing feature quantity extraction in the feature quantity extraction section 402 for the physical quantity obtained from the sensor 58 and recording the result thereof to the feature quantity database 403. Figure 5 The illustrated method of obtaining material information is implemented by using either of a "metal mold with sensor" in which a sensor for measuring a predetermined physical quantity is provided at a predetermined position, and a "metal mold with built-in sensor" in which a sensor for measuring a predetermined physical quantity is built in.

[0106] First, for any material 601, a fixed molding condition, i.e., a reference molding condition 602, is input to an actual injection molding machine 603, whereby a physical quantity at a predetermined position in a metal mold is acquired. Here, the injection molding machine 603 corresponds to the injection molding machine 50 described in Figure 3 . Also, the reference condition corresponds to a condition input to the injection molding machine when the injection molding process 55 described in Figure 1 is executed.

[0107] The physical quantity at the predetermined position in the metal mold is affected by material information of the material itself used, mechanical errors inherent to the metal mold and the injection molding machine, and molding conditions. Therefore, by setting the reference condition for each combination of the metal mold and the injection molding machine, the effects of the mechanical errors and the molding conditions can be suppressed, and the material information inherent to the material can be recorded as a characteristic quantity of the physical quantity in the characteristic quantity database 610. That is, the reference condition can be changed for each combination of the metal mold and the injection molding machine.

[0108] In order to acquire a molding phenomenon in the actual injection molding machine 603, a metal mold in-mold sensor 606 is used. By disposing the metal mold in-mold sensor 606 at an arbitrary position in the metal mold 604, the molding phenomenon in the metal mold 604 can be directly measured, whereby an actual measurement value 608 of the physical quantity related to the material information can be acquired. The quality of the molded product 605 can be acquired by a product quality inspection 607.

[0109] A characteristic quantity is extracted from the acquired physical quantity (609). The acquired physical quantity is acquired as a time change in the injection molding process, and thus it is difficult to directly evaluate. Therefore, in the present embodiment, a characteristic quantity related to the material information is acquired from the time change of the physical quantity, whereby a quantitative evaluation of the material information is performed. By molding under the reference condition common to the materials, the material information between the materials can be compared by comparing the characteristic quantities between the materials.

[0110] Using Figure 6 and Figure 7 , a measured result of an experimental example for verifying optimization of the molding condition described in Figure 4 will be described. Figure 6 shows time series data of a pressure sensor and a resin sensor in the sensor arrangement portion 73 of the runner when three batches P, Q, and R of PP different in delivery period are each injection molded under the reference condition. As shown in Figure 6 , it can be confirmed that even with the same molding condition, the variation in the time series data of each batch is different, and the time series data of the pressure sensor is affected by the material information inherent to the material.

[0111] Figure 7 is fromFigure 6 The pressure sensor and the resin temperature sensor of each material batch in Table 1 extract the peak value, the maximum differential value, the integral value to the peak value, and the integral value from the peak value as the results of the characteristic quantities (the values of the resin temperature sensor are omitted). According to the characteristic quantities, it is possible to confirm that the extracted characteristic quantities are influenced by the material information. Thus, the characteristic quantity extraction process is performed for each batch, and the data set in which the characteristic quantities are associated for each batch is referred to as a characteristic quantity data set, and is recorded in the characteristic quantity database 403. In the present embodiment, as shown in Table 1, 4 dimensions in the pressure sensor and 4 dimensions in the resin temperature sensor are used, and a total of 8 dimensions of the characteristic quantity data set are used. Figure 7 Since the characteristic quantities are biased between the batches, it is possible to confirm that the extracted characteristic quantities are influenced by the material information. Thus, the characteristic quantity extraction process is performed for each batch, and the data set in which the characteristic quantities are associated for each batch is referred to as a characteristic quantity data set, and is recorded in the characteristic quantity database 403. In the present embodiment, as shown in Table 1, 4 dimensions in the pressure sensor and 4 dimensions in the resin temperature sensor are used, and a total of 8 dimensions of the characteristic quantity data set are used. Figure 7

[0112] Next, the site of the metal mold at which the physical quantity is measured, the parameter of the physical quantity related to the resin information, and the characteristic quantity are described.

[0113] First, the site (hereinafter, referred to as a measurement site) in the metal mold at which the physical quantity is measured is described. In any one metal mold structure, it is preferable that the measurement site at least include the sprue land portion or the runner portion from the resin flow inlet in the metal mold to the cavity.

[0114] It is also possible to use the cavity as the measurement site, but when the material information inherent to the material is derived by the above-described steps, it is necessary to consider the loss of each physical quantity from the resin flow inlet to the cavity. Therefore, it is necessary to ensure the resolution accuracy from the resin flow inlet to the cavity. In addition, in the case where the sensor is provided in the cavity to perform the measurement, it is possible that the trace caused by the shape of the sensor remains on the molded product. Therefore, in the place where the appearance quality is required, there is a constraint that the sensor cannot be introduced.

[0115] Therefore, in the present embodiment, by using the sprue land portion or the runner portion near the resin flow inlet as the measurement site, which does not require the appearance quality, it is possible to easily and accurately derive the physical quantity related to the material information inherent to the material.

[0116] In addition to the sprue land portion and the runner portion, it is also possible to use the site such as the gate just below portion, the resin confluence portion (welding portion), the flow end portion, and the like, at which the characteristic flow can be observed, as the measurement site. In this case, it is possible to more accurately derive the physical quantity related to the material information inherent to the material from the physical quantities obtained by the plurality of sensors.

[0117] For example, it is possible to derive the flow rate of the molten resin from the passing time of the flow front of the plurality of measurement sites, and thus it is possible to derive the material information on the velocity of the molten resin. Furthermore, by measuring the pressure and the temperature at that time, it is possible to estimate the viscosity of the molten resin in the metal mold.

[0118] ​Further, the appropriate measurement site differs depending on the metal mold structure and the physical quantity to be measured. In the case of a physical quantity other than the metal mold opening amount, it is preferable to set the sprue land as the measurement site, regardless of the metal mold structure, if possible. In the present specification, the expression "preferable" is used only in the sense that a certain advantageous effect can be expected, and does not mean that the structure is essential.

[0119] In the case of a side gate, a jump gate, a submarine gate, and a banana gate, the sensor is disposed in the runner portion just below the sprue land, the runner portion just before the gate, or the like. In the case of a pin gate, since a 3-plate structure is formed, the sensor needs to be disposed in the runner portion just below the sprue land or the like. In the case of a pin gate, a dummy runner not connected to the cavity can be provided as the measurement site for measurement. By providing a site dedicated to measurement, the degree of freedom in the design of the metal mold is improved. In the case of a pin gate, a dummy runner not connected to the cavity can be provided as the measurement site for measurement. By providing a site dedicated to measurement, the degree of freedom in the design of the metal mold is improved. In the case of a film gate or a fan gate, the sensor is disposed in the runner portion before the gate portion.

[0120] The parameters measured as the above-described physical quantity will be described. In the present embodiment, at least the pressure and the temperature are measured in order to optimize the molding conditions corresponding to the material information. In the measurement of the pressure and the temperature, for example, a metal mold internal pressure sensor, a metal mold surface temperature sensor, a resin temperature sensor, or the like can be used. In the resin temperature sensor, either or both of a contact type temperature sensor such as a thermocouple or a non-contact type temperature sensor such as an infrared radiation thermometer can be used, and the time change in the injection molding process is recorded for either of the physical quantities of the pressure and the temperature.

[0121] The injection molding condition optimization system 1 can acquire the flow front speed, the flow front passage time, in addition to the metal mold opening amount, the temperature, and the pressure. The information on the flow front passage time can be obtained from the sensor for detecting the speed of the flow front and the passage of the flow front, instead of the time change in the injection molding process. In the case where the flow front passage time is acquired, at least two or more sensors are provided, and the passage times of the resin between two points are compared. By detecting the speed of the flow front and the passage time, the injection speed can be more accurately evaluated.

[0122] The characteristic quantity of the above physical quantity will be described. In the present embodiment, for example, the maximum value (peak value of time change) and the integral value of pressure, and the maximum value (peak value of time change) of temperature can be used. In addition, for the time change of pressure, it is also effective to obtain the maximum value of the time differential value. The maximum value of the time differential value of pressure is related to the instantaneous viscosity of the material. The integral value of pressure can also be calculated separately in the injection process and the holding process. The integral value of pressure in the injection process is related to the average viscosity of the material in the injection process.

[0123] In the case of using a resin temperature sensor of infrared radiation type, the maximum value of the time differential value can be obtained for the output value of the temperature sensor in the injection process. This characteristic quantity is related to the flow front speed of the molten resin. In the case of obtaining the flow front speed, it is directly used as a characteristic quantity related to the flow speed. In the case of obtaining the flow front passing time, the flow rate is calculated from the passing times between 2 points to be used as a characteristic quantity. By recording the relationship of the flow rate with respect to the set value of the injection speed (fixed value regardless of the material under the reference condition), the injection speed can be more accurately recorded.

[0124] Using Figure 8 The process data set recorded in the process database 408 will be described. In the process database 408, data relating the molded product quality of the molded product molded by the injection molding process 54 or the injection molding process 55, the molding condition input to the injection molding machine, the predetermined material unit, the combination of the injection molding machine and the metal mold are recorded, and the process data set is recorded for each combination of the injection molding machine and the metal mold.

[0125] In the present embodiment, using the above-described material lots P, Q, R, the molded product weight is set as the molded product quality, and the clamp pressure, the metal mold temperature, the molding nozzle temperature, the injection speed, the holding pressure, the V-P switching position, and the cooling time are set as the parameters of the molding condition. It is also possible to add parameters other than the above-described parameters to the parameters of the molding condition, or to reduce any one or more of the above-described parameters, but as the parameters of the molding condition, it is preferable to include at least one or more parameters related to the molded product quality (molded product weight in the case of the present embodiment).

[0126] Figure 8 is an example of the process database recorded for each combination of the injection molding machine and the metal mold in the present embodiment, relating the molding condition, the material lot, and the molded product weight. Specifically, in the case of #1 of Figure 8 , as the parameters of the molding condition, the clamp pressure is set to 120 [t], the metal mold temperature is set to 30 [°C], the molding nozzle temperature is set to 180 [°C], the injection speed is set to 40 [mm / s], the holding pressure is set to 30 [kg / m2 ]、V-P switching position 10 [mm] and cooling time 35 [s] indicated that the average of the molded product weight when using the batch P as a material batch for multiple injections was 62.83 [g]. Here, the molding conditions and the material batch changed in the trial molding can be determined, for example, based on an experimental design method, or can be determined based on the results of CAE simulation.

[0127] Using Figures 9 to 16 The method of molding condition optimization in the learning and optimization system 4 will be described. The learning and optimization system 4 has two functions of an optimization mode and a learning mode. In the optimization mode, the best molding condition that satisfies the required quality for the second material is generated. On the other hand, in the learning mode, a learned dimension reduction model and a learned regression model used in the optimization mode are generated.

[0128] Using Figures 9 to 13 The learning mode will be described. Figure 9 A flowchart indicating the learning mode. When the learning mode starts, first, the learning and optimization system 4 reads out the feature quantity data set from the feature quantity database 403 (an example of the feature quantity data set is Figure 7 ), and performs a standardization process (S101).

[0129] The standardization process refers to a process of calculating the average value and the standard deviation of the data set for each column as an object, subtracting the average value from the value of the data set, and dividing by the standard deviation. By the standardization process, the average value of the data set of each column is 0 and the standard deviation is 1, so that the influence of the difference in units between the columns can be removed. Generally, by performing the standardization process, it has an effect of improving the learning accuracy of the dimension reduction model and the regression model.

[0130] Next, the learning and optimization system 4 performs learning of the dimension reduction model on the feature quantity data set on which the standardization process has been performed (S102). In the present embodiment, as the dimension reduction model, UMAP is used, and the 8-dimensional feature quantity data set is reduced to 2 dimensions.

[0131] UMAP is a nonlinear dimension reduction model that converts input data into data having a lower dimension than the dimension of the input data while maintaining the positional information between the input data. As other well-known models related to dimension reduction, there are principal component analysis, auto-encoding, and they can also be used as a dimension reduction model for the feature quantity data set.

[0132] Figure 10is a feature quantity dataset (dimension: 8, number of samples 120) obtained by respectively molding each of the material lots P, Q, R 40 times under the reference conditions in the injection molding process 55, converted into 2-dimensional vectors Z (Z1, Z2) using the dimension reduction model UMAP, and plotted on a graph on a 2-dimensional plane.

[0133] Figure 10 The 1 point in corresponds to 1 injection, and the shape of the point differs depending on the material lot. According to Figure 10 , points of the same material lot are close to each other, and points of different material lots are far from each other, and thus it is possible to confirm that the vectors Z after dimension reduction can express the difference in material information between the lots. That is, the learned dimension reduction model generates 2-dimensional data related to the material information from 8-dimensional input data. In the present embodiment, the vectors after dimension reduction are set to 2 dimensions, but as long as the dimension is lower than the input dimension, it can be reduced to any dimension of 1 dimension or more.

[0134] Returning to Figure 9 , the learning and optimization system 4 saves information related to the standardization process and the learned dimension reduction model to the learned dimension reduction model storage section 406 (S103). The information related to the standardization process refers to the average value and the standard deviation of each column of the feature quantity dataset.

[0135] As shown in Figure 11 , the learning and optimization system 4 calculates the average value of the dimension of each vector with respect to the vectors Z after dimension reduction obtained from the dimension reduction model, per lot, and generates the average value in association with the lot as a feature quantity dataset after dimension reduction (S104).

[0136] The learning and optimization system 4 reads out the feature quantity dataset after dimension reduction and the process dataset set saved in the process database 408, generates a learning dataset by combining the lots as a combination key, and records it in the learning database 412 (S105).

[0137] Figure 12 indicates an example of a learning dataset generated by combining the process dataset set shown in Figure 8 with the feature quantity dataset after dimension reduction shown in Figure 11 . According to Figure 12 , it is possible to confirm that the column of the material lot in the process dataset set of Figure 8 is replaced with the feature quantity Z (Z1, Z2) after dimension reduction.

[0138] When a regression model is generated based on the learning data set, in a case where the dimensionality of the learning data set is unnecessarily excessive, the cost of learning becomes high and the accuracy becomes unstable. Therefore, it is preferable to generate the learning data set by combining the process data set with the data set after the dimensionality reduction. However, the learning data set can also be generated by combining the process data set with the feature data set without implementing the dimensionality reduction.

[0139] The learning and optimization system 4 performs a standardization process on each column of the learning data set, as performed in step S101 (S106).

[0140] The learning and optimization system 4 generates a regression model that takes the weight as a target variable (output parameter) and the mold clamping pressure, the metal mold temperature, the molding nozzle temperature, the injection speed, the holding pressure, the V-P switching position, the cooling time, Z1, and Z2 as explanatory variables (input parameters) by machine learning using the data set after the standardization process (S107).

[0141] In the present embodiment, the learned regression model is generated using support vector regression, but the regression model can also use a regression model such as linear regression, ridge regression, support vector machine, neural network, random forest regression, or a regression model that combines them.

[0142] Figure 13 The performance evaluation result of the learned regression model generated by support vector regression in the present embodiment is shown. This performance evaluation is performed in order to confirm the effectiveness of the learned regression model generated in the present embodiment, and it is not necessarily required to perform this performance evaluation in terms of the injection molding condition generation system 1. Hereinafter, the method of the performance evaluation will be described.

[0143] The learning data set shown in the data structure of the above Figure 12 is randomly extracted and divided into a first data set and a second data set. Also, in the present performance evaluation, the first data set is used to generate a learned regression model of support vector regression. In the present performance evaluation, the predicted value of the molded product weight is calculated by inputting the first data set and the explanatory variables of the second data set to the learned regression model.

[0144] Figure 13 is a graph showing the relationship between the measured weight of the molded product and the predicted weight. The horizontal axis represents the measured weight of the molded product measured by the molded product quality inspection unit 57. The vertical axis represents the predicted weight of the molded product based on the learned regression model. The straight line of the dotted line represents a reference line when the measured weight and the predicted weight coincide. The closer each point is to the reference line, the higher the accuracy of the predicted value.

[0145] Typically, the coefficient of determination is used as a representative indicator for quantitatively evaluating the predictive accuracy of a regression model. The closer the coefficient of determination is to 1, the higher the predictive accuracy of the regression model. Figure 13 The results confirm that the coefficients of determination for both the first and second datasets are above 0.9. Furthermore, the coefficient of determination for the second dataset, which was not used in the regression model training, deviates less from that of the first dataset, which was used in the regression model training, thus confirming the good general performance of the learned regression model. This evaluation demonstrates that a good model for predicting the weight of molded products can be obtained by using a training dataset containing a dimension-reduced feature Z related to the molding conditions and material information obtained from the trial molding in injection molding process 54.

[0146] return Figure 9 The learning and optimization system 4 stores the information related to the standardization process and the learned regression model in the learned regression model storage unit 415 (S108) and ends the learning mode.

[0147] use Figures 14 to 16 The optimization mode in the learning and optimization system 4 is explained. The optimization mode generates the optimal molding conditions for the second material to meet the required quality based on the dimensionality reduction model generated in the learning mode, the learned regression model, the feature dataset of the second material, and the optimization algorithm.

[0148] Figure 14 A flowchart illustrating the optimization mode. In the optimization mode, the target weight of the molded product is read from the manufacturing condition determination unit 31 as the required quality information (S201).

[0149] The learning and optimization system 4 uses the second material to perform injection molding under baseline conditions in the injection molding process 55 with a predetermined number of injections (S202). Based on the molding result, the learning and optimization system 4 generates a feature quantity dataset of the second material through the feature quantity extraction unit 402 (S203).

[0150] The learning and optimization system 4 reads the standardized processing information of the feature quantity database generated in the learning mode and the learned dimension reduction model from the learned dimension reduction model storage unit 406 (S204).

[0151] The learning and optimization system 4 performs standardization processing based on the feature database information and dimension reduction based on the learned dimension reduction model on the feature dataset of the second material to generate dimension-reduced features of the second material (S205). In this embodiment, the dimension-reduced features of the second material are (Z1, Z2) = (0.38, -4.0).

[0152] The learning and optimization system 4 reads out the standardized processing information of the learning database and the learned regression model from the learned regression model storage section 415 (S206), and optimizes the molding conditions based on the target weight, the dimensional reduction feature amount data set of the second material, the standardized processing information, the learned regression model, and a known optimization algorithm.

[0153] In the present embodiment, a case where a grid search method is used as the optimization algorithm is described. In the optimization algorithm, a gradient method, a simulated annealing method, a genetic algorithm, and a Bayesian optimization can be used in addition to the grid search method. In addition, these optimization algorithms can be combined to optimize the molding conditions.

[0154] The method of the molding condition optimization based on the grid search method is described using steps S207 to S209 of Figure 14 First, an outline of the grid search method is described.

[0155] The grid search method is an optimization algorithm in which a plurality of candidate values are set for each parameter of the explanatory variable, the predicted value of the target variable is calculated using the learned regression model for all combinations of the explanatory variable composed of the combinations of the candidate values of the parameters, and the combination of the explanatory variable having the smallest deviation between the predicted value and the target value is extracted. The grid search method can robustly find a global optimal solution not only for a problem in which local characteristics are stronger than global characteristics but also for a problem in which the distribution of the target variable in the entire region of the explanatory variable can be grasped.

[0156] In the present embodiment, first, all combinations of the parameters of the explanatory variable of the regression model are generated (S207). In the parameters of the explanatory variable of the regression model in the present embodiment, as shown in the example of the learning data set of Figure 12 there are parameters related to the molding conditions (clamp pressure, metal mold temperature, molding nozzle temperature, injection speed, holding pressure, V-P switching position, cooling time) and parameters related to the material information (dimensional reduction feature amounts Z1 and Z2).

[0157] In the parameters of the explanatory variable, regarding the parameters related to the molding conditions, the maximum value and the minimum value of each parameter are determined, and the candidate values are generated with the setting resolution of each parameter in the injection molding machine as the step size. Regarding the parameters related to the material information, the values of the feature amount data set generated in step S205 are used to generate the combinations of the parameters of the explanatory variable. Regarding the parameters related to the molding conditions, for example, for the clamp pressure, the minimum value is set to 120 [t] and the maximum value is set to 150 [t], and the clamp pressure is changed by 1 [t] each time. Figure 15An example combination of parameters generated in this embodiment is shown. Thus, by fixing parameters related to material information, optimal molding conditions can be generated when using a second material.

[0158] For each combination of parameters, the learning and optimization system 4 uses standardized processing information from the learning database and the learned regression model to calculate the weight of the molded product (S208).

[0159] In order to quantitatively evaluate the deviation between the predicted value and the target value, the learning and optimization system 4 calculates the absolute value of the difference between the predicted value and the target value, sorts the absolute values ​​of the difference in ascending order, and generates the molding conditions that are closest to the target weight as the optimal conditions (S209).

[0160] Regarding the quantitative evaluation of the deviation between the predicted value and the target value, this embodiment sets the absolute value of the difference between the predicted value and the target value, but other indicators representing the deviation between the predicted value and the target value, such as the case value of the difference between the predicted value and the target value, can also be set.

[0161] This embodiment illustrates a case where only the weight of the molded product is required as a quality requirement, but it can also be applied to cases where multiple quality requirements exist, such as the weight and dimensions of the molded product. Any index can be set to represent the deviation between the target value and the predicted value of the object, according to the required quality.

[0162] Figure 16 This indicates the weight of the molded article formed under the baseline conditions in the injection molding process 55 using the second material in this embodiment, and the weight distribution of the molded article formed under the optimized conditions for the generation of the second material.

[0163] To generate optimal molding conditions, the required quality (molded product weight) input to the optimal condition generation unit 417 is 63.89 g. Figure 16 In the diagram, the numbers are represented by dashed lines parallel to the vertical axis. In each distribution, the number of formed items is 40, and the distributions are standardized based on the number of formed items. The vertical axis represents the probability density.

[0164] according to Figure 16 The weight of the molded product is close to the required quality, confirming the appropriateness of the molding conditions. This ensures that even if the material properties of the resin change, the molding conditions can still be optimized to meet the required quality.

[0165] Explanation of reference numerals in the attached figures

[0166] 1: injection molding condition generation system, 2: production management system, 3: manufacturing execution system, 4: learning and optimization system, 5: manufacturing plant, 31: manufacturing condition decision unit, 32: production performance storage unit, 33: production performance acquisition unit, 34: manufacturing execution instruction unit, 35: production performance recording unit, 401: sensor information recording unit, 402: feature quantity extraction unit, 403: feature quantity database, 404: dimension reduction model learning unit, 405: learned dimension reduction model storage unit, 406: dimension reduction model storage unit, 407: process data recording unit, 408: process database, 409: dimension reduction model readout unit, 410: dimension reduction execution unit, 411: linking processing unit, 412: learning database, 413: regression model learning unit, 414: regression model storage unit, 415: learned regression model storage unit, 416: regression model readout unit, 417: optimum condition generation unit, 51: manufacturing execution unit, 52: molding condition generation unit, 57: molded product quality inspection unit, 58: sensor

Claims

1. An injection molding condition generating system that generates an injection molding condition using a computer, characterized by, the computer including a processor and a memory used by the processor, the processor generating an injection molding condition of injection molding using a first resin material based on a target value of a quality parameter related to a quality of a molded product, a first material characteristic value of the first resin material, and a predetermined relational expression when the first material characteristic value is acquired, the predetermined relational expression being an expression representing a relationship among a material characteristic value of a resin material, a plurality of injection molding conditions input to an injection molding machine, and a quality parameter related to a quality of a molded product molded by the injection molding machine based on the material characteristic value and each of the injection molding conditions, the predetermined relational expression being generated based on data accumulated in the memory in association with the material characteristic value of the resin material, the injection molding condition, and the quality parameter, the first material characteristic value being generated based on a feature amount of a measured value of a sensor in a metal mold when molded with a reference injection molding condition common to each predetermined unit among the predetermined units for the first resin material.

2. The injection molding condition generating system according to claim 1, characterized by, the predetermined unit being a delivery unit of a recycled material supplied from a material supplier.

3. The injection molding condition generating system according to claim 1, characterized by, the predetermined unit being a material supplier unit when a raw material of the same kind is supplied from a plurality of different material suppliers.

4. The injection molding condition generating system according to any one of claims 1 to 3, characterized by, the predetermined relational expression being a regression expression that uses a machine learning model, in which a quality parameter related to a quality of a molded product when trial molding is performed while changing an injection molding condition for a trial resin material is used as an output parameter, and the injection molding condition and a material characteristic value of the trial resin material are used as input parameters.

5. The injection molding condition generating system according to claim 4, characterized by, the injection molding condition in the predetermined relational expression including at least any one of a clamp pressure, an injection speed, a temperature of a nozzle portion of an injection molding machine, a holding pressure, a speed-pressure control switching position, and a metal mold temperature.

6. The injection molding condition generating system according to claim 4, characterized by, the quality parameter in the predetermined relational expression including at least any one of an average value of a weight of a molded product, an average value of a size of a molded product, an average value of a warpage amount of a molded product, and a defective rate of a molded product under the predetermined unit.

7. The injection molding condition generating system according to claim 4, characterized by, the feature amount including at least any one of a temperature, a speed, and a pressure.

8. The injection molding condition generating system according to claim 4, characterized by, the feature amount including a flowability of a resin material. ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ The flowability of the resin material is calculated based on at least any one of a peak value of a measured value of the in-mold sensor, an integral value of the measured value from the start of injection to the peak value, an integral value of the measured value from the start of injection to the opening of the mold, and a maximum differential value of the measured value.

9. The injection molding condition generation system according to claim 6, wherein The dimension of the characteristic quantity is reduced, and the reduced characteristic quantity is used as the material property value.

10. An injection molding condition generation method of generating an injection molding condition using a computer, characterized by The computer acquires a first material property value of a first resin material; The computer generates an injection molding condition of injection molding using the first resin material based on a target value of a quality parameter related to the quality of a molded product, the first material property value, and a predetermined relational expression; The computer outputs the generated injection molding condition, The predetermined relational expression is an expression representing a relationship among a material property value of a resin material, a plurality of injection molding conditions input to an injection molding machine, and a quality parameter related to the quality of a molded product molded by the injection molding machine based on the material property value and each of the injection molding conditions, and the predetermined relational expression is generated based on data obtained by accumulating the material property value of the resin material, the injection molding condition, and the quality parameter in association with each other, The first material property value is generated based on a characteristic quantity of a measured value of an in-mold sensor when molded with a reference injection molding condition common to each predetermined unit for the first resin material.

Images