SYSTEM AND METHOD FOR PRODUCE INJECTION PILLING CONDITIONS

Abstract

Injection molding condition generation system (1) for generating an injection molding condition using a computer (10), wherein the computer (10) has a processor (11) and memory (12, 13) used by the processor (11), the processor (11), when it receives a first material property value of a first resin material (601), generates the injection molding condition using the first resin material (601) based on a target value of a quality parameter relating to the quality of a molded part (605), the first material property value and a predefined relational expression, and The predefined relational expression specifies a relationship between the material property value of the resin material (601), several injection molding conditions entered into an injection molding machine (50, 603), and a quality parameter that relates to the quality of the molded part (605) formed by the injection molding machine (50, 603) in a mold (604) based on the material property value and the injection molding conditions, and is generated on the basis of data collected in memory (12, 13) in connection with the material property value of the resin material (601), the injection molding conditions, and the quality parameter. characterized in that the material property value for generating the specified relational expression for each specified unit of the resin material (601) is generated on the basis of a feature of a measured value of a mold internal sensor (58, 606) arranged in the mold (604) when a molding is carried out under a common reference injection molding condition (602) between the specified units.

Description

Technical area

[0001] The present invention relates to a system and a method for generating injection molding conditions. Technical background

[0002] In an injection molding process, various factors, such as multiple injection molding conditions input to a molding machine and variations in the material properties of a resin, affect the quality of a molded part. Therefore, adjusting the molding conditions to stabilize the quality of the molded part is not easy and requires the technology and time of a skilled operator. To solve this problem, an operator-independent method for optimizing an injection molding condition was investigated (JP 2016-527109 A).

[0003] In the method described in JP 2016-527109 A, an injection molding control device that provides a control signal for partially setting the 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 at a first time in an injection molding cycle. A second control signal from the pressure control output is measured at a second time in a subsequent injection molding cycle. A control signal for the pressure at a third time in an injection molding cycle is set based on the comparison result between the first control signal and the second control signal.That is, in JP 2016-527109 A, by measuring a material property change during the injection molding process as a change in the control signal from the controller and adjusting the pressure based on the change in the control signal, the injection molding condition (the injection pressure) can be controlled taking the material property change into account without depending on the operator.

[0004] AT 519 096 A4 discloses an injection molding setting system according to the preamble of present claim 1. Further conventional injection molding setting systems are described in US 2002 / 0 055 806 A1, DE 10 2017 130 997 A1 and DE 10 2020 209 479 A1. Summary of the inventionTechnical problem

[0005] A prerequisite of the method described in JP 2016-527109 A is that the pressure is adjusted sequentially for each cycle during an injection molding process. Therefore, the method described in JP 2016-527109 A requires a dedicated control device for measuring the control signal based on the pressure control output during the injection molding process and executing the pressure control according to the measured value, as well as development for the normal connection of the control device and the injection molding machine, as well as additional development and labor.

[0006] The relationship between molding conditions in injection molding and quality is influenced by the material properties of the resin. For example, because even materials of the same type (e.g., polypropylene) vary in flowability depending on the grade or supplier, the behavior of the resin in a mold varies considerably even when the same molding conditions are applied, and the quality of the molded part also varies.

[0007] In recent years, much attention has been paid to the use of recycled plastic materials due to the problems of marine pollution caused by plastic waste and the ban on the transfer of plastic waste from China and Southeast Asia. Taxation and regulations on the use of virgin materials in some regions of Europe are also being studied, highlighting the urgent need for manufacturers of plastic-based products to use recycled materials. However, because recycled materials exhibit a high degree of variability in material properties compared to virgin materials, due to thermal history during molding, environmental deterioration during use, and contamination by foreign substances, as well as thermal history during recycling, the quality variation of molded parts becomes greater than that of virgin materials.

[0008] Fig. 17 is a graph showing a comparative example for explaining an advantage of later-described embodiments, which is not related to the prior art. Fig. Figure 17 shows the weight distributions of molded parts when molding is carried out using, for example, three recycled materials from Lot A, Lot B and Lot C. The horizontal axis in Fig. 17 represents the weight of the molded part, and the vertical axis in Fig. 17 represents a probability density. Fig. Figure 17 shows the weight distribution of molded parts when three recycled materials with different delivery times are molded with the same mold and the same injection molding machine and under the same molding conditions.

[0009] In the following, a difference due to delivery time is referred to as a lot. Fig. 17, the number of molded parts per batch is 40. The Fig. The weight distributions of molded parts shown in Figure 17 are all normalized by the number of molded parts.

[0010] Fig. Figure 17 shows that the weight variation between batches is significantly larger than the variation within the batch, even when forming is carried out under the same forming conditions. Therefore, it can be confirmed that, for recycled materials, the material property difference between incoming batches affects the relationship between the forming conditions and quality.

[0011] Regarding the influence of the material property of the resin, JP 2016-527109 A measures the change in the material property during the injection molding process as a change in the control signal from the controller. In JP 2016-527109 A, adjustment according to the change in the material property is performed by controlling the pressure of the injection molding condition based on the change in the control signal from the controller. However, when controlling only the pressure, if the temperature property and fluidity of the resin change significantly, the amount of pressure adjustment increases. Therefore, in the method described in JP 2016-527109 A, if the pressure is too low, there is a high probability of short injection molding, and if the pressure is too high, there is a high probability of appearance defects such as burrs occurring.

[0012] The invention has been made in view of the above problems, and an object of the invention is to provide a system and a method for generating injection molding conditions, whereby the quality of injection molding can be improved. Solution to the problem

[0013] To solve the above problems, an injection molding condition generating system having the features defined in claim 1 or a method having the features defined in claim 10 is used. Advantageous effects of the invention

[0014] According to the invention, an injection molding condition suitable for a material property value of a resin material used in an injection molding machine and the quality of a molded part can be obtained. Brief description of the drawings

[0015] They show: Fig. 1 is a functional block diagram of a molding condition optimization system of an injection molding machine, Fig. 2 is a configuration diagram of a computer usable for implementing an injection molding condition generation system, Fig. 3 a concept diagram of the injection molding machine, Fig. 4 is a conceptual diagram of a mold used in an experiment to verify the optimization of a molding condition, Fig. 5 is a block diagram for extracting a feature based on a physical quantity obtained from a sensor and storing the extracted feature in a feature database, Fig. 6 is a graph showing a time series change in data acquired by a pressure sensor and a resin sensor, Fig. 7 a diagram of an example of a feature extraction result, Fig. 8 a diagram of a process data set, Fig. 9 a flowchart of a training mode, Fig. 10 is a graph showing an example of a result of converting a feature data set by UMAP, Fig. 11 a diagram of an example of the feature data set after dimension reduction, Fig. 12 a diagram of an example of a training data set, Fig. 13 a graph for assessing the performance of a trained regression model, Fig. 14 a flowchart of an optimization mode, Fig. 15 a diagram of an example of a combination of explanatory variables in the grid search, Fig. 16 a weight distribution diagram of a molded part to confirm the validity of an optimized molding condition and Fig. 17 shows a weight distribution diagram of a molded part as a comparison example when recycled materials with different delivery times are molded under the same molding condition. Description of embodiments

[0016] An embodiment of the invention will be described below with reference to the drawings. This embodiment provides a technique for reducing the variation in the quality of a molded article due to a variation in a material property. An injection molding condition generation system according to this embodiment typically does not need to be connected to an injection molding machine and can provide the injection molding machine with suitable injection molding conditions.

[0017] The embodiment described below does not limit the invention according to the claims, and not all elements and combinations thereof described in the embodiment are necessarily essential to the solution of the invention.

[0018] The system according to the present embodiment optimizes the injection molding condition according to a material property value of a resin material and the required quality.For example, the system generates in advance a relational expression indicating a relationship between the material property value of the resin material used for injection molding, a plurality of injection molding conditions input to the injection molding machine, and a quality parameter related to the quality of the article molded by the injection molding machine based on the resin material and the injection molding condition, based on collected data related to the material property value, the injection molding condition, and the quality parameter obtained in the past during molding, obtains a first-material property value of a first material, and generates an injection molding condition suitable for the first material based on the first material property value, a target value of the quality parameter, and the predetermined relational expression.

[0019] According to the present embodiment, even if a material property of the resin changes, an optimal molding condition for stabilizing the molding quality can be obtained regardless of the skill of an operator and without requiring real-time control during the injection molding process. [Embodiment 1]

[0020] Embodiment 1 is described with reference to the Fig. 1 to 16 are described. Fig. 1 shows a functional block diagram of a system 1 that generates a molding condition for an injection molding machine.

[0021] The injection molding condition generation system 1 includes, for example, a manufacturing treatment system 2, a manufacturing execution system 3, a training and optimization system 4, and a manufacturing factory 5. Some or all of the functions of the injection molding condition generation system 1 described below can be implemented in software, implemented through the cooperation of software and hardware, or implemented using hardware with a fixed circuit. At least some of the functions of the manufacturing treatment system 2, the manufacturing execution system 3, and the manufacturing factory 5 can be performed by an operator.

[0022] The production processing system 2 processes a production plan and includes at least one production plan processing unit 21. The production plan processing unit 21 generates a production plan including production specifications, quantities, periods, and the like, according to the order receipt status and the inventory status.

[0023] The manufacturing execution system 3 instructs the manufacturing factory 5 to perform manufacturing. The manufacturing execution system 3 determines a manufacturing condition based on the manufacturing plan generated by the manufacturing treatment system 2 and sends a manufacturing instruction including the manufacturing condition to the manufacturing factory 5. The manufacturing condition may include, for example, information for specifying an injection molding machine used for manufacturing (injection molding), information for specifying a mold to be used for manufacturing, information for specifying a material to be used for manufacturing, the number of molded parts to be manufactured, the manufacturing time, and the required quality.

[0024] The manufacturing execution system 3 will be described. The manufacturing execution system 3 includes, for example, a manufacturing condition determination unit 31, a manufacturing result storage unit 32, a manufacturing result acquisition unit 33, a manufacturing execution instruction unit 34, and a manufacturing result recording unit 35.

[0025] The manufacturing condition determination unit 31 determines the aforementioned manufacturing condition based on the manufacturing plan generated by the manufacturing plan processing unit 21 of the manufacturing processing system 2. The manufacturing condition determination unit 31 can transmit information about the manufacturing condition to the training and optimization system 4. The information about the manufacturing conditions can include predetermined information about the mold, the injection molding machine, and the material.

[0026] The predetermined information includes, for example, the mold capacity and the runner configuration. The predetermined information may include, for example, the required quality of the molded parts to be produced, the amount of resin material to be used in a supply unit, and the quantity shipped from each supplier. To optimize the molding condition, the training and optimization system 4 inputs information about the required quality of the molded part received from the manufacturing condition determination unit 31 into an optimal condition generation unit 417 and generates an optimized molding condition.

[0027] The manufacturing result storage unit 32 stores a manufacturing result. According to the present embodiment, the manufacturing result indicates a molding condition under which it has been confirmed that good quality of the molded parts can be obtained with a combination of the injection molding machine, the mold, and the material. Good quality of the molded parts means that the required quality determined in the manufacturing execution system 3 is met during molding.

[0028] The manufacturing result acquisition unit 33 obtains the manufacturing result from the manufacturing result storage unit 32. The manufacturing result acquisition unit 33 obtains the manufacturing result for the shape determined by the manufacturing condition determination unit 31 (hereinafter referred to as a first shape) and a used material determined by the manufacturing condition determination unit 31 (hereinafter referred to as a first material) by reading from the manufacturing result storage unit 32.

[0029] When no manufacturing result is obtained for a combination of the first shape and the first material, the manufacturing result acquisition unit 33 requests the manufacturing execution instruction unit 34 to optimize the molding condition based on the combination of the first shape and the first material or to mold based on the optimization result.

[0030] When requesting molding condition optimization, trial molding is performed in the manufacturing factory 5 under an injection molding condition set in advance as a reference (hereinafter referred to as the reference condition), information related to the trial molding is input to the training and optimization system 4, and the molding condition that meets the required quality is generated. Examples of the information to be input to the training and optimization system 4 will be described later. In addition, molding based on the optimization result means that molding is performed under a molding condition output from the training and optimization system 4, where a favorable result is expected to be obtained.

[0031] On the other hand, when the manufacturing result is obtained by the combination of the first mold and the first material, the manufacturing result acquisition unit 33 requests the manufacturing execution instruction unit 34 to execute molding based on a manufacturing result or to execute trial molding for training (hereinafter referred to as training trial molding).

[0032] Training trial molding here means performing injection molding while changing the molding condition, inputting the obtained information into the training and optimization system 4, and storing the obtained data as data for extracting an optimal molding condition. Details of the training trial molding will be described later with reference to Fig. 8 described.

[0033] The manufacturing execution instruction unit 34 instructs the execution of manufacturing in the manufacturing factory 5. Manufacturing execution may also be referred to as manufacturing. The manufacturing execution instruction includes, for example, the molding condition optimization request input from the manufacturing result acquisition unit 33, the molding request based on the optimization result, the molding request based on the manufacturing result, and the training trial molding request.

[0034] The manufacturing result recording unit 35 stores the molding condition for which it has been confirmed that good molded part quality can be obtained in the manufacturing factory 5 in the manufacturing result storage unit 32. Based on the information indicating the quality of the molded part obtained from a quality inspection unit 57 of the manufacturing factory 5, the manufacturing result recording unit 35 registers in the manufacturing result storage unit 32 the molding condition under which good molded part quality is obtained.

[0035] Manufacturing factory 5 is described. Manufacturing factory 5 receives the manufacturing execution instruction from manufacturing execution system 3 and executes one or more of the injection molding processes 53 to 56. Injection molding may be abbreviated as "IM" hereinafter.

[0036] The manufacturing factory 5 comprises, for example, a production execution unit 51, several injection molding machines 50 (later in Fig. 3), several forms (later in Fig. 3), a molding condition generation unit 52, and a molded part quality inspection unit 57. Hereinafter, the molded part quality inspection unit 57 may be abbreviated as quality inspection unit 57.

[0037] The manufacturing execution unit 51 executes the injection molding processes based on the manufacturing condition input from the manufacturing execution instruction unit 34 of the manufacturing execution system 3. When molding is requested based on the manufacturing result, the manufacturing execution unit 51 executes the injection molding process 53 with respect to a combination of a specified shape and a specified material by inputting the manufacturing result to the injection molding machine. That is, the injection molding process 53 is performed using a combination of a specified shape and a specified material under a molding condition that produces defect-free products.

[0038] When training trial molding is required, the manufacturing execution unit 51 outputs a training trial molding instruction to the molding condition generation unit 52. Upon receiving the training trial molding instruction from the manufacturing execution unit 51, the molding condition generation unit 52 generates a plurality of molding conditions stored in a training database 412 of the training and optimization system 4 with respect to the combination of the specified shape and the specified material. The manufacturing execution unit 51 executes the injection molding process 54 by inputting the plurality of generated molding conditions to the injection molding machine, while changing the generated molding conditions for each predetermined injection molding process unit.That is, in the injection molding process 54, the molding conditions are changed for each predetermined injection molding unit, and trial molding is carried out according to the molding conditions.

[0039] When molding condition optimization is required, the manufacturing execution unit 51 issues an instruction to mold under the reference condition to the molding condition generation unit 52. Upon receiving the instruction to mold under the reference condition from the manufacturing execution unit 51, the molding condition generation unit 52 executes the injection molding process 55 with respect to the instructed combination of mold and material by inputting the prespecified reference condition to the injection molding machine. That is, in the injection molding process 55, injection molding is performed according to the reference condition.

[0040] When molding based on the optimization result is required, the manufacturing execution unit 51 instructs the molding condition generation unit 52 to mold based on the optimization result. When an instruction to mold based on the optimization result is received from the manufacturing execution unit 51, the molding condition generation unit 52 receives the optimized molding condition generated by the optimal condition generation unit 417 from the training and optimization system 4 and executes the injection molding process 56 with respect to the combination of the specified shape and the specified material by inputting the received molding condition to the injection molding machine. That is, in the injection molding process 56, injection molding is performed according to the optimized molding condition.

[0041] The quality inspection unit 57 determines whether the quality of the molded part obtained through the injection molding process is good or poor. The quality of the molded part is assessed based on, for example, dimensions, the amount of warpage, burrs, scratches, gloss, and color. The inspection of the quality of the molded part can be performed automatically, manually by an inspector, or semi-automatically.

[0042] If the quality of the molded part is good, the quality inspection unit 57 outputs the manufacturing condition, the combination of the injection molding machine and the mold, the molding condition, and the molded part quality inspection result in association with each other to the manufacturing result recording unit 35 of the manufacturing execution system 3. In addition, when the injection molding processes 54 to 56 are executed, the manufacturing condition, the combination of the injection molding machine and the mold, the molding condition, and the molded part quality inspection result in association with each other are output to a process data recording unit 407 of the training and optimization system 4.

[0043] According to the present embodiment, information regarding the property of each material used when using the first mold is acquired by measuring a physical quantity at a predetermined position in the mold by sensors 58 mounted on the injection molding machines and the molds pre-stocked in the manufacturing factory 5, and outputting the physical quantity to a sensor information recording unit 401 of the training and optimization system 4. Details of the sensor mounted on the mold will be described later with reference to Fig. 4. Here, the information related to the material property includes, for example, the flowability and physical properties of the material and physical quantities that are correlated with the flowability and physical properties of the material.

[0044] For example, a given position of the injection molding machine is a nozzle tip position. For example, a given position in the mold is a resin inlet opening. The physical quantity is, for example, the pressure of a resin, the temperature of the resin, the speed of the resin, a material property of the resin, and the opening amount of the mold (mold opening amount). The material property is, for example, the density of the resin, the viscosity of the resin, or the fiber length distribution of the resin (in the case of a material with a reinforcing fiber content). Here, the physical quantity most strongly correlated with the flowability of the material is the viscosity of the resin, but other characteristics calculated based on pressure, temperature, and speed that are correlated with flowability can also be used.

[0045] The training and optimization system 4 is described. The training and optimization system 4 generates a suitable injection molding condition when no manufacturing result is obtained by combining the specified mold and the specified resin material. According to the present embodiment, the optimal injection molding condition refers to a suitable injection molding condition.

[0046] The training and optimization system 4 generates, for materials for which no manufacturing result has been obtained by combining the first mold and the first material in the manufacturing result storage unit 32 of the manufacturing execution system 3 and which are to be subjected to condition optimization (hereinafter, the materials are referred to as second material), the optimal molding condition for the second material to meet the required quality by inputting the required quality and the information obtained by the previously executed IM process 54 and the IM process 55 using the second material.

[0047] The training and optimization system 4 comprises, for example, the sensor information recording unit 401, a feature extraction unit 402, a feature database 403, a dimensionality reduction model training unit 404, a trained dimensionality reduction model storage unit 405, a trained dimensionality reduction model buffering unit 406, the process data recording unit 407, a process database 408, a dimensionality reduction model reading unit 409, a dimensionality reduction execution unit 410, a connection unit 411, the training database 412, a regression model training unit 413, a regression model storage unit 414, a trained model buffering unit 415, a regression model reading unit 416, and the optimal condition generation unit 417. on.

[0048] The sensor information recording unit 401 will be described. The sensor information recording unit 401 records the physical quantity at the specified position in the mold detected by the sensor 58 during the injection molding process 55 performed in the manufacturing factory 5. The feature extraction unit 402 performs feature extraction on the physical quantity temporarily recorded in the sensor information recording unit and records the extracted feature in association with a specified material unit and the combination of the injection molding machine and the mold in the feature database 403. Here, the specified material unit means, for example, a unit for distinguishing materials such as a model number or a lot for each material supplier.

[0049] According to the present embodiment, a data group recorded in the feature database 403 is referred to as a feature data set. The processing in the feature extraction unit 402 will be described later with reference to Fig. 6 and Fig. 7. When only the predetermined material unit is changed while the combination of the injection molding machine and the mold is fixed and the molding condition in the injection molding process 55 is fixed, the feature extracted from the physical quantity obtained from the sensor 58 is greatly affected by a change in material information (e.g., flowability or value of a physical property) between the material units.

[0050] Therefore, it is possible to record a variation in material information between material units as a variation between features. Here, because the feature extracted from the physical quantity obtained by the sensor 58 is affected by a difference between the injection molding machine and the mold, the variation in material information between material units is stored as a variation between features. That is, the feature data set is recorded for each combination of an injection molding machine and a mold. For all the feature data sets described below, the combination of the injection molding machine and the mold is fixed.

[0051] The dimensionality reduction model training unit 404 converts a dimension of the feature of the feature data set stored in the feature database into a lower-dimensional vector using a dimensionality reduction model. The dimension after the conversion can be set to any dimension by a user of the training and optimization system. Dimensionality reduction is an unsupervised learning technique in which information about an input vector is extracted into a vector that has a lower dimension than the input vector while preserving the information as much as possible. Examples of dimensionality reduction models include principal component analysis, autoencoders, and UMAP.

[0052] In general, converting a multidimensional vector into a lower-dimensional vector can improve data interpretation or generalization performance during regression model generation. By training the dimensionality reduction model using the feature data set of the feature database 403, a trained dimensionality reduction model is generated. When the feature data recorded in the feature database is input to the trained dimensionality reduction model, a vector converted to a user-specified rank is output.

[0053] The trained dimensionality reduction model storage unit 405 stores the dimensionality reduction model generated by the dimensionality reduction model training unit 404 in the trained dimensionality reduction model buffer unit 406.

[0054] The process data recording unit 407 is described. The process data recording unit 407 stores the quality of the molded part formed by the injection molding process 54 or the injection molding process 55, the molding condition input to the injection molding machine, the specified material unit, and the combination of the injection molding machine and the mold in the process database 408.

[0055] According to the present embodiment, a data group stored in the process database is referred to as a process data set. Similar to the feature data set, the process data set is also stored for each combination of the injection molding machine and the mold. The process data set described later assumes that the combination of the injection molding machine and the mold is fixed.

[0056] The dimensionality reduction model reading unit 409 reads the trained dimensionality reduction model stored in the trained dimensionality reduction model buffering unit 406. The read dimensionality reduction model is output to the dimensionality reduction execution unit 410.

[0057] The dimensionality reduction execution unit 410 reduces the dimension of the feature database using the trained dimensionality reduction model read from the dimensionality reduction model reading unit 409 with respect to the feature data set read from the feature database 403. Details of a dimensionality reduction result will be described later with reference to Fig. 9 to 11. The data set after dimension reduction is output to the connection unit 411.

[0058] With respect to the process data set in which the feature data set after the dimensional reduction are associated with the material information output from the dimensional reduction execution unit 410, the material information obtained from the process database 408, the molding condition, and the quality of the molded part are associated with each other, the connection unit 411 generates a data set in which the molding condition, the quality of the molded part, and the vector after the dimensional reduction correspond to each other (hereinafter, the data set is referred to as a training data set) using the material information as a combination key, and stores the training data set in the training database 412.

[0059] Here, the connection unit 411 does not necessarily use the feature data set after dimensional reduction, and may combine the feature data set related to the material information with the process data set from the feature database 403 instead of the feature data set after dimensional reduction to generate the training data set, and store the training data set in the training database 412. In the following description of the present embodiment, a case where the training data set is generated using the vector after dimensional reduction will be described.

[0060] The regression model training unit 413 obtains the training data set from the training database 412, uses a regression model with a vector and a molding condition after dimensionality reduction as the explanatory variable and with the quality of the molded part as the target variable, trains the regression model to predict the target variable based on the explanatory variable, and generates the trained regression model.

[0061] In general, a regression model refers to a model (y = f(X)) for predicting a target variable (y) based on an explanatory variable (X). Model parameters are determined by training data. In the description of the present embodiment, the term "regression model" refers to a general regression model, and the term "trained regression model" refers to a regression model in which the model parameters are determined by the training data.

[0062] According to the present embodiment, a regression model such as linear regression, ridge regression, support vector machine, neural network, and random forest regression, or a regression model that is a combination thereof, may be used. If the regression model to be used can have one or more target variables, such as in the neural network, one or more properties of the molded part may be selected as the target variable. The specific training of the regression model will be described later with reference to Fig. 9 to 13 described.

[0063] The regression model storage unit 414 stores the trained regression model generated by the regression model training unit 413 in the trained regression model buffer unit 415. The regression model reading unit 416 receives the trained regression model from the trained regression model buffer unit 415 and inputs it to the optimal condition generation unit 417.

[0064] The optimal condition generation unit 417 generates the molding condition (injection molding condition) for achieving the required quality. The optimal condition generation unit 417 generates an optimal molding condition for achieving the required quality by obtaining the required quality of the molded part from the manufacturing condition determination unit 31, obtaining the trained regression model from the regression model reading unit 416, and obtaining the feature data after dimensional reduction for the second material from the joining unit 411. The generation of the optimal molding condition will be described later with reference to Fig. 14 to 16 described.

[0065] Fig. 2 shows a configuration example of a computer 10 that can be used to implement the injection molding condition generation system 1. Here, a case where the injection molding condition generation system 1 is implemented by a single computer 10 will be described, but the invention is not limited to this, and one or more injection molding condition generation systems 1 may be formed by linking multiple computers. As described above, the manufacturing treatment system 2, the manufacturing execution system 3, and the manufacturing factory 5 can implement the injection molding condition generation system 1 by executing some or all of the functions by the operator without using dedicated software or hardware.

[0066] The computer 10 includes, for example, a computing device 11, a memory 12, a storage device 13, an input device 14, an output device 15, a communication device 16, and a media interface unit 17. The devices 11 to 17 are connected by a communication path CN1. The communication path CN1 is, for example, an internal bus, a local area network (LAN), or the like.

[0067] The computing device 11 includes, for example, a microprocessor. The computing device 11 is not limited to a microprocessor and may be, for example, a digital signal processor (DSP) or a graphics processing unit (GPU). The computing device 11 implements the functions 21, 31 to 35, 401 to 417, 51, 52, and 60 as the injection molding condition generation system 1 by reading a computer program stored in the storage device 13 into the memory 12 and executing the computer program.

[0068] The storage device 13 stores the computer program and data and comprises, for example, a rewritable storage medium such as a flash memory or a hard disk. The storage device 13 stores a computer program for implementing a graphical user interface (GUI) unit 60, which provides a GUI to the operator, and the computer program for implementing the above functions 21, 31 to 35, 401 to 417, 51, and 52.

[0069] The operator inputs information into the computer 10 through the input device 14. Examples of the input device 14 include a voice command device and a pointing device such as a keyboard, a touch pad, and a mouse (all not shown). The computer 10 outputs the information through the output device 15. Examples of the output device 15 include a display, a printer, and a speech synthesis device (all not shown).

[0070] The communication device 16 causes an external information processing device to communicate with the computer 10 via a communication path CN2. Examples of the external information processing device (not shown) include an external storage device 19 in addition to a computer. The computer 10 can read data (information for the computer, a production result, and the like) and a computer program stored in the external storage device 19. The computer 10 can also transfer all or part of the computer program and the data stored in the storage device 13 to the external storage device 19 and store them therein.

[0071] The media interface unit 17 is used for reading from and writing to an external recording medium 18. Examples of the external recording medium 18 include a universal serial bus (USB) memory, a memory card, and a hard disk. All or part of the computer program and data from the external recording medium 18 stored in the storage device 13 can be transferred to or stored in the external recording medium 18.

[0072] Fig. 3 is a conceptual diagram of the injection molding machine 50. Processes of an injection molding process are described with reference to Fig. 3. In the present embodiment, a molding phenomenon refers to a series of phenomena occurring during the injection molding process. In the present embodiment, the injection molding process is roughly divided into a measuring and softening process, an injection and pressure-holding process, a cooling process, and a discharging process.

[0073] In the measurement and plasticization process, a screw 502 is retracted using a plasticization motor 501 to provide driving force, and resin pellets 504 are fed from a hopper 503 into a cylinder 505. A resin is softened into a uniform molten state by heating it with a heater 506 and rotating the screw 502. By setting the back pressure and 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 part.

[0074] In the injection and pressure-holding process, the screw 502 is advanced using an injection motor 507 to provide a driving force, and the molten resin is injected into a mold 509 via a nozzle 508. Cooling from the wall surface of the mold 509 and shear heating by a flow act in parallel on the molten resin injected into the mold 509. That is, the molten resin flows in the mold 509 while undergoing a cooling action and a heating action. If the mold clamping force, which is the force for closing the mold 509, is small, a small mold opening occurs after the molten resin solidifies, and the quality of the molded part is affected by the small gap.

[0075] During the cooling process, the molten resin is cooled to a solidification temperature or below by the mold 509, which is maintained at a constant temperature. Residual stresses generated during the cooling process affect the quality of the molded part. These residual stresses are caused by anisotropy of a material property caused by the flow in the mold, the density distribution due to pressure retention, and unevenness in a mold shrinkage ratio.

[0076] In the removal process, the mold 509 is opened by driving a mold clamping mechanism 512 using a motor 511 to provide a driving force to open and close the mold 509. In a case where the solidified molded article is removed from the mold 509 by driving an ejection mechanism 514 using an ejection motor 513 to provide a driving force, if a sufficient ejection force is not uniformly applied to the molded article, residual stresses remain in the molded article, thereby affecting the quality of the molded article.

[0077] In the injection molding machine 50, pressure control is performed so that a pressure value induced by a load cell 510 approaches a pressure value in an input molding condition. The temperature of the cylinder 505 is controlled by a plurality of heaters 506. The shape of the screw 502, the shape of the cylinder 505, and the shape of the nozzle 508 cause a different pressure loss for each injection molding machine. Accordingly, the pressure at a resin inlet port of the mold 509 is lower than the pressure specified by the molding condition input to the injection molding machine. Furthermore, the resin temperature in the resin inlet port of the mold 509 may differ from the resin temperature specified by the molding condition input to the injection molding machine due to the arrangement of the heaters 506 and the shear heating of the resin in a nozzle section.

[0078] The configuration of an injection mechanism (shape of screw 502, shape of cylinder 505, shape of nozzle 508, arrangement of heaters 506, and the like) may be different for each injection molding machine, resulting in machine difference and affecting the quality of the molded part.

[0079] The quality of the molded part is assessed based on molding properties (weight, length, thickness, shrinkage, burr, warpage, and the like), surface properties such as appearance defects (welding, silver, burning, whitening, scratching, air bubbles, peeling, flow marks, splashing, color, gloss, and the like), and mechanical and optical properties (tensile strength, impact resistance, light transmittance, and the like).

[0080] Mold properties have a strong correlation between the mold clamping force and the pressure and temperature history during the injection, pressure-holding, and cooling processes. Regarding surface properties, the factors responsible for a given phenomenon are different. For example, flow marks and jetting have a strong correlation with the temperature and velocity of the resin during the injection process. In the case of tensile strength, for example, an assessment in a fracture test is required, so mechanical and optical properties are often assessed by other quality indicators that correlate with weight or the like.

[0081] For each part of the injection molding process, a corresponding parameter is set in the molding conditions. For the measuring and softening process, the measuring position, a back suction, a back pressure, a back pressure speed, a rotation speed, and the like are set. For the injection and pressure-holding process, the pressure, temperature, time, and speed are set, respectively. For the injection and pressure-holding process, the screw position (VP switching position) at which the injection and pressure are switched, and also the clamping force of the mold 509 are set. For the cooling process, the cooling time after pressure-holding is set. As a temperature-related parameter, the temperature of the plurality of heaters 506, the temperature and flow rate of a coolant for cooling the mold 509, and the like are set.

[0082] Fig. 4 shows an outline of an experiment used in a test for verifying the optimization of the molding condition according to the present embodiment. Fig. 4 shows a top view 70 of a product section, a side view 71 of the product section, and a top view 72 of a runner section. The mold is configured so that the resin flows from the runner section to the product section in a 5-point pin gate manner. In an actual molding experiment, a pressure sensor and a resin sensor (all not shown) were arranged in a sensor arrangement section 73 of a runner, and temporal changes were detected. Polypropylene (PP) was used as the molding material. An electric injection molding machine with a maximum mold clamping force of 150 t and a screw diameter of 44 mm was used as the injection molding machine.

[0083] Fig. 5 is a block diagram of an example of a method for obtaining the material information by extracting features in the feature extraction unit 402 with respect to the physical quantity obtained from the sensor 58 and by storing the extraction result in the feature database 403. In Fig. 5 is acquired using a “sensor-equipped mold” or a “sensor-embedded mold,” wherein a sensor for measuring a predetermined physical quantity is provided at a predetermined position.

[0084] First, for a material 601, a physical quantity is obtained at a given part in the mold by inputting a reference molding condition 602, which is a fixed molding condition, into an actual injection molding machine 603. Here, the injection molding machine 603 corresponds to the one shown in Fig. 3. In addition, the reference condition corresponds to a condition that is input into the injection molding machine when the Fig. 1 illustrated injection molding process 55 is carried out.

[0085] The physical quantity of the specified part in the mold is influenced by information about the material to be used, a machine difference unique to the mold and injection molding machine, and the molding condition. Therefore, by setting the reference condition 602 for each combination of the mold and injection molding machine, the influence of the machine difference and the molding condition can be prevented, and the material information unique to the material can be stored as a feature of the physical quantity in the feature database 610. This means that the reference condition can be changed for each combination of a mold and injection molding machine.

[0086] An in-mold sensor 606 is used to detect a molding phenomenon in the actual injection molding machine 603. By disposing the in-mold sensor 606 at a position in a mold 604, the molding phenomenon in the mold 604 can be directly measured, and an actual measured value 608 of the physical quantity correlated with the material information can be acquired. The quality of the molded part 605 can be obtained through product quality inspection 607.

[0087] The feature is extracted from the obtained physical quantity (609). All obtained physical quantities are detected as temporal changes in the injection molding process and are therefore difficult to directly evaluate. Therefore, according to the present embodiment, quantitative evaluation of the material information is performed by obtaining the feature correlated with the material information from the temporal change of the physical quantity. By performing molding under a common reference condition between the materials, the material information for the respective materials can be compared by comparing the feature.

[0088] A measurement result of an experimental example for checking the optimization of the forming condition as described with reference to Fig. 4, is described with reference to the Fig. 6 and Fig. 7 described. Fig. Figure 6 shows time series data of the pressure sensor and the resin sensor in the sensor arrangement section 73 of the runner when injection molding is carried out under the reference condition for each of three PP material lots P, Q, and R with different feeding times. As shown in Fig. As shown in Figure 6, even under the same forming condition, the variation of the time series data is different for each lot, and it can be confirmed that the time series data of the pressure sensor is affected by the material information unique to the material.

[0089] Fig. 7 shows an example of a result where a peak value, a maximum difference value, an integral value up to the peak value and an integral value from the peak value are obtained as features from the pressure sensor and the resin temperature sensor for each in Fig. 6 shown material lot can be extracted (with regard to the resin temperature sensor, a numerical value is omitted). Because, as in Fig. 7, the feature is differentiated between batches, it can be confirmed that the extracted feature is influenced by the material information. As described above, the feature extraction is performed for each batch, and the data set related to the feature for each batch is referred to as a feature data set and recorded in the feature database 403. According to the present embodiment, as shown in Fig. 7, a feature data set with a total of eight dimensions is used, of which four dimensions are used each for the pressure sensor and the resin temperature sensor.

[0090] A part of the mold where the physical quantity is measured, a parameter of the physical quantity correlated with the resin information, and the feature are described below.

[0091] First, a part in the mold where the physical quantity is measured (hereinafter referred to as the measuring part) will be described. It is preferable for each mold structure that the measuring part includes at least a gate portion or the runner portion extending from the resin inlet port in the mold to the inside of a cavity.

[0092] The measuring part can be located inside the cavity, but the loss of each physical quantity from the resin inlet to the cavity must be considered when deriving the material information unique to the material through the above procedure. Therefore, analysis accuracy must be ensured from the resin inlet to the inside of the cavity. In addition, if the sensor is located inside the cavity to measure the physical quantity, a trace may remain in the molded part due to the shape of the sensor. This creates the limitation that the sensor cannot be installed in a location where an aesthetic appearance is required.

[0093] Therefore, according to the present embodiment, the physical quantity correlated with the material information unique to the material can be obtained easily and with high accuracy by using the gate portion or the runner portion, which is located near the resin inflow port and where no aesthetic appearance is required, as the measuring part.

[0094] In addition to the gate section and the runner section, the measurement part can be a part where a property profile can be observed, such as a section immediately below a gate in the cavity, a resin joining section (a welding section), and a flow connection section. In this case, the physical quantity correlated with the material information unique to the material can be obtained with higher accuracy using physical quantities obtained by multiple sensors.

[0095] Because the flow rate of the molten resin can be determined, for example, based on the transit time of a flow front across multiple measurement parts, material information can be derived from the flow rate of the molten resin. Furthermore, the viscosity of the molten resin in the mold can be estimated by measuring the pressure and temperature at that time.

[0096] A suitable measuring part depends, for example, on the mold structure and the physical quantity to be measured. For each mold structure, it is preferable, if possible, to specify the gate portion as the measuring part for a physical quantity different from the mold opening amount. In this specification, the term "preferred" is used merely in the sense that a beneficial effect is expected, and is not intended to imply that the configuration is essential.

[0097] In a side gate, a jump gate, a submarine gate, and a banana gate, a sensor is arranged at a runner section immediately below the gate section, a runner section immediately before the gate, or the like. In a pin gate, arranging the sensors takes time because it has a three-plate structure, and the sensors are arranged at the runner section or the like immediately below the gate section. In a pin gate, the measuring part is a blind runner that is not connected to the cavity and is used for measurement. By providing a dedicated part for measurement, the degree of freedom for mold design is improved. In the pin gate, the blind runner that is not connected to the cavity can be provided for measurement and used as the measuring part.By providing a dedicated measurement section, the degree of freedom for mold design is improved. In a film gate or fan gate, a sensor is positioned at the runner section before the flow enters a gate section.

[0098] Parameters to be measured as the aforementioned physical quantity will be described. According to the present embodiment, in order to optimize the molding condition according to the material information, at least the pressure and temperature are measured. For measuring the pressure and temperature, for example, an in-mold pressure sensor, a mold surface temperature sensor, a resin temperature sensor, and the like can be used. As the resin temperature sensor, either a contact temperature sensor such as a thermocouple or a non-contact temperature sensor such as an infrared radiation thermometer, or both, can be used. For both the physical quantities of pressure and temperature, the temporal change during the injection molding process is also stored.

[0099] The injection molding condition optimization system 1 can detect the flow front velocity and flow front passage time in addition to the mold opening amount, temperature, and pressure. From the sensor that detects the flow front velocity and flow front passage time, information about the flow front passage time can be obtained instead of the time change in the injection molding process. When the flow front passage time is to be detected, at least two sensors are provided, and the resin passage times between two points are compared. The injection speed can be more accurately judged by detecting the flow front velocity and passage time.

[0100] The characteristic of the aforementioned physical quantity will be described. According to the present embodiment, for example, the maximum value of pressure (peak value of the temporal change), the integral value, and the maximum value of temperature (peak value of the temporal change) can be used. The maximum value of the temporal change of pressure can also be effectively detected. The maximum value of the time derivative of pressure is correlated with the instantaneous viscosity of the material. The integral value of pressure can be calculated separately in the injection process and the pressure-holding process. The integral value of pressure in the injection process correlates with the average viscosity of the material in the injection process.

[0101] When an infrared radiation resin temperature sensor is used, the maximum value of the time derivative can be detected with respect to an output value of the time change of the temperature sensor during the injection process. This feature correlates with the flow front velocity of the molten resin. When the flow front velocity is detected, it is used as it is as a flow rate-correlated feature. When the flow front transit time is detected, the flow velocity is calculated based on the transit time between two points and used as a feature. The injection velocity can be recorded more accurately by storing a relationship between the flow velocity and a setpoint injection velocity (a fixed value independent of the material under the reference condition).

[0102] The process data record stored in the process database 408 is stored with reference to Fig. 8. Data is stored in the process database 408, with the quality of the molded part formed by the injection molding process 54 or the injection molding process 55, the molding condition input to the injection molding machine, the specified material unit, and the combination of the injection molding machine and the mold being associated with each other. The process data set is stored for each combination of an injection molding machine and a mold.

[0103] In the present embodiment, the above-mentioned material lots P, Q, and R were used, the weight of the molded part was set as its quality, and the mold clamping pressure, mold temperature, molding nozzle temperature, injection speed, holding pressure, VP switching position, and cooling time were set as the molding condition parameters. Although parameters other than those described above may be added to the molding condition parameters, or several of the above-mentioned parameters may be omitted, the molding condition parameters preferably include at least one parameter correlated with the quality of the molded part (the weight of the molded part in the present embodiment).

[0104] Fig. Figure 8 shows an example of the process database stored for each combination of the injection molding machine and the mold according to the present embodiment. The molding condition, the material lot, and the weight of the molded part are associated with each other. In particular, #1 in Fig. 8, that as the molding condition parameter, the mold clamping pressure was set to 120 [t], the mold temperature was set to 30 [°C], the molding nozzle temperature was set to 180 [°C], the injection speed was set to 40 [mm / s], the holding pressure was set to 30 [kg / m 2] was set, the VP switching position was set to 10 [mm], and the cooling time was set to 35 [s], and the average value of the weight of the molded part when an injection molding process was performed multiple times using the material lot P was 62.83 [g]. Here, the molding condition and the material lots to be changed during trial molding can be changed based on, for example, an experimental design or determined based on the result obtained by CAE simulation.

[0105] A method for optimizing the forming condition in the training and optimization system 4 is described with reference to the Fig. 9 to 16. The training and optimization system 4 has two functions, including an optimization mode and a training mode. In the optimization mode, the optimal forming condition is generated that meets the required quality for the second material. In contrast, in the training mode, the trained dimensionality reduction model and the trained regression model are generated for use in the optimization mode.

[0106] The training mode 16 is used with reference to the Fig. 9 to 13 described. Fig. Figure 9 shows a flowchart of the training mode. When the training mode is initiated, the training and optimization system 4 first reads a feature data set ( Fig. 7 shows an example of the feature data set) from the feature database 403 and performs standardization (S101).

[0107] Standardization involves calculating the mean and standard deviation of a data group for each column of a target data set, subtracting the mean from the numerical value of the data group, and dividing the mean by the standard deviation. Because standardization makes the mean of the data group for each column 0 and the standard deviation 1, the influence caused by unit differences between columns can be eliminated. Generally, standardization improves the training accuracy of dimensionality reduction models and regression models.

[0108] Next, the training and optimization system 4 trains the dimensionality reduction model with respect to the feature data set on which standardization is performed (S102). According to the present embodiment, the eight-dimensional feature data set is reduced to two dimensions using UMAP as the dimensionality reduction model.

[0109] UMAP is a nonlinear dimensionality reduction model that converts positional information between the input data into data with a lower dimension than the input data while preserving the positional information. Other well-known models related to dimensionality reduction include principal component analysis and autoencoders, which can be used as dimensionality reduction models for the feature dataset.

[0110] Fig. 10 is a graph obtained by converting the respective material lots P, Q and R into a two-dimensional vector Z (Z1, Z2) using UMAP as a dimensional reduction model with respect to a feature data set obtained by molding the respective material lots P, Q and R through 40 injection molding processes under the reference conditions in the injection molding process 55 (rank: eight, number of samples: 120) and plotting it on a two-dimensional plane.

[0111] Everyone in Fig. The point shown in Figure 10 corresponds to an injection molding process, and the shapes of the points depend on the material batch. Fig. 10, it can be confirmed that the vector Z after dimensional reduction can express differences in material information between batches because the distance between points of the same material batch is smaller and the distance between points of different material batches is larger. That is, the trained dimensional reduction model generates two-dimensional data correlated with the material information based on the eight-dimensional input data. According to the present embodiment, the vector after dimensional reduction is set to be two-dimensional, but its dimension can be set to any dimension higher than one as long as it is lower than the input rank.

[0112] It is on Fig. 9. The training and optimization system 4 stores the trained dimensionality reduction model and the information related to standardization in the trained dimensionality reduction model buffer unit 406 (S103). The information related to standardization indicates the mean value and standard deviation of each column of the feature data set.

[0113] As in Fig. 11, the training and optimization system 4 calculates, for each batch, an average value of each dimension of the vectors with respect to the vector Z after the dimensionality reduction obtained by the dimensionality reduction model, associates the average value with the batch, and generates the association as a feature data set after the dimensionality reduction (S104).

[0114] The training and optimization system 4 generates a training data set by reading the feature data set after dimensional reduction and the process data set stored in the process database 408 and combining the feature data set after dimensional reduction with the process data set stored in the process database 408 using the lot as a combination key, and stores the training data set in the training database 412 (S105).

[0115] Fig. 12 shows an example of the combination of the Fig. 8 process data set with the one in Fig. 11 shown feature data set generated after dimension reduction. Fig. 12 can be confirmed that a column of the material lot in Fig. 8 is replaced by the feature Z (Z1, Z2) after the dimension reduction.

[0116] If the regression model is generated based on the training dataset, there is a high probability that the training cost will increase and the accuracy will become unstable if the rank of the training dataset is unnecessarily large. Therefore, it is preferable to generate the training dataset by combining the process dataset with the dataset after dimensionality reduction. However, the training dataset can be generated by combining the process dataset and the feature dataset that has not undergone dimensionality reduction.

[0117] The training and optimization system 4 performs the standardization for each column of the data set as performed in step S101 with respect to the training data set (S106).

[0118] The training and optimization system 4 uses the data set after standardization to generate a regression model by machine learning with the weight as the target variable (output parameter) and with the mold clamping pressure, mold temperature, molding nozzle temperature, injection speed, holding pressure, VP switching position, cooling time, Z1 and Z2 as explanatory variables (input parameters) (S107).

[0119] According to the present embodiment, the trained regression model is generated using support vector regression, and a regression model such as linear regression, ridge regression, support vector machine, neural network, and random forest regression, or a regression model that is a combination thereof may be used.

[0120] Fig. 13 shows a performance evaluation result of the trained regression model generated by support vector regression according to the present embodiment. The performance evaluation is performed to confirm the effectiveness of the trained regression model generated according to the present embodiment and is not necessarily performed to run the injection molding condition optimization system 1. The performance evaluation method will be described below.

[0121] The training dataset, which is a Fig. 12, is divided into a first and a second data set by performing random extraction. In the performance evaluation, the trained regression model of support vector regression is generated using the first data set. In the performance evaluation, a predictive value of the weight of the molded part is calculated by inputting the explanatory variables of the first data set and the second data set into the trained regression model.

[0122] Fig. Figure 13 is a graph showing the relationship between the measured weight and the predicted weight of the molded part. The horizontal axis indicates the weight of the molded part measured by the molded part quality inspection unit 57. The vertical axis indicates the predicted weight of the molded part based on the trained regression model. A dotted straight line indicates a reference line when the measured weight matches the predicted weight. A point closer to the reference line has a higher prediction accuracy.

[0123] Generally, a coefficient of determination is used as a representative index to quantitatively assess the predictive accuracy of the regression model. A coefficient of determination closer to 1 indicates a higher predictive accuracy of the regression model. Fig. 13, it can be confirmed that the determination coefficients of the first data set and the second data set are high and at least 0.9. In addition, the deviation between the determination coefficient with respect to the second data set, which is not used for training the regression model, and the determination coefficient with respect to the first data set used for training the regression model is small. Therefore, it can be confirmed that the generalization performance of the trained regression model is also good. From the evaluation, it can be confirmed that a favorable model for predicting the weight of the molded part can be obtained by using the training data set including the molding condition obtained by trial molding in the injection molding process 54 and the feature Z after dimensional reduction correlated with the material information.

[0124] It is on Fig. 9. The training and optimization system 4 stores the trained regression model and the information related to the standardization in the trained regression model buffer unit 415 (S108) and terminates the training mode.

[0125] The optimization mode in the training and optimization system 4 is determined with reference to the Fig. 14 to 16. In the optimization mode, the optimal forming condition for the second material to meet the required quality is generated based on the dimensionality reduction model and the trained regression model generated by the training mode, the feature data set of the second model, and an optimization algorithm.

[0126] Fig. Figure 14 shows a flowchart of the optimization mode. In the optimization mode, the target weight of the molded part is read from the manufacturing condition determination unit 31 as information on the required quality (S201).

[0127] The training and optimization system 4 uses the second material to perform molding under the reference condition in the injection molding process 55 for a predetermined number of injection molding operations (S202). With respect to the molding result, the training and optimization system 4 generates the feature data set of the second material through the feature extraction unit 402 (S203).

[0128] The training and optimization system 4 reads from the trained dimensionality reduction model buffer unit 406 the trained dimensionality reduction model and standardization information about the feature database generated in the training mode (S204).

[0129] The training and optimization system 4 performs standardization on the feature data set of the second material based on the standardization information in the feature database and dimensionality reduction based on the trained dimensionality reduction model, and generates the feature after dimensionality reduction of the second material (S205). In the example of the embodiment, the feature after dimensionality reduction of the second material is (Z1, Z2) = (0.38, -4.0).

[0130] The training and optimization system 4 reads the trained regression model and the standardization information via the training database from the trained regression model buffer unit 415 (S206), and optimizes the forming condition based on the target weight, the feature data set after dimensional reduction of the second material, the standardization information, the trained regression model, and a known optimization algorithm.

[0131] According to the present embodiment, an example in which a grid search is used as the optimization algorithm will be described. In addition to the grid search, a gradient method, pseudo-annealing, a genetic algorithm, and Bayesian optimization can also be used as the optimization algorithm. Furthermore, the optimization algorithms can be combined to optimize the shaping condition.

[0132] A method for optimizing the forming condition by grid search is presented with reference to the Fig. 14, steps S207 to S209 are described. First, an overview of the grid search is given.

[0133] Grid search is an optimization algorithm that specifies multiple candidate values ​​for each explanatory variable parameter, calculates the predicted value of the target variable using the trained regression model for all combinations of explanatory variables formed by combining the candidate values ​​of each parameter, and extracts the combination of explanatory variables that has the smallest deviation between the predicted value and the target value. Grid search can not only provide a global optimization solution that is more resilient to the problem of a local feature being stronger than a global feature than other general optimization algorithms, but also enables understanding the distribution of the target variable across the entire explanatory variable space.

[0134] According to the present embodiment, all combinations of the parameters of the explanatory variables of the regression model are first generated (S207). The parameters of the explanatory variables of the regression model according to the present embodiment include the parameters related to the molding condition (mold clamping pressure, mold temperature, molding nozzle temperature, injection speed, holding pressure, VP switching position, and cooling time) and the parameters related to the material information (the features Z1 and Z2 after dimensional reduction), as in the example of the model shown in Fig. 12 shown training dataset.

[0135] For the parameters related to the molding condition among the explanatory variable parameters, the maximum and minimum values ​​of each parameter are determined, and a resolution of each parameter set in the injection molding machine is set as the step size to generate the candidate value. For the parameter related to the material information, a combination of explanatory variable parameters is generated using a value of the feature data set generated in step S205. Regarding the parameters related to the molding condition, for example, the mold clamping pressure changes by 1 [t] each time, where 120 [t] is the minimum value and 150 [t] is the maximum value. Fig. Figure 15 shows an example of the combination of parameters generated according to the present embodiment. Accordingly, the optimal molding conditions when using the second material can be generated by fixing the parameters related to the material information.

[0136] The training and optimization system 4 calculates the weight of the molded part using the trained regression model and the standardization information via the training database for each combination of parameters (S208).

[0137] To quantitatively assess the deviation between the predicted value and the target value, the training and optimization system 4 generates the shaping condition closest to the target weight by calculating the absolute values ​​of the difference between the predicted value and the target value and sorting the absolute values ​​of the differences in ascending order as the optimal condition (S209).

[0138] For the quantitative judgment of the deviation between the predicted value and the target value, according to the present embodiment, the absolute values ​​of the differences between the predicted value and the target value are set, but for example, another index representing the deviation between the predicted value and the target value may be set, such as a situation value of the difference between the predicted value and the target value.

[0139] According to the present embodiment, the case where the required quality is only the weight of the molded part is described, and the invention can also be applied to a case where there are multiple required qualities such as the weight of the molded part and a dimension of the molded part. Each index representing the deviation between the target value, which is a target for the required quality, and the prediction can be set.

[0140] Fig. 16 shows, according to the present embodiment, the weight of the molded part molded under the reference condition in the injection molding process 55 using the second material and the weight distribution of the molded part molded under the optimization condition generated for the second material.

[0141] The required quality (the weight of the molded part) entered into the unit 417 to generate optimal conditions is 63.89 [g] and is in Fig. 16 is indicated by a dotted line parallel to the vertical axis. For each distribution, the number of molded parts is 40 and is normalized by the number of each molded part, and the vertical axis indicates the probability distribution.

[0142] According to Fig.16, the weight of the molded part is close to the required quality, and the validity of the generated molding condition can be confirmed. Even if the material properties of the resin vary as described above, the molding condition that meets the required quality can be optimized. List of reference symbols 1 Injection molding condition generation system 2 Manufacturing treatment system 3 Manufacturing Execution System 4 Training and optimization system 5 Manufacturing factory 31 Manufacturing condition determination unit 32 Production result storage unit 33 Production result recording unit 34 Production execution instruction unit 35 Manufacturing result recording unit 401 Sensor information recording unit 402 Feature extraction unit 403 Feature database 404 Dimension Reduction Model Training Unit 405 Trained dimensionality reduction model storage unit 406 Trained Dimensionality Reduction Model Read Unit 407 Process data recording unit 408 Process database 409 Dimension Reduction Model Reading Unit 410 Dimension Reduction Execution Unit 411 connection unit 412 training database 413 Regression Model Training Unit 414 Regression model storage unit 415 Trained Regression Model Cache Unit 416 Regression Model Reading Unit 417 Unit for generating optimal conditions 51 Manufacturing Execution Unit 52 Forming condition generation unit 57 Molded part quality inspection unit 58 Sensor

Claims

[1] Injection molding condition generation system (1) for generating an injection molding condition using a computer (10), wherein the computer (10) has a processor (11) and a memory (12, 13) used by the processor (11), the processor (11), when receiving a first material property value of a first resin material (601), generates the injection molding condition using the first resin material (601) based on a target value of a quality parameter related to the quality of a molded part (605), the first material property value, and a predetermined relational expression, and the predetermined relational expression indicates a relationship between the material property value of the resin material (601), a plurality of injection molding conditions input to an injection molding machine (50, 603), and a quality parameter relating to the quality of the molded part (605) molded by the injection molding machine (50, 603) on the basis of the material property value and the injection molding conditions in a mold (604), and is generated on the basis of data collected in the memory (12, 13) in connection with the material property value of the resin material (601), the injection molding conditions, and the quality parameter, characterized byin that the material property value for generating the predetermined relational expression is generated for each predetermined unit of the resin material (601) on the basis of a characteristic of a measured value of an in-mold sensor (58, 606) arranged in the mold (604) when molding is carried out under a common reference injection molding condition (602) between the predetermined units. [2] The injection molding condition generating system (1) according to claim 1, wherein the predetermined unit is a supply unit of a recycled material supplied from a material supplier. [3] The injection molding condition generating system (1) according to claim 1, wherein the predetermined unit is a material supply unit when the same type of virgin material is supplied from a plurality of different material suppliers. [4] The injection molding condition generation system (1) according to any one of the preceding claims, wherein the predetermined relational expression is a regression equation generated using a machine learning model, wherein the quality parameter refers to the quality of the molded part (605) when trial molding is performed on a trial resin material while changing the injection molding condition as an output parameter, and wherein a material property value of the trial resin material is an input parameter. [5] Injection molding condition generating system (1) according to claim 4, wherein the injection molding condition in the predetermined relational expression the mold clamping pressure, the injection speed, the temperature of a nozzle section of the injection molding machine (50, 603), the holding pressure, the speed-pressure control switching position and / or the mold temperature. [6] Injection molding condition generation system (1) according to claim 4 or 5, wherein the quality parameter in the predetermined relational expression the average weight of the molded part (605), the average value of a dimension of the molded part (605), the average value of the distortion amount of the molded part (605) and / or the defect rate of the molded part (605) in the specified unit. [7] Injection molding condition generating system (1) according to one of claims 4 to 6, wherein the feature comprises temperature, speed and / or pressure. [8] Injection molding condition generating system (1) according to one of claims 4 to 7, wherein the feature comprises the flowability of the resin material (601) and the flowability of the resin material (601) is calculated based on a peak value of the measured value of the in-mold sensor (58, 606), an integral value of the measured value from the start of injection to a peak, an integral value of the measured value from the start of injection to mold opening, and / or a maximum derivative value of the measured value. [9] The injection molding condition generation system (1) according to any one of claims 6 to 8, wherein the rank of the feature is reduced and the feature after the reduction is used as a material property value. [10] A method for generating an injection molding condition using a computer (10), wherein the computer (10) obtains a first material property value of a first resin material (601), generates the injection molding condition using the first resin material (601) on the basis of a target value of a quality parameter related to the quality of a molded part (605), the first material property value, and a predetermined relational expression, and outputs the generated injection molding condition, the predetermined relational expression indicates a relationship between the material property value of the resin material (601), a plurality of injection molding conditions input to an injection molding machine (50, 603), and a quality parameter related to the quality of the molded part (605) molded by the injection molding machine (50, 603) on the basis of the material property value and the injection molding conditions in a mold (604), and is generated on the basis of data collected in connection with the material property value of the resin material (601), the injection molding conditions, and the quality parameter, and the material property value for generating the predetermined relational expression is generated for each predetermined unit of the resin material (601) based on a characteristic of a measured value of an in-mold sensor (58, 606) arranged in the mold (604) when molding is carried out under a common reference molding condition (602) between the predetermined units.

Images