Physical property prediction method, physical property prediction program, physical property prediction device, and manufacturing method of films

Abstract

To provide a physical property prediction method capable of effectively predicting physical properties, when setting an optimum manufacturing condition; and to provide a physical property prediction program, a physical property prediction device and a manufacturing method of films thereof.SOLUTION: A physical property prediction method according to the present invention comprises: an acquisition step of acquiring step data obtained for each of a plurality of processing steps; a prediction step of calculating a physical property predicted by the step data acquired by the acquisition step using a learned model by reading, from a storage unit the learned model obtained by learning the step data according to the plurality of processing steps as explanatory variables, and properties of a film manufactured by the plurality of processing steps as objective variables; an influence degree calculation step of calculating each influence degree of a plurality of explanatory variables in the learned model with respect to the properties calculated in the prediction step; a first extraction step of extracting the number of preset explanatory variables in order of high influence degree among the plurality of explanatory variables; and a display step of displaying the explanatory variables extracted by the first extraction step.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to a physical property prediction method, a physical property prediction program, a physical property prediction device, and a film manufacturing method. [Background technology]

[0002] Films are used in a variety of products because they have excellent heat resistance, flame retardancy, chemical resistance, electrical insulation, moist heat resistance, mechanical strength, dimensional stability, etc. Such films are produced, for example, by biaxial stretching, in which resin chips are melted, stretched in one direction, and then stretched in a direction perpendicular to the one direction (see, for example, Patent Document 1). [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Publication No. 2023-24179 Summary of the Invention [Problem to be solved by the invention]

[0004] Incidentally, in Patent Document 1, optimal manufacturing conditions are found by simulating the stretching (physical properties) of a stretched film using material characteristics and the finite element method. However, the method described in Patent Document 1 requires simulation for each material and manufacturing condition to predict physical properties, and it takes time to set optimal manufacturing conditions.

[0005] The present invention has been made in consideration of the above, and aims to provide a physical property prediction method, a physical property prediction program, a physical property prediction device, and a film manufacturing method that can efficiently predict physical properties when setting optimal manufacturing conditions. [Means for solving the problem]

[0006] In order to solve the above-mentioned problems and achieve the object, the physical property prediction method of the present invention includes an acquisition step of acquiring process data obtained for each of multiple processing steps into which a film manufacturing process is subdivided; a prediction step of reading from a memory unit a trained model that has been trained using the process data for the multiple processing steps as explanatory variables and the physical properties of the film manufactured by the multiple processing steps as objective variables, and using the trained model to calculate predicted physical properties based on the process data acquired by the acquisition step; an influence calculation step of calculating the influence of each of the multiple explanatory variables in the trained model on the physical properties calculated in the prediction step; a first extraction step of extracting a predetermined number of explanatory variables from the multiple explanatory variables in order of decreasing influence; and a display step of displaying the explanatory variables extracted in the first extraction step.

[0007] In addition, in the above invention, the physical property prediction method according to the present invention further includes a second extraction step of extracting explanatory variables from the explanatory variables extracted in the first extraction step based on the number of occurrences in past performance, an anomaly calculation step of calculating the degree of anomaly of the explanatory variables extracted in the second extraction step, and an output step of outputting notification information based on the degree of anomaly calculated in the anomaly calculation step.

[0008] Moreover, in the above-described physical property prediction method according to the present invention, the method further includes a labeling step of assigning a label indicating the presence or absence of a standard to the physical property calculated in the prediction step.

[0009] In the property prediction method according to the present invention, in the above invention, the labeling step not only determines whether a standard exists, but also, if a standard exists, assigns a label indicating whether the standard is within a standard range.

[0010] In the physical property prediction method according to the present invention, in the above invention, the labeling step assigns a label indicating whether or not the result is within a control range when the result is within a specification range.

[0011] In the physical property prediction method according to the present invention, in the above invention, the degree of abnormality calculation step uses the reciprocal of the number of occurrences as the degree of abnormality.

[0012] In addition, the physical property prediction program of the present invention causes a computer to execute the following steps: an acquisition step of acquiring process data obtained for each of multiple processing steps into which a film manufacturing process is subdivided; a prediction step of reading from a memory unit a trained model that has been trained using the process data for the multiple processing steps as explanatory variables and the physical properties of the film manufactured by the multiple processing steps as objective variables, and using the trained model to calculate predicted physical properties based on the process data acquired by the acquisition step; an influence calculation step of calculating the influence of each of the multiple explanatory variables in the trained model on the physical properties calculated in the prediction step; a first extraction step of extracting a predetermined number of explanatory variables from the multiple explanatory variables in order of decreasing influence; and a display step of displaying the explanatory variables extracted in the first extraction step.

[0013] In addition, the physical property prediction device of the present invention includes an acquisition unit that acquires process data obtained for each of a plurality of processing steps that are obtained by subdividing the film manufacturing process; a prediction unit that calculates predicted physical properties based on the process data acquired by the acquisition unit using a trained model that has been trained using the process data for the plurality of processing steps as explanatory variables and the physical properties of the film manufactured by the plurality of processing steps as objective variables; an influence calculation unit that calculates the influence of each of the plurality of explanatory variables in the trained model on the physical properties calculated by the prediction unit; a first extraction unit that extracts a predetermined number of explanatory variables from the plurality of explanatory variables in order of decreasing influence; and a display unit that displays the explanatory variables extracted by the first extraction unit.

[0014] Furthermore, the film manufacturing method of the present invention sets conditions for each processing step based on the physical properties predicted by the physical property prediction method of the present invention, and produces a film according to the set conditions. [Effects of the Invention]

[0015] According to the present invention, physical properties can be efficiently predicted in order to set optimal manufacturing conditions. [Brief explanation of the drawings]

[0016] [Figure 1] FIG. 1 is a diagram showing a schematic configuration of a property prediction system according to an embodiment of the present invention. [Figure 2] FIG. 2 is a diagram for explaining the flow of the property prediction process performed by the property prediction system according to one embodiment of the present invention. [Figure 3] FIG. 3 is a diagram for explaining an example of a manufacturing method using biaxial stretching. [Figure 4] FIG. 4 is a block diagram showing the configuration of a learning device included in a property prediction system according to an embodiment of the present invention. [Figure 5] FIG. 5 is a block diagram showing the configuration of a property prediction device included in a property prediction system according to an embodiment of the present invention. [Figure 6] FIG. 6 is a flowchart showing an outline of the property prediction process performed by the property prediction device according to one embodiment of the present invention. [Figure 7] FIG. 7 is a diagram showing an example of a display of the results of the property prediction process. [Figure 8] FIG. 8 is a block diagram showing the configuration of a property prediction device according to the first modification of the present invention. [Figure 9] FIG. 9 is a flowchart showing an outline of the property prediction process performed by the property prediction device according to the first modification of the present invention. [Figure 10] FIG. 10 is a block diagram showing the configuration of a property prediction device according to the second modification of the present invention. [Figure 11] FIG. 11 is a flowchart showing an outline of the property prediction process performed by the property prediction device according to the second modification of the present invention. [Figure 12] FIG. 12 is a flowchart showing an outline of the abnormality degree calculation process. [Figure 13] FIG. 13 is a block diagram showing the configuration of a property prediction device according to the third modification of the present invention. [Figure 14] FIG. 14 is a flowchart showing an outline of the property prediction process performed by the property prediction device according to the third modification of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0017] Hereinafter, an embodiment of a physical property prediction system according to the present invention will be described in detail with reference to the drawings, although the present invention is not limited to this embodiment.

[0018] (Embodiment) 1 is a diagram showing a schematic configuration of a physical property prediction system according to one embodiment of the present invention. The physical property prediction system 1 includes a learning device 2 that creates learning data and generates a trained model trained using the created learning data, a physical property prediction device 3 that sets blending conditions that will give the user desired physical properties to the resin film using the trained model generated by the learning device 2, a display device 4 that displays information including the setting results of the physical property prediction device 3, and an input device 5.

[0019] The physical properties predicted by the physical property prediction device 3 include mechanical strength, electrical properties, thermal properties, etc., predicted based on data obtained during each processing step in the film manufacturing process.

[0020] The film is, for example, a resin film formed using a resin. Examples of the resin constituting the resin film include thermoplastic resins. Examples of the thermoplastic resin include polyester resins.

[0021] 2 is a diagram illustrating the flow of a property prediction process performed by a property prediction system according to an embodiment of the present invention. A learning device 2 generates a trained model 100 by learning using training data LD. The physical property prediction device 3 inputs data (process data IP) obtained in the film manufacturing process into the trained model 100 to obtain physical property values (predicted values).

[0022] Here, an example of a method for producing a resin film will be described with reference to Fig. 3. Fig. 3 is a diagram for explaining an example of a production method using biaxial stretching. In Fig. 3, the longitudinal direction of the film along which the resin is fed in the treatment step is defined as the X-axis direction, the width direction of the resin film (the same as the rotation axis of the roll that feeds it to the next step) is defined as the Y-axis direction, and the height direction perpendicular to the X-axis and Y-axis directions is defined as the Z-axis direction.

[0023] The resin film manufacturing apparatus 200 includes a manufacturing unit 201 and a control unit 202. The control unit 202 is also electrically connected to the physical property prediction device 3. Furthermore, the physical property prediction device 3 and the manufacturing device 200 constitute a resin film manufacturing system. Note that the manufacturing device 200 is not limited to a manufacturing unit that manufactures a resin film by biaxial stretching, such as the manufacturing unit 201 shown in FIG.

[0024] The manufacturing section 201 includes a vacuum drying section 211 that removes moisture from the pellets (resin), a storage section 212 that stores the pellets after vacuum drying, an extrusion section 213 that melts and extrudes the pellets sent from the storage section 212, a filtration filter 214 that filters the molten resin sent from the extrusion section 213, a die 215 that forms the filtered molten resin into a sheet, a casting drum 216 that wraps the sheet-like resin from the die 215 around and cools and solidifies it, a first stretching section 217 that stretches the sheet-like resin in the X-axis direction, a second stretching section 218 that stretches in the Y-axis direction perpendicular to the stretching direction by the first stretching section 217, a transfer conveying section 219 that conveys the biaxially stretched film that has passed through the second stretching section 218 while cooling it, and a winding section 220 that winds up the stretched resin sent from the transfer conveying section 219.

[0025] The extrusion section 213 mixes and melts the pellets and forces them through a die 215 . The casting drum 216 rotates around an axis in the Y-axis direction, cools and solidifies the resin, and sends the sheet-shaped resin in the film longitudinal direction (X-axis direction).

[0026] The first stretching section 217 stretches the sheet-shaped resin sent from the casting drum 216 in the longitudinal direction of the film (X-axis direction) using a plurality of rolls while heating as necessary, to form a uniaxially stretched film.

[0027] The second stretching section 218 stretches the uniaxially stretched film (film 221) sent from the first stretching section 217 in a direction perpendicular to the longitudinal direction of the film using clips while heating as necessary. Here, the film is stretched in the Y-axis direction (the film width direction, the direction of the roll rotation axis) to form a biaxially stretched film.

[0028] The winding section 220 is a roll that winds up the biaxially stretched film after stretching in the second stretching section 218 via the transfer conveying section 219. Through this series of processes, multiple resin raw materials are mixed, melted, stretched, and formed into a sheet, producing a biaxially stretched resin film that is stretched in two perpendicular directions. Because the biaxially stretched resin film is stretched in different directions, it exhibits higher isotropy than a uniaxially stretched film, and exhibits, for example, excellent dimensional stability.

[0029] The control unit 202 controls the operation of the production unit 201. For example, when production conditions for biaxial stretching, such as the temperature and rotation speed of the rolls, are set, the control unit 202 controls the production unit 201 in accordance with the production conditions. The control unit 202 is a computer configured using one or more hardware components, such as a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), an ASIC (Application Specific Integrated Circuit), and an FPGA (Field Programmable Gate Array), and a memory in which various programs and the like are pre-installed.

[0030] Next, the configuration of the learning device 2 will be described with reference to Fig. 4. The learning device 2 is electrically connected to a property prediction device 3. Fig. 4 is a block diagram showing the configuration of the learning device provided in a property prediction system according to an embodiment of the present invention. The learning device 2 has a learning data generation unit 21, a learning unit 22, a control unit 23, and a storage unit 24.

[0031] The learning data generation unit 21 generates, for example, from data stored in the memory unit 24, data that associates process data obtained for each of a plurality of processing steps into which the film manufacturing process is subdivided with the physical properties of the film manufactured using the process data, as learning data. In other words, the learning data is data on film whose physical properties are known. Note that the learning data generation unit 21 may also extract data according to conditions input via the input device 5 to generate learning data. In the learning data, the process data for each processing step is treated as an explanatory variable, and the physical properties of the film corresponding to the process data are treated as a response variable.

[0032] The process data includes the conditions in the processing steps, the physical properties of the molded product obtained by the processing steps, or a combination thereof. The conditions in the process data include polymer pressure, extruder temperature, casting drum temperature, longitudinal stretching temperature, longitudinal stretching ratio, transverse stretching temperature, transverse stretching ratio, film formation speed, thickness, roll peripheral speed, roll torque, etc. Here, polymer pressure refers to the pressure of the molten polymer between the extruder and the die. Furthermore, the physical properties in the process data include the physical properties of the raw material and the physical properties of the resin film. The physical properties of the raw material include intrinsic viscosity, solution haze (turbidity), and color tone.

[0033] The physical properties of the film include intrinsic viscosity, film haze (turbidity), color tone, thickness, mechanical strength, and thermal shrinkage. The physical properties of the film can be expressed numerically or as an image. Furthermore, when multiple physical properties of the resin film are set, these physical properties may be combined to express the physical properties of a single film. In addition, when the film is produced by biaxial stretching, the mechanical strength includes the strength in each stretching direction. In this case, the learning data is associated with the blending conditions of the raw materials that form the film. The blending conditions of the raw materials may be used as explanatory variables. The blending conditions include the type, physical properties, blending ratio, etc. of the raw materials. The blending conditions, such as the type, physical properties, blending ratio, etc. of the raw materials, can be digitized or visualized. Furthermore, by combining multiple blending conditions and digitizing or visualizing them, they can be expressed as a single blending condition. Here, the blending ratio refers to data accumulated in the past that expresses the composition ratio of the raw materials in the resin film as a weight ratio, and includes the resin ratio and additive ratio as necessary.

[0034] The learning unit 22 performs learning using the learning data to generate a trained model. The learning unit 22 generates a trained model in which process data for each processing step is used as an explanatory variable and the physical properties of the film corresponding to the process data are used as a target variable. A known learning method can be used for the learning performed by the learning unit 22. Examples of statistical models used for learning include a simple linear regression model, Ridge regression, Lasso regression, Elastic Net regression, general additive model, random forest regression, rule fit regression, gradient boosting tree, extra tree, support vector regression, Gaussian process regression, k-nearest neighbor regression, kernel ridge regression, and neural network.

[0035] For example, when the learning unit 22 generates a trained model by learning using regularization, the learning unit 22 provides multiple candidate values for the hyperparameters of the trained model, performs learning for each of the provided candidate values of the hyperparameters, generates a trained model for the objective variable (physical properties of film), and stores the trained model in the storage unit 24. The learning unit 22 then calculates prediction errors by cross-validation or holdout validation using the training data for the models obtained by learning using each candidate value, and selects the trained model that provides the smallest prediction error. The hyperparameters referred to here are parameters that are set in advance by the learning unit 22 for learning, and include, for example, regularization coefficients. In addition, in the case of a trained model using a neural network, the hyperparameters also include, for example, the number of layers of the neural network.

[0036] The control unit 23 controls the overall operation of the learning device 2 .

[0037] The storage unit 24 stores various programs for operating the learning device 2 and data including various parameters necessary for the operation of the learning device 2. The various programs include a training data generation program that generates training data for generating a trained model, and a trained model generation program that generates a trained model by training using the training data. The various parameters include hyperparameters and parameters acquired by the learning unit 22 through training. The storage unit 24 also stores data for constituting the training data (for example, the process data and physical properties described above).

[0038] The storage unit 24 is configured using a ROM (Read Only Memory) in which various programs etc. are pre-installed, a RAM (Random Access Memory) for storing calculation parameters and data for each process, an HDD (Hard Disk Drive), an SSD (Solid State Drive), etc.

[0039] The various programs can also be widely distributed by recording them on computer-readable recording media such as HDDs, flash memories, CD-ROMs, DVD-ROMs, Blu-ray (registered trademark), etc. The communication network referred to here is configured using, for example, an existing public line network, a LAN (Local Area Network), a WAN (Wide Area Network), etc., and may be wired or wireless.

[0040] The learning device 2 having the above functional configuration is a computer configured using one or more pieces of hardware such as a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), an ASIC (Application Specific Integrated Circuit), and an FPGA (Field Programmable Gate Array).

[0041] Next, the physical property prediction device 3 will be described with reference to Fig. 5. Fig. 5 is a block diagram showing the configuration of the physical property prediction device provided in a physical property prediction system according to an embodiment of the present invention. The physical property prediction device 3 is electrically connected to a learning device 2, a display device 4, and an input device 5. The physical property prediction device 3 has a physical property prediction unit 31, a calculation unit 32, an extraction unit 33, a control unit 34, and a memory unit 35.

[0042] The physical property prediction device 3 uses the physical properties to be predicted and the trained model acquired from the learning device 2 to set blending conditions including the types of raw materials and their blending ratios that are suitable for the conditions.

[0043] The physical property prediction unit 31 reads out a trained model from the learning device 2 or the storage unit 35, and inputs process data into the trained model to obtain predicted physical properties (predicted values).

[0044] The calculation unit 32 calculates the explanatory variable influence on the predicted value for each explanatory variable.

[0045] The extraction unit 33 extracts the top 20 explanatory variables with the highest explanatory variable influences from among the multiple explanatory variables based on the calculated explanatory variable influences.

[0046] The control unit 34 comprehensively controls the operation of the physical property prediction device 3. The control unit 34 has a display control unit 341 that causes the display device 4 to display the prediction results of the physical property prediction unit 31 and the extraction results of the extraction unit 33. The display control unit 341 may cause the display device 4 to display information such as raw materials, the raw materials, and their predicted physical properties in addition to the setting results.

[0047] The storage unit 35 stores various programs for operating the property prediction device 3 and data including various parameters necessary for the operation of the property prediction device 3. The various programs include a blending condition setting program executed using a trained model. The storage unit 35 may also store a trained model. In this case, the trained model may be updated in synchronization with the learning device 2. The storage unit 35 is configured using a ROM in which various programs are pre-installed, and a RAM, HDD, SSD, etc. that store calculation parameters and data for each process.

[0048] The various programs can be recorded on computer-readable recording media such as HDDs, flash memories, CD-ROMs, DVD-ROMs, and Blu-ray (registered trademark) and distributed widely. The physical property prediction device 3 can also acquire the various programs via a communications network. The communications network referred to here is configured using, for example, an existing public line network, LAN, WAN, etc., and may be wired or wireless.

[0049] The physical property prediction device 3 having the above-described functional configuration is a computer configured using one or more pieces of hardware such as a CPU, a GPU, an ASIC, and an FPGA.

[0050] The display device 4 is a display made of liquid crystal or organic EL (Electro Luminescence) or the like, and is electrically connected to the physical property prediction device 3. The display device 4 acquires and displays display data output from the physical property prediction device 3 under the control of the display control unit 341. The display device 4 may also have an audio output function such as a speaker.

[0051] The input device 5 receives input of various information including information such as set mechanical property values related to the process of setting blending conditions, and outputs the received information to the learning device 2 and the property prediction device 3. The input device 5 is configured using a user interface such as a keyboard, mouse, microphone, and touch panel.

[0052] The physical property prediction device 3 executes a physical property prediction process using the process data. The physical property prediction according to this embodiment predicts the physical properties of the film obtained by the processing process and extracts parameters that have a large influence on the predicted value. This allows the user (film manufacturer) to efficiently set up a manufacturing process for a film with desired physical properties.

[0053] 6 is a flowchart showing an outline of a physical property prediction process performed by a physical property prediction device according to an embodiment of the present invention. First, the physical property prediction unit 31 acquires process data (step S101: acquisition step). At this time, the physical property prediction unit 31 acquires a plurality of process data with different conditions so as to cover all possible process conditions, for example. The physical property prediction unit 31 acquires process data in which the extruder temperature, casting drum temperature, longitudinal stretching temperature, longitudinal stretching ratio, transverse stretching temperature, transverse stretching ratio, etc. are set.

[0054] After acquiring the process data, the physical property prediction unit 31 acquires the trained model from the learning device 2 and acquires the physical properties of the film according to the process conditions (step S102: prediction step). The physical property prediction unit 31 inputs each piece of process data into the trained model to acquire a predicted value of the physical property.

[0055] Thereafter, the calculation unit 32 calculates the explanatory variable influence on the predicted value (step S103: influence calculation step). The calculation unit 32 calculates the explanatory variable influence by quantifying the contribution of each explanatory variable to the predicted value. The explanatory variable influence can be calculated using a known method, and for example, SHAP (SHapley Additive exPlanations) can be adopted. The explanatory variable influence indicates a positive or negative value. In this case, a positive influence indicates the degree of influence when the dependent variable increases, and a negative influence indicates the degree of influence when the dependent variable decreases.

[0056] After calculating the explanatory variable influence, the extraction unit 33 extracts the top 20 explanatory variables with the highest explanatory variable influence (step S104: first extraction step). The extraction unit 33 extracts the 20 explanatory variables in descending order of the absolute value of the explanatory variable influence of each explanatory variable.

[0057] When the control unit 34 acquires the extraction result from the extraction unit 33, the control unit 34 outputs the extraction result to the display device 4 and performs display control to display the result on the display device 4 (step S105: display step). The display device 4 displays the predicted value of the physical property and the explanatory variable influence of each explanatory variable on the predicted value.

[0058] 7 is a diagram showing an example of the display of the results of the physical property prediction process. The display screen W1 has a first display section W11 that displays information related to variable influence and a second display section W12 that displays information related to physical property values (predicted values). In FIG. 7, the first display section W11 displays a ranking of the explanatory variables extracted by the extraction unit 33 in order of influence. In this ranking, the explanatory variables are arranged in order of the absolute values of positive and negative influence. The second display unit W12 also displays the trend of the physical property values. This trend of the physical property values is expressed as a graph with the horizontal axis representing time-series data (time, lot number, etc.) and the vertical axis representing the value of that time-series data, or as the original data of the graph itself. Here, for example, values predicted up to that point are displayed.

[0059] The explanatory variables extracted by the extraction unit 33 are explanatory variables (parameters) that have a significant effect on the physical properties. The user can efficiently and quickly adjust the physical properties of the film by changing the conditions, etc., that correspond to the extracted explanatory variables.

[0060] Thereafter, for example, a film is produced using the above-described production apparatus 200 according to a production process (conditions) set based on the results of the above-described physical property prediction method. For example, the film is produced by a process in which conditions corresponding to influential explanatory variables are changed or adjusted. Specifically, for example, if the predicted result for the thermal shrinkage rate in the Y direction is higher than expected, indicating that the heat treatment temperature inside the transverse stretching oven has a large influence, the heat treatment temperature of the oven is increased to adjust the thermal shrinkage rate so as to decrease it.

[0061] In the embodiment described above, physical properties are predicted using a trained model that uses process data as explanatory variables and the physical properties of the film produced using this process data as objective variables, and it is possible to identify explanatory variables that have a significant effect on the predicted physical properties. According to this embodiment, parameters that are effective for adjustment in setting optimal manufacturing conditions can be identified appropriately and quickly, thereby enabling efficient prediction of physical properties.

[0062] In the embodiment, the number of explanatory variables extracted by the extraction unit 33 is set to 20, but this is not limiting and the number of extracted variables can be set arbitrarily depending on, for example, the items to be checked in the manufacturing process. Furthermore, the extraction unit 33 may extract either the positive influence degree or the negative influence degree.

[0063] (Variation 1) Next, a first modification of the embodiment will be described with reference to Figures 8 and 9. In the first modification, a property prediction device 3A is provided instead of the property prediction device 3 in the property prediction system according to the above-described embodiment. Note that the same components as those in the property prediction system according to the above-described embodiment are denoted by the same reference numerals.

[0064] 8 is a block diagram showing the configuration of a property prediction device according to Modification 1 of the present invention. Property prediction device 3A according to Modification 1 is electrically connected to learning device 2, display device 4, and input device 5. Property prediction device 3A includes property prediction unit 31, calculation unit 32, extraction unit 33, control unit 34, storage unit 35, and label assignment unit 36. Property prediction device 3A having the above functional configuration is a computer configured using one or more hardware components such as a CPU, GPU, ASIC, FPGA, etc.

[0065] The labeling unit 36 assigns information about standards and the like to the predicted values (or corresponding process data) as labels. Specifically, the labeling unit 36 assigns labels to the predicted values indicating whether the predicted values meet standards, are within the standard range, or are within the control range.

[0066] Fig. 9 is a flowchart showing an outline of the physical property prediction process performed by the physical property prediction device according to Modification 1. First, the physical property prediction unit 31 acquires process data (step S201) in the same manner as steps S101 and S102 shown in Fig. 6, and acquires the physical properties of the film according to the process conditions based on the trained model (step S202).

[0067] After acquiring the film's physical properties, the control unit 34 determines whether or not the film produced by the process for which the predicted values were calculated meets the standards (step S203). The control unit 34 reads information about the standards from the storage unit 35, for example, and determines whether or not the standards exist for the film corresponding to the target process data. If the control unit 34 determines that the standards exist (step S203: Yes), the control unit 34 proceeds to step S205. On the other hand, if the control unit 34 determines that the standards do not exist (step S203: No), the control unit 34 proceeds to step S204.

[0068] In step S204, the labeling unit 36 assigns a label indicating "no standard" to the predicted value.

[0069] Furthermore, in step S205, the control unit 34 determines whether or not a predicted value having a standard falls within a standard range. The control unit 34, for example, reads information about the standard range from the storage unit 35, and determines whether or not the target predicted value falls within the standard range. If the control unit 34 determines that the predicted value falls within the standard range (step S205: Yes), the control unit 34 proceeds to step S207. On the other hand, if the control unit 34 determines that the predicted value does not fall within the standard range (step S205: No), the control unit 34 proceeds to step S206.

[0070] In step S206, the labeling unit 36 assigns a label indicating "out of spec" to the predicted value.

[0071] Furthermore, in step S207, the control unit 34 determines whether or not a predicted value that is within the specification range is within the control range. The control unit 34, for example, reads information about the control range from the storage unit 35, and determines whether or not the target predicted value is within the control range. If the control unit 34 determines that the predicted value is within the control range (step S207: Yes), the control unit 34 proceeds to step S209. On the other hand, if the control unit 34 determines that the predicted value is not within the control range (step S207: No), the control unit 34 proceeds to step S208. The specification range refers to the desired range of performance, etc., set for a product, while the control range refers to the range used to determine whether a process for manufacturing a product that meets the specification range is stable. Specifically, the specification range is the quality standard range specified in the delivery specifications concluded with the customer, and products that fall within this range are delivered. In contrast, the control range is the range of physical property values during stable production, set within the above specification range but narrower than the specification range.

[0072] In step S208, the labeling unit 36 assigns a label indicating "outside control range" to the predicted value.

[0073] In step S209, the calculation unit 32 calculates the explanatory variable influence on the predicted value in the same manner as in step S103 shown in Fig. 6. Then, the extraction unit 33 extracts the top 20 explanatory variables with the highest explanatory variable influence (step S210).

[0074] When the control unit 34 acquires the extraction results from the extraction unit 33, it outputs the extraction results to the display device 4 and performs display control to display the results on the display device 4 (step S211). The display device 4 displays the predicted values of the physical properties and the explanatory variable influence of each explanatory variable on the predicted values. In this case, in this modification 1, the predicted values can be assigned labels such as "not in specification," "out of specification," and "out of control range," so that the quality of the film having the predicted value can be reliably grasped.

[0075] Thereafter, for example, a film is produced using the above-described production apparatus 200 in accordance with the production process (conditions) set based on the results of the above-described physical property prediction method.

[0076] In the above-described first modification, similar to the above-described embodiment, physical properties are predicted using a trained model in which process data is used as explanatory variables and the physical properties of the film produced using this process data are used as objective variables, and explanatory variables that significantly affect the predicted physical properties can be identified. According to the first modification, parameters that are effective for adjustment can be identified appropriately and quickly when setting optimal manufacturing conditions, thereby enabling efficient prediction of physical properties.

[0077] Furthermore, according to the present modification 1, a label indicating quality is assigned to the predicted value, so that the quality of the film having the predicted value can be reliably grasped. In addition to "not specified," "out of specification," and "out of control range," labels such as "within specification range" and "within control range" may be assigned so that the film quality can be understood for all predicted values. Also, at least one label from among "not specified," "out of specification," and "out of control range" may be assigned. In this first modification, steps S204, S206 and S208 correspond to labeling steps.

[0078] (Variation 2) Next, a first modification of the embodiment will be described with reference to Fig. 10 to Fig. 12. In the second modification, a property prediction device 3B is provided in place of the property prediction device 3 in the property prediction system according to the above-described embodiment. Note that the same components as those in the property prediction system according to the above-described embodiment are denoted by the same reference numerals.

[0079] 10 is a block diagram showing the configuration of a physical property prediction device according to Modification 2 of the present invention. Physical property prediction device 3B according to Modification 2 is electrically connected to learning device 2, display device 4, and input device 5. Physical property prediction device 3B has a physical property prediction unit 31, a calculation unit 32, an extraction unit 33, a control unit 34, a storage unit 35, and an anomaly degree calculation unit 37. Physical property prediction device 3B having the above functional configuration is a computer configured using one or more hardware components such as a CPU, GPU, ASIC, FPGA, etc.

[0080] The anomaly degree calculation unit 37 calculates the degree of anomaly for some of the explanatory variables involved in the calculation of the predicted value. In the present modification 2, the degree of anomaly indicates the degree of rarity of the explanatory variable extracted as the explanatory variable influence degree.

[0081] Fig. 11 is a flowchart showing an outline of the physical property prediction process performed by the physical property prediction device according to Modification 2. Similar to steps S101 to S104 shown in Fig. 6, the physical property prediction device 3B obtains predicted values of the film's physical properties for the obtained process data, and extracts the top 20 explanatory variables with the highest explanatory variable influence levels based on the explanatory variable influence levels on the predicted values (steps S301 to S304). Then, upon obtaining the extraction results from the extraction unit 33, the control unit 34 outputs the extraction results to the display device 4 and performs display control to display the results on the display device 4 (step S305).

[0082] In the present modified example 2, after the above processing, the abnormality degree calculation unit 37 calculates the abnormality degree (step S306). Note that this abnormality degree calculation processing may be executed simultaneously with steps S304 and S305, or may be executed before each step.

[0083] 12 is a flowchart showing an outline of the abnormality degree calculation process. First, the abnormality degree calculation unit 37 acquires past performance data for the explanatory variables (step S401). The abnormality degree calculation unit 37 reads out the past performance data for the explanatory variables from the storage unit 35, and acquires the explanatory variables for which the explanatory variable influence levels have been calculated as past performance data.

[0084] Furthermore, the abnormality degree calculation unit 37 acquires film physical property data (step S402). The abnormality degree calculation unit 37 acquires data on the physical properties of the film that was actually produced, which corresponds to the calculated predicted values. Steps S401 and S402 may be executed simultaneously, or step S402 may be executed first.

[0085] Then, the abnormality degree calculation unit 37 acquires data of product lots that fall within the standard range from the acquired film physical property data (step S403). Furthermore, the abnormality degree calculation unit 37 extracts past results of products that fall within the standard range from the acquired past results (step S403). Steps S403 and S404 may be executed simultaneously, or step S404 may be executed first.

[0086] Thereafter, the anomaly degree calculation unit 37 counts the number of times each explanatory variable appears from the extracted past performance of explanatory variable influence (step S405). The anomaly degree calculation unit 37 counts the number of times each explanatory variable appears for multiple explanatory variables related to products within the specification range.

[0087] Then, the anomaly degree calculation unit 37 extracts explanatory variables that have appeared less frequently in the past performance from among the top 20 explanatory variable influence degrees on the predicted value (step S406: second extraction step). At this time, the anomaly degree calculation unit 37 extracts, for example, explanatory variables that have appeared less frequently than a preset threshold value or explanatory variables that have appeared the least frequently.

[0088] After extracting the explanatory variables, the anomaly degree calculation unit 37 calculates the anomaly degree (step S407: anomaly degree calculation step). The anomaly degree calculation unit 37 calculates the reciprocal of the number of occurrences and sets this as the anomaly degree. Therefore, the anomaly degree is a value greater than 0 and less than 1, and the smaller the number of occurrences, the larger the value.

[0089] 11, the control unit 34 executes a notification process according to the calculated abnormality level (step S307: output step). In this notification process, information including the abnormality level is output, for example, by displaying an image or by sound. Alternatively, the notification process may be executed by light or the like.

[0090] Thereafter, for example, a film is manufactured using the above-described manufacturing apparatus 200 in accordance with the manufacturing process (conditions) set based on the results of the above-described physical property prediction method. At this time, in this second modification, explanatory variables with a low appearance frequency are notified as abnormal information as the degree of abnormality. This degree of abnormality is calculated based on the appearance frequency of explanatory variables that are within the top 20 explanatory variable influence degrees and have a relatively high influence among all explanatory variables, but do not appear frequently in past performance. For this reason, notification of the degree of abnormality draws attention to explanatory variables that have not had an impact on physical properties in the past.

[0091] In the second modification described above, similar to the embodiment described above, physical properties are predicted using a trained model in which process data is used as explanatory variables and the physical properties of the film produced using this process data are used as objective variables, and explanatory variables that have a significant effect on the predicted physical properties can be identified. According to the second modification, parameters that are effective for adjustment in setting optimal manufacturing conditions can be identified appropriately and quickly, thereby enabling efficient prediction of physical properties.

[0092] Furthermore, according to this variant example 2, abnormalities are notified regarding explanatory variables that have not previously affected physical properties, so that if a rare explanatory variable has an effect among the explanatory variables (parameters) that affect film production, the user can be made aware of that explanatory variable.

[0093] (Variation 3) Next, a third modification of the embodiment will be described with reference to Fig. 13 and Fig. 14. In the third modification, a property prediction device 3C is provided in place of the property prediction device 3 in the property prediction system according to the above-described embodiment. Note that the same components as those in the property prediction systems according to the above-described embodiment and modification are denoted by the same reference numerals.

[0094] 13 is a block diagram showing the configuration of a physical property prediction device according to Modification 3 of the present invention. Physical property prediction device 3C according to Modification 3 is electrically connected to learning device 2, display device 4, and input device 5. Physical property prediction device 3C includes a physical property prediction unit 31, a calculation unit 32, an extraction unit 33, a control unit 34, a storage unit 35, a label assignment unit 36, and an anomaly degree calculation unit 37. Physical property prediction device 3C having the above functional configuration is a computer configured using one or more hardware components such as a CPU, a GPU, an ASIC, an FPGA, etc.

[0095] Fig. 14 is a flowchart showing an outline of the property prediction process performed by the property prediction device according to Modification 3. First, similar to steps S201 to S211 shown in Fig. 9, the property prediction unit 31 assigns labels relating to the specifications, their ranges, or control ranges to the predicted values predicted for the acquired process data, and extracts and displays explanatory variables with high explanatory variable influence (steps S501 to S511).

[0096] Thereafter, the control unit 34 determines whether the predicted value is labeled as "out of specification" or "out of control range" (step S512). If the control unit 34 determines that the predicted value is not labeled (step S512: No), the control unit 34 ends the process. On the other hand, if the control unit 34 determines that the predicted value is labeled (step S512: Yes), the control unit 34 proceeds to step S513.

[0097] In step S513, the abnormality degree calculation unit 37 calculates the degree of abnormality. The abnormality degree calculation unit 37 calculates the degree of abnormality for predicted values labeled "out of specification" or "out of control range" according to the flowchart shown in Fig. 12.

[0098] The control unit 34 executes a notification process according to the calculated abnormality degree (step S514). By this notification process, information including the abnormality degree is output.

[0099] Thereafter, for example, a film is produced using the above-described production apparatus 200 in accordance with the production process (conditions) set based on the results of the above-described physical property prediction method.

[0100] In the third modification described above, similar to the above-described embodiment, physical properties are predicted using a trained model in which process data is used as explanatory variables and the physical properties of the film produced using this process data are used as objective variables, and explanatory variables that have a significant effect on the predicted physical properties can be identified. According to the third modification, parameters that are effective for adjustment in setting optimal manufacturing conditions can be identified appropriately and quickly, thereby enabling efficient prediction of physical properties.

[0101] Furthermore, according to the third modification, a label indicating quality is assigned to a predicted value, so that the quality of a film having that predicted value can be reliably grasped.

[0102] Furthermore, according to this variant example 3, for predicted values that are judged to be "out of specification" or "out of control range," an abnormality notification is given for explanatory variables that have not previously affected physical properties. Therefore, if a rare explanatory variable has an effect among the explanatory variables (parameters) that affect film production, the user can be made aware of that explanatory variable.

[0103] (Other embodiments) Although the embodiments of the present invention have been described above, the present invention should not be limited to the above-described embodiments. For example, the physical property prediction device may have the function of a learning unit. In this case, the physical property prediction device not only predicts physical properties but also sequentially updates the trained model. [Explanation of symbols]

[0104] 1. Physical property prediction system 2 Learning device 3. 3A~3C Physical property prediction equipment 4 Display device 5 Input Devices 21 Learning data generation unit 22 Learning Department 23, 34, 202 Control section 24, 35 Storage section 31 Physical Properties Prediction Department 32 Calculation section 33 Extraction part 36 Labeling section 37 Abnormality calculation part 200 Manufacturing equipment 201 Manufacturing Department

Claims

1. an acquisition step of acquiring process data obtained for each of a plurality of processing steps obtained by dividing the film manufacturing process; a prediction step of reading out from a storage unit a trained model that has been trained using process data in the plurality of processing steps as explanatory variables and physical properties of the film manufactured by the plurality of processing steps as objective variables, and calculating predicted physical properties based on the process data acquired in the acquisition step using the trained model; an influence calculation step of calculating the influence of each of a plurality of explanatory variables in the trained model on the physical properties calculated in the prediction step; a first extraction step of extracting a predetermined number of explanatory variables from the plurality of explanatory variables in descending order of the degree of influence; a display step of displaying the explanatory variables extracted in the first extraction step; A physical property prediction method including:

2. a second extraction step of extracting explanatory variables from the explanatory variables extracted in the first extraction step based on the number of occurrences in past performance data; an anomaly degree calculation step of calculating an anomaly degree of the explanatory variables extracted in the second extraction step; an output step of outputting notification information based on the degree of abnormality calculated in the degree of abnormality calculation step; The method for predicting physical properties according to claim 1 , further comprising:

3. a labeling step of assigning a label indicating whether or not a standard is met to the physical property calculated in the prediction step; The method for predicting physical properties according to claim 1 , further comprising:

4. the labeling step not only determines whether a standard exists, but also, if a standard exists, assigns a label indicating whether the standard is within a range; The method for predicting physical properties according to claim 3 .

5. the labeling step includes, when the test result is within the specification range, assigning a label indicating whether the test result is within the control range; The method for predicting physical properties according to claim 3 or 4.

6. the abnormality degree calculation step uses the reciprocal of the number of occurrences as the abnormality degree; The method for predicting physical properties according to claim 2 .

7. On the computer, an acquisition step of acquiring process data obtained for each of a plurality of processing steps obtained by dividing the film manufacturing process; a prediction step of reading out from a storage unit a trained model that has been trained using process data in the plurality of processing steps as explanatory variables and physical properties of the film manufactured by the plurality of processing steps as objective variables, and calculating predicted physical properties based on the process data acquired in the acquisition step using the trained model; an influence calculation step of calculating the influence of each of a plurality of explanatory variables in the trained model on the physical properties calculated in the prediction step; a first extraction step of extracting a predetermined number of explanatory variables from the plurality of explanatory variables in descending order of the degree of influence; a display step of displaying the explanatory variables extracted in the first extraction step; A physical property prediction program that executes the above.

8. an acquisition unit that acquires process data obtained for each of a plurality of processing steps obtained by dividing the film manufacturing process; a prediction unit that calculates predicted physical properties based on the process data acquired by the acquisition unit using a trained model that has been trained using process data in the plurality of processing steps as explanatory variables and physical properties of the film manufactured by the plurality of processing steps as objective variables; and an influence calculation unit that calculates the influence of each of a plurality of explanatory variables in the trained model on the physical properties calculated by the prediction unit; a first extraction unit that extracts a predetermined number of explanatory variables from the plurality of explanatory variables in descending order of the degree of influence; a display unit that displays the explanatory variables extracted by the first extraction unit; A physical property prediction device comprising:

9. setting conditions for each processing step based on the physical properties predicted by the physical property prediction method according to claim 1, and producing a film according to the set conditions; Film manufacturing method.

Images