Method and system for predicting change in properties of coated film, method and system for predicting change in manufacturing conditions of coating material, method for manufacturing coating material

Abstract

In the method and system for predicting a change in a property of a coating film according to the present invention, a prescribed artificial intelligence model that takes a change in a manufacturing condition as input and outputs a change in a property of a coating film is obtained through computer-based machine learning, and a change in a property of a coating film is predicted by inputting a change in a manufacturing condition into the artificial intelligence model through a computer. In the method and system for predicting a change in a manufacturing condition according to the present invention, furthermore, an optimal change in a manufacturing condition is determined as an optimal change in a manufacturing condition for achieving an objective based on a plurality of predicted changes in a property of a coating film. In the method for manufacturing a coated object according to the present invention, a predicted change in a manufacturing condition is adjusted, a coating material is applied to an object, a coating film is obtained, and a property of the coating film of the obtained coated object is measured.

Description

Technical Field

[0001] This invention relates to a method and system for predicting changes in the properties of a coating, a method and system for predicting changes in the manufacturing conditions of a coating, and a method for manufacturing a coating. Background Technology

[0002] Traditionally, when applying paint to an object, the process involves first applying the paint under standard or previous conditions, then repeatedly adjusting it to approximate the target coating properties. Historically, standard and previous conditions have been established through a method involving gradual manual operation: historical data such as pre-recorded operation management sheets are used for reference, thereby approximating the target coating properties to a certain extent. In this method, the paint manager manually adjusts manufacturing conditions such as coating pressure, linear speed, spray pressure, voltage, and temperature. By repeating this operation until the target film thickness, color difference from the target coated panel, and gloss difference are all below a certain level, a coating with the specified properties can be obtained.

[0003] The aforementioned adjustment operations rely heavily on the experience of the coating manager and are cumbersome in process. To automate these repetitive adjustment operations, a scheme has been proposed (Patent Documents 1-8) to predict the properties of the coating film obtained by applying paint to an object, using machine learning methods. This allows for the adjustment of manufacturing conditions based on the predicted values.

[0004] Existing technical documents

[0005] Patent documents

[0006] Patent Document 1: Japanese Patent Application Publication No. 2020-039992;

[0007] Patent Document 2: Japanese Patent Application Publication No. 2001-078627;

[0008] Patent document 3: Japanese Patent Application Publication No. 2000-219097;

[0009] Patent document 4: Japanese Patent Application Publication No. 1999-149118;

[0010] Patent document 5: Japanese Patent Application Publication No. 1993-300386;

[0011] Patent document 6: Japanese Patent Application Publication No. 1991-070565;

[0012] Patent document 7: Japanese Patent Application Publication No. 2016-501119;

[0013] Patent document 8: International Publication No. 2019 / 171498. Summary of the Invention

[0014] The problem the invention aims to solve

[0015] However, in the methods of Patent Documents 1 to 8, even when the calculated predicted value is within the range of the standard value of quality, it is determined that no adjustment of the manufacturing conditions is required and therefore no appropriate adjustment of the manufacturing conditions is made, even when the predicted value deviates from the measured value. As a result, sometimes a coating with the desired coating quality cannot be obtained.

[0016] Therefore, the object of the present invention is to provide a method and system for predicting the amount of change in the properties of a coating that can rapidly predict the amount of change in the properties of the coating; a method and system for predicting the amount of change in the manufacturing conditions of a coating that can adjust the manufacturing conditions independently of the range of standard values ​​of quality; and a method for manufacturing a coating that can produce a coating with the desired coating quality.

[0017] Solution for solving the problem

[0018] The main components of this invention are as follows.

[0019] (1) A method for predicting the change in the properties of a coating film, characterized in that,

[0020] The method for predicting the changes in the properties of the above-mentioned coating film includes the following steps:

[0021] The artificial intelligence model acquisition process acquires a defined artificial intelligence model that includes changes in manufacturing conditions as input and changes in the properties of the aforementioned coating as output; and

[0022] The property change prediction process uses a computer to predict the change in the properties of the coating by inputting the change in the manufacturing conditions into the artificial intelligence model.

[0023] (2) A method for predicting the amount of change in manufacturing conditions, characterized in that, when manufacturing a coated object by applying a coating to it, and in the case of manufacturing a coated object with a target property of a coating film by adjusting the manufacturing conditions by changing the manufacturing conditions, the method predicts the amount of change in the aforementioned manufacturing conditions that is optimal in order to achieve the target.

[0024] The method for predicting the changes in the above manufacturing conditions includes the following steps:

[0025] The artificial intelligence model acquisition process acquires a defined artificial intelligence model that includes changes in manufacturing conditions as input and changes in the properties of the aforementioned coating as output; and

[0026] The property change prediction process uses a computer to predict the change in the properties of the coating film by inputting the changes in the manufacturing conditions into the aforementioned artificial intelligence model.

[0027] In the above-mentioned property change prediction process, multiple changes in the aforementioned manufacturing conditions are input, and the changes in the properties of multiple coatings corresponding to each change in the aforementioned manufacturing conditions are predicted.

[0028] The method for predicting changes in the aforementioned manufacturing conditions also includes the following steps:

[0029] The manufacturing condition variation prediction process determines the optimal manufacturing condition variation as the optimal manufacturing condition variation for achieving the objective based on the predicted variations in the properties of the multiple coatings.

[0030] (3) The method according to (1) or (2) above, wherein the method further includes the following steps:

[0031] In the relational data input process, relational data showing the relationship between the changes in the manufacturing conditions and the changes in the properties of the coating film is input into the computer.

[0032] The process of obtaining the aforementioned artificial intelligence model is carried out in the following manner: using the aforementioned computer, the aforementioned relational data input in the aforementioned relational data input process is used as training data for machine learning.

[0033] (4) The method according to any one of (1) to (3) above, wherein the properties of the coating film include any one or more of the following: color, gloss, viscosity, film thickness, smoothness, visual appearance, color variation with angle and coating film abnormalities.

[0034] (5) The method according to any one of (1) to (4) above, wherein the manufacturing conditions include one or more of the following (a) to (d):

[0035] (a) The viscosity of the coating and any one or more of the residual components after heating;

[0036] (b) Any one or more of the roller circumferential speed, the coating pressure on the object, and the flow rate of the coating in the process of applying the coating to the object;

[0037] (c) any one or more of the baking temperature and baking time in the baking process; and

[0038] (d) Any one or more of the production line temperature, humidity and coating liquid temperature.

[0039] (6) The method according to any one of (1) to (5) above, wherein the artificial intelligence model acquisition step is a step of making the artificial intelligence model by machine learning, wherein the machine learning uses a prediction algorithm based on neural networks, an ensemble tree composed of multiple decision trees or partial least squares regression.

[0040] (7) The method described in (6) above, wherein the machine learning method uses a neural network, random forest or gradient boosting.

[0041] (8) The method according to any one of (1) to (7) above, wherein after the above relational data input step, a step of data cleaning of the above relational data is further included.

[0042] (9) According to the method described in (2) above, in the manufacturing condition change prediction step, the following manufacturing condition change is determined as the optimal manufacturing condition change for achieving the target, and the manufacturing condition change is the manufacturing condition change corresponding to the change in the property of the coating that has the smallest difference from the target property among the predicted changes in the properties of the plurality of coatings.

[0043] (10) A method for manufacturing a coating, further comprising the following steps:

[0044] In the post-conditioning manufacturing process, the manufacturing conditions are adjusted according to the change in the manufacturing conditions predicted by the method described in (2) above, and the coating is applied to the object to obtain the coating film; and

[0045] The property testing process measures the properties of the coating film obtained from the above-mentioned coating.

[0046] If the properties of the coating measured in the above-mentioned property measurement process deviate from the expected properties by more than a specified threshold, the above-mentioned property change prediction process, the above-mentioned manufacturing condition change prediction process, the above-mentioned condition adjustment manufacturing process, and the above-mentioned property measurement process are repeated.

[0047] (11) According to the method for manufacturing the coating as described in (10) above, if the condition in the above-described property measurement step deviates from the desired property by a predetermined threshold or more than a predetermined number of times,

[0048] The manufacturing method of the above-mentioned coating includes the following steps:

[0049] The artificial intelligence model obtains the sub-process, takes the changes in the aforementioned manufacturing conditions as input, and outputs the changes in the properties of the aforementioned coating as another artificial intelligence model; and

[0050] The sub-step for predicting the change in properties involves using the aforementioned computer to input the changes in the manufacturing conditions into another artificial intelligence model to predict the changes in the properties of the coating film.

[0051] In the aforementioned property change prediction sub-process, multiple changes in the aforementioned manufacturing conditions, pre-selected as candidates, are input, and the changes in the properties of multiple coating films corresponding to the changes in each of the aforementioned manufacturing conditions are predicted.

[0052] The method for manufacturing the above-mentioned coating also includes the following steps:

[0053] The manufacturing condition variation prediction sub-process, based on the predicted variations in the properties of the aforementioned multiple coatings, determines another optimal variation in the aforementioned manufacturing conditions as the optimal variation in the aforementioned manufacturing conditions for achieving the objective.

[0054] The method for manufacturing the above-mentioned coating also includes the following steps:

[0055] After condition adjustment, in the manufacturing sub-process, the change in the other manufacturing condition predicted by the aforementioned manufacturing condition adjustment is used to apply the aforementioned coating to the aforementioned object, thereby obtaining the aforementioned coated object; and

[0056] The property determination sub-step measures the properties of the coating film obtained from the above-mentioned coating.

[0057] If the properties of the coating measured in the above-mentioned property measurement process deviate from the expected properties by more than a specified threshold, the above-mentioned artificial intelligence model acquisition sub-process, the above-mentioned property change prediction sub-process, the above-mentioned manufacturing condition change prediction sub-process, the above-mentioned condition adjustment manufacturing sub-process, and the above-mentioned property measurement sub-process are repeated.

[0058] (12) The method for manufacturing the coating as described in (10) above, wherein the method for manufacturing the coating further includes the following steps:

[0059] In the relational data update process, relational data showing the relationship between the change in the manufacturing conditions and the change in the properties of the coating measured in the property measurement process is input into the computer, and the relational data is updated.

[0060] (13) A system characterized in that it predicts the amount of change in the properties of a coating obtained by applying a coating to an object.

[0061] The above system has a computer,

[0062] The above-mentioned computer has:

[0063] The artificial intelligence model acquisition unit acquires a prescribed artificial intelligence model that includes changes in manufacturing conditions as input and changes in the properties of the aforementioned coating as output; and

[0064] The property prediction unit calculates the change in the properties of the coating film by inputting the change in the manufacturing conditions into the artificial intelligence model.

[0065] (14) A system characterized in that, when manufacturing a coated object by applying a coating material to the object, in the case of manufacturing a coated object having the desired properties of a coating film by adjusting the manufacturing conditions, the system predicts the optimal amount of change in the manufacturing conditions to achieve the desired properties.

[0066] The above system has a computer,

[0067] The above-mentioned computer has:

[0068] The artificial intelligence model acquisition unit acquires a prescribed artificial intelligence model that includes the changes in the aforementioned manufacturing conditions as input and the changes in the properties of the aforementioned coating as output.

[0069] The property prediction unit predicts the properties of multiple coating films by inputting variations in multiple manufacturing conditions into the aforementioned artificial intelligence model; and

[0070] The manufacturing condition variation prediction unit determines the optimal manufacturing condition variation as the optimal manufacturing condition variation for achieving the objective based on the predicted variation in the properties of the multiple coatings.

[0071] (15) According to the manufacturing method described in (11) above, wherein the manufacturing conditions include process manufacturing conditions for each of a plurality of processes.

[0072] Monitor the manufacturing conditions of each of the above-mentioned processes when the manufacturing conditions are adjusted to the amount of change in manufacturing conditions predicted by the method described in (2) above.

[0073] In cases where the manufacturing conditions of a process are adjusted after being observed to deviate from the specified value by more than a certain amount in any of the aforementioned manufacturing conditions, the deviation of the variation amount from the specified value is observed.

[0074] The above manufacturing method also includes the following steps:

[0075] In the property re-prediction process, in the aforementioned artificial intelligence model, based on a fixed manufacturing condition for the monitored process, multiple variations of the aforementioned manufacturing conditions are input, thereby calculating and predicting the variations in the properties of the aforementioned coatings; and

[0076] The manufacturing condition variation re-prediction process, based on the predicted variations in the properties of the aforementioned coatings, determines again the optimal variation in the manufacturing conditions as the optimal variation in the manufacturing conditions to achieve the objective.

[0077] Based on the results of the re-prediction of the manufacturing condition changes mentioned above, the changes in the manufacturing conditions in subsequent processes are readjusted.

[0078] Furthermore, as one approach to the method described above, it is possible to include the following method: during the learning process in the aforementioned method for predicting changes, the change in coating properties before and after adjustment, obtained by measurement, is added to a fixed, representative absolute value of the coating properties before adjustment, thereby using the changed, representative absolute value for learning. Additionally, as another approach to the method described above, it is also possible to include the following method: when outputting the prediction result, after obtaining the predicted value of the change in coating properties obtained by the aforementioned method, it is added to a fixed, representative absolute value of the coating properties before adjustment, thereby indirectly obtaining the absolute value of the coating properties.

[0079] Invention Effects

[0080] According to the present invention, a method and system for predicting the amount of change in the properties of a coating film that can be rapidly predicted, a method and system for predicting the amount of change in the manufacturing conditions of a coating that can adjust the manufacturing conditions independently of the range of standard values ​​of quality, and a method for manufacturing a coating that can produce a coating having the desired coating quality are provided. Attached Figure Description

[0081] Figure 1 This is a flowchart illustrating a method for predicting changes in the properties of a coating film according to one embodiment of the present invention.

[0082] Figure 2 This is a flowchart illustrating a method for predicting changes in manufacturing conditions according to one embodiment of the present invention.

[0083] Figure 3 This is a flowchart illustrating a method for manufacturing a coating according to one embodiment of the present invention.

[0084] Figure 4 This is a flowchart illustrating a sub-process of a method for manufacturing a coating according to one embodiment of the present invention.

[0085] Figure 5 This is a schematic diagram of CCL (color-coated line).

[0086] Figure 6 This is a block diagram of a system for predicting changes in the properties of a coating film according to one embodiment of the present invention.

[0087] Figure 7 This is a block diagram of a system for predicting changes in manufacturing conditions according to one embodiment of the present invention. Detailed Implementation

[0088] First, the technology related to the prediction method of the present invention will be explained. The tolerance for coating properties varies depending on the application. In examples of color-related coating properties, in automotive repair painting, it is possible to apply a gradient (gradient) to the boundary between the normally painted portion of a faulty vehicle (equivalent to the target paint panel) and the portion where the paint is adjusted for repair after the new paint is applied. Therefore, the color consistency is ΔE < 0.5, which is not strictly required. On the other hand, for example, in roll coating for pre-coating (coating before assembly), sometimes paint panels from different manufacturing batches are assembled adjacently during product manufacturing. Therefore, many products allow for very small color differences due to batch differences, for example, ΔE < 0.1. The color difference level varies depending on the model and conditions of the colorimeter. Therefore, the accuracy of the absolute color value is insufficient, and the following adjustments are required: Prepare a standard panel (as the target coating panel) agreed upon with the customer. At the same time under the same conditions, measure the color difference between the standard panel and the panel during each manufacturing and color matching process, and control the color difference with the standard panel within the specified range.

[0089] The aforementioned adjustment operations rely heavily on the experience of coating managers and are cumbersome in process, thus requiring automation. However, when artificial intelligence learns the relationship between manufacturing conditions and the absolute values ​​of coating properties and makes predictions, even small deviations in prediction accuracy tend to produce directional errors in correction. For example, when taking color as an example, in the case of the target color b... * A value of 3.0 indicates the true color of the paint before adjustment. * With a value of 2.9 and a predicted color value of 3.1 for the unadjusted paint based on manufacturing conditions (i.e., an error of 0.2), in reality, in order to achieve b... * An increase of 0.1 should raise the temperature of the drying oven, but since the absolute value prediction exceeds 0.1 (strong yellowness), the temperature of the drying oven is lowered instead. In absolute value prediction, sometimes the color moves further away from the target color by adjustment, and sometimes fine-tuning is not possible.

[0090] Even when manufacturing conditions are set identically, factors such as minute differences in coating properties, changes in equipment over time, difficulty in quantification, difficulty in measurement, and high measurement time with limited data (high data costs) can accumulate and hinder machine learning from learning. These factors, though small, can still cause variations in coating properties. This is one reason why artificial intelligence in absolute value prediction generates prediction errors. Therefore, in methods such as those described in Patent Documents 1-8, accurate predictions may become difficult under such circumstances.

[0091] Therefore, the object of the present invention is to provide a method and system for predicting the amount of change in the properties of a coating that achieves high efficiency and high prediction accuracy through machine learning, a method and system for predicting the amount of change in the manufacturing conditions of a coating, a method for manufacturing a coating, and a method for manufacturing a coating that achieves high efficiency and enables the coating properties to closely approximate the desired coating properties through machine learning.

[0092] The present invention has the following advantages: By predicting the amount of change, even if the state of the manufacturing conditions at the time of prediction changes compared to the state of the manufacturing conditions when creating artificial intelligence, resulting in an error in the prediction of the absolute value, for example, in roll coating, as long as the amount of change is learned, such as increasing the coating pressure by 50 kgf under certain manufacturing conditions, the film thickness decreases by 1.5 μm, etc., because the unintentional changes in manufacturing conditions are equally applied before and after the intentional changes in manufacturing conditions, the effects are offset, and it can also be practically applied to fine-tuning.

[0093] Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings.

[0094] <Methods for predicting changes in coating properties>

[0095] Figure 1 This is a flowchart illustrating a method for predicting changes in the properties of a coating film according to one embodiment of the present invention. Hereinafter, reference will be made to... Figure 1An embodiment of a method for predicting the amount of change in the properties of a coating film obtained by applying a coating to an object will be illustrated. Furthermore, as an example, the method for predicting the amount of change in the properties of the coating film according to this embodiment can be performed using a coating film property change prediction system according to an embodiment of the present invention described later. Here, the predicted coating film properties preferably include at least the color of the coating, and in addition to color, preferably also the gloss and film thickness. Furthermore, in addition to color, gloss, and film thickness, the predicted coating film properties preferably include, for example, any one or more of the following: contrast, angle-dependent color variation, smoothness, visual appearance, resistivity, contact angle, contamination, solar reflectance, ultraviolet transmittance, durability, viscoelasticity, and coating anomalies. In particular, if the coating pressure is adjusted to change the film thickness, the color and gloss will also change; if a gloss modifier is added to adjust the gloss, the color will also change. Since they influence each other in this way, it is preferable to predict and adjust the color, gloss, and film thickness simultaneously. For example, L... * a * b * L in color space * value, a * value, b * Value (JIS Z8781-4 (2013)). Color can be measured using known color measurement methods. For example, color can be measured as follows: using a CM-512m3 commercially available from Konica Minolta Co., Ltd., with the light-receiving portion perpendicular to the coating at 0°, the color is measured by irradiating the light source at angles of 25°, 45°, and 75°. * value, a * value, b *The value can be measured using an X-Rite MA68II (manufactured by X-Rite). The measurement angle can be adjusted appropriately depending on the purpose or the equipment used. Any other index can be used. Furthermore, any index can be used, such as the intensity of the reflectance spectrum in every 5 nm from 380 nm to 780 nm as a color index. Gloss is not particularly limited, and gloss can be used as an index. Gloss can be measured using known gloss measurement methods. As an example, the 60° gloss of the coating film formed on the test plate can be measured using a specular gloss meter (gloss meter VG 7000 (manufactured by Nippon Denshoku Kogyo Co., Ltd.)) according to JIS K 5600-4-7 (specular gloss). Viscosity can be measured using known viscosity measurement methods. As an example, it can be measured according to JIS K5600-2-2 (flow cup method). Smoothness is preferably measured using wave-scan values. The waveform scan value is preferably one or more of the following: du (wavelength ≤ 0.1 mm), Wa (wavelength 0.1–0.3 mm), Wb (wavelength 0.3–1.0 mm), Wc (wavelength 1.0–3.0 mm), Wd (wavelength 3.0–10.0 mm), We (wavelength 10.0–30.0 mm), Lw (wavelength 1.2–12 mm), and Sw (wavelength 0.3–1.2 mm). Furthermore, the smaller the waveform scan value, the less unevenness on the surface at that wavelength, and the better the appearance quality of the coating. Anisochromia at angles can be measured using a variable angle difference meter, etc. Specific coating abnormalities include inhomogeneity, blistering, cracks, runs, pinholes, and thick edges. These can also be measured using known methods.

[0096] The coating for painting can be prepared as a coating that is usually used for this painting. In the case of preparing the coating, for example, it can be prepared by dispersing pigments, resins, additives and solvents and / or water in an SG mill, adding resins, solvents and additives to a primary color coating prepared by preparing various pigments, etc., dispersing them, and then repeatedly adjusting.

[0097] like Figure 1 As shown, in this embodiment, firstly, relational data showing the relationship between the amount of change in manufacturing conditions and the amount of change in the properties of the coating are input into the computer (step S101: relational data input process). The relational data can be prepared from historical data, etc.

[0098] Here, the aforementioned manufacturing conditions refer to the manufacturing conditions related to obtaining a coating film by applying a coating to an object. As an example, the manufacturing conditions can be one or more characteristics of the coating; specifically, one or more of the coating viscosity and residual heat content are preferred. Furthermore, as another example, the manufacturing conditions can be the conditions of each step in the process of applying the coating to the object. Specifically, in the process of applying the coating to the object, if roller coating is used, these include linear speed, roller circumferential speed, coating pressure, coating flow rate, the state, material, hardness of the coating roller, and type of pickup roller, etc., with one or more of roller circumferential speed, coating pressure on the object, and coating flow rate being preferred. Here, the roller for roller circumferential speed refers to the coating roller, support roller, pickup roller, metering roller, etc., and is not particularly limited, but the coating roller is preferred. Furthermore, as another example, the manufacturing conditions can be the conditions of processes such as baking processes performed between steps, including, for example, baking temperature, baking time, and maximum temperature reached in the baking process, with one or more of baking temperature and baking time being particularly preferred. Furthermore, as another example, manufacturing conditions can refer to the environmental conditions of the production line, or to any one or more of the following: production line temperature, humidity, and paint liquid temperature. In addition, examples include physical and chemical properties such as plate thickness, coating line, undercoat / topcoat adhesion, colorimeter readings, type or color of the undercoat, roughness, and type. For spray coating, examples include spray pressure, spray volume, spray gun type, baking temperature, baking time, temperature, and humidity during the baking process. For electrophoretic coating, examples include coating voltage, liquid temperature, energizing time, and coating direction (horizontal or vertical).

[0099] In summary, as an example, if it is roller coating, the manufacturing conditions can include any one or more of the following (a) to (d): (a) the viscosity of the coating and any one of the heating residual components; (b) any one of the roller circumferential speed, the coating pressure on the object and the flow rate of the coating in the process of applying the coating to the object; (c) any one of the baking temperature and baking time in the baking process; and (d) any one of the temperature, humidity and coating liquid temperature of the production line.

[0100] In addition, in order to make more accurate predictions, it is preferable to update the above relationship data frequently, timely, or periodically.

[0101] Furthermore, after the relational data input step (step S101), a data cleaning step is preferably included. Here, the prepared relational data undergoes processing such as data normalization / standardization, new data generation, and data filtering to remove inappropriate data. For example, regarding the difference between the measured and predicted values ​​of coating quality, data with an average value exceeding ±2σ when the standard deviation is set to σ can be carefully examined, and data suspected of being input errors or recording errors can be deleted. Through such data cleaning and normalization, overlearning in the machine learning process described later can be prevented, enabling more accurate predictions of coating properties. However, in this application, data cleaning is not mandatory and this step can be omitted. Data cleaning can be applied broadly to multiple or even all algorithms (e.g., removing outliers that would be deemed unsuitable by any machine learning algorithm) or specifically to the machine learning algorithms used in practice (e.g., removing data prone to errors in a particular algorithm). The artificial intelligence model creation step described later can also be created by learning from the cleaned data.

[0102] Next, as Figure 1 As shown, a defined artificial intelligence model is created using computer-based machine learning, taking the change in manufacturing conditions as input and the change in the properties of the coating as output (step S102: Artificial Intelligence Model Creation Process). In this example, the artificial intelligence model creation process is performed as follows: using a computer, the relational data input in the relational data input process (step S101) is used as training data for machine learning. Regarding the input, the relational data can be only the change in manufacturing conditions, or it can be the numerical value and change in manufacturing conditions. These can use measured values ​​or set values ​​as data. Regarding the output, the measured value of the coating properties can be used, and the difference between the measured values ​​of the coating properties before and after the change in manufacturing conditions can be calculated as the change. In addition, although an artificial intelligence model is created in this step in this embodiment, it is not necessary to create an artificial intelligence model, and a pre-created artificial intelligence model can be obtained from an external source. The artificial intelligence model can be obtained through, for example, the communication unit of a computer, or it can be obtained through transfer by a person (Artificial Intelligence Model Acquisition Process).

[0103] Machine learning algorithms can use any known algorithm, such as prediction algorithms based on decision trees, linear regression, partial least squares regression, lasso regression, ridge regression, multinomial regression, Gaussian process regression, support vector machines, random forests, gradient boosting, K-nearest neighbor method, neural networks, Bayesian estimation, or ensemble learning of these.

[0104] Machine learning particularly favors prediction algorithms based on neural networks. Alternatively, deep learning with three to several hundred layers can be achieved by simultaneously performing convolutional neural networks and dropout.

[0105] In addition, machine learning frameworks such as TensorFlow, Keras, Caffe, PyTorch, Chainer, and Scikit-learn can be used when creating artificial intelligence.

[0106] Here, machine learning can be supervised learning (including reinforcement learning). In this embodiment, the artificial intelligence model creation process (step S103) is performed as follows: using a computer, the relational data input in the relational data input process (step S102) is used as learning data (supervised data in this example) for machine learning (supervised learning).

[0107] Here, the algorithm is preferably determined through the following method: Multiple artificial intelligence models are created using various machine learning algorithms. The predicted output is compared with the given answer using data with pre-known answers, thereby confirming the accuracy of each AI model. Then, for the algorithm with high accuracy, the optimal hyperparameters are determined through methods such as cross-validation or Bayesian optimization. By relearning all data using the determined hyperparameters, an AI model can be created that takes the changes in manufacturing conditions as input and the changes in the adjusted coating properties as output.

[0108] In this embodiment, the prepared relational data is preprocessed. This preprocessing includes, for example, data normalization / standardization, logarithmic calculation, applying exponentiation (exponentiation processing), generating new data, or data filtering to remove inappropriate data. For instance, data with significant discrepancies between measured and predicted coating properties can be thoroughly examined, and data suspected of being input errors or recording errors can be deleted.

[0109] As described above, various machine learning algorithms can be used in this embodiment, but the preprocessing described above can also be a shared process applied to multiple or even all algorithms (e.g., removing outliers that would be considered inappropriate by any machine learning algorithm), or it can be a process specifically applied to the machine learning algorithm actually used (e.g., removing data that is prone to errors in a particular algorithm).

[0110] Furthermore, it is preferable to add zero-point relation data (zero-point addition) during preprocessing, showing the relationship that the change in coating properties is zero when the change in manufacturing conditions is zero. If the data used for machine learning is, for example, recorded data from adjustments made by production line managers, it does not include data such as "no change in manufacturing conditions." Therefore, the accuracy of predictions may decrease if the machine learning results do not show points where the change in coating properties is zero when manufacturing conditions remain unchanged. Thus, by adding the aforementioned zero-point relation data, this problem can be avoided, and the prediction accuracy can be further improved. Here, a value close to zero can also be used instead of zero.

[0111] Furthermore, when using records of consecutive adjustment operations as data, the results of summing these records can be combined into a single adjustment record to augment the data. For example, in the first and second adjustment records, the first record shows an increase in coating pressure of 50 kgf resulting in a decrease in film thickness of 1.2 μm, and the second record shows an increase in coating pressure of 100 kgf resulting in a further decrease in film thickness of 2.6 μm. By summing these records, the following additional data can be added: initially, the coating pressure was increased by 150 kgf, resulting in a decrease in film thickness of 3.8 μm. This further improves the accuracy of predictions by both learning from the effects of large changes in data and the effects of data augmentation.

[0112] As a preprocessing step, it's also possible to append inferred data between data sets. For example, by deriving approximate straight lines or curves from the data, points on these approximate lines or curves can be used as inferred data. This allows for increasing the amount of data, further improving prediction accuracy. Furthermore, historical data can be connected to other data sets using straight lines or curves, with any number of data points added to these lines.

[0113] Such preprocessing or normalization enables more accurate prediction of coating properties in the machine learning process described later. However, in this application, preprocessing is not necessary and can be omitted.

[0114] It is possible to create a prescribed artificial intelligence model using preprocessed data through the same method described above.

[0115] Next, in this embodiment, the amount of change in the manufacturing conditions is input into the artificial intelligence model via a computer to predict the amount of change in the properties of the coating (step S103: property change prediction process). Furthermore, when the input relational data includes not only the amount of change in the manufacturing conditions but also the numerical value of the manufacturing conditions, it is preferable to input both the numerical value and the amount of change in the manufacturing conditions into the artificial intelligence model to predict the amount of change in the properties of the coating.

[0116] In the process of predicting the amount of change in traits (step S103), the manufacturing conditions input are preferably used to pre-calculate the range by referring to historical data, etc.

[0117] By inputting the variation in manufacturing conditions into the aforementioned artificial intelligence model, the computer can calculate the variation in the properties of the coating. Furthermore, the predicted value of the calculated variation in the coating properties can be a single value or a combination of multiple candidate groups. When the predicted value of the calculated variation in the coating properties consists of multiple candidate groups, it is preferable to further include a step of appropriately selecting a single value from them using a predetermined criterion. This predetermined criterion can be of various types.

[0118] According to the method for predicting the change in the properties of the coating according to this embodiment, the predicted value of the change in the properties of the coating relative to the change in manufacturing conditions can be quickly obtained by computer calculation.

[0119] <Methods for predicting changes in manufacturing conditions>

[0120] Figure 2 This is a flowchart illustrating a method for predicting variations in manufacturing conditions according to one embodiment of the present invention. Hereinafter, reference will be made to... Figure 2 In the case of manufacturing a coating by applying a coating to an object and adjusting the manufacturing conditions to produce a coating film with the desired properties, an embodiment of a method for predicting the optimal amount of change in manufacturing conditions to achieve the desired properties will be described. Furthermore, as an example, the method for predicting the amount of change in manufacturing conditions in this embodiment can be executed using a system for predicting the amount of change in manufacturing conditions according to an embodiment of the present invention described later. Here, with... Figure 1 Similarly, in the implementation method, the properties of the coating preferably include one or more of the following: color, gloss, viscosity, film thickness, smoothness, visual appearance, color variation with angle, and coating abnormalities. Color can be, for example, L... * a * b * L in color space * value, a * value, b * Value (JIS Z8781-4 (2013)). Color can be measured using known color measurement methods. For example, color can be measured as follows: using a CM-512m3 commercially available from Konica Minolta Co., Ltd., with the light-receiving portion perpendicular to the coating at 0°, the color is measured by irradiating the light source at angles of 25°, 45°, and 75°. * value, a * value, b *The value can be measured using an X-Rite MA68II (manufactured by X-Rite). The measurement angle can be adjusted appropriately depending on the purpose or the equipment used. Any other index can be used. Furthermore, any index can be used, such as the intensity of the reflectance spectrum in every 5 nm from 380 nm to 780 nm as a color index. Gloss is not particularly limited, and gloss level can be used as an index. Gloss level can be measured using known gloss measurement methods. As an example, the 60° gloss level of the coating film formed on the test plate can be measured using a specular gloss meter (gloss meter VG 7000 (manufactured by Nippon Denshoku Kogyo Co., Ltd.)) according to JIS K5600-4-7 (specular gloss). Viscosity can be measured using known viscosity measurement methods. As an example, it can be measured according to JIS K5600-2-2 (flow cup method). Smoothness is preferably measured using waveform scan values ​​as an index. The preferred waveform scan value is one or more of the following: du (wavelength ≤ 0.1 mm), Wa (wavelength 0.1–0.3 mm), Wb (wavelength 0.3–1.0 mm), Wc (wavelength 1.0–3.0 mm), Wd (wavelength 3.0–10.0 mm), We (wavelength 10.0–30.0 mm), Lw (wavelength 1.2–12 mm), and Sw (wavelength 0.3–1.2 mm). Furthermore, a smaller waveform scan value indicates fewer irregularities on the surface at that wavelength, resulting in better coating appearance quality. Anisochromia can be measured using a variable angle difference meter, etc. Coating anomalies specifically include unevenness, blistering, cracks, runs, pinholes, and thick edges. These can also be measured using known methods. Additionally, the manufacturing conditions, and... Figure 1 The implementation method preferably includes one or more of the following (a) to (d): (a) the viscosity of the coating and the residual components after heating; (b) the roller circumferential speed, the coating pressure on the object and the flow rate of the coating in the process of applying the coating to the object; (c) the baking temperature and baking time in the baking process; and (d) the temperature and humidity of the production line.

[0121] For further details regarding the above matters, due to... Figure 1 The implementation method is the same, so it will be omitted from the description again.

[0122] like Figure 2As shown, in this embodiment, firstly, relational data showing the relationship between changes in manufacturing conditions and changes in the properties of the coating are input into the computer (step S201: relational data input process). Next, through computer-based machine learning, a predetermined artificial intelligence model is created, which includes changes in manufacturing conditions as input and changes in the properties of the coating as output (step S202: artificial intelligence model creation process). For these steps S201 to S202, since... Figure 1 The implementation method is the same, so detailed descriptions are omitted.

[0123] like Figure 2 As shown, in this embodiment, next, the amount of change in the manufacturing conditions is input into the aforementioned artificial intelligence model using a computer to predict the amount of change in the properties of the coating (step S203: property change prediction process). Additionally, with... Figure 1 Similarly, when the input relational data includes not only the change in manufacturing conditions but also the numerical value of the manufacturing conditions, it is preferable to input both the numerical value and the change in manufacturing conditions into the aforementioned artificial intelligence model, thereby predicting the change in the properties of the coating. In the property change prediction step (step S203), the input manufacturing conditions are preferably narrowed down in advance by referring to historical data, etc.

[0124] Here, in Figure 3 In the property change prediction step (step S203) of the embodiment, the change amounts of multiple manufacturing conditions are input, and the change amounts of the properties of multiple coatings corresponding to the change amounts of each manufacturing condition are predicted. When multiple manufacturing conditions are pre-selected, historical data can be referenced, for example.

[0125] In this embodiment, the optimal change in manufacturing conditions is then determined based on the predicted changes in the properties of multiple coating films, serving as the optimal change in manufacturing conditions for achieving the objective (step S204: Manufacturing Condition Change Prediction Step). Specifically, as an example, in the Manufacturing Condition Change Prediction Step (step S204), the following change in manufacturing conditions is determined as the optimal change in manufacturing conditions for achieving the objective: the change in manufacturing conditions corresponding to the change in the properties of the coating film with the smallest difference from the target property among the predicted changes in the properties of multiple coating films.

[0126] According to the method for predicting the amount of change in manufacturing conditions in this embodiment, the predicted value of the optimal amount of change in manufacturing conditions can be obtained when the properties of the coating change by a predetermined amount. Therefore, the manufacturing conditions can be adjusted to this predicted value. Moreover, this adjustment can be performed regardless of the range of standard quality values, thus avoiding the aforementioned problem where, even if the predicted value deviates from the measured value, it is determined that no adjustment of the manufacturing conditions is necessary, resulting in a coating with the desired coating quality not being obtained.

[0127] In the preparation of coatings using artificial intelligence, when predicting the required changes in manufacturing conditions based on coating properties such as color, it is possible that the solution is not unique and suggests changes in manufacturing conditions that differ significantly from actual records (for example, although the target color can be obtained, it cannot be adjusted in reality).

[0128] Therefore, the process also includes setting numerical generation ranges for the variation amounts of manufacturing conditions. In the manufacturing condition variation amount prediction process (step S204), it is preferable to input multiple variation amounts of the aforementioned manufacturing conditions, which are formed by combining the numerical values ​​of the variation amounts of each manufacturing condition generated within the set numerical generation range. Accordingly, prediction can be performed efficiently.

[0129] Furthermore, it is preferable to determine the manufacturing condition variation within the allowable range from the manufacturing condition variation predicted in the manufacturing condition variation prediction process (step S204) as the manufacturing condition variation for achieving the objective. Accordingly, the manufacturing condition variation can be predicted after the allowable range is preset, thus reducing the possibility of useless formulations and enabling more efficient and accurate prediction results of the manufacturing condition variation.

[0130] The aforementioned numerical range can be calculated based on a certain range from the standard or actual values ​​of pre-set manufacturing conditions. This allows for the generation of predicted values ​​with small deviations from the standard or actual values.

[0131] Here, the multiple manufacturing condition variation amounts input in the manufacturing condition variation prediction process (step S204) are preferably composed of more than one candidate number, which is an exponential number of the number of manufacturing conditions. For example, when the number of manufacturing conditions is 4, it is 10. 4=Total of 1000 combinations. This 1000 corresponds to the following process: when the allowable variation in manufacturing conditions under coating pressure is -200Kgf to +250Kgf, candidate variations in manufacturing conditions are prepared in 10 stages at 50Kgf intervals; variations in the other three manufacturing conditions are also prepared in 10 stages. This is because, in cases where finer color adjustments are required to strictly control color differences, the number of each raw material with a variation of 10 manufacturing conditions is increased by an exponent of 1 (10000 in the case of 4 raw materials). Furthermore, especially in applications such as coil coating, where strict color difference requirements are needed and the allowable range of variations in manufacturing conditions is large, the number of candidates is increased by an exponent of 2 (100000 in the case of a variation of 4), thereby enabling predictions corresponding to such fine-tuning.

[0132] Furthermore, it is preferable to generate numerical values ​​for the variations in each manufacturing condition, and then randomly combine these values ​​to generate candidate variations in the manufacturing condition. These candidate variations are then used to predict the variations in the manufacturing conditions needed to achieve the target. For example, when fine-tuning of coating properties such as film thickness is required, and the range of variations in manufacturing conditions is large, the total number of combinations can sometimes be as high as 1 trillion (e.g., 12 raw materials, 10 stages, or 6 raw materials, 100 stages), making it difficult to obtain a predictive value and impractical. Therefore, by using random combinations as described above, the number can be set to 10 million, reducing the number of combinations and shortening the calculation time for obtaining the predictive value. In other cases, the total number of combinations may sometimes be as high as 100,000, 1 million, or 100 million.

[0133] In the above methods, the numerical value itself can also be randomly generated within the range of numerical value generation.

[0134] Here, it is preferable to determine the amount of change in manufacturing conditions used to achieve the objective by adjusting the amount of change in manufacturing conditions multiple times on standard conditions or previous conditions, by performing a specified calculation on the coating properties, or by performing a specified calculation on the amount of change in manufacturing conditions obtained by using a target value that uses the target coating properties as a reference value.

[0135] As an example of performing a calculation on the predicted change in manufacturing conditions, such as multiplying the predicted value by 70%.

[0136] As a first example of using a gap target value, it can be set to 70% of the target's ΔL*. For example, if ΔL* = 3 and the gap target value is set to 70% of the target's ΔL*, then through three adjustments, the difference from the target value can be reduced to less than 0.1 in the order of 3 → 0.9 → 0.27 → 0.08. As a second example of using a gap target value, it can also be set to 700% of the target's ΔL*, and the resulting predicted value can be multiplied by 10%.

[0137] Furthermore, when using coatings in manufacturing conditions, it is preferable to use a formulation composition with a variation amount of 3% to 99% of the optimal manufacturing conditions to achieve the desired effect, assuming that adjustments for coating film properties are made through multiple additions. This method also prevents over-addition caused by batch-to-batch deviations or prediction errors in the coatings.

[0138] Furthermore, knowing in advance that the prediction accuracy for large changes is higher than that for small changes, the target change value can be set to be larger than the difference from the target.

[0139] <Method for manufacturing coatings>

[0140] Figure 3 This is a flowchart illustrating a method for manufacturing a coating according to one embodiment of the present invention. Hereinafter, reference will be made to... Figure 3 An embodiment of a method for manufacturing a coating with the desired properties of a coating film by adjusting manufacturing conditions when applying a coating to an object is described by way of example. Here, with Figure 1 , Figure 2 Similarly, in the implementation method, the properties of the coating preferably include one or more of the following: color, gloss, viscosity, film thickness, smoothness, visual appearance, color variation with angle, and coating abnormalities. Color can be, for example, L... * a * b * L in color space * value, a * value, b * Value (JIS Z8781-4 (2013)). Color can be measured using known color measurement methods. For example, color can be measured as follows: using a CM-512m3 commercially available from Konica Minolta Co., Ltd., with the light-receiving portion perpendicular to the coating at 0°, the color is measured by irradiating the light source at angles of 25°, 45°, and 75°. * value, a * value, b *The value can be measured using an X-Rite MA68II (manufactured by X-Rite). The measurement angle can be adjusted appropriately depending on the purpose or the equipment used. Any other index can be used. Furthermore, any index can be used, such as the intensity of the reflectance spectrum in every 5 nm from 380 nm to 780 nm as a color index. Gloss is not particularly limited, and gloss level can be used as an index. Gloss level can be measured using known gloss measurement methods. As an example, the 60° gloss level of the coating film formed on the test plate can be measured using a specular gloss meter (gloss meter VG 7000 (manufactured by Nippon Denshoku Kogyo Co., Ltd.)) according to JIS K5600-4-7 (specular gloss). Viscosity can be measured using known viscosity measurement methods. As an example, it can be measured according to JIS K 5600-2-2 (flow cup method). Smoothness is preferably measured using waveform scan values ​​as an index. The preferred waveform scan value is one or more of the following: du (wavelength ≤ 0.1 mm), Wa (wavelength 0.1–0.3 mm), Wb (wavelength 0.3–1.0 mm), Wc (wavelength 1.0–3.0 mm), Wd (wavelength 3.0–10.0 mm), We (wavelength 10.0–30.0 mm), Lw (wavelength 1.2–12 mm), and Sw (wavelength 0.3–1.2 mm). Furthermore, a smaller waveform scan value indicates fewer irregularities on the surface at that wavelength, resulting in better coating appearance quality. Anisochromia can be measured using a variable angle difference meter, etc. Coating anomalies specifically include unevenness, blistering, cracks, runs, pinholes, and thick edges. These can also be measured using known methods. Additionally, the manufacturing conditions, and... Figure 1 The implementation method preferably includes one or more of the following (a) to (d): (a) the viscosity of the coating and the residual components after heating; (b) the roller circumferential speed, the coating pressure on the object and the flow rate of the coating in the process of applying the coating to the object; (c) the baking temperature and baking time in the baking process; and (d) the temperature and humidity of the production line.

[0141] For further details regarding the above matters, due to... Figure 1 , Figure 2 The implementation method is the same, so it will be omitted from the description again.

[0142] like Figure 3As shown, in this embodiment, firstly, relational data showing the relationship between changes in manufacturing conditions and changes in coating properties is input into a computer (step S301: Relationship Data Input Process). Next, using computer-based machine learning, a defined artificial intelligence model is created, incorporating changes in manufacturing conditions as input and changes in coating properties as output (step S302: Artificial Intelligence Model Creation Process). Next, the changes in manufacturing conditions are input into the aforementioned artificial intelligence model using a computer to predict changes in coating properties (step S303: Properties Change Prediction Process). Next, based on the predicted changes in multiple coating properties, the optimal changes in manufacturing conditions are determined as the optimal changes in manufacturing conditions to achieve the objective (step S304: Manufacturing Condition Change Prediction Process). For these steps S301 to S304, since... Figure 2 The implementation methods are the same (for steps S301, S302 and...). Figure 1 The implementation method is the same, so detailed descriptions are omitted.

[0143] like Figure 3 As shown, in this embodiment, the change in manufacturing conditions predicted by the manufacturing condition adjustment is then applied to the object to obtain a coating film (step S305: manufacturing process after condition adjustment). Next, the properties of the obtained coating film are measured (step S306: property measurement process). The measurement can be performed using any measurement method corresponding to the measured property.

[0144] According to the coating manufacturing method of this embodiment, it is possible to predict the amount of change in manufacturing conditions corresponding to the amount of change in properties required to obtain the desired coating properties. Based on the prediction result, the manufacturing conditions are adjusted and the coating is formed again, thereby enabling the manufacture of a coating with the desired coating quality.

[0145] Here, if the properties of the coating measured in the property measurement process deviate from the desired properties by more than a predetermined threshold, it is preferable to have a repeatable property change prediction process, a manufacturing condition change prediction process, a condition adjustment manufacturing process, and a property measurement process. Therefore, even if a deviation occurs in the prediction in a single step, a coating with the desired coating quality can be manufactured through multiple adjustments.

[0146] In this embodiment, a relational data update step is preferably included. In this step, relational data showing the relationship between the change in manufacturing conditions adjusted in the manufacturing process after condition adjustment and the change in the properties of the coating measured in the property measurement process are input into the computer, and the relational data is updated. This update of the relational data enables the creation of an artificial intelligence model based on more data. The update can be performed immediately after obtaining new data, or periodically or as needed. However, it may sometimes lead to overlearning, so this step is not essential and can be omitted.

[0147] Figure 4 This is a flowchart illustrating a sub-process of a method for manufacturing a coating according to one embodiment of the present invention.

[0148] In this embodiment, if the properties of the coating measured in the property measurement process deviate from the desired properties by more than a predetermined threshold a predetermined number of times, the following sub-process is preferably performed.

[0149] like Figure 4 As shown, in this sub-process, firstly, another artificial intelligence model is created using computer-based machine learning, taking the change in manufacturing conditions as input and the change in coating properties as output (step S307: Artificial Intelligence Model Creation Sub-process). Next, the change in manufacturing conditions is input into the other artificial intelligence model using a computer to predict the change in the coating properties (step S308: Property Change Prediction Sub-process). In the property change prediction sub-process, multiple changes in manufacturing conditions are pre-selected as candidates, and the changes in the properties of multiple coatings corresponding to each change in manufacturing conditions are predicted. Next, based on the predicted changes in the properties of multiple coatings, another optimal change in manufacturing conditions is determined as the optimal change in manufacturing conditions to achieve the target (step S309: Manufacturing Condition Change Prediction Sub-process). Next, the manufacturing conditions are adjusted according to the predicted change in the other manufacturing conditions to obtain a coating applied to an object (step S310: Condition Adjustment Manufacturing Sub-process). Next, the properties of the obtained coating are measured (step S311: Property Measurement Sub-process). Then, as Figure 4 As shown, if the properties of the coating measured in the property measurement process deviate from the expected properties by more than a specified threshold, the artificial intelligence model creation sub-process, property change prediction sub-process, manufacturing condition change prediction sub-process, condition adjustment manufacturing sub-process, and property measurement sub-process are repeated.

[0150] In addition to using another artificial intelligence model to obtain another prediction result, each sub-process is able to... Figure 3 The corresponding steps shown in this process are also performed in the same way.

[0151] This sub-process enables the production of coatings with desired properties, even when the initial AI model has low accuracy.

[0152] exist Figure 3 The process shown is this process or Figure 4 In the sub-processes shown, the process ends when the properties of the coating, which can be measured in the property measurement process, deviate from the desired properties by less than a specified threshold. However, in this application, sub-processes are not necessarily required.

[0153] However, the manufacturing methods of coatings also mostly involve multiple steps. Figure 5 This is a schematic diagram of CCL (Color Coating Line) as an example.

[0154] Figure 5 The example illustrated herein consists of multiple processes, which are: a coating preparation process (process (A)); a coating process using rollers (process (B)); a baking process using an oven (process (C)); and a winding process (process (D)). Each process has process manufacturing conditions. In process (A), the process manufacturing conditions are, for example, any one or more of the viscosity of the coating and the residual content after heating; in process (B), any one or more of the roller circumferential speed, the coating pressure on the object, and the flow rate of the coating; in process (C), any one or more of the baking temperature and baking time; and in process (D), any one or more of the temperature of the production line, the humidity, and the coating liquid temperature.

[0155] Here, monitoring is used to adjust manufacturing conditions. Figure 3 The process involves predicting the changes in manufacturing conditions during the manufacturing process, specifically the manufacturing conditions of each step. For example, if it is a coating, the aforementioned properties and characteristics can be measured; if it is a manufacturing condition, measured values ​​or (adjusted) set values ​​can be used; if it is baking temperature and baking time, measured values ​​or (adjusted) set values ​​can be used; and if it is the environmental conditions of the production line, temperature and humidity can be measured.

[0156] Then, in each process manufacturing condition, if the process manufacturing condition is adjusted when a deviation from the predetermined amount of change is observed to exceed a specified value, the following process is performed.

[0157] First, in the artificial intelligence model, the process manufacturing conditions that are adjusted based on the observed deviation of the change from the predetermined value are fixed. Then, the change of multiple other manufacturing conditions is input, and the change of multiple coating properties is calculated to make predictions (property re-prediction process).

[0158] Next, based on the predicted changes in the properties of multiple coatings, the optimal change in manufacturing conditions is determined as the optimal change in manufacturing conditions to achieve the objective (manufacturing condition change prediction process).

[0159] The results of the process are used to re-predict the changes in manufacturing conditions to adjust the changes in manufacturing conditions in subsequent processes.

[0160] When passing Figure 5 When illustrating the above content with examples, for instance, if a coating process (process (B)) using the current roller is performed, and if the deviation of the properties and characteristics of the coating in process (A) is observed, then based on this premise, the optimal change in the conditions of process (C) and / or process (D) following the current process can be predicted. Thus, it is sometimes possible to make appropriate adjustments that take into account the deviation of the adjustment in process (A).

[0161] In this example, an artificial intelligence model is used, with the values ​​of the coating properties and characteristics observed in process (A), where deviations have been observed and adjusted, as fixed values. Based on this, the variations in manufacturing conditions for multiple processes (B) to (D) are input, and the variations in the properties of multiple coating films are calculated and predicted. Based on the predicted variations in the properties of multiple coating films, the optimal variations in manufacturing conditions in processes (C) and / or (D) are determined as the optimal variations in manufacturing conditions to achieve the desired properties. Thus, based on the deviations in process (A), the optimal variations in manufacturing conditions in processes (C) and / or (D) for obtaining the desired coating properties can be predicted. Then, the manufacturing conditions in processes (C) and / or (D) are adjusted based on the re-predicted results.

[0162] The properties of the coating need to be evaluated after a series of processes have been completed and the paint has dried. Therefore, in the event of a deviation from the established standards, time is required before it is determined that readjustment is necessary. During this period, it is possible to produce coatings with a quality standard that does not meet the established standards.

[0163] In contrast, this method allows for the re-prediction of manufacturing conditions during intermediate processes without waiting for the completion of a series of steps, thus minimizing the production of coatings with substandard quality.

[0164] <Prediction System for Changes in Coating Properties>

[0165] Figure 6This is a block diagram of a system for predicting the amount of change in the properties of a coating film according to one embodiment of the present invention. This system predicts the amount of change in the properties of a coating film obtained by applying a coating to an object. Figure 6 As shown, the coating property change prediction system 10 of this embodiment includes a computer 11. The computer 11 has machine learning capabilities. Furthermore, the computer 11 includes an artificial intelligence model creation unit 12 and a property prediction unit 13. The artificial intelligence model creation unit 12 has machine learning capabilities, and it creates a predetermined artificial intelligence model that includes changes in manufacturing conditions as input and changes in coating property as output. The property prediction unit 13 predicts the changes in coating property by inputting changes in manufacturing conditions into the artificial intelligence model. The artificial intelligence model creation unit 12 and the property prediction unit 13 can be processors.

[0166] The aforementioned relational data is input into the computer 11. In this embodiment, the computer 11 has a storage unit 14 (memory) for storing relational data and a communication unit 15 for sending and receiving relational data. The communication unit 15 can send and receive not only relational data but also other data. The artificial intelligence model creation unit 12 preferably has the function of using the input relational data as training data for machine learning. In addition, the computer 11 preferably also has a data cleaning function for relational data. Furthermore, this prediction system 10 preferably has a display unit (display) for displaying prediction results.

[0167] Detailed information regarding manufacturing conditions, coating properties, machine learning, etc. Figure 1 The implementation of the method shown is the same, so it will not be described again.

[0168] According to this prediction system 10, the predicted value of the change in the properties of the coating relative to the change in manufacturing conditions can be quickly obtained through computer calculation.

[0169] <Prediction System for Changes in Manufacturing Conditions>

[0170] Figure 7 This is a block diagram of a system for predicting the amount of change in manufacturing conditions according to one embodiment of the present invention. This system is designed to predict the optimal amount of change in manufacturing conditions to achieve the desired properties when manufacturing a coated object by applying a coating to it, and adjusting the manufacturing conditions to produce a coating film with the desired properties. Figure 7 As shown, the manufacturing condition prediction system 20 of this embodiment is composed of a computer 21. The computer 21 has machine learning capabilities. Furthermore, the computer 21 includes an artificial intelligence model creation unit 22, a trait prediction unit 23, a storage unit 24, and a communication unit 25. These are related to… Figure 6The artificial intelligence model creation unit 12, trait prediction unit 13, storage unit 14 and communication unit 15 in the illustrated embodiment are the same as those described above, so detailed descriptions are omitted.

[0171] The property prediction unit 23 is configured to predict the property changes of multiple coatings corresponding to the input of multiple manufacturing condition changes. The computer 21 also includes a manufacturing condition prediction unit 26, which determines the optimal manufacturing condition change based on the predicted property changes of the multiple coatings, as the optimal manufacturing condition for achieving the target. The manufacturing condition prediction unit 26 can be a processor. Furthermore, the manufacturing condition prediction unit 26 is configured to determine the optimal manufacturing condition change for achieving the target, which is the manufacturing condition change corresponding to the coating property change with the smallest difference from the target coating property among the predicted property changes of the multiple coatings. Additionally, the computer 21 preferably also includes a data cleaning function for relational data. Furthermore, the prediction system 20 preferably includes a display unit 27 (monitor) for displaying the prediction results.

[0172] Detailed information regarding manufacturing conditions, coating properties, machine learning, etc. Figure 2 The implementation of the method shown is the same, so it will not be described again.

[0173] According to this prediction system 20, the predicted value of the optimal change in manufacturing conditions can be obtained when the properties of the coating change by a specified amount. Therefore, the manufacturing conditions can be adjusted according to this predicted value. Moreover, this adjustment can be performed regardless of the range of standard quality values, thus avoiding the aforementioned problem where, even if the predicted value deviates from the measured value, it is determined that no adjustment of the manufacturing conditions is necessary, resulting in a coating with the desired coating quality.

[0174] The following describes embodiments of the present invention, but the present invention is not limited to these embodiments.

[0175] Example

[0176] The invention will be further illustrated by the following embodiments, but the invention is not limited thereto.

[0177] First, the preparation method of the primary color paint used in the preparation of the paints of Examples 1-3 and Comparative Examples 1-2 will be described.

[0178] <Example of preparation of white primary color paint>

[0179] Add 20 parts by weight of acrylic resin as resin and 35 parts by weight of isophorone as organic solvent. After uniformly dissolving the resin using a disperser, mix in 46 parts by weight of titanium dioxide as white pigment. Disperse the pigment using a sand mill (dispersion medium: glass beads) until the maximum particle size of the coarse particles is less than 10 μm to prepare a white primary color coating.

[0180] <Examples of Preparation of Other Primary Color Coatings>

[0181] The types and amounts of each material were changed as shown in Table 1 below. Except for this, black primary color paint 1, black primary color paint 2, yellow primary color paint, and red primary color paint were prepared using the same method as in the preparation example of the white primary color paint described above. The formulations of each primary color paint are shown in Table 1.

[0182] [Table 1]

[0183]

[0184] <Gloss modifier>

[0185] The gloss modifier uses commercially available silica as a matting agent.

[0186] <Viscosity modifier>

[0187] The viscosity modifier uses isophorone, which is used as a solvent.

[0188] <Preparation Example of Coating Composition 1>

[0189] Add 5 parts by weight of acrylic resin, 25 parts by weight of fluororesin, 35 parts by weight of isophorone and 35 parts by weight of cyclohexanone. After the resin is dissolved evenly using a disperser, add 25 parts by weight of white primary color paint, 12 parts by weight of black primary color paint, 69 parts by weight of yellow primary color paint, 4 parts by weight of red primary color paint and 12 parts by weight of gloss modifier. Mix evenly using a disperser to prepare paint composition 1 (solid component concentration: 47% by weight).

[0190] <Preparation Examples of Coating Compositions 2-4>

[0191] By changing the types and amounts of each material as shown in Table 2 below, and otherwise preparing coating compositions 2 to 4 in the same manner as the preparation example of coating composition 1 described above, coating compositions 2 to 4 are prepared. The formulations of each coating composition are shown in Table 2.

[0192] [Table 2]

[0193] Coating composition 1 Coating composition 2 Coating composition 3 Coating composition 4 acrylic resin 5 5 5 5 Fluororubber 25 25 25 25 Isophorone 35 35 35 35 Cyclohexanone 35 35 35 35 White base color paint 25 97 57 2 Black primary color paint 1 2 0 0 77 Black primary color paint 2 0 1 8 0 Yellow primary color paint 69 2 29 7 Red primary color paint 4 0 6 14 Gloss modifier 1 2 1 0 0 Gloss modifier 2 0 0 2 2 Solid component concentration (mass%) 47 47 45 38

[0194] The details of the materials used to prepare the coating composition are as follows.

[0195] • Acrylic resin: PARALOID B44 (manufactured by Rohm and Haas), solids concentration: 100% by mass

[0196] Fluororubber: KYNAR500 (manufactured by Arkema), solids content: 100% by mass

[0197] • White pigment: TI-PURE R-706 (titanium oxide, manufactured by DuPont)

[0198] • Black Pigment 1: Mitsubishi Carbon Black MA-100 (Carbon Black, manufactured by Mitsubishi Chemical Corporation)

[0199] • Black Pigment 2: SUNBLACK X15 (Carbon Black, manufactured by Shiraishi Calcium Co., Ltd.)

[0200] • Yellow pigment: TAROX synthetic iron oxide LL-XLO (yellow iron oxide, manufactured by Titanium Industry Co., Ltd.)

[0201] • Red pigment: TODA COLOR 140ED (iron oxide, manufactured by Toda Kogyo Co., Ltd.)

[0202] • Gloss modifier 1: GASIL HP395 (synthetic silica, manufactured by Enios Silica Company)

[0203] • Gloss modifier 2: SYLYSIA 435 (silica, manufactured by Fuji Silysia Chemical Co., Ltd.)

[0204] • Organic solvent: Isophorone (manufactured by Arkema)

[0205] • Organic solvent: Cyclohexanone (manufactured by Ube Industries, Ltd.)

[0206] Coating preparation method

[0207] The following process is performed in a single step within a coating line. After roller coating (reference film thickness: 5 μm) Fine Tough G primer (primer 1) as a primer onto the raw material (zinc-aluminum alloy plated steel sheet), it is baked for 60 seconds at a maximum temperature of 210°C to form a primer film. Next, after roller coating of paint composition 1 (reference film thickness: 15 μm), it is baked for 60 seconds at a maximum raw material temperature of 250°C, and then immediately cooled, thereby forming a coating film of the untreated paint composition.

[0208] Methods for measuring film thickness

[0209] The film thickness of the primer coating and the coating composition is calculated from the weight and specific gravity of the coating before and after the coating is removed from the cut coating plate by sandblasting.

[0210] Methods for measuring color difference

[0211] The color of the coating was determined using Lab Scan XE (manufactured by Hunter Lie Corporation), and the difference between its color and the color of the target coated panel was used as the color difference.

[0212] AI-generated examples

[0213] • Data input in the creation of artificial intelligence models

[0214] In the coating line, as coating parameters for adjusting the film thickness, color, and gloss, data such as the change in coating roller (AP roller) speed, the change in coating pressure (coating pressure of support roller and AP roller), and the amount of primary color correction additive are input into the computer.

[0215] In addition, for film thickness, color, and gloss, the changes in film thickness, color, gloss, and paint viscosity before and after coating parameter adjustment are also considered.

[0216] • Creation of artificial intelligence models

[0217] Pre-processing steps include data cleaning, zero-point import, combination, exponentiation, and normalization. An AI model is then created in the computer, using changes in film thickness, color, gloss, paint viscosity, AP roller speed, coating pressure, and primary color paint addition as explanatory variables, and changes in film thickness, color, and gloss after coating parameter adjustment as target variables. As machine learning algorithms, gradient boosting, random forest, support vector machine, and neural networks are used with cross-validation to determine the optimal hyperparameters. These determined hyperparameters are then used to relearn all data, thereby creating an AI model that uses changes in manufacturing conditions as explanatory variables and changes in the adjusted coating properties as target variables.

[0218] Similarly, L is used to predict colors. * a * b * Artificial intelligence models of gloss variation values ​​were used to obtain a total of five artificial intelligence models.

[0219] Comparative examples of artificial intelligence manufacturing

[0220] Using the AP roller speed change, coating pressure change, and primary color paint addition amount as the data input and explanatory variables in the creation of the artificial intelligence model, the primary color paint addition amount is the absolute value of the AP roller speed, the absolute value of the coating pressure, and the primary color paint mixing ratio. The pre-film thickness, color, gloss, and paint viscosity are not used as explanatory variables. Thus, the artificial intelligence model is created. Otherwise, it is created in the same way as the artificial intelligence creation example to obtain comparative artificial intelligence model 1.

[0221] Similarly, L is used to predict colors. * a * b * The comparison of the absolute value of gloss using artificial intelligence models yielded a total of five artificial intelligence models.

[0222] Table 3 below shows examples of artificial intelligence creation.

[0223] Next, the qualification criteria for prediction of each embodiment and comparative example will be explained.

[0224] Criteria for Judging the Quality of Coating Properties

[0225] The following conditions will be used as the criteria for determining whether a product is qualified.

[0226] The film thickness and L of the standard plate and the coated plate with adjusted paint were compared. * value, a * value, b * The difference in values ​​is taken as ΔFT, ΔL * , Δa * Δb * .

[0227] • The criteria for acceptance will be the case where the ΔFT is less than 1μm and the color meets all three of the following conditions.

[0228] ·For ΔE=√(ΔL) *2 +Δa *2 +Δb *2 The value of ) is below 0.1 and ΔL * , Δa * Δb * Each is below 0.1.

[0229] Regarding gloss value, the following conditions are considered acceptable: a gloss value below 10 with a difference of less than 0.5 from the standard plate; a gloss value between 10 and 20 with a difference of less than 1 from the standard plate; and a gloss value above 30 with a difference of less than 3 from the standard plate.

[0230] Criteria for determining the number of adjustments

[0231] When the difference between the film thickness of the film coating property that can be automated and adjusted before adjustment and the target film thickness is within 5 μm and within ΔE2, and the number of times of adjusting the film coating property reaches within 3 times to meet the qualified standard, it is regarded as qualified.

[0232] The evaluation results in each of the examples and comparative examples are as described below. In addition, the evaluation results are also shown in Table 4.

[0233] <Example 1>

[0234] In the coating composition 1 and manufacturing conditions with a film coating property having a difference of about 3 μm from the target film thickness and about ΔE = 1, allowable ranges are set for the viscosity change amount, coating pressure change amount, and AP roller circumferential speed change amount. After generating 10 million numbers respectively within these ranges and randomly combining them, together with the film thickness before adjustment, they are input into five artificial intelligence models that have completed gradient boosting learning, obtaining 10 million candidate data for the film thickness change amount, L* value change amount, a* value change amount, b* value change amount, and gloss change amount after adjusting the coating parameters.

[0235] Taking 100% of ΔFT, ΔL * value, Δa * value, Δb * value, and 80% of Δ gloss as the gap target values of the change amounts, one is obtained from 10 million numbers that simultaneously satisfy: the ΔL * value, Δa * value, Δb * value, and the manufacturing conditions of the gloss change amount, and adjustments are made.

[0236] Due to non - compliance, the second adjustment is carried out in the same way with a 100% gap target value. As a result, it is qualified, and the number of adjustment times is 2 times, qualified.

[0237] The average number of adjustment times after such adjustments for coating compositions 2 - 4 is 2.0 times.

[0238] <Example 2>

[0239] In the coating composition 1 and manufacturing conditions with a film coating property having a difference of about 2 μm from the target film thickness and about ΔE = 2.5, allowable ranges are set for the viscosity change amount, coating pressure change amount, primary color paint addition amount, baking temperature change amount, and the change amount of the temperature of the production line. After generating 10 million numbers respectively within the ranges and randomly combining them, together with the film thickness before adjustment, color, gloss, and coating viscosity, they are input into five artificial intelligence models that have completed neural network learning, obtaining 10 million candidate data for the film thickness change amount, L* value change amount, a* value change amount, b* value change amount, and gloss change amount after adjusting the coating parameters.

[0240] 100% of ΔFT and ΔL of the standard plate * Value, Δa * Value, Δb * The point representing 90% of the value and 90% of the Δ gloss is taken as the target value for the difference in variation. From 10 million points, one point is selected that simultaneously satisfies: ΔL = [value missing] which satisfies the target value for the difference. * Value, Δa * Value, Δb * The manufacturing conditions for adjusting the changes in value and gloss were adjusted to ensure the product met the required standards.

[0241] The average number of adjustments made to coating compositions 2 to 4 was 1.5.

[0242] <Comparative Example 1>

[0243] In a coating composition 1 and manufacturing conditions having a film property with a difference of about 2 μm relative to the target film thickness and about ΔE = 2.5, instead of using an artificial intelligence model, a coating manager with more than 3 years of experience adjusted various coating conditions until the film thickness, color and gloss were satisfactory. As a result, it took 3 adjustments to achieve the desired result.

[0244] The average number of adjustments made to coating compositions 2 to 4 was 3.5.

[0245] <Comparative Example 2>

[0246] In a coating composition 1 with a coating film property having a difference of approximately 3 μm relative to the target film thickness and approximately ΔE = 1, and under manufacturing conditions, allowable ranges are set for the absolute values ​​of viscosity, coating pressure, and AP roller speed. Within these ranges, 10 million numbers are generated for each, randomly combined, and input into five trained comparative artificial intelligence models to obtain the absolute values ​​of film thickness and L after adjusting the coating parameters. * absolute value, a * absolute value, b * 10 million candidate data points for absolute values ​​and absolute gloss values.

[0247] 100% of ΔFT and ΔL of the standard plate * Value, Δa * Value, Δb * The changes in 90% of the value and 80% of the Δ gloss are added to the absolute values ​​and used as the target gap value. From 10 million, one value L that simultaneously satisfies the target gap value is selected. * absolute value, a * Absolute value, b *The absolute values ​​of AP roller circumferential speed, coating pressure, and primary color mixing ratio were adjusted. Since the result was unsatisfactory, a second adjustment was performed with a target difference of 100%, but this also failed. Even after three more adjustments (a total of five) in the same manner, the result was still unsatisfactory.

[0248] Such adjustments were made to the four coating compositions 2 to 4, but no adjustments could be made within 5 times for each composition.

[0249] Examples 1 and 2 achieved compliance with the aforementioned compliance criteria through an average of 1.5 and 2.0 adjustments, respectively. On the other hand, in Comparative Example 1, even experienced coating managers required more than 3 adjustments, easily increasing working hours and hindering automation, indicating a reliance on the coating manager's skills. In Comparative Example 2, using an AI model based on a comparative AI manufacturing example to implement film thickness adjustment resulted in the inability to adjust coating parameters. This is because: even when the predicted and measured values ​​deviate, if the predicted value is within the target film thickness range, it is determined that no adjustment is needed, and coating parameters cannot be changed; or, when the absolute value prediction gradually approaches the target, adjustments often result in properties moving further away from the target properties.

[0250] [Table 3]

[0251]

[0252] [Table 4]

[0253]

[0254] Explanation of reference numerals in the attached figures

[0255] 10: Prediction System

[0256] 11: Computer

[0257] 12: Artificial Intelligence Model Making Department

[0258] 13: Trait Prediction Department

[0259] 14: Storage Department

[0260] 15: Ministry of Communications

[0261] 20: Prediction System

[0262] 21: Computer

[0263] 22: Artificial Intelligence Model Making Department

[0264] 23: Trait Prediction Department

[0265] 24: Storage Department

[0266] 25: Ministry of Communications

[0267] 26: Manufacturing Condition Prediction Department

[0268] 27: Display Section

Claims

1. A method for predicting the change in the properties of a coating film, characterized in that, The method for predicting the change in the properties of the coating film includes the following steps: The artificial intelligence model acquisition process involves acquiring a defined artificial intelligence model that includes changes in manufacturing conditions as input and changes in the properties of the coating as output; and The property change prediction process uses a computer to predict the change in the properties of the coating film by inputting the change in manufacturing conditions into the artificial intelligence model. The manufacturing conditions include any one or more of the following (a) to (d): (a) The viscosity of the coating and any one or more of the residual components after heating; (b) Any one or more of the roller circumferential speed, coating pressure and coating flow rate in the process of applying coating to the object; (c) any one or more of the baking temperature and baking time in the baking process; and (d) Any one or more of the following: production line temperature, humidity, and paint liquid temperature. The process of obtaining the artificial intelligence model is the process of creating the artificial intelligence model through machine learning. The machine learning uses prediction algorithms based on decision trees, linear regression, partial least squares regression, lasso regression, ridge regression, multinomial regression, Gaussian process regression, support vector machines, random forests, gradient boosting, K-nearest neighbors, neural networks, Bayesian estimation, or ensemble learning of these.

2. A method for predicting changes in manufacturing conditions, characterized in that, When manufacturing a coated object by applying a coating to an object, and in the case of manufacturing a coated object with a target property of a coating film by adjusting the manufacturing conditions through changes, a method for predicting the amount of change in the manufacturing conditions to achieve the target includes the following steps: The artificial intelligence model acquisition process involves acquiring a defined artificial intelligence model that includes changes in manufacturing conditions as input and changes in the properties of the coating as output; and The property change prediction process uses a computer to predict the change in the properties of the coating film by inputting the change in manufacturing conditions into the artificial intelligence model. In the property change prediction process, multiple changes in the manufacturing conditions are input, and the changes in the properties of multiple coatings corresponding to each change in the manufacturing conditions are predicted. The method for predicting changes in manufacturing conditions also includes the following steps: The manufacturing condition variation prediction process, based on the predicted variations in the properties of the plurality of coatings, determines the optimal variation in the manufacturing conditions as the optimal variation in the manufacturing conditions to achieve the objective. The manufacturing conditions include any one or more of the following (a) to (d): (a) The viscosity of the coating and any one or more of the residual components after heating; (b) Any one or more of the roller circumferential speed, coating pressure and coating flow rate in the process of applying coating to the object; (c) any one or more of the baking temperature and baking time in the baking process; and (d) Any one or more of the following: production line temperature, humidity, and paint liquid temperature. The process of obtaining the artificial intelligence model is the process of creating the artificial intelligence model through machine learning. The machine learning uses prediction algorithms based on decision trees, linear regression, partial least squares regression, lasso regression, ridge regression, multinomial regression, Gaussian process regression, support vector machines, random forests, gradient boosting, K-nearest neighbors, neural networks, Bayesian estimation, or ensemble learning of these.

3. The method according to claim 1 or 2, wherein, The method further includes the following steps: The relational data input process involves inputting relational data into the computer showing the relationship between the changes in the manufacturing conditions and the changes in the properties of the coating. The process of obtaining the artificial intelligence model is carried out in the following manner: the relational data input in the relational data input process is used as training data for machine learning by the computer.

4. The method according to claim 1 or 2, wherein, The properties of the coating include any one or more of the following: color, gloss, viscosity, film thickness, smoothness, visual appearance, color variation with angle, and coating abnormalities.

5. The method according to claim 1 or 2, wherein, The machine learning method used is a prediction algorithm based on neural networks, ensemble trees consisting of multiple decision trees, or partial least squares regression.

6. The method according to claim 5, wherein, The machine learning method used is a neural network, a random forest, or a gradient boosting method.

7. The method according to claim 3, wherein, Following the relationship data input process, a data cleaning process for the relationship data is also included.

8. The method according to claim 2, wherein, In the manufacturing condition variation prediction process, the following manufacturing condition variation is determined as the optimal manufacturing condition variation for achieving the target: the manufacturing condition variation is the variation of the manufacturing condition corresponding to the coating property variation that has the smallest difference from the target property among the predicted variations of the properties of the plurality of coatings.

9. A method for manufacturing a coating, further comprising the following steps: Following the condition adjustment manufacturing process, the manufacturing conditions are adjusted by the amount of change in the manufacturing conditions predicted by the method described in claim 2, and the coating is applied to the object to obtain the coating film; and The property testing process measures the properties of the coating film obtained from the coating. If the properties of the coating measured in the property measurement step deviate from the expected properties by more than a specified threshold, the property change prediction step, the manufacturing condition change prediction step, the condition-adjusted manufacturing step, and the property measurement step are repeated.

10. The method for manufacturing a coating according to claim 9, wherein, If the properties of the coating measured in the property measurement process deviate from the desired properties by more than a specified threshold a specified number of times, The method for manufacturing the coating includes the following steps: An artificial intelligence model obtains a sub-process, and obtains another artificial intelligence model that takes the change in the manufacturing conditions as input and the change in the properties of the coating as output; and The sub-step for predicting the change in properties involves using the computer to predict the change in the properties of the coating film by inputting the change in manufacturing conditions into another artificial intelligence model. In the property change prediction sub-step, multiple changes in manufacturing conditions selected in advance as candidates are input, and the property changes of multiple coatings corresponding to the changes in each of the manufacturing conditions are predicted. The method for manufacturing the coating also includes the following steps: The manufacturing condition variation prediction sub-step, based on the predicted variations in the properties of the plurality of coatings, determines another optimal variation in the manufacturing conditions as the optimal variation in the manufacturing conditions for achieving the objective. The method for manufacturing the coating also includes the following steps: After condition adjustment, the manufacturing sub-process involves adjusting the manufacturing conditions by the predicted change in another manufacturing condition, and then applying the coating to the object to obtain the coated material. as well as The property determination sub-step measures the properties of the coating film obtained from the coating. If the properties of the coating measured in the property measurement process deviate from the expected properties by more than a specified threshold, the artificial intelligence model acquisition sub-process, the property change prediction sub-process, the manufacturing condition change prediction sub-process, the condition adjustment manufacturing sub-process, and the property measurement sub-process are repeated.

11. The method for manufacturing a coating according to claim 9, wherein, The method for manufacturing the coating also includes the following steps: In the relational data update process, relational data showing the relationship between the amount of change in the manufacturing conditions and the amount of change in the properties of the coating measured in the property measurement process is input into the computer to update the relational data.

12. The manufacturing method according to claim 10, wherein, The manufacturing conditions include process manufacturing conditions for each of multiple processes. Monitor the manufacturing conditions of each process step when the manufacturing conditions are adjusted to the amount of change in manufacturing conditions predicted by the method of claim 2. In cases where the manufacturing conditions of a process are adjusted after being observed to deviate from a predetermined value by more than the specified amount of change in each of the aforementioned process manufacturing conditions. The manufacturing method further includes the following steps: In the property re-prediction process, the artificial intelligence model, with the manufacturing conditions of the monitored process fixed, inputs the changes in multiple other manufacturing conditions, thereby calculating the changes in the properties of multiple coating films for prediction. as well as The manufacturing condition variation re-prediction process, based on the predicted variations in the properties of the multiple coatings, determines again the optimal variation in the manufacturing conditions as the optimal variation in the manufacturing conditions to achieve the objective. Based on the results of the re-prediction of the process based on the changes in manufacturing conditions, the changes in manufacturing conditions in subsequent processes are readjusted.

13. A system, characterized in that, Predicting the amount of change in the properties of a coating film obtained after applying paint to an object. The system has a computer. The computer has: An artificial intelligence model acquisition unit acquires a defined artificial intelligence model that includes changes in manufacturing conditions as input and changes in the properties of the coating as output; and The property prediction unit predicts the change in the properties of the coating by inputting the change in manufacturing conditions into the artificial intelligence model. The manufacturing conditions include any one or more of the following (a) to (d): (a) The viscosity of the coating and any one or more of the residual components after heating; (b) Any one or more of the roller circumferential speed, coating pressure and coating flow rate in the process of applying coating to the object; (c) any one or more of the baking temperature and baking time in the baking process; and (d) Any one or more of the following: production line temperature, humidity, and paint liquid temperature. The artificial intelligence model acquisition unit is configured to generate the artificial intelligence model through machine learning. The machine learning uses prediction algorithms based on decision trees, linear regression, partial least squares regression, lasso regression, ridge regression, multinomial regression, Gaussian process regression, support vector machines, random forests, gradient boosting, K-nearest neighbors, neural networks, Bayesian estimation, or ensemble learning of these.

14. A system, characterized in that, When manufacturing a coated object by applying a coating to it, and in the case of manufacturing a coated object with the desired properties of a coating film by adjusting the manufacturing conditions through variations, the optimal amount of variation in the manufacturing conditions to achieve the desired properties is predicted. The system has a computer. The computer has: The artificial intelligence model acquisition unit acquires a defined artificial intelligence model that includes the change in the manufacturing conditions as input and the change in the properties of the coating as output. The property prediction unit predicts the properties of multiple coating films by inputting multiple changes in the manufacturing conditions into the artificial intelligence model. as well as The manufacturing condition variation prediction unit determines the optimal manufacturing condition variation as the optimal manufacturing condition variation for achieving the objective based on the predicted variations in the properties of the plurality of coatings. The manufacturing conditions include any one or more of the following (a) to (d): (a) The viscosity of the coating and any one or more of the residual components after heating; (b) Any one or more of the roller circumferential speed, coating pressure and coating flow rate in the process of applying coating to the object; (c) any one or more of the baking temperature and baking time in the baking process; and (d) Any one or more of the following: production line temperature, humidity, and paint liquid temperature. The artificial intelligence model acquisition unit is configured to generate the artificial intelligence model through machine learning. The machine learning uses prediction algorithms based on decision trees, linear regression, partial least squares regression, lasso regression, ridge regression, multinomial regression, Gaussian process regression, support vector machines, random forests, gradient boosting, K-nearest neighbors, neural networks, Bayesian estimation, or ensemble learning of these.

Images