Control device, control method, and program

Abstract

To provide a control method making it possible to sustain manufacture of products of stable quality.SOLUTION: A control method includes a step of acquiring a prediction model by executing T-method data processing for plural sample data sets, a step of calculating a prediction value of an objective variable by inputting measured values of plural explanatory variables into the prediction model for each product, a step of selecting an explanatory variable of an adjustment object from among the plural explanatory variables when the prediction value falls out of a management range, and a step of determining an adjustment quantity of the explanatory variable of the adjustment object so that the value of the objective variable approaches the center of the management range. The plural explanatory variables are divided into a first group and a second group, a degree of influence on the objective variable of which is smaller than that of the first group, on the basis of SN ratios. The explanatory variable of the adjustment object is selected from the first group.SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] The present disclosure relates to a control device, a control method, and a program.

Background Art

[0002] "Kanagaku, "Multivariate Statistical Process Control", [online], June 2005, [searched on January 4, 2022], Internet <http: / / manabukano.brilliant-future.net / research / report / Report2005_MSPC.pdf>" (Non-Patent Document 1) discloses a technique for monitoring the operating state of a production line.

Prior Art Documents

Non-Patent Documents

[0003]

Non-Patent Document 1

Non-Patent Document 2

Summary of the Invention

Problems to be Solved by the Invention

[0004] The operating state of a production line determines the quality of products produced by the production line. According to the technique disclosed in Non-Patent Document 1, when an abnormality in the operating state of the production line is detected, a process for returning the abnormality to normal is executed. However, conventionally, the process for returning the abnormality to normal has been laborious, such as manual analysis by a person. Therefore, it has been difficult to maintain the production of products with stable quality.

[0005] This disclosure has been made in view of the above-mentioned problems, and its purpose is to provide a control device, a control method, and a program that can maintain the production of products of stable quality. [Means for solving the problem]

[0006] According to one example of this disclosure, a control device for controlling a production line including multiple processes comprises an acquisition unit, a collection unit, a prediction unit, a selection unit, a determination unit, and a control unit. The acquisition unit acquires a prediction model obtained by performing T-method data processing on multiple sample datasets showing the values ​​of multiple explanatory variables relating to the operation of the production line and the value of an objective variable relating to the quality of the products produced by the production line. The prediction model includes a signal-to-noise ratio (SNR) for each of the multiple explanatory variables, representing the prediction accuracy of the value of the objective variable. The collection unit collects measured values ​​of the multiple explanatory variables from the production line. The prediction unit calculates a predicted value of the objective variable for each product by inputting the measured values ​​of the multiple explanatory variables into the prediction model. The selection unit selects an explanatory variable to be adjusted from among the multiple explanatory variables in response to the predicted value falling outside the control range. The determination unit determines the amount of adjustment for the explanatory variable to be adjusted so that the value of the objective variable approaches the center of the control range. The control unit controls the production line to change the value of the explanatory variable to be adjusted by the amount of adjustment. The selection unit divides multiple explanatory variables into a first group and a second group, which have a smaller influence on the dependent variable than the first group, based on the signal-to-noise ratio, and selects the explanatory variables to be adjusted from among the explanatory variables belonging to the first group.

[0007] According to this disclosure, the value of the dependent variable is predicted from the measured values ​​of multiple explanatory variables. If the predicted value falls outside the control range, an explanatory variable to be adjusted is selected from the first group that has a relatively large impact on the dependent variable. Then, the amount of adjustment for the adjusted explanatory variable is determined so that the value of the dependent variable approaches the center of the control range, and the production line is controlled to change the value of the adjusted explanatory variable by the adjustment amount. As a result, the number of defective products produced is reduced, and the production of products of stable quality is maintained.

[0008] In the disclosure described above, the selection unit calculates the degree of variation of each explanatory variable belonging to the first group and selects the first explanatory variable with the greatest degree of variation as the explanatory variable to be adjusted.

[0009] According to the disclosure above, the first explanatory variable, which exhibits the greatest degree of variation, is considered to be the cause of the variation in the dependent variable. Therefore, by selecting the first explanatory variable as the adjustment target, the variation in the dependent variable can be suppressed.

[0010] In the disclosure described above, the determination unit determines the adjustment amount for the first explanatory variable as the amount obtained by subtracting the measured value of the first explanatory variable from the reference value of the first explanatory variable.

[0011] According to the above disclosure, the value of the first explanatory variable can be returned to the baseline value. As a result, the fluctuation of the dependent variable is suppressed.

[0012] In the above disclosure, the selection unit calculates the degree of variation of each explanatory variable belonging to the first group. The selection unit selects a second explanatory variable, which corresponds to a second step that is later than the first step corresponding to the first explanatory variable with the greatest degree of variation among the explanatory variables belonging to the first group, as the explanatory variable to be adjusted.

[0013] According to the above disclosure, the fluctuations in the dependent variable caused by fluctuations in the first independent variable can be canceled out by adjusting the value of the second independent variable. As a result, fluctuations in the dependent variable are suppressed.

[0014] In the disclosure described above, the prediction model predicts the predicted value using the sum of the values ​​obtained by multiplying each measured value of a plurality of explanatory variables by the coefficient corresponding to that explanatory variable. The decision unit determines the adjustment amount for the second explanatory variable as the quotient obtained by subtracting the baseline value of the dependent variable from the predicted value and dividing the result by the coefficient corresponding to the second explanatory variable.

[0015] According to the above disclosure, the adjustment amount for the second explanatory variable can be easily determined using the predictive model.

[0016] In the disclosure described above, the prediction model predicts the predicted value using the sum of the values ​​obtained by multiplying each measured value of a plurality of explanatory variables by the coefficient corresponding to that explanatory variable. The decision unit determines the initial value of the adjustment amount for the second explanatory variable as the quotient obtained by dividing the value obtained by subtracting the reference value of the dependent variable from the predicted value by the coefficient corresponding to the second explanatory variable. Based on the trend of changes in the first explanatory variable for products that are stuck between the first and second processes, the decision unit calculates the correction amount and corrects the adjustment amount by the correction amount for each product.

[0017] For example, the decision unit determines the predicted value by subtracting the baseline value of the first explanatory variable from the measured value of the first explanatory variable for the predicted product, the second variation obtained by subtracting the baseline value of the first explanatory variable from the measured value of the first explanatory variable for the latest product, the initial value, and the number of products that have been held up between the first and second processes. {(Second variation - First variation) / First variation} × Initial value / Number The correction amount is calculated by substituting the values ​​into the formula.

[0018] According to the above disclosure, even for products that remain stuck between the first and second processes, fluctuations in the target variable can be suppressed.

[0019] According to another example of the present disclosure, a control method for controlling a production line including a plurality of processes includes steps 1 to 6. The first step is to obtain a prediction model obtained by performing data processing of the T method on a plurality of sample data sets indicating values of a plurality of explanatory variables related to the operation of the production line and values of an objective variable related to the quality of products produced by the production line. The prediction model includes, for each of the plurality of explanatory variables, a signal-to-noise ratio (SN ratio) representing the prediction accuracy of the value of the objective variable. The second step is to collect measured values of the plurality of explanatory variables from the production line. The third step is to calculate, for each product, a predicted value of the objective variable by inputting the measured values of the plurality of explanatory variables into the prediction model. The fourth step is to select an explanatory variable to be adjusted from among the plurality of explanatory variables in response to the predicted value falling outside the control range. The fifth step is to determine an adjustment amount of the explanatory variable to be adjusted so that the value of the objective variable approaches the center of the control range. The sixth step is to control the production line so that the value of the explanatory variable to be adjusted changes by the adjustment amount. The fourth step includes dividing the plurality of explanatory variables into a first group and a second group having a smaller influence on the objective variable than the first group based on the SN ratio, and selecting an explanatory variable to be adjusted from among the explanatory variables belonging to the first group.

[0020] According to another example of the present disclosure, a program causes a computer to execute the above control method. Also, by these disclosures, production of products with stable quality is maintained.

Effects of the Invention

[0021] According to the present disclosure, production of products with stable quality is maintained.

Brief Description of the Drawings

[0022] [Figure 1] It is a schematic diagram showing a flow of the control method according to the embodiment. [Figure 2] It is a diagram showing a specific example of a system including a control device according to the present embodiment. [Figure 3]It is a block diagram showing an example of the hardware configuration of the control device according to the present embodiment. [Figure 4] It is a block diagram showing an example of the functional configuration of the control device according to the present embodiment. [Figure 5] It is a diagram showing an example of the transition of the value of the target variable predicted by the prediction unit. [Figure 6] It is a diagram for explaining a method of selecting explanatory variables to be adjusted. [Figure 7] It is a flowchart showing the flow of processing of the control device in the learning phase. [Figure 8] It is a flowchart showing the flow of the first processing of the control device in the operation phase. [Figure 9] It is a flowchart showing the flow of the second processing of the control device in the operation phase. [Figure 10] It is a flowchart showing the flow of the third processing of the control device in the operation phase. [Figure 11] It is a diagram showing an example when feedback control is executed according to the flowchart shown in FIG. 8. [Figure 12] It is a diagram showing the first timing of the first example when feedback control is executed according to the flowchart shown in FIG. 9. [Figure 13] It is a diagram showing the second timing of the first example when feedback control is executed according to the flowchart shown in FIG. 9. [Figure 14] It is a diagram showing the first timing of the second example when feedback control is executed according to the flowchart shown in FIG. 9. [Figure 15] It is a diagram showing the first timing of an example when feedback control and feedforward control are executed according to the flowchart shown in FIG. 10. [Figure 16] It is a diagram showing the second timing of an example when feedback control and feedforward control are executed according to the flowchart shown in FIG. 10.

Embodiments for Carrying Out the Invention

[0023] Embodiments of the present invention will be described in detail with reference to the drawings. Note that identical or corresponding parts in the drawings are denoted by the same reference numerals, and their descriptions will not be repeated.

[0024] §1 Examples of Application Referring to Figure 1, an example of a scenario in which the present invention is applied will be described. Figure 1 is a schematic diagram showing the flow of a control method according to an embodiment. Figure 1 shows the flow of a control method for controlling a production line 300 that produces products. The control method shown in Figure 1 is executed by one or more processors. One or more processors may be incorporated into a single control device, or they may be distributed among multiple control devices that can communicate with each other.

[0025] Production line 300 includes multiple processes. Production line 300, shown in Figure 1, includes processes 1 through 4. Various types of equipment are installed in each of processes 1 through 4, and products are produced using these various types of equipment. In process 1, the values ​​of explanatory variables a1, a2, ... related to the operation of the equipment belonging to process 1 are collected. In process 2, the values ​​of explanatory variables b1, b2, ... related to the operation of the equipment belonging to process 2 are collected. Similarly, in process 3, the values ​​of variables related to the operation of the equipment belonging to process 3 are collected. Process 4 is the final process. In process 4, the value of the objective variable related to the quality of the produced product is collected.

[0026] In the fourth step, the value of the target variable is collected, and the quality of the product is determined. If a product is defective, there is a high probability that other products produced under the same conditions will also be defective. It takes a predetermined amount of time for one product to go through the first to fourth steps. Therefore, if a product is determined to be defective in the fourth step, there is a high probability that defective products will be included among the multiple products that have been stuck in the first to fourth steps.

[0027] The control method according to this embodiment comprises steps (1) to (6) in order to minimize the number of defective products produced and to maintain the production of products of stable quality. As shown in Figure 1, the control method includes a learning phase and an operation phase. The operation phase is performed after the learning phase is performed. Step (1) is performed in the learning phase. Steps (2) to (6) are performed in the operation phase.

[0028] Step (1) is the step of obtaining a predictive model. The predictive model is obtained by performing T-method data processing on multiple sample datasets that show the values ​​of multiple explanatory variables related to the operation of production line 300 and the value of the dependent variable related to the quality of the products produced by production line 300. The T-method is one of the multivariate analysis methods provided by statistician Genichi Taguchi, and is a method for predicting the value of one dependent variable from the values ​​of multiple explanatory variables (see Non-Patent Literature 2). In the example shown in Figure 1, the sample dataset shows the values ​​of explanatory variables a1, a2, ..., b1, b2, ... related to the operation of one or more pieces of equipment installed in the first and second processes, and the value of the dependent variable d. Therefore, by using the predictive model, the value of the dependent variable can be predicted for products after the second process is completed.

[0029] The T-method data processing generates a predictive model that calculates the predicted value of the dependent variable by performing simple linear regression for each explanatory variable and then performing a weighted average for each explanatory variable. The predictive model obtained by the T-method data processing is represented by the following equation (1).

[0030]

number

[0031] In equation (1), Xij is the normalized value of the j-th explanatory variable for the i-th product. Normalization is the process of subtracting the mean value of the variable as shown by the sample dataset. ηj represents the signal-to-noise ratio (SNR) of the j-th explanatory variable. The SNR ηj indicates the linearity between the value of the j-th explanatory variable and the value of the dependent variable, and represents the prediction accuracy of the dependent variable. βj represents the proportionality constant of simple regression. Mi is the normalized predicted value of the dependent variable for the i-th product.

[0032] Step (2) is the step of obtaining measured values ​​of explanatory variables a1, a2, ..., b1, b2, ... from the first and second processes of the production line 300.

[0033] Step (3) is the step of calculating the predicted value of the dependent variable d by inputting the measured values ​​of the explanatory variables a1, a2, ..., b1, b2, ... for each product into the prediction model. In equation (1) above, Mi is the predicted value of the dependent variable after normalization processing for the i-th product. Therefore, the predicted value of the dependent variable d is calculated by adding the value of Mi obtained by inputting the measured values ​​of the explanatory variables a1, a2, ..., b1, b2, ... into the prediction model and the mean value of the dependent variable shown by the sample dataset.

[0034] Step (4) is the step of selecting an explanatory variable to adjust from among the explanatory variables a1, a2, ..., b1, b2, ..., in response to the predicted value of the dependent variable d falling outside the control range. In Step (4), based on the signal-to-noise ratio, the explanatory variables a1, a2, ..., b1, b2, ..., are divided into a first group and a second group which has a smaller impact on the dependent variable than the first group. Then, an explanatory variable to adjust is selected from among the explanatory variables belonging to the first group. In the example shown in Figure 1, explanatory variable b2 is selected as the one to adjust.

[0035] Step (5) is the step of determining the amount of adjustment for the explanatory variable to be adjusted so that the value of the dependent variable d approaches the center of the control range.

[0036] Step (6) is the step of controlling the production line so as to change the value of the explanatory variable to be adjusted by the adjustment amount. If explanatory variable b2 is selected as the variable to be adjusted, the equipment of the second process is controlled.

[0037] According to this embodiment, the value of the dependent variable is predicted from the measured values ​​of multiple explanatory variables. If the predicted value falls outside the control range, an explanatory variable to be adjusted is selected from the first group, which has a relatively large impact on the dependent variable. Then, the amount of adjustment for the adjusted explanatory variable is determined so that the value of the dependent variable approaches the center of the control range, and the production line is controlled to change the value of the adjusted explanatory variable by the adjustment amount. As a result, the number of defective products produced is reduced, and the production of products with stable quality is maintained.

[0038] §2 Specific Examples <System Configuration> Figure 2 shows a specific example of a system including a control device according to this embodiment. As shown in Figure 2, system 1 includes a control device 100, an HMI (Human Machine Interface) 200, and a production line 300. The control device 100 and the HMI 200 are communicated via an information network 6. The control device 100 and one or more devices included in the production line 300 are communicated via a control network 4.

[0039] The control device 100 is typically a PLC (Programmable Logic Controller) and controls one or more pieces of equipment included in the production line 300.

[0040] The HMI200 includes functions for presenting information to the user and for receiving user input. In this embodiment, the HMI200 provides the user with information such as the time-dependent changes in the predicted value of the target variable monitored by the control device 100.

[0041] The production line 300 includes one or more pieces of equipment for producing a product. The production line 300 illustrated in Figure 2 produces painted metal plates 400. The production line 300 includes a paint preparation step 310, a painting step 320, a drying step 330, and an inspection step 340.

[0042] The equipment installed in the paint preparation process 310 includes a raw material dispenser 311, a stirrer 312, and a paint storage container 313. The raw material dispenser 311 is a device that dispenses the raw materials of the paint (pigments, resins, additives, solvents, etc.) into a container. The stirrer 312 stirs and mixes the raw materials in the container. The paint storage container 313 stores the paint at an appropriate temperature (paint storage temperature) to maintain the viscosity of the paint within a predetermined range. Explanatory variables related to the operation of the equipment installed in the paint preparation process 310 include the paint dilution ratio, stirring speed, and paint storage temperature.

[0043] The equipment installed in the painting process 320 includes a coating device 321 and a belt conveyor 322. The coating device 321 uses compressed air to spray paint supplied from the paint storage container 313 onto the metal plate 400 being transported by the belt conveyor 322. Explanatory variables related to the operation of the equipment installed in the painting process 320 include the feed rate of the metal plate 400 by the belt conveyor 322, the pressure of the compressed air (air pressure), the amount of paint discharged from the coating device 321, the distance between the coating device 321 and the metal plate 400 (spraying distance), and the ambient temperature around the coating device 321.

[0044] The equipment installed in the drying process 330 includes a dryer 331 and a belt conveyor 332. The dryer 331 dries the paint on the metal plate 400 by heating the metal plate 400 being transported by the belt conveyor 332. Explanatory variables related to the operation of the equipment installed in the drying process 330 include the feed rate of the metal plate 400 by the belt conveyor 332 and the drying temperature.

[0045] The values ​​of explanatory variables related to the operation of the equipment installed in the paint preparation process 310, the painting process 320, and the drying process 330 are collected by the control device 100. The values ​​of the explanatory variables may be measured by sensors included in the equipment, or they may be calculated based on data (such as command values) output from the control device 100 to the equipment.

[0046] Inspection step 340 is a step in which the paint film thickness of the metal plate 400 is measured. The paint film thickness may be measured by an operator or automatically using a film thickness measuring machine. The paint film thickness measured in inspection step 340 corresponds to the objective variable indicating the quality of the product produced by the production line 300.

[0047] <Control device hardware configuration> Figure 3 is a block diagram showing an example of the hardware configuration of the control device according to this embodiment. As shown in Figure 3, the control device 100 includes a processor 102 such as a CPU (Central Processing Unit) or MPU (Micro-Processing Unit), a chipset 104, main memory 106, storage 110, a control system network controller 120, an information system network controller 122, a USB controller 124, and a memory card interface 126.

[0048] The processor 102 reads various programs stored in the storage 110, loads them into the main memory 106, and executes them to perform control calculations for controlling the controlled object. The chipset 104 controls data transmission between the processor 102 and each component.

[0049] The storage 110 stores a system program 112 for performing basic processing, a user program 114 for performing control calculations, and a monitoring program 116 for monitoring the operating status of one or more pieces of equipment included in the production line 300.

[0050] The control system network controller 120 controls the exchange of data with devices via the control system network 4.

[0051] The information network controller 122 controls the exchange of data with the HMI 100 and other devices via the information network 6.

[0052] The USB controller 124 controls the exchange of data with an external device (e.g., a support device) via a USB connection.

[0053] The memory card interface 126 is configured to allow the insertion and removal of the memory card 128, enabling data to be written to the memory card 128 and various types of data (such as user programs) to be read from the memory card 128.

[0054] Figure 3 shows an example configuration in which the processor 102 provides the necessary processing by executing a program. However, some or all of these provided processes may be implemented using dedicated hardware circuits (e.g., ASIC or FPGA). Alternatively, the main part of the control device 100 may be implemented using hardware that conforms to a general-purpose architecture (e.g., an industrial PC based on a general-purpose PC). In this case, virtualization technology may be used to run multiple operating systems with different purposes in parallel, and to run the necessary applications on each OS.

[0055] <Functional Configuration of Control Device> Figure 4 is a block diagram showing an example of the functional configuration of a control device according to this embodiment. As shown in Figure 4, the control device 100 includes an IO processing unit 10, a control calculation unit 11, an analysis unit 12, a prediction unit 13, a selection unit 14, and a decision unit 15. The IO processing unit 10 and the control calculation unit 11 are realized by the processor 102 executing a user program 114. The analysis unit 12, prediction unit 13, selection unit 14, and decision unit 15 are realized by the processor 102 executing a monitoring program 116.

[0056] The I / O processing unit 10 performs data collection and output processing. Data collection is the process of collecting data from one or more devices included in the production line 300. Output processing is the process of outputting the data obtained by the control calculation unit 11 to one or more devices included in the production line 300.

[0057] The data collected by the data collection process includes measured values ​​of multiple explanatory variables related to the operation of one or more pieces of equipment included in the production line 300 (paint dilution ratio, stirring speed, paint storage temperature, feed rate (painting process), air pressure, room temperature, discharge volume, spray distance, feed rate (drying process), drying temperature). Furthermore, the data collected by the data collection process includes measured values ​​of the objective variable (paint film thickness) that indicates the quality of the products produced by the production line 300. The data shows the measured values ​​of the variables and a work ID that identifies individual products. The work ID is, for example, a serial number.

[0058] The IO processing unit 10 collects the latest data from each process of the production line 300 at regular intervals. A certain number of products accumulate at each process and between processes of the production line 300. Therefore, the IO processing unit 10 collects, for example, the measured value data of the explanatory variable "paint storage temperature" for product with work ID "2200" and the measured value data of the explanatory variable "feed rate (painting process)" for product with work ID "2000" at the same interval. The data collected by the IO processing unit 10 is stored.

[0059] The control calculation unit 11 uses the data collected by the data collection process to perform calculations for controlling one or more pieces of equipment included in the production line 300. When the control calculation unit 11 receives an adjustment amount from the determination unit 15, it calculates the operating amount of the equipment corresponding to the explanatory variable selected by the selection unit 14 so as to change the value of the explanatory variable to be adjusted by the adjustment amount. The calculated operating amount is output to the equipment, causing the value of the explanatory variable to be adjusted to change by the adjustment amount.

[0060] The analysis unit 12 performs multivariate analysis on the data collected by the IO processing unit 10 during the learning phase. The learning phase is, for example, the period of mass production experiments or a certain period immediately after the start of mass production. The analysis unit 12 extracts a sample dataset for each product from the data set collected by the IO processing unit 10 during the learning phase. The sample dataset includes data indicating the work ID of the corresponding product, data indicating the values ​​of each of the multiple explanatory variables, and data indicating the value of the objective variable (paint film thickness).

[0061] In this embodiment, the sample dataset includes data showing the values ​​of each of the explanatory variables "paint dilution ratio," "stirring speed," and "paint storage temperature" collected from the paint preparation step 310; the explanatory variables "feed rate (painting step)," "air pressure," "discharge volume," "spraying distance," and "room temperature" collected from the painting step 320; and the objective variable "paint film thickness" collected from the inspection step 340. The sample dataset does not include data showing the values ​​of each of the explanatory variables "feed rate (drying step)" and "drying temperature" collected from the drying step 330.

[0062] The analysis unit 12 performs T-method data processing on multiple sample datasets corresponding to multiple products. Specifically, the analysis unit 12 performs data processing according to the following steps (1) to (4).

[0063] Procedure (1): For each explanatory variable, calculate the mean of the values ​​shown by multiple sample datasets. Similarly, for the dependent variable, calculate the mean of the values ​​shown by multiple sample datasets.

[0064] Step (2): Perform a normalization process for the values ​​shown by multiple sample datasets. That is, for each variable, subtract the mean value of that variable from the values ​​shown by multiple sample datasets.

[0065] Procedure (3): For each explanatory variable, calculate the proportionality constant β and the signal-to-noise ratio η from the relationship between the normalized value of the explanatory variable and the normalized value of the dependent variable. The normalized explanatory variable and dependent variable have a zero-point proportional relationship. A zero-point proportional relationship means a linear relationship passing through the origin zero. The proportionality constant β indicates the slope of this linear relationship. The signal-to-noise ratio η is the reciprocal of the degree of data variability with respect to the zero-point proportional relationship.

[0066] Step (4): Using the proportionality constant β and SB ratio η calculated for each explanatory variable, determine a predictive model to predict the value of the dependent variable. That is, determine equation (1) above.

[0067] The prediction unit 13 uses data collected by the IO processing unit 10 during the operation phase to predict the value of the target variable for each product. Specifically, the prediction unit 13 inputs the measured values ​​of the explanatory variables "paint dilution ratio", "stirring speed", "paint storage temperature", "feeding speed (painting process)", "air pressure", "discharge volume", "spraying distance", and "room temperature" corresponding to the same work ID into the prediction model to calculate the predicted value of the target variable "paint film thickness".

[0068] Figure 5 shows an example of the trend of the value of the target variable predicted by the prediction unit. As shown in Figure 5, the value of the target variable is predicted for each product. The prediction unit 13 may also provide the HMI200 with a graph showing the trend of the predicted value of the target variable, as shown in Figure 5.

[0069] The selection unit 14 selects an explanatory variable to be adjusted from among several explanatory variables in response to the predicted value of the dependent variable falling outside the control range. The control range is predetermined according to the quality required for the product. The selection unit 14 may select an explanatory variable to be adjusted in response to the predicted value corresponding to a certain product falling outside the control range. Alternatively, the selection unit 14 may select an explanatory variable to be adjusted in response to the average of the predicted values ​​corresponding to a predetermined number of consecutive products falling outside the control range. In the example shown in Figure 5, the selection unit 14 selects an explanatory variable to be adjusted at timing T1, when the average of the predicted values ​​corresponding to a predetermined number of consecutive products falls outside the control range.

[0070] Figure 6 illustrates the method for selecting explanatory variables to be adjusted. The selection unit 14 divides multiple explanatory variables into a first group and a second group which has less influence on the target variable than the first group, based on the signal-to-noise ratio (SNR) η of each explanatory variable. For example, the selection unit 14 may designate the top predetermined number of explanatory variables with large SNRs as the first group and the remaining explanatory variables as the second group. Alternatively, the selection unit 14 may designate explanatory variables with an SNR greater than a threshold as the first group and explanatory variables with an SNR below a threshold as the second group. In the example shown in Figure 6, the top four explanatory variables with large SNRs, "feed rate (painting process)", "spray distance", "stirring speed", and "paint storage temperature", belong to the first group.

[0071] Alternatively, the selection unit 14 may consider the proportionality constant β in addition to the signal-to-noise ratio η and divide the multiple explanatory variables into a first group and a second group. For example, the selection unit 14 may designate the top predetermined number of explanatory variables whose values ​​obtained by multiplying the signal-to-noise ratio η by a coefficient corresponding to the absolute value of the proportionality constant β are large as the first group, and the remaining explanatory variables as the second group. Or, the selection unit 14 may designate the explanatory variables whose signal-to-noise ratio η is greater than a first threshold and whose absolute value of the proportionality constant β is greater than a second threshold as the first group, and the explanatory variables whose signal-to-noise ratio η is less than or equal to the threshold as the second group.

[0072] The selection unit 14 calculates the degree of variation of each explanatory variable belonging to the first group based on the trend of the measured value. For example, the selection unit 14 calculates the degree of variation by dividing the difference between the measured value and the reference value by the width of the control range (the difference between the lower limit and the upper limit). Based on the calculated degree of variation, the selection unit 14 selects the explanatory variable to be adjusted.

[0073] Explanatory variables with a large degree of variation are considered to be the cause of the predicted value of the dependent variable falling outside the control range. Therefore, the selection unit 14 may select the first explanatory variable with the largest degree of variation as the explanatory variable to be adjusted. In this case, the variation in the value of the dependent variable can be directly suppressed by adjusting the value of the first explanatory variable. In other words, the control that adjusts the value of the first explanatory variable corresponds to feedback control.

[0074] Alternatively, the selection unit 14 may select a second explanatory variable from among the explanatory variables belonging to the first group that corresponds to a process later than the process corresponding to the first explanatory variable as the explanatory variable to be adjusted. In this case, fluctuations in the value of the target variable caused by fluctuations in the value of the first explanatory variable can be canceled out by adjusting the value of the second explanatory variable. That is, the control that adjusts the value of the second explanatory variable corresponds to feedforward control. It is preferable that the selection unit 14 selects the second explanatory variable with the largest signal-to-noise ratio η from among the explanatory variables corresponding to a process later than the process corresponding to the first explanatory variable as the explanatory variable to be adjusted. This makes it easier to suppress fluctuations in the value of the target variable.

[0075] In the example shown in Figure 6, among the explanatory variables belonging to the first group—"feed rate (painting process)," "spray distance," "stirring rate," and "paint storage temperature"—the measured value of the explanatory variable "paint storage temperature" is fluctuating. Therefore, it is considered that the reason the predicted value of the objective variable "paint film thickness" falls outside the control range is due to the fluctuation in the explanatory variable "paint storage temperature." Accordingly, the selection unit 14 may select the explanatory variable "paint storage temperature" (first explanatory variable) as the explanatory variable to be adjusted.

[0076] The first explanatory variable, "paint storage temperature," corresponds to the paint preparation process 310. The first group includes explanatory variables "feed rate (painting process)" and "spray distance," which correspond to the painting process 320 that occurs after the paint preparation process 310. Therefore, the selection unit 14 may determine either the explanatory variable "feed rate (painting process)" or "spray distance" (second explanatory variable) as the explanatory variable to be adjusted. Preferably, the selection unit 14 selects "feed rate (painting process)," which has a larger signal-to-noise ratio η, as the explanatory variable to be adjusted among the explanatory variables "feed rate (painting process)" and "spray distance."

[0077] The decision unit 15 determines the amount of adjustment for the explanatory variable to be adjusted so that the value of the dependent variable approaches the center of the control range.

[0078] For example, if the first explanatory variable is selected as the explanatory variable to be adjusted, the determination unit 15 determines the adjustment amount for the first explanatory variable as the amount obtained by subtracting the measured value of the first explanatory variable from the reference value of the first explanatory variable. The reference value is, for example, the center of the standard. As a result, the control calculation unit 11 calculates the operation amount of one or more devices included in the production line 300 so that the value of the first explanatory variable approaches the center of the standard. When the calculated operation amount is output to the device, the value of the first explanatory variable approaches the center of the standard, and the value of the target variable approaches the center of the control range.

[0079] If a second explanatory variable is selected as the explanatory variable to be adjusted, the determination unit 15 determines the adjustment amount for the second explanatory variable based on the value obtained by subtracting the reference value from the predicted value of the dependent variable and the coefficient of the second explanatory variable in equation (1) above. The specific method for determining the adjustment amount for the second explanatory variable will be described later.

[0080] <Processing flow of the control device> (Learning Phase) Figure 7 is a flowchart showing the processing flow of the control device during the learning phase. First, the processor 102 acquires a sample dataset for each product (step S1). Next, the processor 102 performs T-method multivariate analysis on multiple sample datasets (step S2). Based on the proportionality constant β and SN ratio η of each explanatory variable obtained by the execution of step S2, the processor 102 generates a predictive model shown by equation (1) (step S3). After step S3, the processing ends.

[0081] (Operational Phase) During the operation phase, the control device 100 executes one of the following first to third processes.

[0082] 《First Process》 Figure 8 is a flowchart showing the first processing flow of the control device during the operation phase. First, the processor 102 acquires measured values ​​of multiple explanatory variables for each product (step S11). Next, the processor 102 predicts the value of the target variable by inputting the measured values ​​of the multiple explanatory variables into a prediction model (step S12).

[0083] The processor 102 determines whether the predicted value of the target variable is outside the control range (step S13). If the predicted value of the target variable is not outside the control range (NO in step S13), the process returns to step S11.

[0084] If the predicted value of the dependent variable falls outside the control range (YES in step S13), the processor 102 calculates the degree of variation of each explanatory variable belonging to the first group with a relatively large signal-to-noise ratio η (step S14). The processor 102 selects the first explanatory variable with the greatest degree of variation as the explanatory variable to be adjusted and determines the adjustment amount for the first explanatory variable (step S15). Specifically, the processor 102 determines the adjustment amount as the amount obtained by subtracting the measured value of the first explanatory variable from the reference value of the first explanatory variable. Next, the processor 102 controls the production line 300 so as to change the value of the first explanatory variable by the adjustment amount (step S16). Step S16 corresponds to feedback control. After step S16, the process ends. The control in step S16 continues at least until the value of the dependent variable for the latest product fed into the process corresponding to the first explanatory variable at the time the control was started is predicted.

[0085] 《Second Process》 Figure 9 is a flowchart showing the second processing flow of the control device during the operation phase. The flowchart in Figure 9 differs from the flowchart in Figure 8 in that it includes steps S17 to S19.

[0086] As shown in Figure 9, after step S14, the processor 102 determines whether feedforward control is possible (step S17). Specifically, the processor 102 determines that feedforward control is possible if there is a second explanatory variable in the first group that corresponds to a process later than the process corresponding to the first explanatory variable.

[0087] If feedforward control is not possible (NO in step S17), steps S15 and S16 are executed.

[0088] If feedforward control is possible (YES in step S17), the processor 102 selects the second explanatory variable as the explanatory variable to be adjusted and determines the amount of adjustment for the second explanatory variable (step S18). Next, the processor 102 controls the equipment included in the production line 300 so that the value of the second explanatory variable changes by the amount of adjustment (step S19). Step S19 corresponds to feedforward control.

[0089] After steps S16 and S19, the process ends. The control in steps S16 and S19 continues at least until the value of the target variable for the latest product fed into the process corresponding to the first explanatory variable at the time the control was started is predicted.

[0090] 《Third process》 Figure 10 is a flowchart showing the third processing flow of the control device during the operation phase. The flowchart in Figure 10 differs from the flowchart in Figure 9 in that step S17 is executed after step S16, not after step S14. That is, in the flowchart in Figure 10, the first explanatory variable is always selected as the variable to be adjusted, and the second explanatory variable is also selected as the variable to be adjusted if feedforward control is possible. Therefore, both feedback control and feedforward control are performed.

[0091] <Example of feedback control> Figure 11 shows an example of when feedback control is performed according to the flowchart shown in Figure 8. In the example shown in Figure 11, at the first timing when painting of product with work ID "2200" is completed in painting process 320, the preparation of paint for product with work ID "2400" is completed in paint preparation process 310, and the inspection of product with work ID "2000" is completed in inspection process 340. That is, products with work IDs "2201" to "2400" are held up from paint preparation process 310 to painting process 320, and products with work IDs "2001" to "2200" are held up from painting process 320 to inspection process 340. In addition, the explanatory variables belonging to the first group are "feed rate (painting process)", "spray distance", "stirring rate", and "paint storage temperature", and the SN ratio η of these explanatory variables is shown in Figure 6. Note that these conditions are the same in the example described later with reference to Figures 12 to 16.

[0092] At the time when painting is completed on product with work ID "2200", all measurement values ​​of the explanatory variables corresponding to that product—"paint dilution ratio", "stirring speed", "paint storage temperature", "feed rate (painting process)", "spraying distance", "air pressure", "discharge volume", and "room temperature"—are acquired. Therefore, at the first timing, the value of the objective variable "paint film thickness" is predicted for product with work ID "2200" (step S12a).

[0093] Depending on whether the predicted value of the dependent variable "paint film thickness" falls outside the control range, the degree of variation of the measured values ​​of the independent variables belonging to the first group is calculated. In the example shown in Figure 11, the degree of variation of the measured value of the independent variable "paint storage temperature" is the largest. Therefore, the first independent variable "paint storage temperature" is selected as the target for adjustment, and the adjustment amount is determined (step S15a). Subsequently, the paint storage unit 313 is controlled to change the value of the first independent variable "paint storage temperature" by the adjustment amount (step S16a). The adjustment amount is the amount obtained by subtracting the measured value from the reference value (standard center) of the independent variable "paint storage temperature". As a result, for products with work ID "2401" and later, the value of the independent variable "paint storage temperature" approaches the standard center. This brings the predicted value of the dependent variable "paint film thickness" within the control range for products with work ID "2401" and later.

[0094] <First example of feedforward control> Figure 12 shows the first timing in the first example when feedback control is performed according to the flowchart shown in Figure 9. In the example shown in Figure 12, at the first timing when painting is completed on product with work ID "2200", all measured values ​​of the explanatory variables corresponding to that product are obtained: "paint dilution ratio", "stirring speed", "paint storage temperature", "feed rate (painting process)", "spraying distance", "air pressure", "discharge volume", and "room temperature". Therefore, at this timing, the value of the objective variable "paint film thickness" is predicted for product with work ID "2200" (step S12b).

[0095] Then, depending on whether the predicted value of the dependent variable "paint film thickness" falls outside the control range, the degree of variation of the measured values ​​of the explanatory variables belonging to the first group is calculated, and the explanatory variable "paint storage temperature" is identified as the first explanatory variable with the largest degree of variation (step S14b). However, within the first group, there are second explanatory variables "feed rate (painting process)" and "spray distance" that correspond to the painting process 320, which is later than the paint preparation process 310 corresponding to the first explanatory variable "paint storage temperature". The signal-to-noise ratio η of the explanatory variable "feed rate (painting process)" is greater than the signal-to-noise ratio η of the explanatory variable "spray distance". Therefore, the second explanatory variable "feed rate (painting process)" is selected as the target for adjustment, and the amount of adjustment is determined.

[0096] The above equation (1), which shows the predictive model, represents the relationship between the variation of each explanatory variable and the variation of the dependent variable. Therefore, in the example shown in Figure 12, the processor 102 extracts the coefficient Kj of the explanatory variable to be adjusted, "feed rate (painting process)," from the above equation (1). This coefficient Kj is expressed by the following equation (2).

[0097]

number

[0098] The adjustment amount is determined in order to bring the value of the objective variable "paint film thickness" closer to the center of the control range. Therefore, the processor 102 determines the adjustment amount for the second explanatory variable "feed rate (painting process)" as the quotient obtained by dividing the value obtained by subtracting the center of the control range of the objective variable "paint film thickness" from the predicted value of the objective variable "paint film thickness" by the coefficient Kj of the second explanatory variable "feed rate (painting process)" that is to be adjusted. For example, if the value obtained by subtracting the center of the control range of the objective variable "paint film thickness" from the predicted value of the objective variable "paint film thickness" is 0.017 μm, and the coefficient Kj of the second explanatory variable "feed rate (painting process)" is -0.81 μm / (m / s), then -0.009 m / s is determined as the adjustment amount.

[0099] Subsequently, for products with work ID "2201" and later, the belt conveyor 322 is controlled to change the value of the second explanatory variable "feed rate (painting process)" by the adjustment amount (step S19b). As a result, fluctuations in the target variable "paint film thickness" caused by fluctuations in the first explanatory variable "paint storage temperature" are canceled out by the adjustment of the value of the second explanatory variable "feed rate (painting process)". Consequently, for products with work ID "2201" and later, the predicted value of the target variable "paint film thickness" falls within the control range.

[0100] Figure 13 shows the second timing in the first example when feedback control is performed according to the flowchart shown in Figure 9. The second timing is when painting is completed on product with work ID "2400" in painting process 320.

[0101] As mentioned above, for products with work ID "2201" and later, the predicted value of the target variable "paint film thickness" falls within the control range when the value of the second explanatory variable "feed rate (painting process)" is adjusted by the adjustment amount. However, if the degree of fluctuation of the first explanatory variable "paint storage temperature," which is the direct cause of the fluctuation of the target variable "paint film thickness," continues to increase, the target variable "paint film thickness" will also fluctuate again.

[0102] In the example shown in Figure 13, the predicted value of the dependent variable "paint film thickness" for product with work ID "2400" falls outside the control range (step S12c). Then, the independent variable "paint storage temperature" is identified again as the first independent variable (step S14c), and the second independent variable "feed rate (painting process)", which corresponds to the painting process 320 that is later than the paint preparation process 310 corresponding to the first independent variable "paint storage temperature", is selected as the adjustment target. If the value obtained by subtracting the center of the control range for the dependent variable "paint film thickness" from the predicted value of the dependent variable "paint film thickness" is 0.017 μm, and the coefficient Kj of the second independent variable "feed rate (painting process)" is -1.81 μm / (m / s), then a further adjustment amount of -0.009 m / s is determined.

[0103] Subsequently, for products with work ID "2401" and later, the belt conveyor 322 is controlled to further change the value of the second explanatory variable "feed rate (painting process)" by an additional adjustment amount (step S19c). As a result, for products with work ID "2401" and later, the value of the objective variable "paint film thickness" falls within the control range.

[0104] <Second example of feedforward control> In the first example of feedforward control described above, the adjustment amount is determined by subtracting the center of the control range for the target variable "paint film thickness" from the predicted value of the target variable "paint film thickness," and then dividing this value by the coefficient Kj of the second explanatory variable "feed rate (painting process)" that is to be adjusted. This keeps the value of the target variable "paint film thickness" within the control range. However, if the degree of fluctuation of the first explanatory variable "paint storage temperature," which is the direct cause of the fluctuation of the target variable "paint film thickness," continues to increase, the target variable "paint film thickness" will also fluctuate again. In the second example, the adjustment amount is determined by taking into account the fluctuation of the first explanatory variable "paint storage temperature." This ensures that the value of the target variable "paint film thickness" remains stable even if the degree of fluctuation of the first explanatory variable "paint storage temperature" continues to increase.

[0105] Figure 14 shows the first timing in a second example when feedback control is performed according to the flowchart shown in Figure 9. Similar to Figure 12, the explanatory variable "feed rate (painting process)" is selected as the adjustment target and the adjustment amount is determined when the predicted value of the objective variable "paint film thickness" for product with work ID "2200" falls outside the control range.

[0106] In the second example, the processor 102 determines the initial adjustment amount for the second explanatory variable "feed rate (painting process)" as the quotient obtained by dividing the predicted value of the objective variable "paint film thickness" by the coefficient Kj of the second explanatory variable "feed rate (painting process)" to be adjusted.

[0107] Furthermore, the processor 102 corrects the adjustment amount for each product based on the trend of fluctuations in the first explanatory variable "paint storage temperature" for products that remain between the paint preparation process 310, which corresponds to the first explanatory variable "paint storage temperature," and the painting process 320, which corresponds to the second explanatory variable "feed rate (painting process)."

[0108] The processor 102 calculates the variation from the specification center of the measured value of the first explanatory variable "paint storage temperature" (hereinafter referred to as the "first variation") for products where the predicted value of the objective variable "paint film thickness" falls outside the control range. Furthermore, the processor 102 calculates the variation from the specification center of the measured value of the first explanatory variable "paint storage temperature" (hereinafter referred to as the "second variation") for the most recent product collected from the paint preparation process 310. In the example shown in Figure 14, the processor 102 calculates a first variation of "0.7°C" and a second variation of "1.4°C".

[0109] The fluctuation in the predicted value of the target variable "paint film thickness" caused by the first fluctuation is canceled out by adjusting the value of the second explanatory variable "feed rate (painting process)" by the initial adjustment amount mentioned above. However, for products that remain between the paint preparation process 310 and the painting process 320, the first explanatory variable "paint storage temperature" fluctuates further. Therefore, simply adjusting the value of the second explanatory variable "feed rate (painting process)" by the initial adjustment amount is insufficient to suppress the fluctuation in the target variable "paint film thickness" for these remaining products. For this reason, the processor 102 corrects the adjustment amount for each product according to the difference between the second fluctuation and the first fluctuation for these remaining products.

[0110] The processor 102 calculates the correction amount for each product according to the following formula (3). Correction amount = {(Second variation amount - First variation amount) / First variation amount} × Initial adjustment amount / Number of products in storage ··Equation (3) In equation (3), the number of products remaining is the number of products remaining between the paint preparation process 310 and the painting process 320. The number of products remaining is calculated from the difference between the work ID corresponding to the latest measurement value collected from the paint preparation process 310 and the work ID corresponding to the latest measurement value collected from the painting process 320. The processor 102 determines the adjustment amount by continuously adding the correction amount calculated for each product to the initial adjustment amount.

[0111] If the initial adjustment amount is 0.009 m / s and the number of stagnant products is 200, then processor 102 will: {(1.4-0.7) / 0.7}×0.009 / 200=0.000045(m / s) This is determined as the correction amount.

[0112] As a result, as shown in Figure 14, the value of the second explanatory variable, "feed rate (painting process)," is adjusted by the initial adjustment amount for product with work ID "2201," and then by (initial adjustment amount + α × correction amount) for product with work ID "2201+α" (step S19d). Consequently, even though the measured value of the first explanatory variable, "paint storage temperature," continues to fluctuate, the predicted value of the objective variable, "paint film thickness," remains stable within the control range.

[0113] <Examples of implementing feedback control and feedforward control> Figure 15 shows the first timing in an example where feedback control and feedforward control are performed according to the flowchart shown in Figure 10. In the example shown in Figure 15, at the first timing when painting of product with work ID "2200" is completed, all measured values ​​of the explanatory variables corresponding to that product, namely "paint dilution ratio", "stirring speed", "paint storage temperature", "feed rate (painting process)", "spraying distance", "air pressure", "discharge volume", and "room temperature", have been acquired (step S12e).

[0114] Then, depending on whether the predicted value of the objective variable "paint film thickness" falls outside the control range, the degree of variation of the measured values ​​of the explanatory variables belonging to the first group is calculated (step S14e), and the first explanatory variable "paint storage temperature," which has the largest degree of variation, is selected as the target for adjustment. Subsequently, the paint storage unit 313 is controlled to change the value of the first explanatory variable "paint storage temperature" by the adjustment amount (step S16e). The adjustment amount is the amount obtained by subtracting the measured value from the reference value (standard center) of the explanatory variable "paint storage temperature."

[0115] Furthermore, the second explanatory variable "feed rate (painting process)," which corresponds to the painting process 320 that occurs after the paint preparation process 310 corresponding to the first explanatory variable "paint storage temperature," is selected as the adjustment target, and the adjustment amount is determined. Subsequently, the equipment included in the painting process 320 is controlled to change the value of the second explanatory variable "feed rate (painting process)" by the adjustment amount (step S19e). The adjustment amount is determined, for example, according to the method described in the <Second Example of Feedforward Control> above.

[0116] Figure 16 shows the second timing in an example where feedback control and feedforward control are performed according to the flowchart shown in Figure 10. As explained with reference to Figure 15, at the first timing, the belt conveyor 322 is controlled to change the value of the second explanatory variable "feed rate (painting process)" by the adjustment amount. As a result, as shown in Figure 16, the value of the second explanatory variable "feed rate (painting process)" is adjusted by the initial adjustment amount for product with work ID "2201", and then adjusted by (initial adjustment amount + α × correction amount) for product with work ID "2201+α". As a result, for products with work ID "2201" and later, the predicted value of the objective variable "paint film thickness" stabilizes within the control range.

[0117] Furthermore, at the first timing, the paint storage unit 313 is controlled to change the value of the first explanatory variable "paint storage temperature" by an adjustment amount. As a result, for products with work ID "2401" and later, the value of the target variable "paint film thickness" falls within the control range. Therefore, for products with work ID "2401" and later, there is no change in the predicted value of the target variable "paint film thickness" caused by the change in the first explanatory variable "paint storage temperature". For this reason, the processor 102 stops adjusting the value of the second explanatory variable "feed rate (painting process)" for products with work ID "2401" and later.

[0118] §3 Addendum As described above, this embodiment includes the following disclosures.

[0119] (Composition 1) A control device (100) for controlling a production line (300) that includes multiple processes (310, 320, 330, 340), The system includes an acquisition unit (12, 102) that acquires a predictive model obtained by performing T-method data processing on multiple sample datasets representing the values ​​of multiple explanatory variables relating to the operation of the production line (300) and the value of the objective variable relating to the quality of the products (400) produced by the production line (300). The prediction model includes, for each of the plurality of explanatory variables, a signal-to-noise ratio (SNR) representing the prediction accuracy of the value of the target variable. The control device (100) further, A collection unit (10, 102) collects measured values ​​of the multiple explanatory variables from the production line (300), For each product, a prediction unit (13,102) calculates a predicted value of the target variable by inputting the measured values ​​of the multiple explanatory variables into the prediction model, A selection unit (14,102) selects an explanatory variable to be adjusted from among the multiple explanatory variables in response to the predicted value falling outside the control range, A determination unit (15,102) that determines the amount of adjustment for the explanatory variable to be adjusted so that the value of the objective variable approaches the center of the control range, The production line is controlled by a control unit (11, 102) that controls the production line so as to change the value of the explanatory variable to be adjusted by the adjustment amount, The selection unit (14,102) divides the plurality of explanatory variables into a first group and a second group which has less influence on the target variable than the first group, based on the signal-to-noise ratio, and selects the explanatory variable to be adjusted from among the explanatory variables belonging to the first group, the control device (100).

[0120] (Configuration 2) The aforementioned selection units (14, 102) are For each explanatory variable belonging to the first group, calculate the degree of variation of that explanatory variable. The control device (100) according to configuration 1, which selects the first explanatory variable with the greatest degree of variation as the explanatory variable to be adjusted.

[0121] (Composition 3) The aforementioned determination unit (15,102) A control device (100) according to configuration 2, which determines the amount obtained by subtracting the measured value of the first explanatory variable from the reference value of the first explanatory variable as the adjustment amount of the first explanatory variable.

[0122] (Composition 4) The aforementioned selection units (14, 102) are For each explanatory variable belonging to the first group, calculate the degree of variation of that explanatory variable. The control device (100) according to configuration 1, which selects as the explanatory variable to be adjusted the second explanatory variable that corresponds to a second process (320) that is later than the first process (310) that corresponds to the first explanatory variable with the greatest degree of variation among the explanatory variables belonging to the first group.

[0123] (Composition 5) The prediction model predicts the predicted value using the sum of the values ​​obtained by multiplying each of the measured values ​​of the plurality of explanatory variables by the coefficient corresponding to that explanatory variable. The aforementioned determination unit (15,102) The control device (100) according to configuration 4, which determines the adjustment amount of the second explanatory variable as the quotient obtained by dividing the value obtained by subtracting the reference value of the objective variable from the predicted value by the coefficient corresponding to the second explanatory variable.

[0124] (Composition 6) The prediction model predicts the predicted value using the sum of the values ​​obtained by multiplying each of the measured values ​​of the plurality of explanatory variables by the coefficient corresponding to that explanatory variable. The aforementioned determination unit (15,102) The quotient obtained by subtracting the reference value of the dependent variable from the predicted value and dividing it by the coefficient corresponding to the second explanatory variable is determined as the initial value of the adjustment amount of the second explanatory variable. Based on the trend of the change in the first explanatory variable for the product that remains between the first process (310) and the second process (320), a correction amount is calculated. A control device (100) according to configuration 4, which corrects the adjustment amount by the correction amount for each of the aforementioned products (400).

[0125] (Composition 7) The determination unit (15,102) determines the first variation obtained by subtracting the reference value of the first explanatory variable from the measured value of the first explanatory variable for the product for which the predicted value was obtained, the second variation obtained by subtracting the reference value of the first explanatory variable from the measured value of the first explanatory variable for the latest product, the initial value, and the number of products that have been held up between the first and second processes. {(Second variation - First variation) / First variation} × Initial value / Number The control device (100) described in configuration 6 calculates the correction amount by substituting it into the formula.

[0126] (Composition 8) A control method for controlling a production line (300) that includes multiple processes (310, 320, 330, 340), The method includes the step of obtaining a predictive model obtained by performing T-method data processing on multiple sample datasets that show the values ​​of multiple explanatory variables relating to the operation of the production line and the value of an objective variable relating to the quality of the products produced by the production line. The prediction model includes, for each of the plurality of explanatory variables, a signal-to-noise ratio (SNR) representing the prediction accuracy of the value of the target variable. The control method further includes, The steps include collecting measured values ​​of the multiple explanatory variables from the production line, For each product, the steps include: calculating the predicted value of the target variable by inputting the measured values ​​of the plurality of explanatory variables into the prediction model; The steps include selecting an explanatory variable to be adjusted from among the multiple explanatory variables in response to the predicted value falling outside the control range, The steps include determining the amount of adjustment for the explanatory variable to be adjusted so that the value of the objective variable approaches the center of the control range, The system includes the step of controlling the production line so as to change the value of the explanatory variable to be adjusted by the adjustment amount, A control method comprising the step of selecting, based on the signal-to-noise ratio, dividing the plurality of explanatory variables into a first group and a second group having less influence on the target variable than the first group, and selecting the explanatory variable to be adjusted from among the explanatory variables belonging to the first group.

[0127] (Composition 9) A program (114,116) that causes the computer to execute the control method described in Configuration 8.

[0128] While embodiments of the present invention have been described, the embodiments disclosed herein should be considered in all respects to be illustrative and not restrictive. The scope of the present invention is defined by the claims, and all modifications within the meaning and scope equivalent to the claims are intended to be included. [Explanation of Symbols]

[0129] 1 System, 4 Control Network, 6 Information Network, 10 IO Processing Unit, 11 Control Calculation Unit, 12 Analysis Unit, 13 Prediction Unit, 14 Selection Unit, 15 Decision Unit, 100 Control Device, 102 Processor, 104 Chipset, 106 Main Memory, 110 Storage, 112 System Program, 114 User Program, 116 Monitoring Program, 120 Control Network Controller, 122 Information Network Controller, 124 USB Controller, 126 Memory Card Interface, 128 Memory Card, 300 Production Line, 310 Paint Preparation Process, 311 Raw Material Input Machine, 312 Mixer, 313 Paint Storage Unit, 320 Painting Process, 321 Coating Equipment, 322, 332 Belt Conveyor, 330 Drying Process, 331 Dryer, 340 Inspection Process, 400 Metal Plate.

Claims

1. A control device for controlling a production line that includes multiple processes, The system includes an acquisition unit that acquires a predictive model obtained by performing T-method data processing on multiple sample datasets representing the values ​​of multiple explanatory variables related to the operation of the production line and the value of a target variable related to the quality of the products produced by the production line. The prediction model includes, for each of the plurality of explanatory variables, an S / N ratio representing the prediction accuracy of the value of the target variable, The control device further, A collection unit that collects measured values ​​of the multiple explanatory variables from the production line, For each product, a prediction unit calculates a predicted value of the target variable by inputting the measured values ​​of the plurality of explanatory variables into the prediction model. A selection unit that selects an explanatory variable to be adjusted from among the multiple explanatory variables in response to the predicted value falling outside the control range, A determination unit that determines the amount of adjustment for the explanatory variable to be adjusted so that the value of the objective variable approaches the center of the control range, The production line is controlled by a control unit that changes the value of the explanatory variable to be adjusted by the adjustment amount, The selection unit divides the plurality of explanatory variables into a first group and a second group which has less influence on the target variable than the first group, based on the signal-to-noise ratio, and selects the explanatory variable to be adjusted from among the explanatory variables belonging to the first group. The aforementioned selection unit is For each explanatory variable belonging to the first group, calculate the degree of variation of that explanatory variable. A control device that selects the first explanatory variable with the greatest degree of variation as the explanatory variable to be adjusted.

2. The aforementioned determination unit, The control device according to claim 1, wherein the amount obtained by subtracting the measured value of the first explanatory variable from the reference value of the first explanatory variable is determined as the adjustment amount of the first explanatory variable.

3. A control device for controlling a production line that includes multiple processes, The system includes an acquisition unit that acquires a predictive model obtained by performing T-method data processing on multiple sample datasets representing the values ​​of multiple explanatory variables related to the operation of the production line and the value of a target variable related to the quality of the products produced by the production line. The prediction model includes, for each of the plurality of explanatory variables, an S / N ratio representing the prediction accuracy of the value of the target variable, The control device further, A collection unit that collects measured values ​​of the multiple explanatory variables from the production line, For each product, a prediction unit calculates a predicted value of the target variable by inputting the measured values ​​of the plurality of explanatory variables into the prediction model. A selection unit that selects an explanatory variable to be adjusted from among the multiple explanatory variables in response to the predicted value falling outside the control range, A determination unit that determines the amount of adjustment for the explanatory variable to be adjusted so that the value of the objective variable approaches the center of the control range, The production line is controlled by a control unit that changes the value of the explanatory variable to be adjusted by the adjustment amount, The selection unit divides the plurality of explanatory variables into a first group and a second group which has less influence on the target variable than the first group, based on the signal-to-noise ratio, and selects the explanatory variable to be adjusted from among the explanatory variables belonging to the first group. The aforementioned selection unit is For each explanatory variable belonging to the first group, calculate the degree of variation of that explanatory variable. A control device that selects, among the explanatory variables belonging to the first group, a second explanatory variable corresponding to a second process that occurs later than the first process corresponding to the first explanatory variable with the greatest degree of variation, as the explanatory variable to be adjusted.

4. The prediction model predicts the predicted value using the sum of the values ​​obtained by multiplying each of the measured values ​​of the plurality of explanatory variables by the coefficient corresponding to that explanatory variable. The aforementioned determination unit, The control device according to claim 3, wherein the quotient obtained by subtracting the reference value of the objective variable from the predicted value and dividing it by the coefficient corresponding to the second explanatory variable is determined as the adjustment amount of the second explanatory variable.

5. The prediction model predicts the predicted value using the sum of the values ​​obtained by multiplying each of the measured values ​​of the plurality of explanatory variables by the coefficient corresponding to that explanatory variable. The aforementioned determination unit, The quotient obtained by subtracting the reference value of the dependent variable from the predicted value and dividing it by the coefficient corresponding to the second explanatory variable is determined as the initial value of the adjustment amount of the second explanatory variable. Based on the trend of the change in the first explanatory variable for the product that remains between the first and second processes, a correction amount is calculated. The control device according to claim 3, wherein the adjustment amount is corrected by the correction amount for each of the aforementioned products.

6. The determination unit calculates the predicted value by subtracting the reference value of the first explanatory variable from the measured value of the first explanatory variable for the product for which the predicted value was obtained, subtracting the reference value of the first explanatory variable from the measured value of the first explanatory variable for the latest product, the initial value, and the number of products that have been held up between the first and second processes, using the formula: {(second variation - first variation) / first variation} × initial value / number The control device according to claim 5, which calculates the correction amount by substituting into the formula.

7. A control method for controlling a production line that includes multiple processes, The method includes the step of obtaining a predictive model obtained by performing T-method data processing on multiple sample datasets that show the values ​​of multiple explanatory variables relating to the operation of the production line and the value of an objective variable relating to the quality of the products produced by the production line. The prediction model includes, for each of the plurality of explanatory variables, an S / N ratio representing the prediction accuracy of the value of the target variable, The control method further includes, The steps include collecting measured values ​​of the multiple explanatory variables from the production line, For each product, the steps include: calculating the predicted value of the target variable by inputting the measured values ​​of the plurality of explanatory variables into the prediction model; The steps include selecting an explanatory variable to be adjusted from among the multiple explanatory variables in response to the predicted value falling outside the control range, The steps include determining the amount of adjustment for the explanatory variable to be adjusted so that the value of the objective variable approaches the center of the control range, The system includes the step of controlling the production line so as to change the value of the explanatory variable to be adjusted by the adjustment amount, The selection step includes dividing the plurality of explanatory variables into a first group and a second group having less influence on the dependent variable than the first group, based on the signal-to-noise ratio, and selecting the explanatory variable to be adjusted from among the explanatory variables belonging to the first group. The step of selecting the explanatory variable to be adjusted from among the explanatory variables belonging to the first group is: For each explanatory variable belonging to the first group, the step of calculating the degree of variation of the explanatory variable, A control method comprising the step of selecting the first explanatory variable with the greatest degree of variation as the explanatory variable to be adjusted.

8. A control method for controlling a production line that includes multiple processes, The method includes the step of obtaining a predictive model obtained by performing T-method data processing on multiple sample datasets that show the values ​​of multiple explanatory variables relating to the operation of the production line and the value of an objective variable relating to the quality of the products produced by the production line. The prediction model includes, for each of the plurality of explanatory variables, an S / N ratio representing the prediction accuracy of the value of the target variable, The control method further includes, The steps include collecting measured values ​​of the multiple explanatory variables from the production line, For each product, the steps include: calculating the predicted value of the target variable by inputting the measured values ​​of the plurality of explanatory variables into the prediction model; The steps include selecting an explanatory variable to be adjusted from among the multiple explanatory variables in response to the predicted value falling outside the control range, The steps include determining the amount of adjustment for the explanatory variable to be adjusted so that the value of the objective variable approaches the center of the control range, The system includes the step of controlling the production line so as to change the value of the explanatory variable to be adjusted by the adjustment amount, The selection step includes dividing the plurality of explanatory variables into a first group and a second group having less influence on the dependent variable than the first group, based on the signal-to-noise ratio, and selecting the explanatory variable to be adjusted from among the explanatory variables belonging to the first group. The step of selecting the explanatory variable to be adjusted from among the explanatory variables belonging to the first group is: For each explanatory variable belonging to the first group, the step of calculating the degree of variation of the explanatory variable, A control method comprising the step of selecting a second explanatory variable to be adjusted from among the explanatory variables belonging to the first group, which corresponds to a second process that occurs later than the first process that corresponds to the first explanatory variable with the greatest degree of variation.

9. A program that causes a computer to execute the control method described in claim 7 or 8.

Images