Predictor of defective product generation rate of molded product

Abstract

To provide a predictor which can accurately predict a defect rate of a formed product formed by press molding.SOLUTION: A predictor 20 comprises: a learning part 22 which performs machine learning of calculation method of a defect rate of formed products F formed based on a manufacturing condition using as teacher data, a plurality of past manufacturing conditions that consists of coating feed rate of wash oil for a metal sheet S, blasting pressure of the wash oil for the metal sheet S, transportation speed of the metal sheet S, and surface roughness of the metal sheet S, and the defect rate of the formed product F in each of the plurality of the past manufacturing conditions; and a prediction part 23 which predicts the defect rate of the formed product F where the defect rate of the formed product F is calculated through the learning model having the machine learning in the learning part 22 into which a manufacturing condition of forming product is entered.SELECTED DRAWING: Figure 4

Description

Technical Field

[0001] The present invention relates to a prediction device for the defective occurrence rate of molded products.

Background Art

[0002] As molded products, for example, molded products such as the outer panels of vehicles are generally manufactured by press-forming a metal sheet such as a steel sheet or an aluminum sheet. For example, Patent Document 1 discloses a manufacturing method of such a molded product.

[0003] In the manufacturing method of Patent Document 1, first, while applying cleaning oil to both sides of the metal sheet, the metal sheet is sandwiched between a pair of brush rolls to clean the metal sheet. The cleaned metal sheet is sandwiched between oil-cut rolls while being conveyed, thereby draining the cleaning oil adhering to the metal sheet. After draining the oil, the metal sheet is formed by press-forming to produce a molded product.

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0005] By the way, in the molded product formed by press-forming, bulges, scratches, etc. may occur due to foreign matters, and it may be treated as a defective product. However, it is difficult to accurately predict the defective occurrence rate of such molded products.

[0006] The present invention has been made in view of such problems, and an object thereof is to provide a prediction device capable of accurately predicting the defective occurrence rate of a molded product formed by press-forming.

Means for Solving the Problems

[0007] As a result of intensive studies, the inventors have obtained new findings that the cleaning conditions for cleaning the metal sheet have a great influence on the quality of the molded product when molding by press forming. The present invention is based on such new findings. The prediction device for the defective occurrence rate of the molded product according to the present invention sandwiches a metal sheet with a pair of brush rolls while applying cleaning oil to both sides of the metal sheet to clean the metal sheet, and after draining the cleaning oil adhering to the metal sheet with an oil cut roll for the cleaned metal sheet, it is a prediction device for predicting the defective occurrence rate of the molded product obtained by press forming the metal sheet. As a plurality of past manufacturing conditions, the application flow rate of the cleaning oil to the metal sheet, the spraying pressure of the cleaning oil to the metal sheet, the conveyance speed of the metal sheet, and the surface roughness of the metal sheet, a learning unit that machine-learns the calculation method of the defective occurrence rate of the molded product formed under the manufacturing conditions using, as teacher data, the defective occurrence rate of the molded product press-formed for each of the plurality of manufacturing conditions, and having a learning model machine-learned by the learning unit, as manufacturing conditions for the planned molding, the application flow rate of the cleaning oil to the metal sheet, the spraying pressure of the cleaning oil to the metal sheet, the conveyance speed of the metal sheet, and the surface roughness of the metal sheet, by inputting the manufacturing conditions into the learning model, causing the learning model to calculate the defective occurrence rate of the molded product, and a prediction unit that predicts the calculated defective occurrence rate as the defective occurrence rate of the molded product to be molded, characterized by comprising.

[0008] As described above, the manufacturing conditions composed of the application flow rate of the cleaning oil to the metal sheet, the spraying pressure of the cleaning oil to the metal sheet, the conveyance speed of the metal sheet, and the surface roughness of the metal sheet are parameters that contribute to the occurrence of defects in the molded product.

[0009] For example, if a certain level of spraying pressure of the cleaning oil for cleaning the metal sheet cannot be ensured, the cleaning oil cannot be stably adhered to the metal sheet. However, if the spraying pressure is too high, the sprayed cleaning oil will be reflected by the metal sheet. As a result, since the adhesion of the cleaning oil is unstable, it is difficult to sufficiently remove foreign matter adhered to the metal sheet by the brush roll.

[0010] In addition to this, it is also considered that the higher the application flow rate of the cleaning oil, the higher the cleaning efficiency. However, if the application flow rate of the cleaning oil is too high, after cleaning, the cleaning oil cannot be sufficiently drained by the oil cut roll, which may not only affect the surface properties of the molded product, but also foreign matter is likely to adhere before performing press forming.

[0011] Thus, when cleaning with a brush roll while applying the cleaning oil, it is preferable that an appropriate amount of cleaning oil adheres to the surface of the metal sheet. Therefore, it has been found that even if the spraying pressure and the application flow rate of the cleaning oil are adjusted, this adhesion amount is also greatly affected by the surface roughness of the metal sheet and the conveyance speed of the metal sheet.

[0012] Therefore, according to the present invention, as manufacturing conditions, a combination of the application flow rate of the cleaning oil to the metal sheet, the spraying pressure of the cleaning oil to the metal sheet, the conveyance speed of the metal sheet, the surface roughness of the metal sheet, and the past manufacturing conditions composed of these manufacturing conditions, and the defective occurrence rate of the molded product molded corresponding to this past manufacturing condition is used as teacher data. Specifically, using this teacher data, the learning unit can machine-learn a calculation method (algorithm) of the defective occurrence rate of the molded product molded under the past manufacturing conditions and construct a machine-learned learning model.

[0013] The prediction unit has a learning model constructed by machine learning in the learning unit. Therefore, for the learning model of the prediction unit, if manufacturing conditions of any forming plan, which are composed of the application flow rate of the cleaning oil to the metal sheet, the spraying pressure of the cleaning oil to the metal sheet, the conveying speed of the metal sheet, and the surface roughness of the metal sheet, are input, the defective rate of the formed product of the forming plan can be accurately predicted.

Effect of the Invention

[0014] According to the present invention, the defective rate of the formed product formed by press forming can be accurately predicted.

Brief Description of the Drawings

[0015]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Embodiments for Carrying Out the Invention

[0016] Hereinafter, with reference to FIGS. 1 to 5, a prediction device for the defective rate of a formed product according to an embodiment of the present invention will be described.

[0017] First, while referring to FIG. 1, a molding method for molding a molded product F from a metal sheet S by press molding will be described. FIG. 1 is a schematic diagram for explaining a manufacturing method of a molded product F for which the defect occurrence rate is predicted, and is a diagram for explaining the cleaning of the metal sheet S and the press molding after cleaning.

[0018] As shown in FIG. 1, in the present embodiment, the molded product F is molded using a metal coil CA around which a metal sheet S such as a steel sheet or an aluminum sheet is wound. However, for example, a metal sheet S cut to a predetermined size may also be used. In the present embodiment, while transporting the metal sheet S, the metal sheet S is cleaned by a cleaning device 90, and press molding is performed on the cleaned metal sheet S.

[0019] Specifically, the metal coil CA around which the metal sheet S is wound is unwound, and the unwound metal sheet S is passed through the cleaning device 90. Specifically, it is sandwiched by a pair of pinch rolls 91 and transported into the cleaning device 90 at a predetermined transport speed.

[0020] While applying cleaning oil from both sides to the transported metal sheet S through a nozzle 92a with a preset spraying pressure and application flow rate, the metal sheet S is sandwiched by a pair of brush rolls 92 to clean the metal sheet S.

[0021] Cleaning oil adheres to the metal sheet S cleaned by the pair of brush rolls 92. Therefore, for the cleaned metal sheet S, oil draining of the cleaning oil adhering to the metal sheet S is performed using a pair of oil cut rolls 93 and 94. After performing the oil draining, the metal sheet S is transported to a press device 95, and the metal sheet S is sandwiched between an upper die 95a and a lower die 95b to perform press molding and manufacture the molded product F.

[0022] By the way, when cleaning the metal sheet S, it is often considered that the higher the spraying pressure of the cleaning oil and the larger the application flow rate of the cleaning oil, the higher the cleaning efficiency of the metal sheet S.

[0023] However, if the spraying pressure of the cleaning oil for cleaning the metal sheet S is too high, the sprayed cleaning oil will be reflected by the metal sheet. As a result, the adhesion of the cleaning oil to the metal sheet S is unstable, so it is difficult to sufficiently remove foreign matter adhering to the metal sheet S by the brush roll 92. Furthermore, if the application flow rate of the cleaning oil is too large, after cleaning, the oil drainage of the cleaning oil cannot be sufficiently performed by the oil cut rolls 93 and 94, which may not only affect the surface properties of the molded product, but also foreign matter is likely to adhere before performing press forming.

[0024] Thus, when cleaning with the brush roll 92 while applying the cleaning oil, it is preferable that an appropriate amount of cleaning oil adheres to the surface of the metal sheet S. Therefore, even if the spraying pressure and the application flow rate of the cleaning oil are adjusted, this adhesion amount is also greatly affected by the surface roughness of the metal sheet S and the conveyance speed of the metal sheet S, so the defective rate of the molded product F is difficult to predict.

[0025] From such a point, in the present embodiment, a prediction system 1 including a prediction device 20 for predicting the defective rate of the molded product F shown in FIG. 2 is used. FIG. 2 is a schematic diagram of a prediction system 1 including a prediction device 20 for predicting the defective rate of the molded product F according to the present embodiment.

[0026] The prediction system 1 according to the present embodiment includes an input device 10, a prediction device 20, and an output device 30. The input device 10 is composed of a keyboard or the like, and is a device for inputting teacher data, manufacturing conditions of a planned molding, etc. to the prediction device 20, and is connected to the prediction device 20.

[0027] The prediction device 20 is composed of an arithmetic device (not shown) such as a CPU and a storage device (not shown) such as a RAM and a ROM. Teacher data and a learning model to be described later are stored in the storage device, and the arithmetic device reads data from the storage device, constructs a learning model by machine learning, and performs processing for utilizing the learned learning model. The prediction device 20 is connected to the output device 30.

[0028] The output device 30 is a device configured with a display or the like, and displays the result of the defect occurrence rate predicted by the prediction device 20. Note that the input device 10 and the output device 30 may form a touch panel display, or the input device 10, the prediction device 20, and the output device 30 may form a tablet terminal, a mobile terminal, or the like.

[0029] Hereinafter, with reference to FIGS. 3 and 4, the software of the prediction device 20 will be described. FIG. 3 is a block diagram of the prediction device 20 shown in FIG. 2. FIG. 4(a) is a diagram for explaining machine learning by the learning unit 22 shown in FIG. 3, and FIG. 4(b) is a diagram for explaining the utilization of the learning model by the prediction unit 23 shown in FIG. 3.

[0030] The prediction device 20 includes, as software, a storage unit 21, a learning unit 22, and a prediction unit 23. As described above, it has been clarified by the inventors' study that the defect occurrence rate of the molded product F molded by press molding greatly contributes to the following manufacturing conditions. Therefore, the storage unit 21 stores data in which the manufacturing conditions in the past molding and the defect occurrence rate of the molded product molded under the manufacturing conditions are associated as teacher data for causing the learning unit 22 to perform machine learning.

[0031] Specifically, as shown in FIG. 4(a), the storage unit 21 stores a plurality of manufacturing conditions A, B, C... as manufacturing conditions composed of (1) the application flow rate of the cleaning oil to the metal sheet S, (2) the spraying pressure of the cleaning oil to the metal sheet S, (3) the conveyance speed of the metal sheet S, and (4) the surface roughness of the metal sheet S. Further, the storage unit 21 also stores the result of the defect occurrence rate of the (5) molded product F molded under these manufacturing conditions for each of the manufacturing conditions A, B, C...

[0032] Here, the defective rate of the molded product F is the ratio of the number of defective products to the total number of molded products F manufactured for each metal coil CA. However, if all the metal coils CA are stored under the same conditions, manufactured under the same conditions, and have the same physical property values, the defective rate of the molded products F molded for all these metal coils CA may also be used.

[0033] Here, (1) the application flow rate of the cleaning oil to the metal sheet S is the flow rate of the cleaning oil per unit time applied from all the nozzles 92a. The application flow rate of the cleaning oil can be obtained from the sum of the measured values of the flow meters attached near the nozzles 92a, or from the discharge amount of a pump (not shown) that pumps the cleaning oil. In addition, if a part of the cleaning oil is applied from the nozzles 92a to the brush roll 92 for the metal sheet S, the application flow rate of that cleaning oil may also be included. However, for the application flow rate of the cleaning oil to the metal sheet S, values measured by the same method are used.

[0034] (2) The spraying pressure of the cleaning oil on the metal sheet S may be the pressure of the cleaning oil when it is discharged from all the nozzles 92a. Therefore, the spraying pressure of the cleaning oil can be obtained from the measured value of the pressure gauge attached near the nozzles 92a, or from the discharge pressure of a pump (not shown) that pumps the cleaning oil. However, for the spraying pressure of the cleaning oil on the metal sheet S, values measured by the same method are used.

[0035] (3) The conveyance speed of the metal sheet S is the speed of the metal sheet S conveyed to the cleaning device 90, and can be obtained, for example, from the rotation speeds of the pinch roll 91, the brush roll 92, or the oil cut rolls 93, 94. However, for the conveyance speed of the metal sheet S, values measured by the same method are used.

[0036] (4) The surface roughness of the metal sheet S may be determined using the physical property data attached at the time of receipt of the metal coil CA that constitutes the metal sheet S, or may be measured using a laser type or stylus type surface roughness meter. The surface roughness may be the center line average roughness Ra or the ten-point average roughness Rz. However, for the surface roughness of the metal sheet S, values measured by the same evaluation index and the same method are used.

[0037] As shown in FIG. 4(a), the learning unit 22 performs machine learning using, as teacher data, a plurality of past molding conditions A, B, C,... and the defective occurrence rate of the molded product F molded by press molding for each of the plurality of molding conditions A, B, C. Specifically, using these teacher data, the learning unit 22 learns the calculation method of the defective occurrence rate of the molded product F molded under the molding conditions through machine learning to construct a learning model (AI model). In this way, the learning unit 22 performs a learning phase by machine learning.

[0038] This calculation method is an algorithm, and the learning model is not particularly limited as long as it can perform such learning. For example, it may be a generalizable learning model such as a neural network (a network structure composed of specific neurons).

[0039] The teacher data is composed of an explanatory variable and an objective variable. The explanatory variable is a variable that constitutes a plurality of molding conditions A, B, C,.... Specifically, the explanatory variables are the application flow rate of the cleaning oil to the metal sheet S, the spraying pressure of the cleaning oil to the metal sheet S, the conveyance speed of the metal sheet S, and the surface roughness of the metal sheet S. The objective variable is the defective occurrence rate of the molded product F molded by press molding for each of the plurality of molding conditions A, B, C,....

[0040] The prediction unit 23 has a learning model (a learned learning model) machine-learned by the learning unit 22. As shown in FIG. 4(b), the prediction unit 23 inputs the manufacturing conditions of the planned molding into the learning model via the input device 10, causing the learning model to calculate the defect occurrence rate of the molded product F. The prediction unit 23 predicts the defect occurrence rate calculated by the learning model as the defect occurrence rate of the molded product F to be molded, and transmits this result to the output device 30.

[0041] Here, the manufacturing conditions of the planned molding are: (1) the application flow rate of the cleaning oil to the metal sheet S when performing the planned press molding, (2) the spraying pressure of the cleaning oil to the metal sheet S when performing the planned press molding, (3) the conveyance speed of the metal sheet S when performing the planned press molding, and (4) the surface roughness of the metal sheet S to be used in the future.

[0042] In this way, according to the present embodiment, the combination of the past manufacturing conditions composed of the application flow rate of the cleaning oil to the metal sheet S, the spraying pressure of the cleaning oil to the metal sheet S, the conveyance speed of the metal sheet S, and the surface roughness of the metal sheet S, and the defect occurrence rate of the molded product F molded corresponding to this past manufacturing condition is used as teacher data. Therefore, in the learning unit 22, it is possible to accurately machine-learn the calculation method (algorithm) of the defect occurrence rate of the molded product F molded under arbitrary manufacturing conditions and construct a machine-learned learning model. As a result, in the prediction unit 23, if arbitrary manufacturing conditions of the planned molding are input to this learned learning model, the defect occurrence rate of the molded product to be molded can be accurately predicted.

[0043] In the present embodiment, the teacher data (1) to (5) described above is used as the teacher data for performing machine learning. Further, if conditions related to the press device 95 (for example, the clearance between the upper die 95a and the lower die 95b), conditions of other facilities, disturbance conditions, etc. are used as teacher data as necessary, the calculation accuracy of the defect occurrence rate of the molded product F can be further improved.

[0044] The following describes a prediction method using the prediction device 20 with reference to the flowchart shown in FIG. 5. First, in step S51, a plurality of past manufacturing conditions and the defective rate of the molded product F corresponding thereto are input from the input device 10 to the prediction device 20. Thereby, it is input as teacher data to the storage unit 21 of the prediction device 20.

[0045] Next, in step S52, the plurality of past manufacturing conditions stored in the storage unit 21 and the defective rate of the molded product F corresponding thereto are used as teacher data, and the learning unit 22 machine-learns the calculation method of the defective rate of the molded product F under arbitrary manufacturing conditions (learning phase). Thereby, a learning model machine-learned by the learning unit 22 is constructed, and this constitutes the prediction unit 23.

[0046] Next, in step S53, the manufacturing conditions of the molded product F scheduled to be molded are input from the input device 10 to the prediction device 20. Thereby, the manufacturing conditions of the molded product F scheduled to be molded are input to the learned learning model of the prediction unit 23.

[0047] Finally, in step S54, the prediction unit 23 calculates the defective rate of the molded product F scheduled to be molded, and outputs the result as the predicted defective rate of the molded product F to the output device 30. Here, based on the defective rate output to the output device 30, if the manufacturing conditions are changed, the occurrence of defective products of the molded product F can also be suppressed.

[0048] As described above in detail with respect to the embodiments of the present invention, the present invention is not limited to the above-described embodiments, and various design changes can be made without departing from the spirit of the present invention described in the claims.

Description of Reference Numerals

[0049] 20: Prediction device, 22: Learning unit, 23: Prediction unit, 92: Brush roll, 93, 94: Oil cut roll, F: Molded product

Claims

【Claim 1】 A prediction device for predicting the defective rate of a molded product obtained by press-molding a metal sheet, which cleans the metal sheet by sandwiching it between a pair of brush rolls while applying cleaning oil to both sides of the metal sheet, and then draining the cleaning oil adhering to the metal sheet with an oil-cut roll. The device comprises: A learning unit that performs machine learning on a method for calculating the defective rate of a molded product formed under the manufacturing conditions, using, as teacher data, a plurality of past manufacturing conditions for the metal sheet, including the application flow rate of the cleaning oil to the metal sheet, the spraying pressure of the cleaning oil on the metal sheet, the conveyance speed of the metal sheet, and the surface roughness of the metal sheet, and the defective rate of the molded product press-molded for each of the plurality of manufacturing conditions. A prediction unit that has a learning model machine-learned by the learning unit, inputs, as manufacturing conditions for a planned molding, the application flow rate of the cleaning oil to the metal sheet, the spraying pressure of the cleaning oil on the metal sheet, the conveyance speed of the metal sheet, and the surface roughness of the metal sheet into the learning model, causes the learning model to calculate the defective rate of the molded product, and predicts the calculated defective rate as the defective rate of the molded product for the planned molding. A prediction device for predicting the defective rate of a molded product, characterized by comprising the above components.

Images