Classification methods

Abstract

To provide a classification method capable of stably classifying a shape pattern of a product.SOLUTION: A classification method is a classification method executed by a computer, and includes: a measuring step of measuring the outer periphery of a product to obtain a measurement result; and a classification step for classifying the product into any of the predetermined shape patterns based on the measurement result using a machine learning model whose parameters are set by previously executed learning.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to a classification method. [Background technology]

[0002] Since the manufacturing process is affected by a wide variety of factors, such as the state of the materials used and environmental conditions, adjustments are often made based on the product. Patent Document 1 discloses a method for manufacturing a light-emitting device, including: a preparation step of preparing a plurality of light sources each having an upper surface with a light-emitting portion, a lower surface opposite the upper surface and having an external connection terminal, and a side surface between the upper surface and the lower surface, and ranked by at least one of luminous flux and chromaticity, an extraction step of extracting a plurality of light sources belonging to a desired rank from the plurality of light sources, and a bonding step of bonding the side surfaces together via a bonding member so that the upper surfaces and the lower surfaces of the extracted plurality of light sources are exposed from a bonding member and the external connection terminals are spaced from the bonding member. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Publication No. 2022-097370 Summary of the Invention [Problem to be solved by the invention]

[0004] The invention described in Patent Document 1 leaves room for improvement in the method for classifying products. [Means for solving the problem]

[0005] A classification method according to a first aspect of the present invention is a computer-implemented classification method comprising: Approximately cylindrical of manufactured goods Distance from the center to the periphery. Measure Angle Measurements Using a measurement process to obtain the above and a machine learning model whose parameters have been set by previously performed learning, The angle measurementand a classification step of classifying the products into any of the predetermined shape patterns based on the In the classification step, a standardized measurement value, which is a value based on the angle measurement value, is calculated for each angle, and classification is performed based on an adjacent difference, which is a difference between the standardized measurement values ​​for adjacent angles. The periphery is classified into divided areas, which are two or more continuous areas. In the classification step, small area classification, which is classification based on the adjacent difference, is performed for each divided area, and the divided area is classified into one of the shape patterns based on a combination of the small area classifications. Each of the divided areas is set based on an influence that the product receives from the outside. . [Effects of the Invention]

[0006] According to the present invention, shape patterns of manufactured products can be stably classified. [Brief explanation of the drawings]

[0007] [Figure 1] Overall configuration diagram of the production system including the computing device [Figure 2] A diagram showing an example of a classification correspondence table [Figure 3] Schematic diagram of the production line [Figure 4] FIG. 10 shows an example of measurement in the sixth step. [Figure 5] Figure showing the measurement in the seventh step [Figure 6] An example of measurement results using a shape measuring device [Figure 7] Overview of the processing performed by the calculation unit [Figure 8] Flowchart showing the processing of the arithmetic unit [Figure 9] FIG. 10 is a diagram showing a drying process in the second embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0008] -First embodiment- A first embodiment of a classification method according to the present invention will be described below with reference to FIGS.

[0009] FIG. 1 is an overall configuration diagram of a production system 1 including a computing device 2. The production system 1 manufactures a honeycomb structure 80. The production system 1 includes a production line 400, the computing device 2, and an output device 490. The output device 490 is, for example, a liquid crystal display. The output device 490 presents information to an operator 9.

[0010] The production line 400 includes a molding device 401, a cutting device 402, a stand 403, a drying oven 404, a finishing device 405, an appearance inspection device 406, and a shape measurement device 407. The configuration of the production line 400 will be described in detail later with reference to the following Figure 2. The operator 9 changes the settings of the molding device 401 based on information obtained from the output device 490.

[0011] The calculation device 2 includes a receiving unit 21, a calculation unit 22, a storage unit 23, and an output unit 24. The receiving unit 21 is a communication module that communicates with the shape measurement device 407. The receiving unit 21 receives measurement results from the shape measurement device 407 and outputs the received measurement results to the calculation unit 22. The calculation unit 22 is composed of, for example, a central processing unit, RAM, and ROM (not shown). The calculation unit 22 performs calculations (described below) by loading a program stored in the ROM into the RAM and executing it, and outputs the calculation results to the output unit 24.

[0012] The storage unit 23 is a non-volatile storage device, for example, a flash memory. The storage unit 23 stores a first determinator 231, a second determinator 232, and a classification correspondence table 233 in advance. The first determinator 231 and the second determinator 232 are trained machine learning models. The first determinator 231 receives a non-contact difference 9N, which is a combination of numbers, and outputs a non-contact class, which is a shape classification. The second determinator 232 receives a contact difference 9T, which is a combination of numbers, and outputs a contact class, which is a shape classification. The non-contact difference 9N and the contact difference 9T will be described later.

[0013] The classification correspondence table 233 is data showing the classification of the shape of the honeycomb structure 80 (hereinafter referred to as "cylindrical classification") corresponding to the combination of the non-contact class output by the first determinator 231 and the contact class output by the second determinator 232.

[0014] FIG. 2 is a diagram showing an example of the classification correspondence table 233. In the example shown in FIG. 2, there are four types of contact classes, T1 to T4, and four types of non-contact classes, N1 to N4. The classification of the shape of the honeycomb structure 80 corresponding to the combination of the contact class and the non-contact class, i.e., the cylindrical classification, is nine types, P1 to P9. This correspondence is predetermined based on human sensibility. For example, when the contact class is T1, the shape is classified into P4 regardless of whether the non-contact class is any of N1 to N3. Examples of the cylindrical classification include perfect circle, left flat, right flat, heart, etc.

[0015] FIG. 3 is a schematic diagram of a production line 400. The honeycomb structure 80 is produced through the first process shown on the far left in FIG. 2 to the seventh process shown on the right. The X, Y, and Z axes shown at the bottom of FIG. 2 are mutually orthogonal coordinate axes and are shown for convenience of explanation. In this embodiment, the honeycomb structure 80 moves in the positive direction of the X axis as the process progresses. The Y axis corresponds to the depth direction from the viewpoint of FIG. 2. The Z axis is the direction of gravity.

[0016] In the first step, a cylindrical continuum 81 with a honeycomb formed inside is formed and extruded from the forming device 401 in the X-axis direction. Hereinafter, the peripheral region of this cylindrical shape will be referred to as the outer periphery of the honeycomb structure 80. The forming device 401 allows adjustment of manufacturing parameters by operation of the operator 9. Specifically, the forming device 401 can perform adjustments such as temperature adjustment, jig replacement, die rotation, and resistor insertion. Temperature adjustment is performed by controlling the operation of the heater provided in the forming device 401. Jig replacement is performed by replacing a jig provided in the forming device 401 that has worn out during use with a new jig. Die rotation is performed by rotating the die provided in the forming device 401 by any angle. Resistor insertion is performed by inserting a resistor at any position on the inner wall of the forming device 401 to slow down the movement of the material.

[0017] In the second step, the continuous body 81 is cut into a plurality of cut bodies 82 of a predetermined length by a cutting device 402. The cut bodies 82 are supported from below by a stand 403. Here, it is assumed that the lower half of the periphery of the cut body 82, i.e., the entire surface on the negative Z-axis side, is in contact with the stand 403. In the honeycomb structure 80, the outer periphery that was in contact with the stand 403 is called the contact periphery 80T, and the outer periphery that was not in contact with the stand 403 is called the non-contact periphery 80N. In the cut body 82 in the second step, the cylindrical axis is parallel to the X-axis.

[0018] In the third step, the cut body 82 is rotated 90 degrees around the Y axis together with the stand 403, so that the axis of the cylinder becomes parallel to the Z axis. At this time, the stand 403 is disposed on the negative side of the X axis. In the fourth step, the stand 403 is removed from the cut body 82, and the cut body 82 is dried in a drying furnace 404 to become a dried body 83. In the fifth step, the dried body 83 is cut in the axial direction to a predetermined finishing dimension by a finishing device 405 to become a finished body 84.

[0019] In the sixth step, the appearance of the finished body 84 is inspected by an appearance inspection device 406. In the seventh step, the shape of the finished body 84 is measured by a shape measurement device 407. If the appearance inspection determines that there are no problems with the finished body 84, it becomes a finished product. Details of the sixth and seventh steps will be described later. The measurement results by the shape measurement device 407 are fed back to the manufacturing process of the honeycomb structure 80. The honeycomb structure 80 is a collective term for the continuous body 81, the cut body 82, the dried body 83, and the finished body 84. The continuous body 81, the cut body 82, the dried body 83, and the finished body 84 all have an approximately cylindrical shape.

[0020] The central axis of the honeycomb structure 80 is parallel to the X-axis in the first and second steps, and shifts to the Z-axis from the third step onwards. In Fig. 2, the central axis of the honeycomb structure 80 temporarily becomes parallel to the X-axis in the fifth step, but the fifth step may be performed with the central axis of the honeycomb structure 80 remaining parallel to the Z-axis.

[0021] In the sixth step, the appearance of the finished body 84 is inspected using a simple method. It is desirable that the side of the finished body 84 be a perfect circle, and the appearance inspection device 406 determines whether it is within the allowable dimensional range. For example, the finished body 84 is photographed from the positive side of the Z axis, and it is confirmed that the outer shape of the finished body 84 is between the minimum allowable circle L1 and the maximum allowable circle L2.

[0022] FIG. 4 shows an example of measurement in the sixth step. For example, as shown on the left side of FIG. 4, a first example of the outer shape of the finished product 84, designated by reference numeral 84EX1, is located between the minimum allowable circle L1 and the maximum allowable circle L2 along its entire circumference. Therefore, in this first example, the visual inspection device 406 judges the photographed finished product 84 to be acceptable. Also, as shown on the right side of FIG. 4, a second example of the outer shape of the finished product 84, designated by reference numeral 84EX2, has a smaller diameter than the minimum allowable circle L1 in the upper left portion of the figure. Therefore, in this second example, the visual inspection device 406 judges the photographed finished product 84 to be unacceptable. In this embodiment, regardless of the judgment result of the visual inspection device 406 in the sixth step, the process proceeds to the seventh step.

[0023] FIG. 5 is a diagram showing the measurement in the seventh step. The shape measuring device 407 is, for example, a plurality of laser range finders, which are arranged to surround the finished workpiece 84. Specifically, as shown on the left side of FIG. 5, the laser range finders are arranged at a predetermined height H, and as shown on the right side of FIG. 5, they are arranged at 45-degree intervals around the finished workpiece 84. A laser is emitted toward the finished workpiece 84 in the XY plane. The laser range finders calculate the distance to the finished workpiece 84 based on the time it takes for the laser to reflect off the obstacle, the finished workpiece 84, and return. The shape measuring device 407 uses the distances measured by each laser range finder and the known distances between each laser range finder and the center position of the finished workpiece 84 to output the distance from the center to the periphery at every 45 degrees at the height H of the finished workpiece 84 as measurement results.

[0024] Figure 6 is a diagram showing an example of the measurement results obtained by the shape measuring device 407. Circles 1 to 8 in Figure 6 schematically show the measurement target positions and the measurement results. The dashed circle in the left diagram of Figure 6 indicates the ideal shape of the finished body 84, i.e., the design value, and the posture of the finished body 84 is the same as that in the right diagram of Figure 5. For convenience, the position with the largest Y coordinate value is set to 0 degrees, and angles are defined clockwise.

[0025] In the example shown in Figure 6, it is shown that at the first point, 0 degree position, the actual measurement value of the finished body 84 was larger than the design value. Furthermore, at the second point, 45 degree position, the actual measurement value of the finished body 84 matched the design value, and at the third point, 90 degree position, the actual measurement value of the finished body 84 was slightly larger than the design value. If the honeycomb structure 80 was moved parallel to the positive side of the X axis from the third step onwards, it would be in contact with the frame 403 in the left half of the figure, i.e., in the range of 180 degrees to 360 degrees. Therefore, 0 degrees to 180 degrees is the non-contact perimeter 80N, and 180 degrees to 360 degrees is the contact perimeter 80T.

[0026] The right side of Figure 6 is a graph showing the standardized measurement values ​​arranged in order of angle. However, for ease of explanation, the first point, 0 degree, is also listed on the right side as the 360 ​​degree value. Standardized measurement values ​​are values ​​processed so that the average value of all measurements for a single finished body 84 is "0" and the variance is "1." For example, to set the average value of eight measurements for a given finished body 84 from 0 degrees to 315 degrees to "0," the same offset value is added to these eight measurements, and then the standardized measurement values ​​are set by multiplying them by the same gain so that the variance of each offset-added measurement value is "1."

[0027] 7 is a schematic diagram showing the processing of the calculation unit 22. When the calculation unit 22 receives the measurement values ​​from the shape measuring device 407, it calculates the outer diameter at every 45-degree angle at the height H of the finished body 84 using the known distances between each laser distance meter and the center position of the finished body 84. Next, the calculation unit 22 calculates the standardized measurement values ​​by dividing each calculated outer diameter by the known design value of the finished body 84. The standardized measurement values ​​for each angle are shown at the top of FIG. 7.

[0028] Next, the calculation unit 22 calculates the absolute value of the difference between the standardized measurement values ​​for each adjacent angle as the adjacent difference. Hereinafter, the adjacent difference at the non-contact periphery 80N will be referred to as the non-contact difference 9N, and the adjacent difference at the contact periphery 80T will be referred to as the contact difference 9T. For example, if the standardized measurement values ​​from 0 degrees to 180 degrees at the non-contact periphery 80N are "1.6, 0, 0.8, -0.8, 0," the non-contact difference 9N will be "1.6, 0.8, 1.6, 0.8."

[0029] The calculation unit 22 inputs the non-contact difference 9N to the first determinator 231 to identify the non-contact class, and inputs the contact difference 9T to the second determinator 232 to identify the contact class. Furthermore, the calculation unit 22 refers to the classification correspondence table 233 to identify a shape pattern corresponding to the combination of the contact class and the non-contact class, and outputs it to the output unit 24.

[0030] 8 is a flowchart showing the processing of the calculation device 2. First, in step S301, the receiving unit 21 of the calculation device 2 receives the measurement results from the shape measuring device 407. Specifically, data for a total of eight points measured every 45 degrees is received. In the following step S302, the calculation unit 22 calculates standardized measurement values ​​for one finished body 84 based on the angle measurement values ​​so that the average value is "0" and the variance is "1".

[0031] In the following step S303, the calculation unit 22 calculates an adjacent difference, which is the absolute value of the difference between the standardized measurement values ​​for each adjacent angle. Next, the calculation unit 22 executes steps S304 and S305 in parallel. However, parallel execution is not essential, and either step may be executed first. In step S304, the calculation unit 22 inputs the contact difference 9T, which is the adjacent difference at the contact perimeter 80T, to the second determiner 232 to obtain a contact class.

[0032] In step S305, the calculation unit 22 inputs the non-contact difference 9N, which is the adjacent difference in the non-contact outer periphery 80N, to the first determinator 231 to obtain a non-contact class. After both S304 and step S305 are completed, the calculation unit 22 refers to the classification correspondence table 233 in step S306 to identify a cylindrical classification corresponding to the identified contact class and non-contact class. In the following step S307, the output unit 24 notifies the operator 9 of the identified classification using the output device 490, and the processing shown in FIG. 8 ends.

[0033] Each of the contact outer periphery 80T and the non-contact outer periphery 80N is a continuous region, and is also called a "divided region" which is a region obtained by dividing the outer periphery of the honeycomb structure 80. Step S301 is also called a "measurement step." Steps S302 to S306 are also called a "classification step." Step S304 is also called a "first classification step," S305 is also called a "second classification step," and S306 is also called a "third classification step." The fourth step is also called a "drying line."

[0034] According to the first embodiment described above, the following advantageous effects can be obtained. (1) The classification method executed by the calculation device 2 includes a measurement step (S301) of measuring the outer periphery of the honeycomb structure 80 to obtain the measurement results, and a classification step (S302 to S306) of classifying the product into one of the predetermined shape patterns based on the measurement results using a machine learning model whose parameters have been set by previously executed learning. Therefore, the shape patterns of the products can be stably classified.

[0035] (2) In the measurement step, the outer circumference of the product is measured at each predetermined angle. In the classification step, a standardized measurement value, which is a value based on the angle measurement value, is calculated for each angle (S302), and the product is classified based on the adjacent difference, which is the difference between the standardized measurement values ​​for adjacent angles (S303).

[0036] (3) The periphery of the honeycomb structure 80 is classified into divided regions, which are two or more continuous regions. Specifically, it is classified into a contact periphery 80T and a non-contact periphery 80N. In the classification process, each divided region is classified into small regions based on the difference between adjacent regions (S304, S305), and each divided region is classified into one of the shape patterns based on the combination of the small region classifications (S306).

[0037] (4) Each of the divided regions is set based on the influence that the product receives from the outside, specifically, whether or not the product is in contact with the stand 403 .

[0038] (5) The honeycomb structure 80 is transported on a stand 403. The periphery of the honeycomb structure 80 is classified into a contact periphery 80T, which is a portion that contacts the stand 403, and a non-contact periphery 80N, which is a portion that does not contact the stand 403. The classification process includes a first classification process (S304) in which a first classification is performed based on the adjacent difference in the contact periphery 80T, a second classification process (S305) in which a second classification is performed based on the adjacent difference in the non-contact periphery 80N, and a third classification process (S306) in which the honeycomb structure 80 is classified into one of the shape patterns based on the results of the first classification and the second classification.

[0039] (6) In the first classification process, the images are classified into four types, in the second classification process, they are classified into four types, and in the third classification process, they are classified into one of nine types of shape patterns predetermined based on human sensitivity, i.e., a number less than 4x4.

[0040] (Variation 1) In the first embodiment described above, the shape measuring device 407 is composed of multiple laser distance meters. However, the shape measuring device 407 may be a combination of a single laser distance meter and a rotary stage. In this case, the finishing body 84 is positioned so that its center coincides with the center of rotation of the rotary stage, and the laser distance meter measures the distance to the finishing body 84 every time it rotates 45 degrees.

[0041] (Variation 2) In the first embodiment described above, the first classification step classifies the objects into four types, and the second classification step classifies the objects into four types. However, the number of classifications in the first classification step and the second classification step do not have to be the same, and may be 2 or more. Furthermore, the number of classifications in the third classification step is not limited to 9. When the number of classifications in the first classification step is P and the number of classifications in the second classification step is Q, the number of classifications R in the third classification step may be smaller than the product of P and Q.

[0042] (Variation 3) In the first embodiment described above, the outer periphery of the honeycomb structure 80 was measured at intervals of 45 degrees in the measurement step. However, the measurement interval is not limited to 45 degrees. For example, measurements may be made at intervals of 30 degrees, 15 degrees, 10 degrees, etc.

[0043] (Variation 4) In the first embodiment described above, the cylinder classification, which is the result of calculation by calculation unit 22, is visually confirmed by operator 9 via output device 490, and the settings of molding apparatus 401 are changed based on the judgment of operator 9. However, calculation unit 22 may change the settings of molding apparatus 401 based on the cylinder classification. For example, a correspondence table between cylinder classification and changes to settings of molding apparatus 401 may be further stored in storage unit 23, and calculation unit 22 may change the settings of molding apparatus 401, i.e., adjust the manufacturing parameters, based on this correspondence table instead of step S307 shown in FIG. 8 .

[0044] According to the fourth modification, the following effects can be obtained. (7) The classification method executed by the arithmetic unit 2 includes an adjusting step of adjusting the manufacturing parameters of the honeycomb structure 80 based on the shape patterns classified in the classification step.

[0045] (Variation 5) The method for calculating the standardized measurement values ​​in step S302 is not limited to the method used in the first embodiment. The standardized measurement values ​​may be calculated based on the angle measurement values ​​according to certain rules. For example, the standardized measurement values ​​may be the difference between the angle measurement values ​​and the design values, or the difference between the angle measurement values ​​and the design values ​​divided by the design values. Furthermore, the standardized measurement values ​​may be calculated so that the maximum value for each finished body 84 is "1" and the minimum value is "0."

[0046] --Second embodiment-- A second embodiment of the classification method will be described with reference to FIG. 9. In the following description, the same components as those in the first embodiment are denoted by the same reference numerals, and differences will be mainly described. Points that are not particularly described are the same as those in the first embodiment. This embodiment differs from the first embodiment mainly in that the degree of dryness is taken into consideration instead of contact with the pedestal 403.

[0047] 9 is a diagram showing details of the fourth step in the second embodiment. In the fourth step, a plurality of cut pieces 82 are simultaneously dried in a drying furnace 404. Specifically, a plurality of cut pieces 82, for example, five pieces, are lined up in the Y-axis direction and moved little by little in the X-axis direction by a conveyor. When the cut pieces 82 reach the outlet of the drying furnace 404, i.e., the end on the positive side of the X-axis, they become dried pieces 83.

[0048] However, since the drying furnace 404 has the characteristic that the outer periphery dries more easily, the degree of drying of the dried bodies 83 varies depending on the position. Specifically, for the dried bodies 83 arranged at both ends in the Y-axis direction, the side that does not face the other dried bodies 83 is more likely to dry. That is, as shown on the right side of Figure 9, if five dried bodies 83 are lined up in the Y-axis direction, the middle three and the insides of the top and bottom ends will be less dry, and the outsides of the top and bottom ends will be more dry. Hereinafter, the region with a high degree of dryness will be referred to as the strong drying region 80S, and the region with a low degree of dryness will be referred to as the weak drying region 80W.

[0049] That is, in the first embodiment, the periphery of the honeycomb structure 80 is classified into two regions, a non-contact periphery 80N and a contact periphery 80T, but in the present embodiment, the periphery of the honeycomb structure 80 is classified into a strong drying region 80S and a weak drying region 80W. Therefore, the configuration of the calculation device 2 and the processing of the calculation unit 22 in the present embodiment differ from those in the first embodiment in the following points.

[0050] The first determinator 231 receives the adjacent difference in the strong drying region 80S as input and outputs the classification of the strong drying region 80S. The second determinator 232 receives the adjacent difference in the weak drying region 80W as input and outputs the classification of the weak drying region 80W. That is, step S304 in FIG. 8 is changed to the classification of the strong drying region 80S, and step S305 is changed to the classification of the weak drying region 80W. In addition, the classification correspondence table 233 is data showing the classification of the shape of the honeycomb structure 80 corresponding to the combination of the classification of the strong drying region 80S and the classification of the weak drying region 80W.

[0051] According to the second embodiment described above, the following advantageous effects can be obtained. (8) The honeycomb structure 80, which is the finished product, is subjected to a drying process in the fourth process, which is the drying line, and then a measurement process is carried out. The outer periphery of the honeycomb structure 80 is classified into either a weakly dried region 80W, which is weakly dried, or a strongly dried region 80S, which is strongly dried, based on the strength of the drying in the drying line. The classification process includes a first classification process in which a first classification is performed based on the adjacent difference in the strongly dried region 80S, a second classification process in which a second classification is performed based on the adjacent difference in the weakly dried region 80W, and a third classification process in which the honeycomb structure 80 is classified into one of the shape patterns based on the results of the first classification and the second classification. Therefore, the shape of the honeycomb structure 80 can be classified based on the influence of the drying process.

[0052] In each of the above-described embodiments and modifications, the functional block configurations are merely examples. Some functional configurations shown as separate functional blocks may be configured as an integrated unit, or a configuration shown in a single functional block diagram may be divided into two or more functions. Furthermore, some of the functions of each functional block may be provided by other functional blocks.

[0053] In the above-described embodiments and modifications, the program is stored in a ROM (not shown), but the program may be stored in the storage unit 23. Furthermore, the arithmetic unit 2 may be provided with an input / output interface (not shown), and the program may be loaded from another device as needed via the input / output interface and a medium available to the arithmetic unit 2. The medium here refers to, for example, a storage medium detachable from the input / output interface, or a communication medium, i.e., a wired, wireless, or optical network, or a carrier wave or digital signal propagating through the network. Furthermore, some or all of the functions realized by the program may be realized by a hardware circuit or FPGA.

[0054] The above-described embodiments and modifications may be combined with each other. Although various embodiments and modifications have been described above, the present invention is not limited to these. Other embodiments conceivable within the scope of the technical concept of the present invention are also included within the scope of the present invention. [Explanation of symbols]

[0055] 2: Arithmetic device 9N: Non-contact difference 9T: Contact difference 22: Arithmetic section 80: Honeycomb structure 80N: Non-contact outer circumference 80S: Strongly dry area 80T: Contact outer circumference 80W: weak dry area 81: Continuum 82: cut body 83 :Dry body 84: Finished body 231: 1st judgment machine 232:Second judgment machine 233: Classification correspondence table 400: Production line 401: Molding equipment 402: Cutting device 403: Mounting stand 404 :Drying oven 405: Finishing equipment 406: Visual inspection equipment 407: Shape measuring device

Claims

1. 1. A computer-implemented classification method comprising: a measuring step of measuring the distance from the center to the outer periphery of a substantially cylindrical product at predetermined angles to obtain angle measurements; a classification step of classifying the product into any one of predetermined shape patterns based on the angle measurement values ​​using a machine learning model whose parameters have been set by previously executed learning, In the classification step, calculating a standardized measurement value for each angle, the standardized measurement value being a value based on the angle measurement value; Classifying based on adjacent differences, which are differences in the standardized measurements for adjacent angles; The periphery is divided into divided regions, which are two or more consecutive regions; In the classification step, small region classification is performed for each divided region based on the adjacent difference, and the divided region is classified into any one of the shape patterns based on a combination of the small region classifications for each divided region. A classification method in which each of the divided regions is set based on an external influence that the product receives.

2. 2. The classification method according to claim 1, The product is transported on a platform, The outer periphery is classified into a contact outer periphery, which is a portion that contacts the pedestal, and a non-contact outer periphery, which is a portion that does not contact the pedestal, each of the contact perimeter and the non-contact perimeter is the divided region; The classification step includes: a first classification step of performing a first classification based on the adjacent difference on the contact periphery; a second classification step of performing a second classification based on the adjacent difference in the non-contact outer periphery; and a third classification step of classifying the pattern into one of the shape patterns based on the results of the first classification and the second classification.

3. 3. The classification method according to claim 2, In the first classification step, the particles are classified into P types (P is a natural number), In the second classification step, the particles are classified into Q types (Q is a natural number), In the third classification step, the shape patterns are classified into one of R types (R is a natural number smaller than the product of P and Q) that are predetermined based on human sensitivity.

4. 2. The classification method according to claim 1, The classification method further includes an adjustment step of adjusting manufacturing parameters of the product based on the shape pattern classified by the classification step.

5. 2. The classification method according to claim 1, The product is a honeycomb structure.

6. 5. The classification method according to claim 4, A classification method, wherein the adjusting step is a molding step of the product.

7. 2. The classification method according to claim 1, the product is subjected to a drying process in a drying line and then the measuring step is carried out; The periphery of the product is classified into either a weakly dried region where the product is weakly dried or a strongly dried region where the product is strongly dried based on the intensity of drying at the drying line; each of the strong drying region and the weak drying region is the divided region, The classification step includes: a first classification step of performing a first classification based on the adjacent difference in the very dry region; a second classification step of performing a second classification based on the adjacent difference in the slightly dry region; and a third classification step of classifying the pattern into one of the shape patterns based on the results of the first classification and the second classification.

Images