Quality identification device, product manufacturing method, and quality identification method

Abstract

This quality identification device is provided with: a quality defect position acquisition unit (101) that acquires the occurrence position of each of a plurality of quality defects; a relationship estimation unit (102) that estimates a network model representing the relationships between the occurrence positions of the plurality of quality defects; an isolated defect identification unit (103) that identifies quality defects that are isolated defects on the basis of the network model, and identifies, as cluster defects, quality defects that have not been identified as isolated defects; and an identification result presentation unit (104) that assigns identification information indicating an isolated defect or a cluster defect to each of the plurality of quality defects, and presents, to a user, information about the plurality of quality defects to which identification information has been assigned.

Description

Quality identification device, product manufacturing method, and quality identification method

[0001] The present disclosure relates to a quality identification device, a product manufacturing method, and a quality identification method, and more particularly to a quality identification device, a product manufacturing method, and a quality identification method that identify quality defects that occur in a product as isolated defects that occur at a position distant from other quality defects, or cluster defects that occur in a cluster close to other quality defects.

[0002] Quality control of steel products is carried out by detecting quality defects that occur inside or on the surface of the product using inspection equipment. Quality defects that occur on the surface of a product are called surface defects. To detect surface defects, the surface inspection equipment takes an image of the defect, calculates the location of the defect within the product, and measures the feature values ​​that represent the shape of the defect (a typical example is the aspect ratio of the defect).

[0003] Furthermore, the cause or process of defect occurrence may be estimated from the acquired defect information, and, depending on the estimated cause or process of defect occurrence, maintenance of the manufacturing equipment or changes to operating conditions may be carried out.

[0004] For example, Patent Document 1 discloses a support system that assists an operator in estimating and searching for defect positions when estimating the process in which a surface defect occurred. The support system in Patent Document 1 calculates the distance between surface scratches, and performs processing to group together surface scratches whose calculated distance is equal to or smaller than a threshold value obtained from the equipment specifications.

[0005] Japanese Patent Application Laid-Open No. 2014-203117

[0006] Here, quality defects such as surface defects can be classified into isolated defects, which occur at a distance from surrounding quality defects, and cluster defects, which occur in a small area. However, the distance between quality defects constituting a cluster defect (cluster defect) varies. Multiple cluster defects may exist within the same product, and the distance between quality defects may vary depending on the cluster defect. Furthermore, the distance between quality defects that is not considered a cluster in one product may be considered the distance between quality defects in another product. The support system of Patent Document 1 uniformly groups defects whose distance between quality defects is equal to or less than a predetermined threshold (Rmax = 1 m). Therefore, the support system of Patent Document 1 may not be able to properly identify cluster defects.

[0007] The purpose of the present disclosure, made in consideration of the above circumstances, is to provide a quality identification device, a product manufacturing method, and a quality identification method that can accurately identify quality defects as cluster defects and isolated defects by taking into account the occurrence status of the entire product.

[0008] (1) A quality identification device according to one embodiment of the present disclosure is a quality identification device that identifies whether a quality defect occurring in a product is an isolated defect or a cluster defect based on the location of the quality defect, and includes: a quality defect location acquisition unit that acquires the location of each of the multiple quality defects; a relationship estimation unit that estimates a network model that represents the relationship between the locations of the multiple quality defects; an isolated defect identification unit that identifies the quality defects as isolated defects based on the network model and identifies the quality defects that are not identified as isolated defects as cluster defects; and an identification result presentation unit that assigns identification information indicating the isolated defect or the cluster defect to each of the multiple quality defects and presents information on the multiple quality defects to which the identification information has been assigned to a user.

[0009] (2) As one embodiment of the present disclosure, in (1), the location of the quality defect is identified by two variables indicating coordinates on a coordinate plane set for the product or three variables indicating coordinates in a coordinate space set for the product.

[0010] (3) As an embodiment of the present disclosure, in (1) or (2), the relationship estimation unit constructs the network model by treating each of the multiple quality defects as a node and determining whether to connect the nodes based on the calculated inter-node distance.

[0011] (4) As an embodiment of the present disclosure, in (3), the relationship estimation unit sets a threshold based on statistical information of the calculated inter-node distance, and connects nodes whose calculated inter-node distance is less than the threshold.

[0012] (5) As an embodiment of the present disclosure, in (3) or (4), the relationship estimation unit creates a histogram of the calculated inter-node distances, classifies each element into two clusters by clustering, and sets the average value of the maximum value of elements belonging to one cluster and the minimum value of elements belonging to the other cluster as the threshold.

[0013] (6) As an embodiment of the present disclosure, in any one of (1) to (5), the relationship estimation unit draws a circle centered on each of the nodes, connects each of the nodes to other nodes present within the circle with edges, and gradually increases the radius of the circle until at least one edge is connected to all of the nodes.

[0014] (7) As one embodiment of the present disclosure, in any one of (1) to (6), the isolated defect identification unit searches for a joint point in the network model and identifies the quality defect corresponding to the searched joint point as an isolated defect.

[0015] (8) As an embodiment of the present disclosure, in any one of (1) to (7), the identification result presentation unit generates a quality defect list, which is a list in which each of the multiple quality defects is assigned identification information indicating an isolated defect or a cluster defect, and a network diagram in which the identification information indicating an isolated defect or a cluster defect is visually assigned to the location where each of the multiple quality defects occurs, and presents the quality defect list and the network diagram to a user.

[0016] (9) As one embodiment of the present disclosure, in any one of (1) to (8), the product is a steel plate, and the quality defect is a surface defect of the steel plate or a defect present inside the steel plate.

[0017] (10) As an embodiment of the present disclosure, in any one of (1) to (9), the identification result presentation unit determines how to dispose of the product based on the identification information of an isolated defect and a cluster of defects.

[0018] (11) A method for manufacturing a product according to an embodiment of the present disclosure includes processing the product based on the product processing decision made by the quality identification device of (10).

[0019] (12) A quality identification method according to one embodiment of the present disclosure is a quality identification method executed by a quality identification device that identifies whether a quality defect occurring in a product is an isolated defect or a cluster defect based on the location of the quality defect, and includes: a quality defect position acquisition step that acquires the location of occurrence of each of a plurality of the quality defects; a relationship estimation step that estimates a network model that represents the relationship between the locations of occurrence of the plurality of the quality defects; an isolated defect identification step that identifies the quality defects that are isolated defects based on the network model and identifies the quality defects that are not identified as isolated defects as cluster defects; and an identification result presentation step that assigns identification information indicating the isolated defect or the cluster defect to each of the plurality of the quality defects and presents information on the plurality of the quality defects to which the identification information has been assigned to a user.

[0020] According to the present disclosure, it is possible to provide a quality identification device, a product manufacturing method, and a quality identification method that can accurately identify quality defects as cluster defects or isolated defects by taking into account the occurrence status of the entire product.

[0021] Fig. 1 is a schematic diagram showing an example configuration of a quality identification device according to an embodiment of the present disclosure. Fig. 2 is a diagram illustrating quality defect locations in a steel plate. Fig. 3 is a diagram illustrating a histogram. Fig. 4 is a diagram illustrating a network model corresponding to the quality defect locations in Fig. 2. Fig. 5 is a diagram illustrating identified isolated defects and clustered defects. Fig. 6 is a diagram illustrating a quality defect list and a network diagram. Fig. 7 is a diagram for explaining another method of constructing a network model.

[0022] Hereinafter, a quality identification device, a product manufacturing method, and a quality identification method according to an embodiment of the present disclosure will be described with reference to the drawings.

[0023] FIG. 1 is a schematic diagram showing an example of the configuration of a quality identification device according to this embodiment. The quality identification device is used in a manufacturing process for producing products. In this embodiment, the product is described as a steel product, but the product is not limited to steel products. The quality identification device identifies whether a quality defect is an isolated defect or a cluster of defects based on the location of the defect occurring on the product. The quality defects identified by the quality identification device according to this embodiment are not limited to specific ones, and may be any quality defect whose location on the product can be identified. Here, for example, a cluster of defects having a specific location, orientation, or size may occur due to certain operating conditions. By accurately identifying whether a quality defect is an isolated defect or a cluster of defects and identifying the trend of the cluster of defects, it becomes possible to quickly identify the cause of the quality defect.

[0024] 1 also represents the processing (step) in each block. The processing flow indicated by the arrows in Fig. 1 represents the quality identification method executed by the quality identification device according to this embodiment, i.e., the method for identifying quality defects in products in the manufacturing process.

[0025] The quality identification device according to this embodiment includes a quality defect location acquisition unit 101, a relationship estimation unit 102, an isolated defect identification unit 103, and an identification result presentation unit 104. The quality defect location acquisition unit 101 acquires the location of occurrence of each of multiple quality defects. The relationship estimation unit 102 estimates a network model that represents the relationship between the locations of occurrence of the multiple quality defects acquired by the quality defect location acquisition unit 101. Specifically, the relationship estimation unit 102 constructs a network model that uses quality defects as nodes (junctions) and indicates the relationship between the quality defects using edges (branches) or the like using a technique described below. The isolated defect identification unit 103 identifies whether a quality defect is an isolated defect or a clustered defect based on the network model constructed by the relationship estimation unit 102. In this embodiment, the isolated defect identification unit 103 identifies quality defects that are isolated defects and identifies quality defects that are not identified as isolated defects as clustered defects. The identification result presenting unit 104 assigns identification information indicating whether each of the multiple quality defects is an isolated defect or a cluster of defects to each of the multiple quality defects, and presents information on the multiple quality defects (identification results) to the user. The user is a person who uses the quality identification device, but may be, for example, a manager who manages the manufacturing process. The user may identify trends in clustered defects and identify the causes of the quality defects based on the presented information on the multiple quality defects.

[0026] The quality identification device may be configured as hardware, for example, a computer. The computer may be a server computer or a portable computer such as a laptop or tablet. The quality identification device may be located at the product manufacturing site or may be located remotely from the manufacturing site, but is capable of communicating with a process computer that manages the product manufacturing process. For example, product quality defects may be measured by an inspection device, and the process computer may store and manage information such as the location of the quality defect in a database. The quality identification device may then obtain information such as the location of the quality defect from the process computer, which is a higher-level computer. The quality identification device may also output the identification results to the process computer or to a display viewed by a user at the manufacturing site.

[0027] The quality identification device may also have the following software configuration: One or more programs used to control the operation of the quality identification device are stored in a storage device (such as a memory) of the computer. The programs stored in the storage device of the computer are loaded by a processor. Then, processing by the quality defect location acquisition unit 101 (quality defect location acquisition step), processing by the relationship estimation unit 102 (relationship estimation step), processing by the isolated defect identification unit 103 (isolated defect identification step), and processing by the identification result presentation unit 104 (identification result presentation step) are executed. Furthermore, a computer functioning as a quality identification device may execute the quality defect location acquisition step, relationship estimation step, isolated defect identification step, and identification result presentation step in this order as a quality identification method.

[0028] As described above, the quality defect position acquisition unit 101 acquires the position where each of the multiple quality defects occurs. The position is specified by coordinates. More specifically, the position where a quality defect occurs is specified by two variables indicating coordinates on a coordinate plane set for the product or three variables indicating coordinates in a coordinate space set for the product. As shown in FIG. 2 , in this embodiment, x and y coordinates are defined parallel to the length direction (longitudinal direction) and width direction (short direction) of the steel plate, which is the product. The position of a defect (quality defect) in the steel plate is then determined by a set of two variables (x and y) corresponding to the coordinate plane. ｉ , y ｉ ) The coordinate plane may be set so that a specific position (for example, one corner) on the steel plate is the origin. Here, the subscript i is an integer between 1 and n, where n is the total number of quality defects. Each quality defect is assigned a unique identifier (ID), and in the example of FIG. 2, the ID is an integer between 1 and n.

[0029] The quality defect position acquisition unit 101 acquires the position of the quality defect having an ID of i, for example (x ｉ , y ｉ ) is obtained, and the location of the quality defect with ID j is obtained. ｊ , y ｊIn this way, the quality defect location acquisition unit 101 acquires the locations of all quality defects with IDs 1 to n. In the following, the distance between two quality defects is represented by d. For example, the distance from the quality defect with ID i to the quality defect with ID j is d. ｉｊ d ｉｊ Had ｊｉ It is equal to (the distance from the quality defect with ID j to the quality defect with ID i).

[0030] In this embodiment, the defects are surface defects, but defects that are not limited to the surface (that may exist inside the steel plate) may also be determined as quality defects. In such cases, the z-coordinate corresponding to the thickness direction of the steel plate is further introduced, and the position where the quality defect occurs can be determined by the set of three variables (x ｉ , y ｉ , z ｉ ) may be identified.

[0031] As described above, the relationship estimation unit 102 estimates the occurrence positions (x １ , y １ ), ..., (x ｉ , y ｉ ), ..., (x ｊ , y ｊ ), ..., (x ｎ , y ｎ In this embodiment, the relationship estimation unit 102 constructs the network model using graph theory. That is, the relationship estimation unit 102 constructs the network model by treating each of the multiple quality defects as a node and determining whether to connect the nodes based on the calculated inter-node distance. The inter-node distance may be, for example, Euclidean distance, and is expressed as the above d ｉｊ (i=1, 2, ..., n, j=1, 2, ..., n) where the distance between nodes is zero when i and j are equal.

[0032] In this embodiment, the relationship estimation unit 102 generates a matrix (distance matrix D) whose components are all the inter-node distances. That is, the distance matrix D is the distance d between any two quality defects (two defects with IDs i and j). ｉｊas its (i, j) components. The distance matrix D is an n×n symmetric matrix. The distance matrix D is expressed by the following formula:

[0033]

[0034] In this embodiment, the relationship estimation unit 102 constructs a network model from the distance matrix D by creating edges between nodes whose components are not 0 in the distance matrix D, with each of the multiple quality defects being a node. To specifically describe the method of creating edges, the relationship estimation unit 102 sets a threshold Z, and calculates a distance d ｉｊ 0, the distance d less than Z ｉｊ is converted to 1.

[0035]

[0036] Then, the distance matrix D whose components have been converted to 0 or 1 becomes the network model. In the distance matrix D whose components have been converted to 0 or 1, 1 corresponds to connection by an edge, and 0 corresponds to no connection by an edge. Therefore, the distance matrix D whose components have been converted to 0 or 1 indicates the presence or absence of edges between all nodes, and becomes a network model that represents the relationships between the locations where multiple quality defects (nodes) occur.

[0037] Here, the configuration of the network model is determined by the threshold Z. ｉｊ The process of converting σ to 1 (corresponding to connection by an edge) is equivalent to drawing a circle with a radius of Z centered on each quality defect and connecting the quality defects within the circle with an edge (see FIG. 7). Here, if the value of threshold Z is too small, the quality defects may not be connected at all, whereas if the value of threshold Z is too large, the quality defects may be connected excessively. In this embodiment, the relationship estimation unit 102 determines threshold Z as follows so that the threshold Z is an appropriate value.

[0038] The relationship estimation unit 102 calculates n ２ components (distance d ｉｊ) is used to create a histogram. Here, if elements not included in the product (for example, distances for quality defects located beyond the entire length of the product in the longitudinal direction) are included as components of the distance matrix D, the histogram may be created excluding such elements. FIG. 3 is a diagram illustrating a histogram. The horizontal axis (Distance) indicates distances, which are components of the distance matrix D. The vertical axis (Frequency) indicates the frequency of each distance. In this embodiment, the relationship estimation unit 102 applies unsupervised clustering, in which the number of clusters is set to two, to such a histogram. The clustering method may be a known method and is not limited to a specific method, but k-means++, for example, can be used. By clustering, each element is classified into cluster A or cluster B. The relationship estimation unit 102 calculates the maximum value of elements belonging to cluster A as d Ａ ｍａｘ , the minimum value of elements belonging to cluster B is d Ｂ ｍｉｎ The average value of these is (d Ａ ｍａｘ +d Ｂ ｍｉｎ ) / 2 is set as the threshold Z.

[0039] Fig. 4 is a diagram illustrating a network model corresponding to the quality defect locations in Fig. 2. The network model shown in the lower diagram of Fig. 4 is constructed by the above-described processing by the relationship estimation unit 102 (generation of distance matrix D, setting of threshold Z based on statistical information of distance matrix D, and binarization of the components of distance matrix D based on threshold Z). Here, the upper diagram of Fig. 4 is the same as Fig. 2, with multiple quality defects (nodes) indicated by black circles. Furthermore, in the lower diagram of Fig. 4, edge connections are indicated by solid lines.

[0040] The isolated defect identifying unit 103 identifies quality defects that are isolated defects based on the network model. Furthermore, the isolated defect identifying unit 103 classifies quality defects that are not identified as isolated defects as clustered defects. In this embodiment, the isolated defect identifying unit 103 searches for articulation points in the network model and identifies quality defects corresponding to the searched articulation points as isolated defects. An articulation point is a vertex in a graph (network model), and is a vertex that, when removed from the graph, causes the graph to become disconnected. The algorithm for searching for articulation points may be a known method and is not limited to a specific method, but a depth-first search algorithm, for example, may be used.

[0041] The isolated defect identifying unit 103 converts the network model into a tree structure and detects articulation points using a depth-first search algorithm. FIG. 5 is a diagram illustrating the identified isolated defects and clustered defects. By detecting the articulation points by the isolated defect identifying unit 103, a network diagram in which isolated defects and clustered defects are identified is generated, as shown in the lower diagram of FIG. 5. Isolated defects are indicated by white circles. Clustered defects are indicated by diagonally shaded circles. The upper diagram of FIG. 5 is the same as the diagram of the network model in FIG. 4.

[0042] Here, the articulation point detection method using the depth-first search algorithm is executed, for example, as follows. First, the isolated defect identification unit 103 calculates ord[u] and low[u] from a certain vertex (u) of the tree structure by depth-first search. ord[u] indicates the ordinal number of the vertex (u) searched in the depth-first search. Furthermore, low[u] indicates the minimum value of ord[w] for a vertex (w) that can be reached from the vertex (u) using a retreat edge at most once. The isolated defect identification unit 103 determines, as an articulation point, a vertex (u) that satisfies the following first or second condition. The first condition is that the vertex (u) is the root of the tree structure and has two or more children. The second condition is that the vertex (u) is not the root of the tree structure, and for a vertex (v) that is a child of the vertex (u), ord[u]≦low[v] is satisfied.

[0043] The identification result presentation unit 104 presents the quality defect identification results obtained by the isolated defect identification unit 103 to the user. In this embodiment, the identification result presentation unit 104 generates the quality defect list and network diagram shown in FIG. 6 and presents them to the user. The quality defect list and the network diagram may be switched, for example, by the user's selection, and one of them may be presented. The quality defect list is a list in which each of multiple quality defects is assigned identification information indicating an isolated defect or a cluster of defects (the "isolated" or "cluster" classification in the upper diagram of FIG. 6). In the example of FIG. 6, each of the multiple quality defects listed in the quality defect list is distinguished by an ID and coordinate information is also indicated. The network diagram is a diagram in which identification information indicating an isolated defect or a cluster of defects is visually assigned to the location where each of the multiple quality defects occurs. In the example of FIG. 6, the network diagram is the same as the lower diagram of FIG. 5, except that isolated defects are visually identified by white circles and clustered defects are visually identified by diagonally-lined circles. The user can use the presented quality defect list or network diagram to accurately identify whether the quality defect is an isolated defect or a cluster of defects, and identify the trend of cluster defects, thereby enabling early identification of the cause of the quality defect.

[0044] The identification result presentation unit 104 may perform a product quality assessment based on the identification information of isolated defects and clustered defects, and determine what to do with the product. Here, product disposition decisions include stopping product shipment, rejecting the product in the next process (removing a portion of the product where the quality defect exists), and shipping the product as is. By utilizing the identification information of isolated defects and clustered defects to improve the accuracy of determining the type and quality grade of the quality defect, it is possible to make an appropriate product disposition decision. This prevents the shipment of products with quality defects, enabling the production of high-quality products.

[0045] As described above, the quality identification device, product manufacturing method, and quality identification method according to the present embodiment, with their configurations described above, can accurately distinguish between clustered defects and isolated defects for quality defects such as surface defects on products, taking into account the occurrence of defects across the entire product using a histogram or the like. For example, when clustered defects are identified using a uniform threshold value as in the past, the user may mistakenly include isolated defects in clustered defects, preventing the user from correctly grasping the trends of clustered defects and resulting in time-consuming analysis of the cause of the quality defects. The quality identification device, product manufacturing method, and quality identification method according to the present embodiment accurately distinguish between isolated defects and clustered defects, which helps to quickly identify the cause of the quality defects.

[0046] Although the embodiments of the present disclosure have been described based on the drawings and examples, it should be noted that those skilled in the art would easily be able to make various modifications or alterations based on the present disclosure. Therefore, it should be noted that these modifications and alterations are included within the scope of the present disclosure. For example, the functions included in each component or step can be rearranged so as not to cause logical inconsistencies, and multiple components or steps can be combined or divided into one. The embodiments of the present disclosure can also be realized as a program executed by a processor included in an apparatus or a storage medium on which a program is recorded. It should be understood that these are also included within the scope of the present disclosure.

[0047] In the above embodiment, the relationship estimation unit 102 constructs a network model by setting the threshold Z based on the statistical information of the distance matrix D, but another method may be used. For example, as shown in FIG. 7 , the relationship estimation unit 102 may draw a circle with each quality defect at its center and set the radius of the circle to the distance d ｉｊmay be gradually changed from the minimum value to the maximum value. In the example of FIG. 7 , the radius of the circle increases from r1 to r2. Then, the network grows by connecting quality defects present within the circle with edges (the number of edges in the network model increases). The relationship estimation unit 102 sets the radius of the circle at the point when at least one edge is connected for all quality defects as the threshold Z. In other words, the relationship estimation unit 102 grows the network and stops the growth when at least one edge is connected for all quality defects, thereby constructing a network model. Construction of such a network model is within the scope of the present disclosure.

[0048] In the above embodiment, the isolated defect identification unit 103 distinguishes between isolated defects and clustered defects by detecting articulation points using a depth-first search algorithm. However, other methods may be used. For example, instead of a method (first identification method) in which isolated defects such as articulation points are identified and the remaining nodes are treated as clustered defects, a method (second identification method) in which nodes that are candidates for clustered defects are identified and a set of identified nodes that reaches a predetermined size is treated as a clustered defect may be adopted. When the first identification method is adopted, the Girvan-Newman algorithm may be used instead of the depth-first search algorithm in the above embodiment. Furthermore, when the second identification method is adopted, a method such as the Ravasz algorithm, a greedy algorithm, or the Louvain algorithm, which are known as agglomerative methods, may be used. Distinguishing between isolated defects and clustered defects using any of these methods is within the scope of the present disclosure.

[0049] Here, the quality identification device (e.g., a computer) may not be a single device, but may be composed of multiple devices located in multiple locations and capable of sending and receiving data to and from each other via a network. In other words, multiple devices connected via a network may function as a quality identification device as a whole. Therefore, for example, the quality identification device may be composed of a single computer in terms of hardware configuration, or may be composed of multiple computers connected via a network. When composed of multiple computers, a shared memory accessible by each computer may be used to share data or programs.

[0050] 101 Quality defect position acquisition unit 102 Relationship estimation unit 103 Isolated defect identification unit 104 Identification result presentation unit

Claims

1. A quality identification device that identifies whether a quality defect that has occurred in a product is an isolated defect or a cluster defect based on the location of the quality defect, comprising: a quality defect location acquisition unit that acquires the location of each of a plurality of the quality defects; a relationship estimation unit that estimates a network model that represents the relationship between the locations of the plurality of quality defects; an isolated defect identification unit that identifies the quality defects that are isolated defects based on the network model and identifies quality defects that are not identified as isolated defects as cluster defects; and an identification result presentation unit that assigns identification information indicating whether the quality defects are isolated defects or cluster defects to each of the plurality of quality defects and presents information on the plurality of quality defects to which the identification information has been assigned to a user.

2. The quality identification device of claim 1, wherein the location of the quality defect is identified by two variables indicating coordinates on a coordinate plane set for the product or three variables indicating coordinates in a coordinate space set for the product.

3. The quality identification device according to claim 1 or 2, wherein the relationship estimation unit constructs the network model by treating each of the multiple quality defects as a node and determining whether to connect the nodes based on the calculated inter-node distance.

4. The quality identification device according to claim 3, wherein the relationship estimation unit sets a threshold based on statistical information of the calculated node-to-node distance, and connects nodes whose calculated node-to-node distance is less than the threshold.

5. The quality identification device of claim 3 or 4, wherein the relationship estimation unit creates a histogram of the calculated inter-node distances, classifies each element into two clusters by clustering, and sets the average value of the maximum value of elements belonging to one cluster and the minimum value of elements belonging to the other cluster as the threshold.

6. A quality identification device as described in any one of claims 1 to 5, wherein the relationship estimation unit draws a circle centered on each of the nodes, connects them to other nodes within the circle with edges, and gradually increases the radius of the circle until at least one edge is connected to all of the nodes.

7. A quality identification device according to any one of claims 1 to 6, wherein the isolated defect identification unit searches for joint points in the network model and identifies the quality defect corresponding to the searched joint point as an isolated defect.

8. A quality identification device as described in any one of claims 1 to 7, wherein the identification result presentation unit generates a quality defect list, which is a list in which each of the multiple quality defects is assigned identification information indicating an isolated defect or a cluster of defects, and a network diagram in which identification information indicating an isolated defect or a cluster of defects is visually assigned to the location where each of the multiple quality defects occurs, and presents the quality defect list and the network diagram to a user.

9. A quality identification device according to any one of claims 1 to 8, wherein the product is a steel plate, and the quality defect is a surface defect of the steel plate or a defect present inside the steel plate.

10. The quality identification device according to any one of claims 1 to 9, wherein the identification result presentation unit makes a decision on the disposal of the product based on the identification information of isolated defects and clustered defects.

11. A method for manufacturing a product, comprising: performing a product treatment based on the product treatment decision made by the quality identification device according to claim 10.

12. A quality identification method executed by a quality identification device that identifies whether a quality defect occurring in a product is an isolated defect or a cluster defect based on the location of the quality defect, the quality identification method comprising: a quality defect location acquisition step that acquires the location of occurrence of each of a plurality of the quality defects; a relationship estimation step that estimates a network model that represents the relationship between the locations of occurrence of the plurality of the quality defects; an isolated defect identification step that identifies the quality defects that are isolated defects based on the network model and identifies the quality defects that are not identified as isolated defects as cluster defects; and an identification result presentation step that assigns identification information indicating an isolated defect or a cluster defect to each of the plurality of the quality defects and presents information on the plurality of the quality defects to which the identification information has been assigned to a user.

Images