Causal relationship inference system, production support system, and program
Patent Information
- Authority / Receiving Office
- JP · JP
- Patent Type
- Applications
- Current Assignee / Owner
- JTEKT CORP
- Filing Date
- 2025-01-08
- Publication Date
- 2026-07-21
AI Technical Summary
Existing causal relationship inference systems, such as SCORE, fail to incorporate prior information, leading to inaccuracies in inferring causal structures for multivariate data.
A causal relationship inference system that includes a storage device and processor, utilizing algorithms to infer topological order and generate directed edges based on prior information, with sparse processing to reduce erroneous edges.
Improves the accuracy of causal relationship inference by suppressing incorrect inferences and reducing unnecessary directed edges, enabling better understanding of causal structures and supporting production processes.
Smart Images
Figure 2026119850000001_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to a causal relationship inference system, a production support system, and a program.
Background Art
[0002] Conventionally, as a technique for estimating a causal structure in analysis target data by structural equation modeling means and generating a non-linear regression model based on the estimated result, Patent Document 1 (Japanese Patent Application Laid-Open No. 2006-099482) is known.
Prior Art Documents
Patent Documents
[0003]
Patent Document 1
Summary of the Invention
Problems to be Solved by the Invention
[0004] Conventionally, as a method for inferring a causal relationship for multivariate data, for example, SCORE is known. This SCORE can accurately generate a causal graph for data including a plurality of nodes representing a plurality of variables.
[0005] However, SCORE cannot incorporate the prior information possessed by the operator into the inference process. Therefore, there is room for improvement in the accuracy of inferring the causal relationship.
[0006] The present invention has been made in view of such problems, and aims to provide a causal relationship inference system, a production support system, and a program with improved inference accuracy of the causal structure relationship.
Means for Solving the Problems
[0007] A first aspect of the present invention is including a storage device and a processor, The aforementioned storage device is Data containing multiple nodes representing multiple variables, An algorithm for inferring the topological order for each of the aforementioned multiple nodes, from leaf nodes that have no child nodes to root nodes that have no parent nodes, Store it, The aforementioned processor, A first prior information acquisition unit acquires first prior information, which is information that a causal relationship exists between two nodes included in the plurality of nodes, and that one of the two nodes is the cause node and the other is the effect node. A topological order inference unit that infers the topological order for each of the plurality of nodes by applying the algorithm to the plurality of nodes, Based on the first prior information, a cause node exclusion unit excludes the cause node corresponding to the result node from the data until the result node is inferred by the topological order inference unit, A directed edge generation unit generates directed edges between the plurality of nodes based on the topological order inferred by the topological order inference unit, The causal relationship inference system includes a sparse processing unit that reduces the number of directed edges after processing by the directed edge generation unit.
[0008] A second aspect of the present invention is: Including a storage device and a processor, The aforementioned storage device is Data containing multiple nodes representing multiple variables, An algorithm for inferring the topological order for each of the aforementioned multiple nodes, from leaf nodes that have no child nodes to root nodes that have no parent nodes, Store it, The aforementioned processor, A second prior information acquisition unit acquires second prior information, which is information that there are two nodes in the aforementioned data that do not have a causal relationship. A topological order inference unit that infers the topological order for each of the plurality of nodes by applying the algorithm to the plurality of nodes, A directed edge generation unit generates directed edges between the plurality of nodes based on the topological order inferred by the topological order inference unit, After processing by the directed edge generation unit, a directed edge removal unit removes the directed edge formed between two nodes where no causal relationship exists, based on the second prior information. The causal relationship inference system includes a sparse processing unit that reduces the number of directed edges after processing by the directed edge removal unit.
[0009] A third aspect of the present invention is: The relational reasoning system described above, An input unit for inputting the aforementioned data, An output unit outputs the quality result of the workpiece, the cause of any abnormality in the workpiece, or the machine setting conditions or machining conditions of the machine tool, based on the input data and the causal graph generated by the processing by the sparse processing unit. It is a production support system equipped with these features.
[0010] A fourth aspect of the present invention is: A first prior information acquisition procedure for acquiring first prior information, which is information that a causal relationship exists between two nodes in data containing multiple nodes representing multiple variables, and that one of the two nodes is the cause node and the other is the effect node, A topological order inference procedure for each of the aforementioned plurality of nodes, by applying an algorithm for inferring the topological order to each of the plurality of nodes in order from leaf nodes that do not have child nodes to root nodes that do not have parent nodes, Based on the first prior information, a cause node exclusion procedure for excluding the cause node corresponding to the result node from the data until the result node is inferred by the topological order inference procedure; Based on the topological order inferred by the topological order inference procedure, a directed edge generation procedure for generating a directed edge between the plurality of nodes; After the directed edge generation procedure, a sparse processing procedure for reducing the number of the directed edges; is in a program for causing a computer to execute.
[0011] A fifth aspect of the present invention is a second prior information acquisition procedure for acquiring second prior information which is information that there are two nodes having no causal relationship among data including a plurality of nodes representing a plurality of variables; For each of the plurality of nodes, by applying an algorithm for inferring a topological order in order from a leaf node having no child node to a root node having no parent node, from the leaf node in order to the root node having no parent node, a topological order inference procedure for inferring a topological order for each of the plurality of nodes; Based on the topological order inferred by the topological order inference procedure, a directed edge generation procedure for generating a directed edge between the plurality of nodes; After the directed edge generation procedure, a directed edge removal procedure for removing the directed edge formed between the two nodes having no causal relationship based on the second prior information; After the directed edge removal procedure, a sparse processing procedure for reducing the number of the directed edges; is in a program for causing a computer to execute.
Advantages of the Invention
[0012] According to the first and fourth aspects of the present invention, it is suppressed that the topological order of the cause node is inferred before the result node. Thereby, it is suppressed that the causal relationship between the cause node and the result node is erroneously inferred, so that the accuracy of the causal relationship inference system can be improved.
[0013] According to the second and fifth aspects of the present invention, the directed edge generated between two nodes having no causal relationship can be surely removed. Thereby, it is suppressed that a directed edge is erroneously generated between two nodes having no causal relationship, so that the accuracy of the causal relationship inference system can be improved.
[0014] According to the third aspect of the present invention, based on the causal structure data, it is possible to obtain the quality result of the workpiece, the cause of the abnormality when an abnormality occurs in the workpiece, or the machine setting conditions or machining conditions of the machine tool. Thereby, it is possible to improve the quality of the workpiece or improve the productivity of the workpiece.
Brief Description of Drawings
[0015] [Figure 1] The block diagram which shows the production support system which concerns on Embodiment 1. [Figure 2] The block diagram which shows the memory | storage device which concerns on Embodiment 1. [Figure 3] The block diagram which shows the processor which concerns on Embodiment 1. [Figure 4] The schematic diagram which shows an example of the causal graph which concerns on Embodiment 1. [Figure 5] The flowchart which shows the process of the causal relationship inference system which concerns on Embodiment 1. [Figure 6] The flowchart which shows the topological order inference process which concerns on Embodiment 1. [Figure 7] The schematic diagram which shows the state where data was acquired in Embodiment 1. [Figure 8]A schematic diagram showing the state in Embodiment 1 where, after the causal node is excluded based on first prior information, leaf nodes are inferred, and then directed edges are generated for the leaf nodes. [Figure 9] This schematic diagram shows the state in Embodiment 1 where leaf nodes are inferred with the causal node excluded, and then directed edges are generated for the leaf nodes. [Figure 10] This schematic diagram shows the state in Embodiment 1 where a causal node is added, a directed edge is then generated from the causal node to the result node, and the topological order is inferred with the causal node still added. [Figure 11] A schematic diagram showing the state in Embodiment 1 where the penalty for the cause node is set to 0. [Figure 12] A schematic diagram showing the state after processing by the sparse processing unit has been performed in Embodiment 1. [Figure 13] A block diagram showing a storage device according to Embodiment 2. [Figure 14] A block diagram showing a processor according to Embodiment 2. [Figure 15] A flowchart illustrating the processing of the causal relationship inference system according to Embodiment 2. [Figure 16] A schematic diagram showing the state in Embodiment 2 where directed edges without causal relationships have been removed based on the second prior information. [Modes for carrying out the invention]
[0016] (Embodiment 1) Referring to Figures 1 to 4, the production support system 1 according to this embodiment 1 will be described. As shown in Figure 1, the production support system 1 comprises a causal relationship inference system 2, an input unit 3, an output unit 4, a machine tool 5, and a server 6. The causal relationship inference system 2, the input unit 3, the output unit 4, the machine tool 5, and the server 6 are connected to a network 7 and configured to communicate with each other. However, there may be two or more machine tools 5, or the machine tool 5 may be omitted.
[0017] Server 6 is equipped with any storage medium, such as a hard disk drive or semiconductor memory. However, Server 6 may be a virtual server on the internet or an intranet.
[0018] The machine tool 5 performs a predetermined machining operation on a workpiece (not shown). The machine tool 5 is not particularly limited and may include, for example, a lathe, machining center, milling machine, gear cutting machine, boring machine, etc.
[0019] The causal relationship inference system 2 comprises a storage device 10, a processor 20, and a display unit 30. In this embodiment 1, the storage device 10, the processor 20, and the display unit 30 constitute a computer.
[0020] The storage device 10 can be any storage device 10 as appropriate, such as a hard disk drive, semiconductor memory, or USB (Universal Serial Bus) memory. As shown in Figure 2, the storage device 10 stores data 11, algorithm 12, causal graph 13, program 14, and first prior information 16. However, the storage device 10 may be configured to be located on the server 6.
[0021] Data 11 includes multiple nodes N representing multiple variables. Data 11 includes machine setting conditions or machining conditions of the machine tool 5 that perform a predetermined process on the workpiece, and the quality result of the workpiece after the predetermined process has been performed on the workpiece. However, Data 11 may also be configured to include data other than those described above.
[0022] Algorithm 12 is an algorithm for estimating the causal relationships of data 11. For each of the multiple nodes N, Algorithm 12 infers the topological order 15 sequentially, from leaf nodes that have no child nodes to root nodes that have no parent nodes.
[0023] The process is explained in detail below. First, we infer which of the multiple nodes N contained in all the data 11 are leaf nodes. Next, we remove the leaf nodes from the data 11 and place the removed leaf nodes at the end of the set in topological order 15. Then, we infer which of the multiple nodes N contained in the data 11 from which the leaf nodes were removed are leaf nodes. We perform the above process sequentially until we reach the root node, which has no parent node. As a result, the set of data 11 becomes an empty set, and the set in topological order 15 becomes a state where multiple nodes N are placed in the order of topological order 15.
[0024] Topological order 15 is the order of multiple nodes N such that no node N is placed before any of its ancestor nodes. An ancestor node refers to the parent node of a given node N, or all of the parent nodes N that precede that parent node.
[0025] Algorithm 12 is not particularly limited and can be appropriately selected from, for example, SCORE, RESIT, etc. In this embodiment 1, SCORE is selected as algorithm 12.
[0026] The causal graph 13 connects two of multiple nodes N with a directed edge DE, showing the causal relationship between the multiple nodes N. The causal relationship includes the causal direction, which is the direction of causality. The causal relationship may also include weights. However, a directed edge DE is defined as an edge that connects two nodes N and has a direction.
[0027] Figure 4 shows an example of a causal graph 13. Figure 4 shows multiple nodes N (35 in this embodiment 1) represented by numbers enclosed in elliptical shapes, and directed edges DE connecting the nodes N. In this embodiment 1, the nodes N are assigned numbers from 0 to 34. However, the numbers are assigned to distinguish each node N and are unrelated to the topological order 15.
[0028] The topological order 15 described above is not uniquely determined. The topological order 15 is correct as long as a node N does not appear earlier in the sequence than its ancestor nodes. Therefore, in the causal graph 13 of Figure 4, Node 1 → Node 10 → Node 5 → Node 20 → Node 2 → ... You could set a topological order of 15, or, Node 10 → Node 1 → Node 11 → Node 5 → Node 20 → ... You can also set a topological order of 15.
[0029] In Figure 4, node 34 is a leaf node with no child nodes. Also, nodes 1, 10, 5, 15, 4, 12, 9, 6, 8, 7, 2, 18, 21, 22, 0, and 3 are root nodes with no parent nodes. The causal graph 13 may have one leaf node or two or more leaf nodes. Also, the causal graph 13 may have one root node or two or more root nodes.
[0030] Program 14 causes the computer to execute the following steps: a first prior information acquisition procedure (S1), a topological order inference procedure (S4), a cause node exclusion procedure (S41), a directed edge generation procedure (S43), and a sparse processing procedure (S6) (see Figures 5 and 6).
[0031] The first prior information 16 is information that a causal relationship exists between two nodes N included in data containing multiple nodes N representing multiple variables, and that one of the two nodes N is the cause node and the other is the effect node. However, the first prior information 16 may be a case where the cause node and the effect node have a direct causal relationship, or it may be a case where one or more nodes N are intervening between the cause node and the effect node to have a causal relationship.
[0032] The first pre-information acquisition procedure (S1) acquires the first pre-information 16.
[0033] The topological order inference procedure (S4) infers the topological order 15 for each of the multiple nodes N by applying an algorithm to infer the topological order 15 sequentially to each of the multiple nodes N, from leaf nodes that have no child nodes down to root nodes that have no parent nodes.
[0034] The cause node exclusion procedure (S41) excludes cause nodes corresponding to result nodes from the data based on the first prior information 16 until the result node is inferred by the topological order inference procedure.
[0035] The directed edge generation procedure (S43) generates directed edges DE between multiple nodes N based on the topological order 15 inferred by the topological order inference procedure (S4).
[0036] The sparse processing procedure (S6) reduces the number of directed edges DE after the directed edge generation procedure (S43).
[0037] As shown in Figure 3, the processor 20 includes a first pre-information acquisition unit 21, a topological order inference unit 22, a cause node exclusion unit 23, a directed edge generation unit 24, a penalty processing unit 25, and a sparse processing unit 26.
[0038] The first pre-information acquisition unit 21 acquires first pre-information 16, which is information that a causal relationship exists between two nodes N included in a plurality of nodes N, and that one of the two nodes N is the cause node and the other is the effect node.
[0039] The topological order inference unit 22 infers the topological order 15 for each of the multiple nodes N by applying the algorithm to the multiple nodes N.
[0040] The cause node exclusion unit 23 excludes cause nodes corresponding to result nodes from the data 11 based on the first prior information 16 until the result node is inferred by the topological order inference unit 22.
[0041] The directed edge generation unit 24 generates directed edges DE between multiple nodes N based on the topological order 15 inferred by the topological order inference unit 22. Based on the topological order 15, the directed edge generation unit 24 generates directed edges DE between the leaf node, which is the last node in the topological order 15 among the multiple nodes N, and a node N that is different from the leaf node. These directed edges DE have a direction from the node N that is different from the leaf node toward the leaf node.
[0042] After processing by the directed edge generation unit 24, the penalty processing unit 25, based on the first prior information 16, sets the penalty in the processing by the sparse processing unit 26 to 0 only for cause nodes that have a causal relationship with the result node.
[0043] The sparse processing unit 26 reduces the number of directed edges DE after processing by the directed edge generation unit 24. The method for reducing the number of directed edges DE is not particularly limited, and any method such as Causal Additive Model (CAM), Least Absolute Shrinkage and Selection Operator (Lasso), or Group Lasso can be selected.
[0044] Returning to Figure 1, the display unit 30 displays the causal graph 13 generated by the processing performed by the sparse processing unit 26. By looking at the causal graph 13 displayed on the display unit 30, the operator can understand the causal relationships in the processing of the workpiece by the machine tool 5.
[0045] The display unit 30 can be a known display device such as a liquid crystal display. The display unit 30 may be configured to be installed in a computer, in a machine tool 5, or as a portable terminal carried by an operator.
[0046] The input unit 3 is not particularly limited, and any input device such as a keyboard, mouse, touch panel, joystick, or voice input device can be selected. The operator inputs the data 11 to the causal relationship inference system 2 via the input unit 3.
[0047] The output unit 4 is not particularly limited, and any output device such as an LCD display, tablet terminal, or printer can be selected.
[0048] The output unit 4 outputs the causal graph 13 estimated by the causal relationship inference system 2 and notifies the operator.
[0049] Furthermore, the output unit 4 outputs the quality result of the workpiece, the cause of any abnormalities in the workpiece, or the machine setting conditions or machining conditions of the machine tool 5, based on the input data 11 and the causal graph 13 generated by the processing by the sparse processing unit 26. However, the information output by the output unit 4 is not limited to the above.
[0050] For example, if data 11 regarding a malfunction of machine tool 5 is input, the system can output the predicted workpiece quality result due to the malfunction of machine tool 5 based on the input data 11 and the causal graph 13.
[0051] Furthermore, for example, if data 11 regarding an abnormality that occurred in a workpiece is input, the cause of the abnormality that occurred in the workpiece, such as vibration or deterioration of the machine tool 5, can be output based on the input data 11 and the causal graph 13.
[0052] Furthermore, for example, if a desired numerical value is entered for the dimensions of a workpiece, the machine setting conditions or machining conditions for the machine tool 5 for manufacturing a workpiece with the dimensions specified by the entered numerical value can be output based on the entered data 11 and the causal graph 13.
[0053] Next, the processing of this embodiment 1 will be described with reference to Figure 5. Figure 5 shows a flowchart of the processing of the causal relationship inference system 2 according to this embodiment 1.
[0054] When the causal relationship inference system 2 is activated, the processor 20 acquires data 11 from the storage device 10 or the input unit 3 (S1).
[0055] Next, in step S2, the processor 20 performs initialization. This initialization clears the set of topological order 15, making it an empty set.
[0056] Next, the first pre-information acquisition unit 21 acquires the first pre-information 16 (S3). The first pre-information 16 may be acquired from a server where the first pre-information 16 is pre-stored, from a storage device 10 where the first pre-information 16 is pre-stored, or by an operator inputting it via the input unit 3.
[0057] Next, in step S4, a topological order inference process is performed on multiple nodes N. The topological order inference process is an example of a topological order inference procedure.
[0058] Next, after processing by the directed edge generation unit 24, the penalty processing unit 25, based on the first prior information 16, sets the penalty in the processing by the sparse processing unit 26 to 0 only for cause nodes that have a causal relationship with the result node (S5).
[0059] In step S6, the sparse processing unit 26 reduces the number of directed edges DE. As a result, the causal relationship inference system 2 generates the causal graph 13.
[0060] With this, the processing of the causal relationship inference system 2 is completed.
[0061] Next, we will explain the topological order inference process (S4) with reference to Figure 6. Figure 6 shows a flowchart of the topological order inference process (S4).
[0062] When the topological order inference process (S4) is executed, the cause node exclusion unit 23 excludes the cause node for the result node from the data 11 based on the first prior information 16 (S41).
[0063] Next, the topological order inference unit 22 infers leaf nodes by applying SCORE to the data 11 from which the causal node has been excluded (S42). The directed edge generation unit 24 generates directed edges DE toward the leaf nodes from the nodes N included in the data 11 from which the causal node has been excluded, excluding the leaf nodes (S43).
[0064] Next, in step S44, leaf nodes are removed from data 11.
[0065] Next, in step S45, the topological order inference unit 22 places the leaf node at the end of the set in topological order 15.
[0066] Next, in step S46, the cause node exclusion unit 23 determines whether node N, which was inferred to be a leaf node, was a result node in the first prior information 16. If node N, which was inferred to be a leaf node, was a result node (S46:Y), the cause node exclusion unit 23 adds the cause node corresponding to the result node to the data 11 based on the first prior information 16 (S47).
[0067] Next, in step S48, it is determined whether data 11 is an empty set. If data 11 is an empty set (S48:Y), the topological order inference process ends.
[0068] If data 11 is not an empty set (S48:N), return to S42 in Figure 6.
[0069] In step S46, if the leaf node was not designated as a result node in the first prior information 16 (S46:N), then step S48 in Figure 6 is executed.
[0070] The topological order inference process is now complete.
[0071] Next, the operation of the causal relationship inference system 2 will be explained with reference to Figures 7 to 12.
[0072] Figure 7 shows the state after the processing of S1 in Figure 5 is completed. That is, data 11 contains multiple nodes N1 to N5 (five in this embodiment 1) indicated by letters enclosed in circles. However, the number of nodes N in data 11 is not limited to five; it may be one to four, or six or more.
[0073] In this embodiment 1, the first prior information 16 shows a case where a direct causal relationship exists between node C(N3) and node D(N4), and node C(N3) is the cause node and node D(N4) is the result node.
[0074] Figure 8 shows that the causative node C(N3) has been removed from data 11, and directed edges DE1, DE2, and DE3 have been generated for the leaf node E(N5) from nodes A(N1), B(N2), and D(N4), excluding node E(N5). This prevents the incorrect generation of directed edges DE from nodes E(N5) and D(N4) towards node C(N3).
[0075] Figure 9 shows a state in which directed edges DE4 and DE5 are generated from nodes A(N1) and B(N2) for node D(N4), which has a higher topological order 15 than node E(N5).
[0076] Next, node C(N3), which is the cause node for node D(N4), is added to data 11. Then, based on the first prior information 16, a directed edge DE8 is generated from the cause node C(N3) to the result node D(N4). Subsequently, directed edges DE6 and DE7 are generated from nodes A(N1) and B(N2), which have a higher topological order 15 than node C(N3), toward node C(N3) (see Figure 10).
[0077] Next, a directed edge DE9 is generated from node B(N2) to node A(N1) in the order of topological order 15. Then, based on the first prior information 16, the penalty for node C(N3), which is the cause node, is set to 0 for the directed edge DE8 going from node C(N3) to node D(4) (see Figure 11). This prevents the directed edge DE8 going from node C(N3) to node D(N4) from being deleted by the sparse processing unit 26.
[0078] Referring to Figure 12 and Figure 11 together, we can see that in Figure 12, the directed edges DE1, DE5, DE3, DE4, and DE5 that were shown in Figure 11 have been removed. As a result, a causal graph 13 is completed with a reduced number of directed edges DE.
[0079] Next, the effects of this embodiment 1 will be described. The causal inference system according to this embodiment 1 includes a storage device 10 and a processor 20. The storage device 10 stores data 11 including multiple nodes N representing multiple variables, and an algorithm for inferring the topological order 15 for each of the multiple nodes N, in order from leaf nodes that do not have child nodes to root nodes that do not have parent nodes.
[0080] The processor 20 includes a first pre-information acquisition unit 21, a topological order inference unit 22, a cause node exclusion unit 23, a directed edge generation unit 24, and a sparse processing unit 26.
[0081] The first prior information acquisition unit 21 acquires first prior information 16, which is information that a causal relationship exists between two nodes N included in a plurality of nodes N, and that one of the two nodes N is the cause node and the other is the effect node. The topological order inference unit 22 infers the topological order 15 for each of the plurality of nodes N by applying the algorithm 12 to the plurality of nodes N.
[0082] The cause node exclusion unit 23 excludes cause nodes corresponding to result nodes from the data 11 based on the first prior information 16 until the result node is inferred by the topological order inference unit 22. The directed edge generation unit 24 generates directed edges DE between multiple nodes N based on the topological order 15 inferred by the topological order inference unit 22. The sparse processing unit 26 reduces the number of directed edges DE after processing by the directed edge generation unit 24.
[0083] According to this embodiment 1, the inference of the topological order 15 of the causal node before the result node is suppressed. This suppresses the inference of incorrect causal relationships between the causal node and the result node, thereby improving the accuracy of the causal relationship inference system.
[0084] The processor 20 according to this embodiment 1 further includes a penalty processing unit 25 that, after processing by the directed edge generation unit 24, sets the penalty in the processing by the sparse processing unit 26 to 0 only for cause nodes that have a causal relationship with the result node, based on the first prior information 16.
[0085] According to this embodiment 1, it is possible to prevent directed edges DE formed between a cause node and an effect node, whose causal relationship is known in advance by the first prior information 16, from being incorrectly reduced by the sparse processing unit 26.
[0086] The algorithm 12 according to this embodiment 1 is SCORE. According to this embodiment, causal relationships can be accurately inferred for data 11 that includes multiple nodes N.
[0087] The causal relationship inference system 2 according to this embodiment 1 further includes a display unit 30 that displays a causal graph 13 generated by processing by the sparse processing unit 26.
[0088] The causal graph 13 generated after processing by the sparse processing unit 26 has unnecessary directed edges DE removed, making it easier for the operator to understand the causal structure. According to this embodiment 1, a causal graph 13 that is easy for the operator to understand can be displayed, thereby improving the operator's work efficiency.
[0089] The data 11 in this embodiment 1 includes machine setting conditions or machining conditions of a machine tool 5 that performs a predetermined process on a workpiece, and data 11 relating to the quality result of the workpiece after the predetermined process has been performed on the workpiece.
[0090] According to this embodiment 1, the causal relationship inference system 2 can be applied to a machine tool 5 that processes a workpiece.
[0091] The production support system 1 according to this embodiment 1 comprises the above-described causal structure system, an input unit 3 for inputting data 11, and an output unit 4.
[0092] Based on the input data 11 and the causal graph 13 generated by processing by the sparse processing unit 26, the output unit 4 outputs the quality result of the workpiece, the cause of any abnormalities in the workpiece, or the machine setting conditions or machining conditions of the machine tool 5.
[0093] According to this embodiment 1, based on the causal graph 13, the quality result of the workpiece, the cause of any abnormality in the workpiece, or the machine setting conditions or machining conditions of the machine tool 5 can be obtained. This makes it possible to improve the quality of the workpiece or the productivity of the workpiece.
[0094] The program 14 according to this embodiment 1 comprises a first prior information acquisition procedure (S1), a topological order inference procedure (S4), a cause node exclusion procedure (S41), a directed edge generation procedure (S43), and a sparse processing procedure (S6).
[0095] The first prior information acquisition procedure (S1) acquires first prior information 16, which is information that a causal relationship exists between two nodes N included in data 11 containing multiple nodes N representing multiple variables, and that one of the two nodes N is the cause node and the other is the effect node. The topological order inference procedure (S4) infers the topological order 15 for each of the multiple nodes N by applying an algorithm that infers the topological order 15 sequentially to each of the multiple nodes N, from leaf nodes that have no child nodes to root nodes that have no parent nodes.
[0096] The cause node exclusion procedure (S41) excludes cause nodes corresponding to result nodes from the data 11 based on the first prior information 16 until the result node is inferred by the topological order inference procedure (S4). The directed edge generation procedure (S43) generates directed edges DE between multiple nodes N based on the topological order 15 inferred by the topological order inference procedure (S4). The sparse processing procedure (S6) reduces the number of directed edges DE after the directed edge generation procedure (S43).
[0097] (Embodiment 2) Next, Embodiment 2 will be described with reference to Figures 6 and 13-14. Note that, unless otherwise specified, reference numerals used in Embodiment 2 and later that are the same as those used in the previously described embodiments represent the same components as those in the previously described embodiments.
[0098] As shown in Figure 13, the storage device 10 according to this second embodiment differs from that of the first embodiment in that it stores second prior information 17. The second prior information 17 is information that there are two nodes N in the data 11 that do not have a causal relationship.
[0099] As shown in Figure 14, the processor 20 according to this second embodiment differs from the first embodiment in that it includes a second pre-information acquisition unit 27 and a directed edge removal unit 28 instead of a penalty processing unit 25.
[0100] The second pre-information acquisition unit 27 acquires second pre-information 17, which is information that there are two nodes N in the data 11 that do not have a causal relationship. The second pre-information 17 may be data 11 in which there is no causal relationship between two nodes N whose topological order 15 is consecutive, or it may be data 11 in which there is no causal relationship between two nodes N whose topological order 15 is not consecutive. However, two nodes N whose topological order 15 is not consecutive means a state in which one or more nodes N are interposed between the topological order 15 of one node N and the topological order 15 of the other node N.
[0101] The directed edge removal unit 28 removes directed edges DE formed between two nodes N that do not have a causal relationship, based on the second prior information 17, after processing by the directed edge generation unit 24. This ensures that directed edges DE generated between two nodes N that do not have a causal relationship are reliably removed.
[0102] Furthermore, the program according to this second embodiment is to be executed by the computer as follows: a second prior information acquisition procedure (S7), a topological order inference procedure (S4), a directed edge generation procedure (S43 in Figure 6), a directed edge removal procedure (S8), and a sparse processing procedure (S6).
[0103] The second prior information acquisition procedure (S7) is the information that, in the data 11 which contains multiple nodes N representing multiple variables, there are two nodes N that do not have a causal relationship. The topological order inference procedure (S4) applies an algorithm to each of the multiple nodes N to infer the topological order 15 sequentially, from leaf nodes that have no child nodes to root nodes that have no parent nodes, thereby inferring the topological order 15 for each of the multiple nodes N sequentially, from leaf nodes to root nodes that have no parent nodes.
[0104] The directed edge generation procedure (S43 in Figure 6) generates directed edges DE between multiple nodes N based on the topological order 15 inferred by the topological order inference procedure. The directed edge removal procedure (S8), after the directed edge generation procedure (S43 in Figure 6), removes directed edges DE formed between two nodes N that do not have a causal relationship, based on the second prior information 17. The sparse processing procedure (S6), after the directed edge removal procedure (S8), reduces the number of directed edges DE.
[0105] The process of this embodiment 2 will now be described with reference to Figure 15. Figure 15 shows a flowchart of this embodiment 2. Since steps S1 to S4 and S6 in Figure 15 are the same as steps S1 to S4 and S6 in Figure 5, redundant explanations will be omitted.
[0106] The second pre-information acquisition unit 27 acquires the second pre-information 17 (S7). The second pre-information 17 may be acquired from a server in which the second pre-information 17 is pre-stored, from a storage device in which the second pre-information 17 is pre-stored, or by an operator inputting it via the input unit.
[0107] After the topological order inference process (S4) is performed, the directed edge removal unit 28 removes the directed edge DE formed between two nodes N that do not have a causal relationship, based on the second prior information 17 (S8).
[0108] Next, the operation of this second embodiment will be explained with reference to Figures 11 and 16. Figure 16 shows the data 11 after the processing of S8 in Figure 15 has been executed. Figure 16 shows an example in which the second prior information 17 included information that there is no causal relationship between the directed edge DE4 connecting node A(N1) and node D(N4). Referring to Figure 16 and Figure 11 together, it can be seen that the directed edge DE4 that connected node A(N1) and node D(N4) in Figure 11 has been deleted. This prevents the incorrect inference that a causal relationship exists between node A(N1) and node D(N4).
[0109] Next, the effects of this second embodiment will be described. The processor 20 according to this second embodiment includes a second prior information acquisition unit 27 that acquires second prior information 17, which is information that there are two nodes N in the data 11 that do not have a causal relationship, and a directed edge removal unit 28 that, after processing by the directed edge generation unit 24, removes a directed edge DE formed between the two nodes N that do not have a causal relationship based on the second prior information 17. The sparse processing unit 26 reduces the number of directed edges DE after processing by the directed edge removal unit 28.
[0110] According to this second embodiment, directed edges DE generated between two nodes N that do not have a causal relationship can be reliably removed. This suppresses the erroneous generation of directed edges DE between two nodes N that do not have a causal relationship, thereby improving the accuracy of the causal relationship inference system.
[0111] The causal relationship inference system 2 according to this embodiment 2 includes a storage device 10 and a processor 20. The storage device 10 stores data 11 including multiple nodes N representing multiple variables, and an algorithm 12 for inferring the topological order 15 for each of the multiple nodes, in order from leaf nodes that do not have child nodes to root nodes that do not have parent nodes.
[0112] The processor 20 includes a second pre-information acquisition unit 27, a topological order inference unit 22, a directed edge generation unit 24, a directed edge removal unit 28, and a sparse processing unit 26.
[0113] The second prior information acquisition unit 27 acquires second prior information 17, which is information that there are two nodes N in the data 11 that do not have a causal relationship. The topological order inference unit 22 infers the topological order 15 for each of the multiple nodes N by applying an algorithm to the multiple nodes N.
[0114] The directed edge generation unit 24 generates directed edges DE between multiple nodes N based on the topological order 15 inferred by the topological order inference unit 22. After processing by the directed edge generation unit 24, the directed edge removal unit 28 removes directed edges DE formed between two nodes N that do not have a causal relationship, based on the second prior information 17. After processing by the directed edge removal unit 28, the sparse processing unit 26 reduces the number of directed edges DE.
[0115] According to this second embodiment, directed edges DE generated between two nodes N that do not have a causal relationship can be reliably removed. This suppresses the erroneous generation of directed edges DE between two nodes N that do not have a causal relationship, thereby improving the accuracy of the causal relationship inference system.
[0116] The program 14 according to this second embodiment includes a second prior information acquisition procedure (S7), a topological order inference procedure (S4), a directed edge generation procedure (S43 in Figure 6), a directed edge removal procedure (S8), and a sparse processing procedure (S6), which are to be executed by the computer.
[0117] The second prior information acquisition procedure (S7) acquires second prior information 17, which is information that there are two nodes N in the data 11 containing multiple nodes N representing multiple variables that do not have a causal relationship. The topological order inference procedure infers the topological order 15 for each of the multiple nodes N by applying an algorithm that infers the topological order 15 sequentially from leaf nodes that have no child nodes down to root nodes that have no parent nodes, starting from the leaf nodes down to the root nodes that have no parent nodes.
[0118] The directed edge generation procedure (S43 in Figure 6) generates directed edges DE between multiple nodes N based on the topological order 15 inferred by the topological order inference procedure (S4). The directed edge removal procedure (S8), after the directed edge generation procedure (S43 in Figure 6), removes directed edges DE formed between two nodes N that do not have a causal relationship, based on second prior information 17. The sparse processing procedure (S6), after the directed edge removal procedure (S8), reduces the number of directed edges DE.
[0119] The present invention is not limited to the embodiments described above, and can be applied to various embodiments without departing from its spirit. [Explanation of Symbols]
[0120] 1: Production support system, 2: Causal relationship inference system, 3: Input unit, 4: Output unit, 5: Machine tool, 10: Memory device, 11: Data, 12: Algorithm, 13: Causal graph, 14: Program, 15: Topological order, 16: First prior information, 17: Second prior information, 20: Processor, 21: First prior information acquisition unit, 22: Topological order inference unit, 23: Cause node exclusion unit, 24: Directed edge generation unit, 25: Penalty processing unit, 26: Sparse Processing unit, 27: Second pre-information acquisition unit, 28: Directed edge removal unit, 30: Display unit, DE, DE1, DE2, DE3, DE4, DE5, DE6, DE7, DE8, DE9: Directed edges, N, N1, N2, N3, N4, N5: Nodes, S1: First pre-information acquisition procedure, S4: Topological order inference procedure, S41: Cause node exclusion procedure, S43: Directed edge generation procedure, S6: Sparse processing procedure, S7: Second pre-information acquisition procedure, S8: Directed edge removal procedure
Claims
1. Including a storage device and a processor, The aforementioned storage device is Data containing multiple nodes representing multiple variables, An algorithm for inferring the topological order for each of the aforementioned multiple nodes, from leaf nodes that have no child nodes to root nodes that have no parent nodes, Store it, The aforementioned processor, A first prior information acquisition unit acquires first prior information, which is information that a causal relationship exists between two nodes included in the plurality of nodes, and that one of the two nodes is the cause node and the other is the effect node. A topological order inference unit that infers the topological order for each of the plurality of nodes by applying the algorithm to the plurality of nodes, Based on the first prior information, a cause node exclusion unit excludes the cause node corresponding to the result node from the data until the result node is inferred by the topological order inference unit, A directed edge generation unit generates directed edges between the plurality of nodes based on the topological order inferred by the topological order inference unit, A causal relationship inference system comprising: a sparse processing unit that reduces the number of directed edges after processing by the directed edge generation unit.
2. The aforementioned processor further, The causal relationship inference system according to claim 1, further comprising a penalty processing unit that, after processing by the directed edge generation unit, sets the penalty in the processing by the sparse processing unit to 0 only for the cause node that has a causal relationship with the result node, based on the first prior information.
3. The aforementioned processor further, A second prior information acquisition unit acquires second prior information, which is information that there are two nodes in the aforementioned data that do not have a causal relationship. The system includes a directed edge removal unit that, after processing by the directed edge generation unit, removes the directed edge formed between two nodes where no causal relationship exists, based on the second prior information. The causal relationship inference system according to claim 1, wherein the sparse processing unit reduces the number of directed edges after processing by the directed edge removal unit.
4. Including a storage device and a processor, The aforementioned storage device is Data containing multiple nodes representing multiple variables, An algorithm for inferring the topological order for each of the aforementioned multiple nodes, from leaf nodes that have no child nodes to root nodes that have no parent nodes, Store it, The aforementioned processor, A second prior information acquisition unit acquires second prior information, which is information that there are two nodes in the aforementioned data that do not have a causal relationship. A topological order inference unit that infers the topological order for each of the plurality of nodes by applying the algorithm to the plurality of nodes, A directed edge generation unit generates directed edges between the plurality of nodes based on the topological order inferred by the topological order inference unit, After processing by the directed edge generation unit, a directed edge removal unit removes the directed edge formed between two nodes where no causal relationship exists, based on the second prior information. A causal relationship inference system comprising: a sparse processing unit that reduces the number of directed edges after processing by the directed edge removal unit.
5. The causal relationship inference system according to claim 1 or 4, wherein the algorithm is SCORE.
6. Furthermore, the causal relationship inference system according to claim 1 or 4 further comprises a display unit that displays a causal graph generated by the processing by the sparse processing unit.
7. The relational inference system according to claim 1 or 4, wherein the data includes machine setting conditions or machining conditions of a machine tool that performs a predetermined process on a workpiece, and data relating to the quality result of the workpiece after the predetermined process has been performed on the workpiece.
8. The relational inference system described in claim 7, An input unit for inputting the aforementioned data, An output unit outputs the quality result of the workpiece, the cause of any abnormality in the workpiece, or the machine setting conditions or machining conditions of the machine tool, based on the input data and the causal graph generated by the processing by the sparse processing unit. A production support system equipped with the following features.
9. A first prior information acquisition procedure for acquiring first prior information, which is information that a causal relationship exists between two nodes in data containing multiple nodes representing multiple variables, and that one of the two nodes is the cause node and the other is the effect node, A topological order inference procedure for each of the aforementioned plurality of nodes, by applying an algorithm for inferring the topological order to each of the plurality of nodes in order from leaf nodes that do not have child nodes to root nodes that do not have parent nodes, Based on the first prior information, a cause node exclusion procedure excludes the cause node corresponding to the result node from the data until the result node is inferred by the topological order inference procedure, A directed edge generation procedure that generates directed edges between the plurality of nodes based on the topological order inferred by the topological order inference procedure, After the directed edge generation procedure, a sparse processing procedure is performed to reduce the number of directed edges, A program that causes a computer to execute something.
10. A procedure for obtaining second prior information, which is the information that there are two nodes in a dataset containing multiple nodes representing multiple variables that do not have a causal relationship, A topological order inference procedure for each of the aforementioned plurality of nodes, in which an algorithm for inferring the topological order is applied sequentially from the leaf node which has no child nodes to the root node which has no parent nodes, thereby inferring the topological order for each of the plurality of nodes sequentially from the leaf node to the root node which has no parent nodes, A directed edge generation procedure that generates directed edges between the plurality of nodes based on the topological order inferred by the topological order inference procedure, Following the directed edge generation procedure, a directed edge removal procedure is performed to remove the directed edge formed between two nodes where no causal relationship exists, based on the second prior information. After the directed edge removal procedure, a sparse processing procedure is performed to reduce the number of directed edges, A program that is executed by a computer.