Defect analysis device, processing system, defect analysis method, and processing method

Abstract

The processing defect analysis device (1) is characterized by comprising: a cutting point information acquisition unit (10) that acquires cutting point information indicating positions of cutting points at which a tool installed in a machine tool cuts a workpiece, the cutting points corresponding to respective ones of a plurality of front end points included in information indicating a movement path of a front end point of the tool, i.e., a tool path; a characteristic quantity acquisition unit (20) that acquires a characteristic quantity indicating a characteristic of processing, the characteristic quantity being calculated in correspondence with the respective ones of the plurality of front end points; and a processing defect analysis unit (110) that determines, based on the cutting point information and the characteristic quantity, a region of the workpiece in which a processing defect occurs, i.e., a processing defect region, and a factor of the processing defect in the processing defect region.

Description

Technical Field

[0001] This disclosure relates to a defect analysis apparatus, a machining system, a defect analysis method, and a machining method for analyzing defects that occur during the machining of an object using a machine tool. Background Technology

[0002] Machine tool drive motors and other actuators move the workpiece or tool, thereby machining the workpiece into the desired shape. At this time, a machining program is used, which describes the movement instructions indicating the tool path. The machining program is input to the CNC device that controls the machine tool. The CNC device analyzes the machining program and generates instruction positions for the drive axes that interpolate the tool path calculated based on the movement instructions for each predetermined control cycle. The CNC device meticulously manages the position of the drive axes according to the instruction positions while simultaneously controlling the actuators.

[0003] However, in the processed workpiece, due to various factors, machining errors occur relative to the target shape. These machining errors lead to defects such as failure to meet dimensional accuracy requirements and damage to the machined surfaces. In such cases, rework, including redesign and reprocessing, occurs, reducing production efficiency.

[0004] Patent Document 1 discloses a technique that converts the actual feed speed of the drive axis of a computer tool into an image, and uses the image data to determine whether the machining is good or bad through machine learning. By using the technique disclosed in Patent Document 1, it is possible to automatically determine whether machining defects have occurred.

[0005] Existing technical documents

[0006] Patent documents

[0007] Patent Document 1: Japanese Patent Application Publication No. 2021-068225 Summary of the Invention

[0008] However, according to the technology disclosed in Patent Document 1, although it is possible to determine whether the machining defect is caused by the actual feed rate, there is a problem that the analysis work of determining which part of the workpiece is affected by the machining defect and determining the factors that cause the machining defect is labor-intensive.

[0009] This disclosure was made in view of the above circumstances, and its purpose is to provide a processing defect analysis device that can reduce the labor of operators in the analysis operation of determining the factors that cause processing defects.

[0010] To solve the aforementioned problems and achieve the objectives, the machining defect analysis apparatus disclosed herein is characterized by comprising: a cutting point information acquisition unit that acquires cutting point information indicating the position of a cutting point, wherein the cutting point is a point on a machine tool mounted on a machine tool that cuts into a workpiece, and the cutting point corresponds to each of a plurality of front-end points containing information representing the movement path of the tool's front end point, i.e., the tool trajectory; a feature quantity acquisition unit that acquires feature quantities representing machining characteristics, wherein the feature quantities are calculated corresponding to each of the plurality of front-end points; and a machining defect analysis unit that, based on the cutting point information and the feature quantities, determines the area on the workpiece where machining defects occur, i.e., the machining defect area, and the factors contributing to machining defects within the machining defect area.

[0011] According to this disclosure, it has the effect of reducing the labor of workers performing analytical tasks to identify factors that cause processing defects. Attached Figure Description

[0012] Figure 1 This is a diagram showing the functional structure of the processing defect analysis device according to Embodiment 1.

[0013] Figure 2 It shows that based on storage in Figure 1 The diagram shows an example of a tool whose tool information is stored in the tool information storage section.

[0014] Figure 3 It shows the storage in Figure 1 A perspective view of an example of the processing target shape in the processing target shape storage section.

[0015] Figure 4 yes Figure 3 The side view of the shape of the processing target is shown.

[0016] Figure 5 yes Figure 3 A top view of the shape of the machining target shown.

[0017] Figure 6 It is shown Figure 1 The image shows a three-dimensional view of an example of a tool trajectory generated by the tool trajectory calculation unit.

[0018] Figure 7 It is shown Figure 1 The image shows a top view of an example of a tool trajectory generated by the tool trajectory calculation unit.

[0019] Figure 8 It is a diagram used to illustrate the configuration of tools and machining surfaces.

[0020] Figure 9 It is a diagram used to illustrate the relationship between the tool's tip and the cutting point.

[0021] Figure 10 It is a diagram used to illustrate the cutting point when the tool is separated from the machined surface.

[0022] Figure 11 It is a diagram of the cutting point used to illustrate the interference between the tool and the machined surface.

[0023] Figure 12 It is shown that... Figure 8 The diagram shows the calculated cutting points corresponding to each front-end point.

[0024] Figure 13 Figure 1 is used to illustrate the method for finding the endpoint adjacent to any endpoint contained in the tool trajectory.

[0025] Figure 14 This is Figure 2, which illustrates the method for finding the endpoint adjacent to any endpoint contained in the tool trajectory.

[0026] Figure 15 Figure 3 illustrates a method for finding the endpoint adjacent to any endpoint contained in the tool trajectory.

[0027] Figure 16 It is shown Figure 1 A perspective view of an example of the display screen shown by the defect analysis device.

[0028] Figure 17 It is shown Figure 1 The top view of an example of the display screen shown by the defect analysis device.

[0029] Figure 18 This is an example of a screen that shows the display results of the machined surface of the target shape, which is displayed using display colors representing characteristic quantities, and the display results of each defective area, which are displayed in different ways for each factor of the defect, are overlaid.

[0030] Figure 19 It is used for explanation Figure 1 The flowchart shows the operation of the defect analysis device.

[0031] Figure 20 This is a diagram illustrating an example of the structure of a computer system implementing the defect analysis apparatus of Embodiment 1.

[0032] Figure 21 This is a diagram showing the functional structure of a variation of Embodiment 1.

[0033] Figure 22 This is a diagram showing the functional structure of the processing defect analysis device according to Embodiment 2.

[0034] Figure 23This is a diagram illustrating an example of the structure of a model building section using machine learning.

[0035] Figure 24 Is with Figure 22 The flowchart shown relates to the learning process performed by the defect analysis device.

[0036] Figure 25 It is shown Figure 22 The diagram shows the functional structure of the inference device in the defect analysis unit.

[0037] Figure 26 It is used for explanation Figure 25 The flowchart shows the operation of the inference device.

[0038] Figure 27 This is a diagram showing the functional structure of the processing defect analysis device according to Embodiment 3.

[0039] Figure 28 It is shown Figure 27 The diagram shows an example of a neural network used in the model construction.

[0040] (Explanation of reference numerals in the attached diagram)

[0041] 1, 1A, 1B, 1C: Machining defect analysis device; 2: Information processing device; 10, 10A: Cutting point information acquisition unit; 20, 20A: Feature quantity acquisition unit; 81: Control unit; 82: Input unit; 83: Storage unit; 84: Display unit; 85: Communication unit; 86: Output unit; 87: System bus; 101: Operation information storage unit; 102: Tool trajectory calculation unit; 103: Tool trajectory storage unit; 104: Tool information storage unit; 105: Machining target shape storage unit; 106: Cutting point calculation unit; 107: Cutting point storage unit; 108: Feature quantity calculation unit; 109: Feature quantity storage unit; 110, 110B: Machining defect analysis unit; 111: Display unit; 112: Position designation unit; 113: Factor pattern storage unit; 114, 114B: Machining defect area extraction unit. ; 115, 115B: Defective Factor Judgment Unit; 116, 116C: Model Construction Unit; 117: Data Accumulation Unit; 118: Tag Generation Unit; 201: Learning Data Acquisition Unit; 202: Model Generation Unit; 203: Learned Model Storage Unit; 204: Inference Device; 205: Data Acquisition Unit; 206: Inference Unit; CP, CP1~CP7: Cutting Point; CPS1: Cutting Point Group; D8: Travel Direction; L1a, L1b, L2a, L2b: Distance; M1: Machining Target Shape; P, P1~P12: Lead-edge Point; PL8: Plane; Q1, Q2: Defective Area; R1, R2: Intersection Point; S, S1~S3: Machining Surface; T1: Tool; T1α: Offset Tool; TP1: Tool Trajectory; V, V1~V7: Tool Direction; W: Machining Object. Detailed Implementation

[0042] Hereinafter, with reference to the accompanying drawings, the processing defect analysis apparatus, processing system, processing defect analysis method, and processing method of the present disclosure will be described in detail.

[0043] Implementation method 1.

[0044] Figure 1 This diagram illustrates the functional structure of the machining defect analysis device 1 according to Embodiment 1. The machining defect analysis device 1 includes an operation information storage unit 101, a tool trajectory calculation unit 102, a tool trajectory storage unit 103, a tool information storage unit 104, a machining target shape storage unit 105, a cutting point information acquisition unit 10 with a cutting point calculation unit 106 and a cutting point storage unit 107, a feature quantity acquisition unit 20 with a feature quantity calculation unit 108 and a feature quantity storage unit 109, a machining defect analysis unit 110, a display unit 111, a position designation unit 112, and a factor pattern storage unit 113. The machining defect analysis unit 110 includes a machining defect area extraction unit 114 and a machining defect factor determination unit 115.

[0045] The machining defect analysis device 1 has the function of supporting analysis operations, in which the areas where machining defects occur and the factors causing the defects are determined during machining performed using a CNC machine tool.

[0046] First, an example of the data used by the processing defect analysis device 1 will be explained using the accompanying drawings.

[0047] The operation information storage unit 101 stores operation information, which represents the operating status of the machine tool when it is in operation. The operation information includes information obtained from the machine tool, the CNC device controlling the machine tool, or sensors installed on the machine tool. More specifically, the operation information may include, for example, position data of each of the multiple drive axes of the machine tool, the load and current value of the machine tool's spindle, the internal temperature of the machine tool, the machining program and machining conditions used when the machine tool is in operation, and status data such as the parameters of the CNC device. The aforementioned position data, spindle load, current value, and internal temperature of the machine tool are time-series data synchronized according to time intervals. Furthermore, the drive axis position data is information generated by the CNC device, including at least one of the command position for each of the multiple drive axes of the machine tool in each predetermined control cycle, and the detection position detected by the position detectors of each of the multiple drive axes for each control cycle. The operation information can be information obtained by actually operating the machine tool, or information generated by simulating the operation of the CNC device and the machine tool using simulation or the like.

[0048] The tool information storage unit 104 stores tool information, which defines the shape of the tool used to machine the workpiece. Tool information is information that generates the tool's shape, including, for example, the tool type, tool diameter, and tool length. In the case of rotary tools such as end mills, the tool information may also include the tool's central axis and external outline. In the case of asymmetrical shapes such as turning tools, the tool information may also include parameter information.

[0049] Figure 2 It shows that based on storage in Figure 1 The diagram shows an example of tool T1, which contains tool information in the tool information storage unit 104. Figure 2 The tool T1 shown is a ball end mill with a spherical front end.

[0050] The machining target shape storage unit 105 stores shape information representing the target shape of the object to be machined. The machining target shape includes the machining surface, which is the surface to be machined using tool T1.

[0051] Figure 3 It shows the storage in Figure 1A perspective view of an example of the processing target shape M1 of the processing target shape storage unit 105 shown. Figure 4 yes Figure 3 The side view of the machining target shape M1 is shown. Figure 5 yes Figure 3 The diagram shows a top view of the machining target shape M1. The machining target shape M1 includes machining surfaces S1 to S3. Furthermore, unless otherwise specified, machining surfaces S1 to S3 are simply referred to as machining surface S.

[0052] The operation information, tool information, and shape information described above are obtained from outside the defect analysis device 1. The operation information, tool information, and the target shape M1 can be read from a storage medium stored external to the defect analysis device 1, obtained via a communication path, or input by the operator using an input unit such as a keyboard. The shape information can be generated by data transformation from CAD data or by graphical input by the operator using a keyboard.

[0053] The tool trajectory calculation unit 102 generates a tool trajectory, which serves as the trajectory of the tool's leading edge, based on the operation information stored in the operation information storage unit 101, and stores the generated tool trajectory in the tool trajectory storage unit 103. Here, the tool trajectory calculation unit 102 performs coordinate transformation on the position data of the time-series data contained in the operation information to determine the position of the tool's leading edge and the tool orientation at that position. Furthermore, the position data can be either a commanded position or a detected position. The choice of which position data to use can be predetermined or selected by the operator. At this time, the tool trajectory calculation unit 102 performs coordinate transformation on the position data using information such as the structure of the machine tool's drive axis, tool length, coordinate system, etc., contained in the parameters of the CNC device, and their relative relationship with the coordinate system of the machining target shape M1. Furthermore, regarding the relative relationship between the coordinate system contained in the CNC device parameters and the coordinate system of the machining target shape M1, for example, an offset can be pre-stored or specified by the operator. The position of the leading edge point thus determined is mapped to each time-series data and stored as the tool trajectory.

[0054] Figure 6 It is shown Figure 1 A perspective view of an example of a tool trajectory TP1 generated by the tool trajectory calculation unit 102 shown. Figure 7 It is shown Figure 1 A top view of an example of a tool trajectory TP1 generated by the tool trajectory calculation unit 102 shown. Figure 6 as well as Figure 7 The image shows the tool path TP1 and the machining target shape M1. Additionally, Figure 8This is a diagram illustrating the configuration of tool T1 and the machining surface S. Figure 8 The diagram shows the positions of the leading endpoints P1 to P7 within the tool trajectory TP1, and the tool directions V1 to V7 when tool T1 is located at each leading endpoint P1 to P7. Figure 8 The diagram shows a cross-sectional view of tool T1 passing over the machining surface S3 of the target shape M1. At this time, the configuration relationship between tool T1 and machining surface S3 is determined according to the positions of the leading edge points P1 to P7 and the tool directions V1 to V7 at those positions. Furthermore, hereinafter, without specifically distinguishing between the leading edge points P1 to P7, each leading edge point will be simply referred to as leading edge point P, and without specifically distinguishing between the tool directions V1 to V7, each tool direction will be simply referred to as tool direction V.

[0055] The cutting point calculation unit 106 of the cutting point information acquisition unit 10 calculates the position of the point where the tool T1 installed on the machine tool cuts the workpiece W, i.e., the position of the cutting point, based on the tool trajectory TP1 stored in the tool trajectory storage unit 103, the tool information stored in the tool information storage unit 104, and the workpiece target shape M1 stored in the workpiece target shape storage unit 105. The cutting point will be explained here. Figure 9 This diagram illustrates the relationship between the tip point P of tool T1 and the cutting point CP. The tip point P is the point located at the front end of tool T1, and its position on tool T1 remains constant. The cutting point CP is the point where tool T1 cuts the workpiece W; ideally, it is the point where tool T1 meets the machined surface S of workpiece W. Figure 9 The diagram shows the state where the cutting point CP is connected to the machined surface S.

[0056] The cutting point calculation unit 106 calculates the position of the cutting point CP corresponding to each of the plurality of front-end points P contained in the tool path TP1. Specifically, the cutting point calculation unit 106 calculates the cutting point CP of tool T1 on the machining surface S of the machining target shape M1 based on the position of the front-end point P and the tool direction V for each of the plurality of front-end points P contained in the tool path TP1. The position of the front-end point P and the tool direction V are relative positions and directions with respect to the machining target shape M1. Therefore, the relative positional relationship between tool T1 and the machining surface S of the machining target shape M1 is determined according to the position of the front-end point P and the tool direction V. Ideally, this positional relationship is such that tool T1 and machining surface S are in contact. Therefore, in such a case... Figure 9 When the tool T1 and the machining surface S are in contact, the cutting point calculation unit 106 calculates the point on the machining surface S where the tool T1 and the machining surface S are in contact as the cutting point CP.

[0057] However, in reality, errors exist in the position of the tool's tip P and the tool direction V, so the tool T1 and the machining surface S sometimes do not connect. For example, consider the case where the tool T1, positioned at the tip P calculated based on the operating information, is separated from the machining surface S. Figure 10 This is a diagram illustrating the cutting point CP when tool T1 is separated from the machined surface S. (In...) Figure 10 When the tool T1 is separated from the machining surface S, the cutting point calculation unit 106 can determine a point on the machining surface S where the distance between the tool T1 and the machining surface S is the shortest, and set that point as the cutting point CP.

[0058] In addition, considering the interference between the tool T1, which is positioned at the front end point P calculated based on the operation information, and the machining surface S. Figure 11 This diagram illustrates the cutting point CP when tool T1 interferes with the machined surface S. (In...) Figure 11 When the tool T1 interferes with the machined surface S, the cutting point calculation unit 106 can determine the point where the biased tool T1α contacts the machined surface S, thus achieving a contact state, and set this point as the cutting point CP. Furthermore, while an example of a ball end mill is shown here, the cutting point CP can be calculated using the same method even if the tool T1 has a special shape such as a radius end mill, a flat end mill, a barrel tool, or a turning tool.

[0059] The cutting point calculation unit 106 stores the cutting point information, which corresponds to the positions of multiple cutting points CP obtained by the method described above and each front end point P, in the cutting point storage unit 107. Furthermore, depending on the relationship between the tool T1 and the machining surface S, sometimes multiple cutting points CP are calculated for one front end point P. In this case, it is sufficient to store the multiple cutting points CP corresponding to one front end point P.

[0060] Figure 12 It is shown that... Figure 8 The diagram shows the cutting points CP1 to CP7 calculated corresponding to each of the front-end points P1 to P7. For simplicity, in... Figure 12 The symbols are omitted. Figure 12 The black dots shown are Figure 8 The front points P1 to P7 are shown. Figure 12 The arrow shown is Figure 8 The tool directions shown are V1 to V7. Figure 12The diagram shows the cutting points CP1 to CP7 of tool T1 on the machining surface S3 of the target shape M1 as tool T1 moves along multiple front point P1 to P7 contained in the tool trajectory TP1 and tool directions V1 to V7. The cutting points CP1 to CP7 corresponding to each front point P1 to P7 are calculated.

[0061] The feature quantity calculation unit 108 of the feature quantity acquisition unit 20 calculates the machining feature quantities corresponding to each of the multiple front end points P included in the tool path TP1, based on the operation information stored in the operation information storage unit 101, the tool path TP1 stored in the tool path storage unit 103, and the cutting point information stored in the cutting point storage unit 107. The feature quantity calculation unit 108 stores the calculated feature quantities in the feature quantity storage unit 109, associating them with the cutting point CP. Alternatively, the feature quantity calculation unit 108 can directly associate the feature quantities with the cutting point CP, and by associating them with the front end points P, the feature quantities corresponding to the common front end point P can be used as the feature quantities corresponding to the cutting point CP.

[0062] A feature quantity is a quantity that represents a feature of the machining process. Feature quantities include, for example, at least one of the following: machining error quantity (the distance between the machining target shape M1 and the tool T1 positioned at the front end point P); velocity of the front end point P; acceleration of the front end point P; jerk of the front end point P; velocity of the cutting point CP; acceleration of the cutting point CP; jerk of the cutting point CP; position of each of the multiple drive axes of the machine tool; velocity of each of the multiple drive axes of the machine tool; acceleration of each of the multiple drive axes of the machine tool; jerk of each of the multiple drive axes of the machine tool; and reverse position of each of the multiple drive axes of the machine tool.

[0063] Here, the machining error can be calculated as the shortest distance between the position of the cutting point CP corresponding to the tip point P of the tool T1 and the shape surface of the tool T1 configured according to the position of the tip point P and the tool direction V.

[0064] In addition, the velocity, acceleration, and jerk of the leading point P of tool T1 can be calculated as described below. When the position of the leading point P at a certain time t is set as PT(t), and the position of the leading point P at time t+Δt, which is a predetermined control period time from time t, is set as PT(t+Δt), the velocity VT(t) of the leading point P at time t is obtained by dividing the distance between the positions of the two leading points P by the predetermined control period time, and is expressed by the following mathematical formula (1).

[0065]

Mathematical Formula 1

[0066]

[0067] Let mathematical formula (2) represent the acceleration AT(t) of the front point P at time t.

[0068]

Mathematical Formula 2

[0069]

[0070] Let mathematical formula (3) represent the jerk JT(t) of the front end point P at time t.

[0071]

Mathematical Expression 3

[0072]

[0073] In addition, the velocity, acceleration, and jerk of the cutting point CP can be calculated as described below. When the position of the cutting point CP corresponding to the front end point P at a certain time t is set as PC(t), and the position of the cutting point CP corresponding to the front end point P at time t+Δt, which is the time to advance a predetermined control cycle from time t, is set as PC(t+Δt), the velocity VC(t) of the cutting point CP corresponding to the front end point P at time t is expressed by the following mathematical formula (4).

[0074]

Mathematical Expression 4

[0075]

[0076] In addition, the acceleration AC(t) of the cutting point CP corresponding to the front end point P at time t is expressed by mathematical formula (5).

[0077]

Mathematical Expression 5

[0078]

[0079] In addition, the jerk JC(t) of the cutting point CP corresponding to the front end point P at time t is expressed by mathematical formula (6).

[0080]

Mathematical Expression 6

[0081]

[0082] Furthermore, it is possible to separately calculate the position, velocity, acceleration, and jerk of each drive axis of the computer tool, as described below. It is also possible to obtain the position PM1(t) of the first drive axis corresponding to the leading edge point P at a certain time t from the time-series data of the operation information.

[0083] Furthermore, when the position of the first drive shaft corresponding to the front point P at time t+Δt, which is the time when the control cycle is advanced from time t, is set to PM1(t+Δt), the following mathematical formula (7) is used to represent the speed VM1(t) of the first drive shaft corresponding to the front point P at time t.

[0084]

Mathematical Expression 7

[0085]

[0086] In addition, the acceleration AM1(t) of the first drive shaft corresponding to the front end point P at time t is expressed by mathematical formula (8).

[0087]

Mathematical Expression 8

[0088]

[0089] In addition, the jerk JM1(t) of the first drive shaft corresponding to the front end point P at time t is expressed by mathematical formula (9).

[0090]

Mathematical Expression 9

[0091]

[0092] Furthermore, the position, velocity, acceleration, and jerk can also be calculated using the same method for drive shafts other than the first drive shaft.

[0093] Furthermore, the reverse position of each of the multiple drive axes of the computer tool can be determined as follows. Using the method described above, the velocity VM1(t) of the first drive axis corresponding to the leading edge point P at a certain time t and the velocity VM1(t+Δt) of the first drive axis corresponding to the leading edge point P at a time t+Δt after advancing a predetermined control cycle from time t are calculated. Then, the signs of the velocities VM1(t) and VM1(t+Δt) are compared, and the position corresponding to the time when the signs reverse is taken as the reverse position of the first drive axis. Furthermore, the reverse positions of drive axes other than the first drive axis can also be determined using the same method.

[0094] Furthermore, the feature quantity calculation unit 108 can also use the difference between the feature quantities of two adjacent front-end points P as a feature quantity. In this case, two adjacent front-end points P refer to the group of two front-end points P that are closest to each other on two adjacent tool paths. For example, in a CAD / CAM system, among tool paths such as so-called scanline machining or contour machining that are parallel to a plane and generated at a certain interval, and so-called surface machining that is generated at a certain interval according to the contour of the machining surface S of the machining target shape M1, the nearest front-end point P on the adjacent tool path that passes through the position of a certain front-end point P away from the distance of that interval is selected, thereby finding two adjacent front-end points P.

[0095] Here, use Figures 13 to 15 This describes a method for finding the endpoint P adjacent to any endpoint P contained in the tool trajectory TP1. Figure 13 Figure 1 is used to illustrate the method for finding the leading endpoint P that is adjacent to any leading endpoint P contained in the tool trajectory TP1. Figure 14 This is Figure 2, which illustrates the method for finding the leading endpoint P that is adjacent to any leading endpoint P contained in the tool trajectory TP1. Figure 15 Figure 3 illustrates the method for finding the endpoint P adjacent to any endpoint P contained in the tool trajectory TP1. Figure 13 This illustrates the case where the leading endpoint P8 of the tool trajectory TP1 and a portion of its surrounding tool trajectory are included. Consider the case where the leading endpoint P adjacent to the leading endpoint P8 is determined. For example... Figure 14 As shown, firstly, assume that the travel direction D8 at the front point P8 is taken as the plane PL8 with normal vector. Here, the travel direction D8 can be determined based on the relationship between the front point P8 and the preceding and following front endpoints P, or it can be another predetermined direction. Next, the feature calculation unit 108 calculates the intersection point of the assumed plane PL8 and the tool trajectory. At this time, the intersection point R1 between the front points P9 and P10 and the intersection point R2 between the front points P11 and P12 are calculated.

[0096] Next, as Figure 15As shown, the feature quantity calculation unit 108 calculates the distance L1a between intersection point R1 and front point P9, and the distance L1b between intersection point R1 and front point P10. By comparing distances L1a and L1b, the feature quantity calculation unit 108 designates the front endpoint P10, which is closer to intersection point R1 than the front point P9, as a candidate for the adjacent front endpoint P. Similarly, the feature quantity calculation unit 108 calculates the distance L2a between intersection point R2 and front point P11, and the distance L2b between intersection point R2 and front point P12. By comparing distances L2a and L2b, the feature quantity calculation unit 108 designates the front endpoint P11, which is closer to intersection point R2 than the front point P11, as a candidate for the adjacent front endpoint P. The feature quantity calculation unit 108 can choose either the front endpoint P10 or P11 that is closer to the front point P8 among the multiple candidate front endpoints calculated here, or it can choose a front point that exists in a specific direction relative to the travel direction D8. Here, it is assumed that the front endpoint P11, which is adjacent to the front endpoint P8, is selected. In this case, the feature quantity calculation unit 108 can use the difference between the feature quantity obtained by subtracting the feature quantity of the front endpoint P11 from the feature quantity of the front endpoint P8 as the new feature quantity of the front endpoint P8.

[0097] Furthermore, the feature quantity detailed above is just one example. The feature quantity calculation unit 108 can calculate and store physical information such as the load, current value, and internal temperature of the machine tool's spindle at each front-end point P in the feature quantity storage unit 109. Additionally, the feature quantity calculation unit 108 can calculate one type of feature quantity or simultaneously calculate and store two or more types of feature quantities. Moreover, the feature quantity calculation unit 108 can also output the feature quantities as image data or time-series data.

[0098] Return to Figure 1 The defective region extraction unit 114 of the defective analysis unit 110 extracts the defective region, i.e., the area that causes the defect in the workpiece W, based on the cutting point information stored in the cutting point storage unit 107 and the feature quantity stored in the feature quantity storage unit 109. The defective region extraction unit 114 can extract one or more defective regions.

[0099] The defect analysis unit 110 extracts defective regions based on the position of the cutting point CP represented by the cutting point information and the feature quantity of the leading edge point P corresponding to the cutting point CP. Regarding the method for extracting defective regions, one possible approach is to extract the cutting point CP corresponding to a feature quantity exceeding a predetermined threshold, and then extract a single defective region as the area on the machining surface S where multiple extracted cutting points CP are approximately located. Alternatively, multiple defective regions can be extracted at this time.

[0100] In addition, as another method for extracting defective areas, there is the following method: calculate the change in characteristic quantity between a certain cutting point CP and its surrounding cutting points CP, successively find the boundary between cutting points CP where the characteristic quantity changes drastically, and extract the area of ​​cutting points CP that exist inside the area surrounded by the found boundary as a defective area.

[0101] In this case, when using machining error as a feature, regions where the machining error exceeds a predetermined threshold can be extracted as defective regions, or regions surrounded by boundaries where the machining error changes drastically can be extracted as defective regions. Similarly, when using the acceleration of the front point P as a feature, regions where the acceleration of the front point P exceeds a predetermined threshold can be extracted as defective regions, or regions surrounded by boundaries where the acceleration of the front point P changes drastically can be extracted as defective regions. Furthermore, defective regions can be extracted based on one feature or multiple feature values.

[0102] Another method for extracting defective areas is as follows: An image is generated that "displays each cutting point of multiple cutting points CP using a representation of feature quantities corresponding to the cutting point CP," i.e., an image that distinguishes and displays cutting points CP based on feature quantities. This image is then processed through image processing such as correction, transformation, filtering, and feature extraction to determine the range within the image. Regions containing cutting points CP within this determined range are extracted as one defective area. For example, when using the method where "the larger the feature quantity corresponding to the cutting point CP, the more concentrated the color is used to display each cutting point of multiple cutting points CP," by determining the range within the image where the color concentration exceeds a threshold, regions where the feature quantity exceeds the threshold can be extracted.

[0103] Furthermore, the defective area extraction unit 114 can also extract defective areas according to the designated location specified by the location designation unit 112. The defective area extraction unit 114 can extract an area including at least a portion of the designated location as a defective area. The location designation unit 112 accepts location designations from operators using an input unit such as a positioning device, and outputs the accepted designated location to the defective area extraction unit 114. The designated location accepted by the location designation unit 112 can be a single point, multiple points, or a continuous area.

[0104] Next, the defect analysis unit 110's defect factor determination unit 115 determines the factors causing the defects in the defective areas based on the cutting point information stored in the cutting point storage unit 107 and the defective areas extracted by the defective area extraction unit 114. When multiple defective areas are extracted by the defective area extraction unit 114, the defect factor determination unit 115 can determine the factors causing the defects for each defective area.

[0105] The machining defect determination unit 115 can determine the factors of machining defects based on the extracted machining defect area and the feature quantity corresponding to the cutting point CP contained in the machining defect area. For example, the machining defect determination unit 115 can easily determine the factors of machining defects by taking the type of feature quantity used to extract the machining defect area as the factors of machining defects.

[0106] In addition, as another method, the machining defect factor determination unit 115 calculates the values ​​inside and outside the machining defect area, the difference value of the characteristic quantity near the boundary of the machining defect area, etc., regarding the characteristic quantity corresponding to the cutting point CP existing inside the machining defect area. For each characteristic quantity, it investigates the correlation with the machining defect area, thereby identifying the type of characteristic quantity with high correlation as a factor of machining defect. For example, when determining the factor of machining defect for a certain machining defect area, if the characteristic quantity of the cutting point CP is the reverse position of the first drive axis of the machine tool inside the machining defect area, but not the reverse position of the first drive axis of the machine tool outside the machining defect area, then the machining defect factor of the machining defect area can be determined as the reverse position of the first drive axis of the machine tool.

[0107] As another method, the defect determination unit 115 can also generate an image showing the defective area and an image showing the cutting point CP existing inside the defective area through a representation method of characteristic quantities. After performing image processing such as correction, transformation, filtering, and feature extraction on these two generated images, the two images are compared to determine the defective factor. In this case, the defect determination unit 115 can determine the feature quantity with the highest correlation between the two images as the defective factor.

[0108] Furthermore, factor patterns can be pre-stored in the factor pattern storage unit 113. These factor patterns are information that maps defective factors to patterns of characteristic quantities exhibited when those factors are present. The defective factor determination unit 115 determines the defective factors in the defective region based on the extracted defective region, characteristic quantities, and factor patterns. Factor patterns can include, for example, patterns exhibited by cutting points CP existing inside, at the boundary, or outside the defective region; correlations between patterns exhibited by multiple characteristic quantities; and the relationship between regions where characteristic quantities exceed predetermined values ​​and the defective region. Furthermore, the shape and size of the defective region can also be associated.

[0109] More specifically, for example, regarding factor pattern #A, it is assumed that the machining error amount inside the defective area and the acceleration of the front point P are above a predetermined value and are positively correlated. Whether factor pattern #A is met can be determined based on the actual defective area and the characteristic quantity. Furthermore, when multiple factor patterns can be determined, the defective factor determination unit 115 can also calculate the determination probability of each factor pattern.

[0110] The display unit 111 displays the defective area extracted by the defective analysis unit 110 in different ways for each factor of the defective area.

[0111] Here, an example is given of a method by which the display unit 111 displays the defective area using different representations for each factor of the defect. For example, the display unit 111 may predetermine a color for each factor of the defect and use the predetermined color for the defective area to display the cutting point CP contained in the defective area. Alternatively, the display unit 111 may predetermine a display mark for each factor of the defect and use the predetermined display mark for the defective area to display the cutting point CP contained in the defective area. Furthermore, the display unit 111 may predetermine the display density and size for each factor of the defect and use a predetermined size of dots, etc., to display the cutting point CP contained in the defective area with the predetermined display density for the defective area.

[0112] Furthermore, the display unit 111 can pre-generate line segments representing the boundaries of the defective area and display these line segments according to a predetermined display color, solid line, or dashed line type. Additionally, the display unit 111 can also completely cover the interior of the defective area's boundary. By using these display methods, the display unit 111 can display the defective area using different display methods for each factor of the defect.

[0113] Figure 16 It is shown Figure 1 A perspective view of an example of the display screen shown by the defect analysis device 1. Figure 17 It is shown Figure 1 The top view shows an example of the display screen of the defect analysis device 1. Figure 16 as well as Figure 17 In this process, the set of cutting points CP calculated for each of the multiple front-end points P contained in the tool path TP1, namely the cutting point group CPS1, is used to display the cutting points CP in each defective region overlaid on the machining target shape M1, using different representation methods for each factor of the defect. Here, the display unit 111 sets the display color of the cutting points CP existing inside the defective region Q1 to a first display color different from the display color of the cutting points CP existing outside the defective region Q1, and displays the line segment representing the boundary of the defective region Q1 with a dashed line of the first display color. In addition, the display unit 111 sets the display color of the cutting points CP existing inside the defective region Q2, which is different from the display color of the CP existing outside the defective region Q2, to a second display color different from the display color of the CP existing outside the defective region Q2, and displays the line segment representing the boundary of the defective region Q2 with a dashed line of the second display color. For example, the first display color can be red, and the second display color can be green. Here, the display unit 111 can also be used by the operator to specify the display method for each factor of processing defects, thereby using the specified display method.

[0114] Furthermore, when there are multiple candidates for factors indicating defects in a single defective area, the display unit 111 can also display the determination probability for each factor of defect in each defective area. In this case, the determination probability for each factor of defect, calculated in the defect analysis unit 110, can be used. More specifically, for example, if the probability that a defective factor in a certain defective area is factor pattern #A is 70%, and the probability that it is factor pattern #B is 30%, the display unit 111 can display these probabilities. Furthermore, the display unit 111 can also display the characteristic quantities that form the basis for determining a defective area as a defective factor.

[0115] The display unit 111 can also overlay the display results of the machining surface S of the machining target shape M1, which is a method of representing the characteristic quantity of the front end point P corresponding to each of the multiple cutting points CP, and the display results of the machining defect area, which is displayed in different ways for each factor of machining defect.

[0116] Figure 18This is an example of a screen that shows the display results of the machining surface S of the machining target shape M1 using display colors representing characteristic quantities, and the display results of the machining defect areas Q1 and Q2 displayed in different ways for each factor of machining defect, which are superimposed on the screen.

[0117] Figure 19 It is used for explanation Figure 1 The flowchart shows the operation of the machining defect analysis device 1. The tool trajectory calculation unit 102 of the machining defect analysis device 1 generates a tool trajectory TP1 based on the operation information stored in the operation information storage unit 101, and stores the generated tool trajectory TP1 in the tool trajectory storage unit 103 (step S101). The cutting point calculation unit 106 calculates the position of the cutting point CP corresponding to each of the multiple front-end points P of the tool trajectory TP1 stored in the tool trajectory storage unit 103, and stores the cutting point information representing the calculated position of the cutting point CP in the cutting point storage unit 107 (step S102). The feature quantity calculation unit 108 calculates the feature quantity corresponding to each of the multiple front-end points P of the tool trajectory TP1 based on the operation information stored in the operation information storage unit 101, the tool trajectory TP1 stored in the tool trajectory storage unit 103, and the cutting point information stored in the cutting point storage unit 107, and stores the calculated feature quantity in the feature quantity storage unit 109 (step S103).

[0118] The defective area extraction unit 114 extracts the defective area based on the cutting point information stored in the cutting point storage unit 107 and the feature quantity stored in the feature quantity storage unit 109 (step S104).

[0119] The defect determination unit 115 determines the defective factors in the defective areas extracted by the defective area extraction unit 114 (step S105). The display unit 111 displays the extracted defective areas using different display methods for each determined defective factor (step S106).

[0120] Figure 20 This is a diagram illustrating an example of the structure of a computer system implementing the defect analysis apparatus 1 of embodiment 1. (See diagram below.) Figure 20 As shown, the computer system includes a control unit 81, an input unit 82, a storage unit 83, a display unit 84, a communication unit 85, and an output unit 86, which are connected via a system bus 87.

[0121] exist Figure 20In this system, the control unit 81 is, for example, a CPU (Central Processing Unit). The control unit 81 executes analysis programs that describe the various processes implemented by the defect analysis device 1. The input unit 82 is, for example, composed of a touch sensor, keyboard, mouse, etc., and is used by the computer system user to input various information. In the above embodiment, when receiving input from the operator, the operator can use the input unit 82 to input. The storage unit 83 includes various memory devices such as RAM (Random Access Memory), ROM (Read Only Memory), and hard disks, and stores the programs to be executed by the control unit 81, necessary data obtained during processing, etc. In addition, the storage unit 83 is also used as a temporary storage area for programs. The display unit 84 is composed of an LCD (Liquid Crystal Display) or the like, and displays various screens to the computer system user. The communication unit 85 is a communication circuit that performs communication processing. The communication unit 85 may also be composed of multiple communication circuits corresponding to multiple communication methods. Output section 86 is an output interface that outputs data to external devices such as printers and external storage devices.

[0122] also, Figure 20 As an example, the architecture of a computer system is not limited to Figure 20 For example, the computer system may not have an output unit 86. Furthermore, in the case where the defect analysis device 1 is implemented through multiple computer systems, not all of these computer systems may be... Figure 20 The computer system shown. For example, some computer systems may not have this feature. Figure 20 At least one of the display unit 84, output unit 86, and input unit 82 shown.

[0123] Here, an example of the operation of a computer system up to the point where it can execute an analysis program describing the processing of the defect analysis device 1 will be described. In a computer system employing the above structure, for example, an analysis program describing the operation of the defect analysis device 1 is loaded into the storage unit 83 from a CD-ROM or DVD-ROM drive (not shown) located on a CD-ROM or DVD-ROM drive. Furthermore, when the analysis program is executed, the analysis program read from the storage unit 83 is stored in the area of ​​the main storage device that becomes the storage unit 83. In this state, the control unit 81 executes the processing of the defect analysis device 1 according to the analysis program stored in the storage unit 83.

[0124] Furthermore, in the above description, CD-ROM or DVD-ROM is used as the recording medium, and a program describing the processing in the defect analysis device 1 is provided. However, it is not limited to this. Depending on the structure of the computer system, the capacity of the provided program, etc., for example, a program provided via the Internet or other transmission medium through the communication unit 85 may be used.

[0125] The analysis program of this embodiment causes a computer to perform the following steps: obtaining cutting point information representing the position of a cutting point CP, wherein the cutting point CP is a point on a machine tool mounted on a machine tool that cuts the workpiece W, and the cutting point CP corresponds to each of a plurality of front-end points P containing information representing the movement path of the tool's front end point P, i.e., the tool trajectory; obtaining a feature quantity representing a feature of the machining process, wherein the feature quantity is calculated corresponding to each of the plurality of front-end points P; and determining, based on the cutting point information and the feature quantity, the area that causes machining defects in the workpiece W, i.e., the machining defect area and the factors causing machining defects in the machining defect area.

[0126] Figure 1 The operation information storage unit 101, tool trajectory storage unit 103, tool information storage unit 104, machining target shape storage unit 105, cutting point storage unit 107, feature quantity storage unit 109, and factor pattern storage unit 113 shown are Figure 20 This is a portion of the storage unit 83 shown. It is implemented using a control unit 81, an input unit 82, a storage unit 83, and a display unit 84. Figure 1 Each of the tool trajectory calculation unit 102, cutting point calculation unit 106, position designation unit 112, machining defect area extraction unit 114, machining defect factor determination unit 115, and display unit 111 shown.

[0127] also, Figure 1 The functional division in the defect analysis device 1 shown is an example. As long as the defect analysis device 1 can perform the above-mentioned operations, the method of dividing each functional part is not limited to... Figure 1 The example shown. Additionally, in Figure 1 The defect analysis device 1 performs all the actions, but multiple devices can also be used to achieve the same function. For example, in Figure 1 In this process, the defect analysis device 1 also has the function of generating information used by the defect analysis unit 110, but it is also possible, as shown in the following variation, to separate the device with the defect analysis function and the device with the function of calculating the cutting point CP and the characteristic quantity.

[0128] <Variation Example>

[0129] Figure 21This diagram illustrates the functional structure of a modified example of Embodiment 1. In this modified example, the machining defect analysis device 1A includes a cutting point information acquisition unit 10A, a feature quantity acquisition unit 20A, a machining defect analysis unit 110, a display unit 111, a position designation unit 112, and a factor pattern storage unit 113. The information processing device 2 is a server or similar device, different from the machining defect analysis device 1A, and includes an operation information storage unit 101, a tool trajectory calculation unit 102, a tool trajectory storage unit 103, a tool information storage unit 104, a machining target shape storage unit 105, a cutting point calculation unit 106, a cutting point storage unit 107, a feature quantity calculation unit 108, and a feature quantity storage unit 109.

[0130] The cutting point information acquisition unit 10 of the defective machining analysis device 1 has a cutting point calculation unit 106 and a cutting point storage unit 107, and has the function of generating cutting point information. In contrast, the cutting point information acquisition unit 10A of the defective machining analysis device 1A acquires cutting point information from the information processing device 2. Furthermore, the feature quantity acquisition unit 20 of the defective machining analysis device 1 has a feature quantity calculation unit 108 and a feature quantity storage unit 109, and has the function of calculating feature quantities. In contrast, the feature quantity acquisition unit 20A of the defective machining analysis device 1A acquires feature quantities from the information processing device 2. Additionally, [further details are missing from the original text]. Figure 1 The functions of the functional parts with the same symbols are the same as those of the defect analysis device 1, so detailed descriptions are omitted.

[0131] Furthermore, regarding the defect analysis device 1A and the information processing device 2, similarly to the defect analysis device 1, one or more such devices can also be used. Figure 20 The computer system shown is used for implementation. In this case, both the cutting point information acquisition unit 10A and the feature quantity acquisition unit 20A can be implemented using the communication unit 85. Furthermore, Figure 21 The functional division of the defect analysis device 1A and the information processing device 2 shown is an example. As long as the system as a whole can achieve the same function as the defect analysis device 1, the method of dividing each functional unit is not limited to [specific method]. Figure 21 The example shown. For example, the functions of the defective area extraction unit 114 and the defective factor determination unit 115 can be implemented using different hardware. The functions of the information processing device 2 can also be implemented using multiple hardware components. Furthermore, a portion of the functions of the defective analysis device 1A and the functions of the information processing device 2 can be implemented using a cloud system. In a cloud system, the hardware of the computer system and the partitioning of servers and other devices for each function can be arbitrarily configured. For example, one computer system can function as multiple devices, or multiple computer systems can function as one device.

[0132] As explained above, the defect analysis apparatus 1 and 1A of Embodiment 1 are characterized by comprising: a cutting point information acquisition unit 10 and 10A, which acquires cutting point information indicating the position of a cutting point CP, wherein the cutting point CP is a point on a machine tool mounted on a machine tool that cuts the workpiece W, and the cutting point CP corresponds to each of a plurality of front point points P that contain information indicating the movement path of the tool's front end point P, i.e., the tool trajectory; a feature quantity acquisition unit 20 and 20A, which acquires feature quantities indicating the characteristics of the machining process, wherein the feature quantities are calculated corresponding to each of the plurality of front point points P; and a defect analysis unit 110, which determines, based on the cutting point information and the feature quantities, the area where a defect occurs in the workpiece W, i.e., the defect area, and the factors causing the defect in the defect area. The defect analysis unit 110 comprises: a defect area extraction unit 114, which extracts the defect area based on the cutting point information and the feature quantities; and a defect factor determination unit 115, which determines the factors causing the defect in the defect area based on the feature quantities and the extracted defect area. By employing the aforementioned structure, the defect analysis devices 1 and 1A can determine which part of the workpiece W is defective, reducing the workload of personnel analyzing the factors causing the defect. Therefore, appropriate countermeasures can be implemented to address the defect, reducing rework and improving productivity.

[0133] Furthermore, the defect analysis devices 1 and 1A also include a display unit 111 that displays the extracted defective areas using different representation methods for each factor of the defect. This allows operators to easily understand the factors contributing to the defects in the defective areas, further reducing the workload of the operators in the analysis process. Moreover, when multiple defective areas are extracted, the defect analysis unit 110 determines the factors contributing to the defects for each area, and the display unit 111 displays the defective areas using different representation methods for each factor of the defects.

[0134] The display unit 111 can also display characteristic quantities that serve as the basis for determining the factors contributing to defects in the extracted defective areas. This function of the display unit 111 further reduces the analytical workload of the operator. Furthermore, when multiple factors contributing to defects exist for a single defective area, the display unit 111 can display the probability of determining each factor. This function allows the operator to understand the multiple potential factors contributing to defects and the probability of determining each factor, further reducing the analytical workload of the operator.

[0135] Furthermore, the defect analysis devices 1 and 1A also include a position designation unit 112 that receives a designated location on the processing target shape M1 of the workpiece W, and the defect analysis unit 110 can extract the area including the designated location as a defective area. With the functions described above, the defect analysis devices 1 and 1A allow operators to observe, for example, the finished workpiece W to grasp the approximate location of the defective area and then designate that area to analyze the factors contributing to the defect. Therefore, the operator's intentions can be reflected in the displayed results, further reducing the labor required for analysis.

[0136] Furthermore, the defect analysis devices 1 and 1A also include a factor pattern storage unit 113 that stores information corresponding to the patterns of defective factors and characteristic quantities present when those factors are present. The defect analysis unit 110 can determine the defective factors in the extracted defective area based on the extracted defective area, characteristic quantity, and factor pattern. Because the defect analysis devices 1 and 1A have this function, when the correspondence between defective factors and characteristic quantity patterns is known in advance based on past analysis results, the past analysis results can be utilized. Therefore, the labor required for analysis work by operators can be further reduced.

[0137] Implementation method 2.

[0138] Figure 22 This diagram illustrates the functional structure of the machining defect analysis device 1B according to Embodiment 2. The target shape for machining in a machine tool varies depending on the machine and the user, and the factors causing machining defects differ for each target shape. However, it is common for a particular user to machine similar target shapes, making it effective to flexibly apply past machining results and simulation results to the analysis of machining defects.

[0139] The machining defect analysis device 1B includes an operation information storage unit 101, a tool trajectory calculation unit 102, a tool trajectory storage unit 103, a tool information storage unit 104, a machining target shape storage unit 105, a cutting point information acquisition unit 10 with a cutting point calculation unit 106 and a cutting point storage unit 107, a feature quantity acquisition unit 20 with a feature quantity calculation unit 108 and a feature quantity storage unit 109, a machining defect analysis unit 110B with a machining defect area extraction unit 114B and a machining defect factor determination unit 115B, a display unit 111, a position designation unit 112, a model construction unit 116, and a data accumulation unit 117.

[0140] The defect analysis device 1B replaces the factor pattern storage unit 113 of the defect analysis devices 1 and 1A, and has a model construction unit 116 and a data accumulation unit 117. The defect analysis unit 110B uses the learned model constructed by the model construction unit 116 to analyze defects, which differs from Embodiment 1. Hereinafter, the parts that are different from Embodiment 1 will be mainly described, and detailed descriptions of the parts that are the same as in Embodiment 1 will be omitted.

[0141] The data accumulation unit 117 accumulates data related to machine tool machining. For example, each time a machining or simulation is performed, the cutting point CP and characteristic quantities are calculated. Even for the same machining target shape M1, different characteristic quantities are calculated depending on the machining method and the surrounding environment. The data accumulation unit 117 associates the machining target shape, tool information, operation information, and the cutting point CP and characteristic quantities calculated based on these, and records them as accumulated information for each past machining or simulation trial. Furthermore, the cutting point CP and characteristic quantities can be recalculated using only the machining target shape M1, tool information, and operation information as accumulated information. When the cutting point CP and characteristic quantities are still accumulated, the processing time for recalculation can be shortened.

[0142] The model building unit 116 uses the accumulated information recorded in the data accumulation unit 117 to generate a learned model for judging factors of poor processing. The model building unit 116 changes the model parameters, hyperparameters, or prediction model of the prediction model used to predict factors of poor processing. Here, an example of using machine learning to change the model parameters of the prediction model will be described.

[0143] Figure 23 This is a diagram illustrating an example of the structure of a model building unit 116 using machine learning. The model building unit 116 includes a learning data acquisition unit 201, a model generation unit 202, and a learned model storage unit 203.

[0144] The learning data acquisition unit 201 acquires at least one feature quantity as learning data based on the accumulation information of the data accumulation unit 117. If the accumulation information includes the feature quantity, the learning data acquisition unit 201 acquires the feature quantity itself from the data accumulation unit 117. If the accumulation information does not include the feature quantity, the learning data acquisition unit 201 can calculate the feature quantity based on the accumulation information and acquire the calculated feature quantity as learning data. The learning data acquisition unit 201 outputs the acquired learning data to the model generation unit 202.

[0145] Furthermore, the learning data acquisition unit 201 can acquire the feature values ​​of each cutting point CP as learning data. The learning data can be data from a subset of cutting points CP, or data from all cutting points CP. Additionally, regarding the learning data, the feature values ​​of a single cutting point CP can be grouped into one set, or the feature values ​​of multiple adjacent points, or points preceding and following each other in a time series, can be grouped into one set. Moreover, the learning data acquisition unit 201 can acquire all types of feature values, or only a subset of them. Furthermore, in addition to feature values, the learning data acquisition unit 201 can also acquire information such as the machining target shape M1 and tool information as learning data.

[0146] The model generation unit 202 learns the factors causing machining defects based on the learning data output from the learning data acquisition unit 201. That is, it generates a learned model that infers the factors causing machining defects based on feature quantities generated from machine tool operation information and tool trajectory.

[0147] The learning algorithm used in the model generation unit 202 can employ well-known algorithms such as supervised learning, unsupervised learning, and reinforcement learning. As an example, the application of k-means as an unsupervised learning method will be explained. Unsupervised learning refers to a method that learns features from learning data (excluding results) by providing the learning device with that data.

[0148] The model generation unit 202, for example, learns poorly processed factors through so-called unsupervised learning by using a grouping method based on the k-means method.

[0149] k-means is a non-hierarchical clustering algorithm that uses the average of clusters to classify a given number of clusters into k.

[0150] Specifically, the k-means method follows this process: First, each data point xi is randomly assigned to a cluster. Next, based on the assigned data, the center Vj of each cluster is calculated. Then, the distance between each xi and each Vj is calculated, and xi is reassigned to the cluster with the nearest center. Furthermore, if the cluster assignments of all xi remain unchanged during the above process, or if the change is less than a pre-set threshold, the process is considered convergent and terminated.

[0151] In this embodiment, based on learning data created using at least one feature obtained by the learning data acquisition unit 201, factors that cause poor processing are learned through so-called unsupervised learning.

[0152] The model generation unit 202 generates a learned model by performing the learning process described above, and outputs the generated learned model. Here, the operator can set labels for each cluster to represent the processing defects corresponding to each cluster in the learned model.

[0153] The learned model storage unit 203 stores the learned models output from the model generation unit 202. The learned models stored in the learned model storage unit 203 are utilized by the processing defect analysis unit 110B.

[0154] Next, use Figure 24 This explains the learning process of Model Construction Department 116. Figure 24 Is with Figure 22 The flowchart shows the learning process performed by the defect analysis device 1B.

[0155] The learning data acquisition unit 201 performs data acquisition processing (step S201) to acquire learning data including at least one feature. Furthermore, when the learning data acquisition unit 201 acquires multiple feature values, the multiple feature values ​​can be associated and input together. Each of the multiple feature values ​​can be acquired at different time intervals, or they can be acquired simultaneously.

[0156] The model generation unit 202 learns from the learning data including at least one feature obtained by the learning data acquisition unit 201 through so-called unsupervised learning, learns the factors of poor processing, and performs learning processing to generate a learned model (step S202).

[0157] The learned model storage unit 203 stores the learned models generated by the model generation unit 202 (step S203).

[0158] Figure 25 It is shown Figure 22 The diagram shows the functional structure of the inference device 204 included in the defect analysis unit 110B. The defect analysis unit 110B has the function of inference device 204 that uses a learned model to infer the factors causing defects.

[0159] The inference device 204 includes a data acquisition unit 205 and an inference unit 206.

[0160] The data acquisition unit 205 acquires at least one feature from the feature acquisition unit 20. Here, the data acquisition unit 205 acquires input data of the same kind as the learning data used when generating the learned model used in the inference unit 206.

[0161] The inference unit 206 infers the factors of poor processing obtained from the learned model stored in the learned model storage unit 203. That is, by inputting at least one feature value obtained by the data acquisition unit 205 into the learned model, the inference unit 206 can infer which cluster the feature value belongs to and output the inference result as a factor of poor processing.

[0162] Furthermore, in this embodiment, it is described that the factors of processing defects are output using a learned model learned by the model generation unit 202 of the processing defect analysis device 1B. However, it is also possible to obtain a learned model generated by a learning device other than the processing defect analysis device 1B and output the factors of processing defects based on the learned model.

[0163] In this way, the inference unit 206 can obtain the factors of poor processing based on at least one feature quantity.

[0164] The defect analysis unit 110B, equipped with such an inference device 204, can use the function of the inference device 204 to determine the factors causing the defect. For example, using the same method as in Embodiment 1, the defect area extraction unit 114B of the defect analysis unit 110B can extract the defect area, and the defect factor determination unit 115B can use the function of the inference device 204 to determine the factors causing the defect at the cutting point CP contained in the extracted defect area. At this time, the defect factor determination unit 115B can set the defect factors set in the cluster where the multiple cutting points CP contained in the extracted defect area are most frequently classified as the defect factors of that defect area. In addition, the defect factor determination unit 115B can also extract multiple candidate factors for defect based on the number of points in the processing area where the cutting points CP are classified into each cluster, and calculate the determination probability for each factor.

[0165] Alternatively, the defect analysis unit 110B, within the defect factor determination unit 115B, utilizes the function of the inference device 204 to classify all cutting points CP into clusters of defect factors. The defect area extraction unit 114B then, based on these clusters, identifies a region comprising multiple adjacent cutting points CP classified into the same cluster as a single defect area. In this case, the clusters classified by the inference device 204 also include clusters without defects.

[0166] Next, use Figure 26 , indicating that it is used for Figure 25 The inference device 204 shown is used to process factors that cause poor processing. Figure 26 It is used for explanation Figure 25 The flowchart shows the operation of the inference device 204. The data acquisition unit 205 performs data acquisition processing to acquire at least one feature from the feature acquisition unit 20 (step S301).

[0167] The inference unit 206 inputs at least one feature quantity into the learned model stored in the learned model storage unit 203 to obtain the factors of poor processing (step S302). The inference unit 206 performs data output processing to output the factors of poor processing obtained by the learned model to the display unit 111 and the like (step S303).

[0168] Furthermore, this section describes the application of unsupervised learning in the learning algorithms used in the model generation unit 202 and the inference unit 206, but is not limited to this. Regarding learning algorithms, in addition to unsupervised learning, reinforcement learning, supervised learning, or semi-supervised learning can also be applied.

[0169] In addition, the learning algorithm used in the model construction section 116 can be deep learning, which learns by extracting the feature quantity itself, or other known methods.

[0170] In implementing unsupervised learning in this embodiment, it is not limited to non-hierarchical clustering using the k-means method as described above, but can be other known methods that can be clustered. For example, it can also be hierarchical clustering such as the shortest distance method.

[0171] In this embodiment, the functions of the model building unit 116 and the inference unit 206 can, for example, be implemented using a device independent of the defect analysis device 1B, connected to the defect analysis device 1B via a network. Alternatively, the functions of the model building unit 116 and the inference unit 206 can also be implemented on a cloud server.

[0172] Furthermore, the model generation unit 202 can learn the factors causing machining defects based on learning data generated for multiple machine tools. Moreover, the model generation unit 202 can obtain learning data from multiple machine tools used in the same area, or it can use learning data collected from multiple machine tools operating independently in different areas to learn the factors causing machining defects. Additionally, the machine tool from which the learning data was collected can be added to the object or removed from the object midway. Furthermore, the learning device that has learned the factors causing machining defects for a certain machine tool can be applied to the machining defect analysis of a different machine tool, and the factors causing machining defects can be learned and updated again for that different machine tool.

[0173] Furthermore, the function of the defect analysis device 1B in Embodiment 2 is the same as that in Embodiment 1, and it can also be used... Figure 20 The computer system shown is used for implementation. Alternatively, similar to Embodiment 1, the functions of generating cutting point information and calculating feature quantities can be implemented using a device different from the machining defect analysis unit 110B. Furthermore, a portion of the functions of the machining defect analysis device 1B can be implemented using a cloud system.

[0174] As explained above, the defect analysis apparatus 1B of Embodiment 2 includes: a learning data acquisition unit 201 that acquires learning data; and a model generation unit 202 that uses the learning data to generate a learned model for inferring factors of defect based on at least one feature quantity. Therefore, it is possible to use past processing or simulation data to generate a learned model for predicting factors of defect.

[0175] Furthermore, the defect analysis unit 110B can use a learned model to determine the factors contributing to defects. Moreover, the defect analysis unit 110B can use the learned model to classify cutting points CP as factors contributing to defects, and group multiple adjacent cutting points CP classified as having the same factor into one defect region. With this structure, the defect analysis device 1B can utilize past machining or simulation results to determine the factors contributing to defects, enabling more accurate identification of these factors. Additionally, by acquiring learning data for each user, or learning data based on machining or simulation results of a machining target shape M1 similar to the user's used machining target shape M1, it is possible to determine the factors contributing to defects in a way that is suitable for the user.

[0176] Implementation method 3.

[0177] Figure 27 This diagram illustrates the functional structure of the defect analysis device 1C according to Embodiment 3. The defect analysis device 1C includes an operation information storage unit 101, a tool trajectory calculation unit 102, a tool trajectory storage unit 103, a tool information storage unit 104, a machining target shape storage unit 105, a cutting point information acquisition unit 10 with a cutting point calculation unit 106 and a cutting point storage unit 107, a feature quantity acquisition unit 20 with a feature quantity calculation unit 108 and a feature quantity storage unit 109, a defect analysis unit 110B with a defect area extraction unit 114B and a defect factor determination unit 115B, a display unit 111, a position designation unit 112, a model construction unit 116C, a data accumulation unit 117, and a label generation unit 118.

[0178] The following mainly describes the parts that are different from the defect analysis device 1B in Embodiment 2, and detailed descriptions of the parts that are the same as those in the defect analysis device 1B are omitted.

[0179] The label generation unit 118 generates labels used when the model construction unit 116C constructs the learned model. The labels represent factors causing machining defects and are generated based on input information from the operator. The label generation unit 118 generates labels assigned to factors causing machining defects based on input information from the operator using a keyboard, buttons, touch sensors, etc. For example, the operator may specify a region of machining defect for the machining target shape M1 of the machining object W displayed on the screen, and input the factors causing the machining defects in that region. The input factors causing machining defects can be selected from pre-prepared factors or entered by the operator using a keyboard or other text input method. The label generation unit 118 can also obtain the input information of machining defect factors for each cutting point in table form.

[0180] The label generation unit 118 generates labels based on the input information from the operator. For example, the label generation unit 118 assigns values ​​corresponding to the set factors of machining defects to all cutting points CP that are included in the area specified by the operator. At this time, the label generation unit 118 assigns the same value to the same factor of machining defect. Furthermore, the label generation unit 118 can also assign values ​​indicating no machining defect to cutting points CP that are not included in the area specified by the operator. By using the input information from the operator, the label generation unit can learn the factors of machining defects determined by the operator.

[0181] The model construction unit 116C generates a learned model that takes at least one feature as input and the label generated by the label generation unit 118 as output.

[0182] Similar to implementation method 2, using Figure 23 The detailed functional structure of the model construction unit 116C is explained below. The model construction unit 116C includes a learning data acquisition unit 201, a model generation unit 202, and a learned model storage unit 203.

[0183] The learning data acquisition unit 201 acquires at least one feature and uses the label generated by the label generation unit 118 as learning data.

[0184] The model generation unit 202 learns labels representing factors of machining defects based on learning data generated from a combination of at least one feature quantity output from the learning data acquisition unit 201 and labels. That is, the model generation unit 202 generates a learned model that infers labels representing factors of machining defects based on feature quantities generated from machine tool operation information and tool trajectory. Here, the learning data is data that correlates at least one feature quantity with labels.

[0185] The learning algorithm used by the model generation unit 202 can employ well-known algorithms such as supervised learning and reinforcement learning. As an example, we will illustrate the application of neural networks.

[0186] The model generation unit 202, for example, learns labels representing poorly processed factors through so-called supervised learning, based on a neural network model. Here, supervised learning refers to a method that, by providing a learning device with sets of input and output data, learns the features present in these learning data and infers the results based on the input.

[0187] A neural network consists of an input layer composed of multiple neurons, an intermediate layer composed of multiple neurons, and an output layer composed of multiple neurons. The intermediate layer is also called a hidden layer, and can have one or more layers.

[0188] Figure 28 It is shown Figure 27 The diagram shows an example of the neural network used in the model construction section 116C. If it is as follows... Figure 28 The three-layer neural network shown in the diagram multiplies multiple inputs to the input layers X1-X3 by weights W1 (w11-w16) and inputs the result to the intermediate layers Y1-Y2. The result is then multiplied by weights W2 (w21-w26) and output from the output layers Z1-Z3. The output varies depending on the values ​​of weights W1 and W2.

[0189] In this embodiment, the neural network learns labels representing poorly processed factors through so-called supervised learning, based on learning data created by combining features and labels obtained by the learning data acquisition unit 201.

[0190] That is, the neural network learns by adjusting weights W1 and W2 in a way that the output from the output layer approximates the labels representing poorly processed factors, in order to input features into the input layer. During learning, backpropagation is used to infer the neural network's behavior so that the output values ​​approximate the training data. The structure of the neural network is not limited to... Figure 28 The example shown.

[0191] The model generation unit 202 generates a learned model by performing the learning process described above, and outputs the generated learned model.

[0192] The learned model storage unit 203 stores the learned model output from the model generation unit 202.

[0193] Furthermore, as the learning algorithm used in the model generation unit 202, it is possible to use deep learning that learns by extracting the feature quantity itself, or to perform machine learning according to other known methods, such as genetic programming, functional logic programming, support vector machines, etc.

[0194] Alternatively, as another example, the model construction unit 116C can also generate image data for each feature quantity and obtain learning data for the image data, including the color information (RGB) of the pixels in the image data and the labels generated by the label generation unit 118, and use such learning data for learning. The image data can also be a black and white image. In learning using image data, convolutional neural networks are generally used, and the error propagation method is used to infer the parameters. Other known methods can also be used to infer the parameters.

[0195] Furthermore, the function of the defect analysis device 1C in Embodiment 3 is the same as that in Embodiment 1, and it can also be used... Figure 20 The computer system shown is used for implementation. Furthermore, similar to Embodiment 1, the functions of generating cutting point information and calculating feature quantities can be implemented using a device different from the machining defect analysis unit 110B. Moreover, as long as the input information based on the tag generation unit 118 can be sent to the model construction unit 116C, the model construction unit 116C and the data accumulation unit 117 can also be implemented using a device different from the machining defect analysis device 1C. Additionally, a portion of the functions of the machining defect analysis device 1C can be implemented using a cloud system.

[0196] As explained above, the defect analysis device 1C according to Embodiment 3, in addition to the functions of the defect analysis device 1B of Embodiment 2, also includes: a label generation unit 118 that generates labels representing factors of defective processing based on input information from the operator; and a model construction unit 116C that generates a learned model that outputs the labels generated by the label generation unit 118. Therefore, a learned model reflecting the operator's judgment for predicting factors of defective processing can be constructed. Thus, the labor required for operators analyzing factors of defective processing can be reduced.

[0197] The structure shown in the above embodiments is an example that can be combined with other known technologies, can be combined with each other, and can omit or change a part of the structure without departing from the spirit of the matter.

[0198] For example, the above describes machining defect analysis devices 1, 1A to 1C for analyzing machining defects generated during machining using a machine tool. However, a machining system including a machine tool for machining the workpiece W using a tool and machining defect analysis devices 1, 1A to 1C can also be provided. Furthermore, a machining defect analysis method including steps for obtaining cutting point information, obtaining feature quantities, and determining the machining defect area and the factors contributing to the machining defect within that area can also be provided. Additionally, a machining method including, in addition to the steps of the machining defect analysis method, the steps of machining the workpiece W using a machine tool can also be provided. Moreover, the machining defect analysis process can be performed after the machining or simulation of the workpiece W using the machine tool has ended, or it can be performed in parallel with the machining or simulation of the workpiece W using the machine tool.

[0199] Furthermore, the machining defect analysis devices 1, 1A to 1C can be installed near the machine tool and the CNC device that controls the machine tool, or they can be installed in a location separate from the machine tool and the CNC device to analyze the machining process of the machine tool. Moreover, the machining defect analysis devices 1, 1A to 1C can also be provided as a function of the CNC device.

Claims

1. A defective processing analysis device, characterized in that, have: The cutting point information acquisition unit acquires cutting point information indicating the position of the cutting point, wherein the cutting point is the point on the machine tool where the tool is cut to process the workpiece, and the cutting point corresponds to each of the multiple front-end points included in the information representing the movement path of the tool's front end point, i.e., the tool trajectory. The feature acquisition unit acquires feature quantities representing the processed features, which are calculated corresponding to each of the plurality of said front-end points; The machining defect analysis unit determines, based on the cutting point information and the feature quantity, the area that causes the machining defect in the workpiece, namely the machining defect area, and the factors causing the machining defect in the machining defect area; as well as The display unit displays the defective area in a different manner for each of the factors, and displays the determination probability for each factor when there are multiple candidates for one defective area.

2. The processing defect analysis device according to claim 1, characterized in that, The display unit displays the characteristic quantity that serves as the basis for determining the factors related to the defective processing area.

3. The processing defect analysis device according to claim 1 or 2, characterized in that, It also includes a location designation unit that accepts a designated location, which is a designated position on the processing target shape of the workpiece. The defect analysis unit extracts the area including the specified location as the defective area.

4. The processing defect analysis device according to claim 1 or 2, characterized in that, It also includes a factor pattern storage unit, which stores information that maps poorly processed factors to patterns of the characteristic quantities present when those factors are present. The defective processing analysis unit determines the factors of the defective processing area based on the extracted defective processing area, the characteristic quantity, and the factor pattern.

5. The processing defect analysis device according to claim 1, characterized in that, The defect analysis unit has: The defective area extraction unit extracts the defective area based on the cutting point information and the feature quantity; and The defect determination unit determines the factors contributing to the defects in the defective areas based on the characteristic quantities and the extracted defective areas.

6. The processing defect analysis device according to claim 1 or 2, characterized in that, It also has: The learning data acquisition unit acquires learning data including at least one of the aforementioned features; and The model generation unit uses the learning data to generate a learned model for inferring the factor based on at least one of the feature quantities.

7. The defect analysis device according to claim 6, characterized in that, It also includes a label generation unit that generates a label representing the factor for each cutting point or defective area based on input information. The model generation unit generates the learned model that takes at least one of the features as input and outputs the label representing the factor.

8. The processing defect analysis device according to claim 1, characterized in that, The defect analysis unit uses a learned model to infer the factor based on at least one of the feature quantities to determine the factor based on the feature quantity.

9. The processing defect analysis device according to claim 8, characterized in that, The defective processing analysis unit uses the learned model, which takes at least one of the features as input and a label as output, to determine the factor based on the features, wherein the label is candidate information representing a pre-determined factor.

10. The processing defect analysis device according to claim 8 or 9, characterized in that, The defect analysis unit uses the learned model to classify the cutting points as factors, and treats multiple adjacent cutting points classified as the same factor as one defect area.

11. The processing defect analysis device according to claim 1 or 2, characterized in that, The feature acquisition unit acquires the feature values ​​as image data or time series data.

12. The processing defect analysis device according to claim 1, characterized in that, It also includes a tool trajectory calculation unit that generates the tool trajectory based on operation information, which represents the operating state of the machine tool when it is operated. The cutting point information acquisition unit includes a cutting point calculation unit, which generates the cutting point information by calculating the position of the cutting point corresponding to each of the plurality of front end points included in the tool trajectory based on the tool trajectory, tool information that defines the shape of the tool, and the machining target shape of the workpiece. The feature quantity acquisition unit has a feature quantity calculation unit, which calculates the feature quantity corresponding to each of the plurality of front end points included in the tool trajectory based on the operation information, the tool trajectory and the cutting point information.

13. A processing system, characterized in that, have: Machine tools are used to process objects; and A machining defect analysis device analyzes machining defects that occur during machining using the machine tool. The defect analysis device includes: The cutting point information acquisition unit acquires cutting point information indicating the position of the cutting point, wherein the cutting point is the point at which the tool mounted on the machine tool cuts the workpiece, and the cutting point corresponds to each of the multiple front-end points included in the information representing the movement path of the tool's front end point, i.e., the tool trajectory; The feature acquisition unit acquires feature quantities representing the processed features, which are calculated corresponding to each of the plurality of said front-end points; The machining defect analysis unit determines, based on the cutting point information and the feature quantity, the area that causes the machining defect in the workpiece, namely the machining defect area, and the factors causing the machining defect in the machining defect area; as well as The display unit displays the defective area in a different manner for each of the factors, and displays the determination probability for each factor when there are multiple candidates for one defective area.

14. A method for analyzing processing defects, characterized in that, include: The step of obtaining cutting point information representing the position of the cutting point, wherein the cutting point is the point on the machine tool where the tool is cut to process the workpiece, and the cutting point corresponds to each of the multiple front-end points contained in the information representing the movement path of the tool's front end point, i.e., the tool trajectory; The step of obtaining a feature quantity representing a feature of the processing, wherein the feature quantity is calculated corresponding to each of the plurality of said front-end points; The steps of determining the area of ​​machining defects that causes the workpiece to be machined, i.e. the machining defect area, and the factors causing machining defects in the machining defect area, based on the cutting point information and the feature quantity; as well as The process involves displaying the defective area in a different manner for each of the aforementioned factors, and, in the case where there are multiple candidates for the aforementioned factors for one defective area, displaying the probability of determination for each of the aforementioned factors.

Images