Learning method, center hole position determining system, and center hole position determining method
A machine learning-based center hole positioning system accurately determines crankshaft hole positions, addressing cost and accuracy issues in existing methods by generating a trained model from external shape and eccentricity data, ensuring precise alignment with the principal axis of inertia.
Patent Information
- Application Number
- JP2020137624
- Authority / Receiving Office
- JP · JP
- Patent Type
- Patents
- Current Assignee / Owner
- Filing Date
- 2020-08-17
- Publication Date
- 2024-11-14
- Estimated Expiration
- 2040-08-17
AI Technical Summary
Existing methods for determining the center hole position in crankshafts, such as using a mass centering machine or three-dimensional shape data, either increase costs or suffer from inaccuracies due to dimensional errors, making precise and cost-effective positioning challenging.
A center hole positioning system and method utilizing machine learning to generate a trained model using external shape data and eccentricity data, allowing for accurate determination of center hole positions on crankshafts by measuring and calculating data at multiple reference points on the crankshaft's outer contour.
The system enables precise and cost-effective determination of center hole positions on crankshafts, reducing the need for large and costly equipment while improving accuracy by using machine learning to align the rotation axis with the principal axis of inertia.
Smart Images

Figure 0007583431000001 
Figure 0007583431000002 
Figure 0007583431000003
Abstract
Description
[Technical field]
[0001] The present invention relates to a learning method, a center hole position determining system, and a center hole position determining method. [Background technology]
[0002] A crankshaft is one of the components of an engine. If a rotational imbalance occurs in a raw crankshaft, which is also a component of an engine, problems such as vibrations will occur when the engine is rotating. The following Patent Document 1 discloses a method for measuring the amount of deviation (offset distance) between the rotation axis and the center of mass of a raw crankshaft using a mass centering machine. Then, based on the amount of deviation, the position of a center hole to be formed in an end face (for example, both end faces) of the raw crankshaft is determined so that the center hole is located on the principal axis of inertia (so that the rotation axis and the principal axis of inertia coincide).
[0003] Furthermore, the following Patent Document 2 discloses a method for assuming the position of a center hole of a material crankshaft based on three-dimensional shape data of the material crankshaft. The rotational unbalance of the material crankshaft after machining simulation is calculated based on the assumed center hole position. If the amount of rotational unbalance is within a predetermined allowable range, the assumed center hole position is determined as the center hole position during actual machining. [Prior art documents] [Patent documents]
[0004] [Patent Document 1] Patent No. 5994184 [Patent Document 2] Patent No. 4791577 Summary of the Invention [Problem to be solved by the invention]
[0005] However, the method of measuring the deviation between the rotation axis and the center of mass of a material crankshaft using a mass centering machine as in Patent Document 1 may lead to an increase in the size and cost of the entire apparatus. On the other hand, in the method of Patent Document 2, the position of the center hole of the material crankshaft is determined based only on the three-dimensional shape data (i.e., design data) of the material crankshaft. Although this method can eliminate the increase in costs as described above, since the eccentricity data (imbalance) of the material crankshaft is obtained based only on the geometric three-dimensional shape data, if a dimensional error such as a processing error occurs between the material crankshaft and the three-dimensional shape data, the accuracy of the eccentricity data is poor. In this case, there is a risk that the position of the center hole cannot be measured with high accuracy. In addition, there are various restrictions, such as the need for the customer to provide the three-dimensional shape data, which is the design data.
[0006] That is, it is desired to determine the position of the center hole with high accuracy while reducing costs. Therefore, one object of the present invention is to provide a center hole position determination system and a center hole position determination method that can determine the center hole position with high accuracy while reducing costs. Another object of the present invention is to provide a learning method that can determine the center hole position with high accuracy. [Means for solving the problem]
[0007] The present invention provides a learning method including: an acquisition step (S1) of acquiring a data set including outer shape data (P) for a plurality of predetermined reference positions (A1 to A60) on the outer shape of a material crankshaft (9) in which a center hole (10A, 10B) is to be formed, and eccentricity data (Q) of the material crankshaft; and a learning step (S2) of generating, by machine learning, a trained model (104) for estimating the position of the center hole corresponding to the outer shape of the material crankshaft, using the outer shape data as input data and the eccentricity data as training data.
[0008] As in one embodiment of the present invention, the eccentricity data may include actual measurement data obtained when the material crankshaft in which the center hole is to be formed is measured using a mass centering machine. As in another embodiment of the present invention, the eccentricity data may include calculation data obtained when the material crankshaft in which the center hole is to be formed is calculated using a calculation.
[0009] The present invention also provides a center hole position determination system (7), including: a measurement means (2) that measures outer shape data (P) for a plurality of predetermined reference positions (A1 to A60) on the outer contour of a material crankshaft (9) in which center holes (10A, 10B) are to be formed; and an estimation means (3) that estimates a position of the center hole corresponding to the measured outer shape data by using a trained model (104) that is generated in advance by machine learning using the outer shape data as input data and eccentricity data (Q) of the material crankshaft as training data.
[0010] In one embodiment of the present invention, the external shape data includes distance data which is the distance of the reference position from a predetermined distance reference position. In one embodiment of the present invention, the material crankshaft includes a plurality of counterweights (CW1 to CW6), and the reference position is set on an outer periphery of the counterweight.
[0011] In one embodiment of the present invention, the reference position is set at at least one of an inflection portion, a top portion, and a bottom portion on the outer periphery of the counterweight. In one embodiment of the present invention, the reference position is a circumferential position on each of the outer peripheries of the plurality of counterweights. throughout theA plurality of eccentricity data may be set within an angle range. The outer shape data may include position information of a reference position. The raw crankshaft may have a first end face in which the first center hole is to be formed, and a second end face opposite to the first end face in which the second center hole is to be formed. The eccentricity data may include eccentricity data of one surface side of the first end face and eccentricity data of a second surface side of the second end face of the raw crankshaft. The learning model may be a learning model using the eccentricity data of the one surface side and the eccentricity data of the second surface side as teacher data. The estimation means may include means for estimating positions of the first center hole and the second center hole. The present invention also provides a center hole position determination method including a measurement step (S4) of measuring outer shape data (P) for a plurality of predetermined reference positions (A1-A60) on the outer shape of a material crankshaft (9) in which center holes (10A, 10B) are to be formed, and an estimation step (S7) of estimating positions of the center holes corresponding to the measured outer shape data by using a trained model (104) generated in advance by machine learning using the outer shape data as input data and eccentricity data (Q) of the material crankshaft as training data. The estimation step may include a step of estimating positions of the first center hole and the second center hole.
[0012] In one embodiment of the present invention, the external shape data includes distance data which is the distance of the reference position from a predetermined distance reference position (C1). In one embodiment of the present invention, the material crankshaft includes a plurality of counterweights (CW1 to CW6), and the reference position is set on an outer periphery of the counterweight.
[0013] In one embodiment of the present invention, the reference position is set at at least one of an inflection portion, a top portion, and a bottom portion on the outer periphery of the counterweight. In the above, the numbers in parentheses indicate reference symbols of corresponding components in the embodiments described below, but these reference symbols are not intended to limit the scope of the claims. Effect of the Invention
[0014] According to the present invention, a trained model for estimating the position of a center hole corresponding to the outer shape of a material crankshaft is generated by machine learning using a plurality of outer shape data as input data and eccentricity data as training data. Furthermore, the position of the center hole corresponding to the measured outer shape data is estimated using a trained model generated in advance by such machine learning. Therefore, the position of the center hole located on the principal axis of inertia (the position of the center hole where the rotation axis and the principal axis of inertia coincide) can be determined with high accuracy. Therefore, it is possible to provide a center hole position determination system and a center hole position determination method that can determine the position of the center hole with high accuracy while reducing costs. Also, it is possible to provide a learning method that can determine the position of the center hole with high accuracy.
[0015] According to an embodiment of the present invention, the measured outer shape data is used as input data for the generated trained model to determine the center hole position, so that the center hole position can be obtained more accurately. This allows the center hole position to be determined with even greater precision. According to another embodiment of the present invention, by increasing the number of data points by increasing the resolution, the resolution for determining the center hole position can be improved, and the accuracy of the center hole position can be improved. This allows the center hole position to be determined with even greater accuracy.
[0016] According to an embodiment of the present invention, since the contour data includes distance data, the contour data can be acquired as one-dimensional data, which makes it easier to handle the contour data compared to the case where the contour data is three-dimensional data. According to an embodiment of the present invention, the center of mass of the crankshaft is highly dependent on the outer peripheral shape (radius, existence angle, axial position) of each counterweight. Therefore, by using data of the reference position set on the outer periphery as the outer peripheral shape data, the position of the center hole can be determined with even higher accuracy so that the center hole is located on the principal axis of inertia.
[0017] According to one embodiment of the present invention, since the reference position is set at an inflection point on the outer periphery of the counterweight, the reference position can be easily specified. [Brief description of the drawings]
[0018] [Figure 1] FIG. 1 is a block diagram showing the configuration of a hole forming system to which a center hole position determining system according to one embodiment of the present invention is applied. [Diagram 2] FIG. 2 is a diagram for explaining a material crankshaft that is an object to be drilled by the hole forming system. [Diagram 3] FIG. 3(A) is a view of the first counterweight seen from the left side of FIG. 2, and FIG. 3(B) is a view of the third counterweight seen from the left side of FIG. [Figure 4] FIG. 4 is a diagram for explaining eccentricity data of the material crankshaft. [Diagram 5] FIG. 5 is a diagram showing a state in which the material crankshaft is held by a hole drilling machine. [Figure 6] FIG. 6 is a diagram for explaining the functional configuration of the center hole position determining device shown in FIG. [Figure 7] FIG. 7 is a diagram showing a model of a neural network that is machine-learned by the machine learning unit shown in FIG. [Figure 8] FIG. 8 is a diagram of the external shape data measuring device as viewed in the horizontal direction. [Figure 9] FIG. 9 is a flow diagram illustrating machine learning by the machine learning unit. [Figure 10]FIG. 10 is a flow chart showing the contents of the center hole position determination and hole drilling process executed in the hole forming system. [Figure 11] FIG. 11 is a flow chart showing the procedure for measuring the external shape data by the external shape data measuring device. [Figure 12] FIG. 12 is a diagram showing changes in distance data accompanying changes in the rotational angle position of the material crankshaft. [Figure 13] FIG. 13 is a flow chart showing the contents of the determination of the center hole position (S5 in FIG. 10). [Figure 14] FIG. 14 is a diagram for explaining another embodiment of the present invention. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0019] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0023] The present invention will be described in detail below with reference to the accompanying drawings, in which: Fig. 1 is a block diagram showing the configuration of a hole forming system 1 to which a center hole position determining system 7 according to an embodiment of the present invention is applied. The hole forming system 1 includes an outer shape data measuring machine (measuring means) 2, a center hole position determining device (estimating means) 3, a main control device 4, a hole processing machine 5, and a mass centering machine 6.
[0020] The outer shape data measuring machine 2 includes a control unit 21, a storage unit 22, and a communication unit 23. The control unit 21 includes a CPU, RAM, and ROM. The storage unit 22 is configured with a storage device such as a non-volatile memory or a hard disk drive. The communication unit 23 is communicatively connected to the main control device 4 via a communication line (not shown). The communication unit 23 may be communicatively connected to the center hole position determining device 3, the hole processing machine 5, the mass centering machine 6, and the like. The outer shape data measuring machine 2 measures outer shape data related to the positions of the outer contours of the material crankshaft 9 to be formed with center holes 10A, 10B (see FIG. 5).
[0021] The center hole position determining device 3 includes a control unit 31, a storage unit 32, and a communication unit 33. The center hole position determining device 3 is configured by a computer. The control unit 31 includes a CPU (processor) 34, a RAM 35, and a ROM 36. In addition to an operating system, computer programs such as a neural network program 37 are installed in the ROM 36. The storage unit 32 is configured by a storage device such as a non-volatile memory or a hard disk drive. The communication unit 33 is connected to the main control device 4 via a communication line (not shown) so as to be able to communicate with it. The communication unit 33 may be provided so as to be able to communicate with the external shape data measuring machine 2, the hole processing machine 5, etc.
[0022] The center hole position determining device 3 determines (estimates) the positions of the center holes 10A, 10B corresponding to the measured material crankshaft 9 (see FIG. 2) by using a trained model 104 (see FIG. 6). The trained model 104 is generated in advance by machine learning using a plurality of external shape data P (see FIG. 7) of a crankshaft of the same type as the material crankshaft 9 as input data and eccentricity data Q (see FIG. 7) of the crankshaft as training data.
[0023] The main control device 4 includes a control unit 41, a storage unit 42, and a communication unit 43. The control unit 41 includes a CPU, RAM, and ROM. The storage unit 42 is configured with a storage device such as a non-volatile memory or a hard disk drive. The communication unit 43 is provided so as to be able to communicate with the external data measuring machine 2, the center hole position determining device 3, the hole processing machine 5, the mass centering machine 6, etc. The main control device 4 is, for example, a PLC.
[0024] The center hole position determining device 3, the outer shape data measuring machine 2, and the main control device 4 are included in the center hole position determining system 7. The center hole position determining system 7 measures the deviation between the rotation axis and the mass axis of the raw crankshaft 9, and determines the positions of the center holes 10A, 10B to be formed on the end faces 9a, 9b of the raw crankshaft 9 so that the center holes 10A, 10B are located on the principal axis of inertia (i.e., so that the rotation axis and the principal axis of inertia coincide). In other words, the center hole position determining system 7 has the same function as the mass centering machine 6.
[0025] The hole processing machine 5 includes a control unit 51, a storage unit 52, and a communication unit 53. The control unit 51 includes a CPU, RAM, and ROM. The storage unit 52 is configured with a storage device such as a non-volatile memory or a hard disk drive. The communication unit 53 is communicably connected to the main control device 4 via a communication line (not shown). The communication unit 53 may be provided so as to be communicable with the external shape data measuring machine 2, the center hole position determining device 3, etc. The hole processing machine 5 drills center holes 10A, 10B at positions determined by the center hole position determining system 7 in the end faces 9a, 9b of the material crankshaft 9.
[0026] Furthermore, the center hole position determining device 3, the mass centering machine 6, and the main control device 4 are included in a learning system 8. The learning system 8 generates a trained model 104 for estimating the positions of the center holes 10A, 10B corresponding to the outer shape of the material crankshaft 9. In this embodiment, the mass centering machine 6 is mainly used for machine learning (generation of the trained model 104). The mass centering machine 6 has a known configuration as shown in Japanese Patent No. 5994184. Therefore, a description of the mass centering machine 6 will be omitted.
[0027] The main control device 4 controls the operation of the hole forming system 1. Machine learning by the learning system 8 is performed by the center hole position determining device 3 and the mass centering machine 6 based on control commands from the main control device 4. The determination of the positions of the center holes 10A, 10B by the center hole position determining system 7 is performed by the external shape data measuring machine 2 and the center hole position determining device 3 based on control commands from the main control device 4. The formation of the center holes 10A, 10B by the hole processing machine 5 is also performed based on control commands from the main control device 4.
[0028] The hole forming system 1 further includes a rotor transport device (not shown). Fig. 2 is a diagram for explaining a raw crankshaft 9 which is the target of drilling center holes 10A, 10B by the hole forming system 1. Fig. 3(A) is a diagram showing the first counterweight CW1 as viewed from the left in Fig. 2, and Fig. 3(B) is a diagram showing the third counterweight CW3 as viewed from the left in Fig. 2. The raw crankshaft 9 will be described taking as an example a raw crankshaft for an in-line three-cylinder engine.
[0029] The material crankshaft 9 includes a first crank journal J1, a second crank journal J2, a third crank journal J3, a fourth crank journal J4, a first crank pin P1, a second crank pin P2, a third crank pin P3, a first counterweight CW1, a second counterweight CW2, a third counterweight CW3, a fourth counterweight CW4, a fifth counterweight CW5, and a sixth counterweight CW6. The crank journals J1 to J4 are arranged coaxially. The crank journals J1 to J4 are rotatably supported by a cylinder block and a crank case provided in the engine via bearings (not shown). The crank pins P1 to P3 are arranged at positions radially spaced from the crank journals J1 to J4, and one end of a connecting rod provided in the engine is rotatably attached to the crank pins P1 to P3. A piston is attached to the other end of each connecting rod.
[0030] The crank pins (crank pins P1 to P3) are sandwiched in the axial direction of the raw crankshaft 9 by two adjacent counterweights among the counterweights CW1 to CW6. These two counterweights are sandwiched between two adjacent crank journals (crank journals J1 to J4). The shapes (radius, existence angle) of a corresponding pair of counterweights are almost the same. That is, the shapes of the counterweights CW1 and CW2 are almost the same, the shapes of the counterweights CW3 and CW4 are almost the same, and the shapes of the counterweights CW5 and CW6 are almost the same.
[0031] In the example of Figures 2 and 3, the number of counterweights CW1 to CW6 is six, but the number of counterweights CW1 to CW6 may be more than six or less than six. Also, in the example of Figures 2 and 3, a full counterweight configuration in which counterweights are attached to both sides of the crank pin is shown as an example, but at least some of the configurations may include a semi-counterweight configuration in which counterweights are attached to only one side.
[0032] Specifically, the first counterweight CW1 is disposed between the first crank journal J1 and the first crank pin P1. The second counterweight CW2 is disposed between the first crank pin P1 and the second crank journal J2. The third counterweight CW3 is disposed between the second crank journal J2 and the second crank pin P2. The fourth counterweight CW4 is disposed between the second crank pin P2 and the third crank journal J3. The fifth counterweight CW5 is disposed between the third crank journal J3 and the third crank pin P3. The sixth counterweight CW6 is disposed between the third crank pin P3 and the fourth crank journal J4. For example, when viewed in the axial direction of the material crankshaft 9 shown in FIG. 2, the three crank pins P1 to P3 are disposed radially outward from the crank journals J1 to J4 and are shifted by 120° around the crank journals J1 to J4.
[0033] The outer circumferential surface of each of the counterweights CW1 to CW6 has a shape with some or multiple protrusions or recesses in the circumferential direction. As shown in Figures 3(A) and 3(B), the outer periphery of each of the counterweights CW1 to CW6 is smoothly curved. Therefore, the shortest distance between the counterweights CW1 to CW6 and the arc C1 centered on the axis C changes depending on the rotational angle position of the counterweights CW1 to CW6 (the rotational angle position of the material crankshaft 9).
[0034] Crankshafts have become lighter in recent years, making it difficult to balance the crankshaft while maintaining the specified weight using counterweights. This also applies to the raw crankshaft 9 in which center holes 10A, 10B are formed. For this reason, the center of mass of the raw crankshaft 9 is highly dependent on the outer peripheral shape (radius, existence angle, axial position) of each of the counterweights CW1-CW6.
[0035] The outer shape data measuring machine 2 is disposed in a measuring station (not shown). The outer shape data measuring machine 2 measures a plurality of outer shape data related to the position of the outer contour of the material crankshaft 9 in which the center holes 10A, 10B are to be formed. Specifically, this outer shape data is position data of a plurality of reference positions (for example, 60 positions in total) on the outer periphery of the plurality of counterweights CW1-CW6 of the material crankshaft 9. This outer shape data includes distance data of the shortest distance to the arc C1 (distance from the distance reference position).
[0036] In this embodiment, the external data measuring device 2 measures position data of a total of 60 reference positions that are predetermined on the outer circumference of the counterweights CW1 to CW6. In this embodiment, 10 reference positions are set for each of the counterweights CW1 to CW6. As shown in FIG. 3(A), 10 reference positions A1 to A10 are set on the outer circumference surface of the first counterweight CW1. As shown in FIG. 12(A), 10 reference positions A11 to A20 are set on the outer circumference surface of the second counterweight CW2. As shown in FIG. 3(B), 10 reference positions A21 to A30 are set on the outer circumference surface of the third counterweight CW3. As shown in FIG. 3(B), 10 reference positions A31 to A40 (not shown) are set on the outer circumference surface of the fourth counterweight CW4. Ten reference positions A41-A50 (not shown) are set on the outer peripheral surface of the fifth counterweight CW5. Ten reference positions A51-A60 (see FIG. 7 for only reference positions A58-A60) are set on the outer peripheral surface of the sixth counterweight CW6.
[0037] Specifically, as shown in Fig. 3(A), the reference positions A2 and A7 are the apex of the outer periphery of the first counterweight CW1. The reference positions A1, A3 to A6, and A8 to A10 are inflection points of the outer periphery of the first counterweight CW1. 3B, the reference positions A21 and A26 are the apex of the outer periphery of the third counterweight CW3. The reference positions A22 to A25 and A27 to A30 are inflection points of the outer periphery of the third counterweight CW3.
[0038] The bottom of the outer periphery of each of the counterweights CW1 to CW6 may be set as the reference position. Fig. 4 is a diagram for explaining eccentricity data for the material crankshaft 9. Fig. 5 is a diagram showing a state in which the material crankshaft 9 is held by the hole machining machine 5. 2, it is necessary to machine center holes 10A, 10B so that a design imbalance remains in the raw crankshaft 9. In this case, as eccentricity data for the raw crankshaft 9, eccentricity data (eccentricity data on the first surface side) of the first end face 9a of the raw crankshaft 9 and eccentricity data (eccentricity data on the second surface side) of the raw crankshaft 9 are required. In addition, the eccentricity data for the raw crankshaft 9 is two-dimensional data including a vertical component and a horizontal component.
[0039] The eccentricity data at the first end face 9a of the raw crankshaft 9 is defined as the first eccentricity data (V1, H1), and the eccentricity data at the second end face 9b of the raw crankshaft 9 (eccentricity data on the two faces) is defined as the second eccentricity data (V2, H2). "H1" and "H2" are the horizontal coordinates of the first and second eccentricity data, respectively. "V1" and "V2" are the vertical coordinates of the first and second eccentricity data.
[0040] 5, the hole drilling machine 5 further includes a support part 54 that supports the raw crankshaft 9, a moving means (not shown) such as an actuator that moves the support part 54 in order to control the attitude of the raw crankshaft 9, and a tool 55 that drills a center hole in the raw crankshaft 9 after the attitude control has been completed. The tool 55 includes a drill or the like. The hole drilling machine 5 is disposed in a processing station (not shown) separate from the measurement station.
[0041] A first center hole 10A is formed in a position corresponding to the first eccentricity data (H1, V1) in a first end face 9a of the raw crankshaft 9. A second center hole 10B is formed in a position corresponding to the second eccentricity data (H2, V2) in a second end face 9b of the raw crankshaft 9. This causes the rotation axis and the principal axis of inertia in the raw crankshaft 9 to coincide with each other.
[0042] FIG. 6 is a diagram for explaining the functional configuration of the center hole position determining device 3. As shown in FIG. The machine learning unit 371 is realized by executing the neural network program 37 installed in the ROM 36 of the control unit 31. The storage unit 32 includes a dataset storage unit 321 for storing a dataset used as teacher data in machine learning by the machine learning unit 371, and a trained model storage unit 322 for storing the trained model 104 after machine learning by the machine learning unit 371.
[0043] FIG. 7 is a diagram showing a model of the neural network 100 that is machine-learned by the machine learning unit 371. The neural network 100 in the example of Fig. 7 has one input layer 101, one output layer 102, and one or more hidden layers 103. In the example of the model in Fig. 7, there are two hidden layers 103, but there may be one hidden layer 103, or there may be three or more hidden layers 103. It may also be a convolutional neural network (CNN) having multiple hidden layers 103.
[0044] 7, the input layer 101 has 60 neurons 101a. The 60 neurons 101a correspond one-to-one to the 60 reference positions A1 to A60. 7, the output layer 102 has four neurons 102a. The four neurons 102a correspond one-to-one to the horizontal coordinate (H1) of the first eccentricity data, the vertical coordinate (V1) of the first eccentricity data, the horizontal coordinate (H2) of the second eccentricity data, and the vertical coordinate (V2) of the second eccentricity data, respectively.
[0045] Each hidden layer 103 has a large number of neurons 103a. The number of neurons included in one hidden layer 103 is, for example, 30 to 50. FIG. 8 is a diagram of the external shape data measuring instrument 2 viewed in the horizontal direction. The external shape data measuring instrument 2 further includes a workpiece holding and rotating mechanism 24. The workpiece holding and rotating mechanism 24 holds the workpiece crankshaft 9 to be measured in a horizontal position, and rotates the workpiece crankshaft 9 about a predetermined horizontal axis.
[0046] The material holding and rotating mechanism 24 is, for example, a chuck that comes into contact with the material crankshaft 9 and clamps the material crankshaft 9. The material holding and rotating mechanism 24 includes a turntable 24A, a holding unit 24B that is provided to rotate together with the turntable 24A and holds the material crankshaft 9, a clamping unit 24C that clamps the material crankshaft 9 between itself and the holding unit 24B, and a material rotation unit 25 that is coupled to the turntable 24A and rotates the turntable 24A to rotate the material crankshaft 9 about a horizontal rotation axis (a predetermined horizontal axis). The material rotation unit 25 includes, for example, a spin motor 26 (also see FIG. 1).
[0047] An encoder 27 is coupled to the spin motor 26, which detects the rotation angle of the output shaft of the spin motor 26. When the spin motor 26 rotates the output shaft, the raw material crankshaft 9 rotates about the rotation axis by an amount corresponding to the rotation angle of the output shaft. Therefore, by detecting the rotation angle with the encoder 27, the rotation angle position of the raw material crankshaft 9 can be detected.
[0048] The external shape data measuring device 2 further includes an external shape measuring sensor 28 and a sensor moving mechanism 29 . The outer shape measuring sensor 28 includes a distance sensor 29A that measures the distance to a measurement point on the outer circumferential surface of the counterweights CW1 to CW6 of the raw crankshaft 9 held by the raw material holding and rotating mechanism 24. The distance sensor 29A is, for example, a non-contact type displacement sensor. The distance sensor 29A includes a light-emitting unit including a light source, a light-receiving unit that receives light from the light-emitting unit, and a sensor housing that accommodates the light-emitting unit and the light-receiving unit. The distance sensor 29A irradiates light onto the outer peripheral surfaces of the counterweights CW1 to CW6 and receives the light reflected by the outer peripheral surfaces of the counterweights CW1 to CW6. The light source includes a laser light source, an infrared light source, an LED light source, etc. That is, the distance sensor 29A includes a laser displacement sensor, an infrared displacement sensor, an LED type displacement sensor, etc.
[0049] The sensor moving mechanism 29 moves the distance sensor 29A in the horizontal and vertical directions. The sensor moving mechanism 29 includes a motor and a ball screw mechanism, and / or a cylinder. The sensor moving mechanism 29 moves the distance sensor 29A along the axial direction of the raw material crankshaft 9 held by the raw material holding and rotating mechanism 24. FIG. 9 is a flow chart illustrating machine learning by machine learning unit 371.
[0050] The machine learning by the machine learning unit 371 is performed prior to the center hole position determination and drilling process described below. In machine learning, a plurality of sets of data sets of outer shape data P and eccentricity data Q are stored in the data set storage unit 321 (step S1: acquisition step). The outer shape data P is outer shape data of a learning crankshaft of the same type as the material crankshaft 9. The outer shape data P is position data corresponding to 60 reference positions A1 to A60 on the crankshaft. If the mass centering machine 6 is equipped with a configuration equivalent to an outer shape measurement sensor 28 and a sensor movement mechanism 29 (described later) of the outer shape data measuring machine 2, the outer shape data P can be actually measured by the mass centering machine 6. In this case, the outer shape data P from the mass centering machine 6 is provided to the center hole position determining device 3.
[0051] The eccentricity data Q is eccentricity data of the learning crankshaft (the crankshaft that is the measurement target of the external shape data P). The eccentricity data Q includes first eccentricity data which is eccentricity data on one surface side and second eccentricity data which is eccentricity data on a second surface side. The eccentricity data Q of the learning crankshaft is measured in the mass centering machine 6. Then, the eccentricity data Q from the mass centering machine 6 is provided to the center hole position determining device 3. The control unit 31 of the center hole position determining device 3 receives multiple combinations of the external shape data P and the eccentricity data Q for the learning crankshaft and stores them in the data set storage unit 321.
[0052] On the other hand, if the mass centering machine 6 is not equipped with the outer shape measurement sensor 28 and the sensor moving mechanism 29, for example, the outer shape data measuring machine 2 acquires the outer shape data P of the learning crankshaft in a manner similar to the outer shape data measurement (S4) described below. Then, the acquired outer shape data P is stored in the data set storage unit 321 via the main control device 4. Even in this case, the eccentricity data Q is acquired from the mass centering machine 6.
[0053] After acquiring the dataset, the machine learning unit 371 executes machine learning of the neural network 100 (step S2: learning process). The machine learning by the machine learning unit 371 is performed by a supervised learning method. The machine learning unit 371 reads the outer shape data P (position data at reference positions A1 to A60) and the eccentricity data Q corresponding to the outer shape data P stored in the dataset storage unit 321, and performs machine learning using these as input data and teacher data, respectively. The machine learning by the machine learning unit 371 is performed using all the datasets stored in the dataset storage unit 321.
[0054] In machine learning by the machine learning unit 371, multiple pieces of external shape data P are input to the input layer 101. Specifically, position data of each of reference positions A1 to A60 (distance data from a predetermined distance reference position) is input to a corresponding one of the 60 neurons 101a in the input layer 101. 7, values corresponding to a plurality of external shape data P are output from the output layer 102. Specifically, the horizontal coordinate (H1) of the first eccentricity data, the vertical coordinate (V1) of the first eccentricity data, the horizontal coordinate (H2) of the second eccentricity data, and the vertical coordinate (V2) of the second eccentricity data are output from corresponding neurons 102a among the four neurons 102a of the output layer 102. The machine learning is performed so that the error between the coordinates (V1, H1; V2, H2) of the output data and the coordinates (V1, H1; V2, H2) of the eccentricity data Q, which is a data set, is minimized. Through machine learning, the weighting values between neurons 101a in the input layer 101 and individual neurons 103a in the hidden layer 103, the weighting values between individual neurons 103a in the hidden layer 103 and neurons 102a in the output layer 102, the thresholds of individual neurons 103a in the hidden layer 103, and the thresholds of neurons 102a in the output layer 102 are adjusted.
[0055] The trained neural network 100, i.e., the trained model 104, is a trained model for estimating the positions of the center holes 10A, 10B corresponding to the outer shape of the material crankshaft 9, using the multiple pieces of outer shape data P as input data and the eccentricity data Q as teacher data. This trained model 104 is stored in the trained model storage unit 322 (step S3).
[0056] Fig. 10 is a flow chart showing the contents of the center hole position determination and hole drilling process executed in the hole forming system 1. Fig. 11 is a flow chart showing the contents of the measurement of the outer diameter data P by the outer diameter data measuring machine 2. Fig. 12 is a diagram showing the change in distance data accompanying the change in the rotational angle position of the material crankshaft 9. Fig. 13 is a flow chart showing the contents of the determination of the positions of the center holes 10A, 10B (S8 in Fig. 10).
[0057] Hereinafter, the center hole position determination and drilling process will be described with reference to Figures 1 to 3, 8 and 10. Figures 11 to 13 will also be referred to as appropriate. The main control device 4 causes the outer shape data measuring machine 2 to measure the outer shape data of the material crankshaft 9 (S4 in FIG. 10: measuring step). Specifically, as shown in Fig. 11, the raw material crankshaft 9 to be measured is carried into the contour data measuring machine 2 by a rotor transport device (not shown) (step S11 in Fig. 11). The raw material crankshaft 9 is delivered to the material holding and rotating mechanism 24 of the contour data measuring machine 2 in a predetermined angular attitude, and is held by the material holding and rotating mechanism 24 (step S12 in Fig. 11). After being held by the material holding and rotating mechanism 24, the raw material crankshaft 9 is clamped by the clamping unit 24C. This makes it possible for the raw material crankshaft 9 to rotate as the material holding and rotating mechanism 24 rotates.
[0058] After the rotor transport device (not shown) has been retracted, the control unit 21 of the external shape data measuring device 2 controls the workpiece rotation unit 25 to increase the rotation speed of the workpiece holding and rotating mechanism 24 to a predetermined measurement speed (for example, about 120 rpm) and maintains it at that measurement speed (Step S13 in FIG. 11: Start of rotation of workpiece crankshaft 9). As the workpiece crankshaft 9 rotates, the counterweights CW1 to CW6 of the workpiece crankshaft 9 also rotate about the horizontal rotation axis.
[0059] Moreover, the control unit 21 of the external shape data measuring machine 2 starts monitoring the output of the encoder 27. That is, it starts detecting the rotation angle position of the material crankshaft 9 (step S14 in FIG. 11). After that, the control unit 21 of the external data measuring device 2 controls the sensor moving mechanism 29 to move the distance sensor 29A from the standby position and place it at a predetermined initial measurement position. With the distance sensor 29A at the initial measurement position, the sensor unit of the distance sensor 29A faces the outer periphery of the first counterweight CW1 (more specifically, the central part in the axial direction of the outer periphery of the first counterweight CW1). The initial measurement position is a position close to the outer periphery of the rotating first counterweight CW1 to an extent that the distance can be detected by the distance sensor 29A.
[0060] Then, when the rotation angle position of the first counterweight CW1 reaches a predetermined start angle position, the control unit 21 starts outer circumference measurement by the distance sensor 29A. The fact that the rotation angle position of the first counterweight CW1 reaches the predetermined start angle position is detected by the coincidence of the rotation angle position of the material crankshaft 9 from the rotation angle detection device. The distance between the outer circumference of the rotating first counterweight CW1 and the distance sensor 29A is measured by the distance sensor 29A. The detection output from the distance sensor 29A is provided to the control unit 21. In addition, the detection output from the encoder 27 is also provided to the control unit 21. The control unit 21 simultaneously transmits the detection output of the distance sensor 29A and the detection output from the encoder 27 to the main control device 4.
[0061] Next, the control unit 21 controls the sensor moving mechanism 29 to place the distance sensor 29A at a position facing the outer periphery of the next counterweight, that is, at a position facing the outer periphery of the second counterweight CW2 (more specifically, at the center in the axial direction of the outer periphery of the second counterweight CW2) (step S15 in FIG. 11). After that, the outer periphery measurement is performed by the distance sensor 29A (step S16 in FIG. 11). The steps of S15 and S16 are repeated until the outer periphery measurements for all counterweights CW1 to CW6 are completed. That is, the outer periphery measurements for the six counterweights CW1 to CW6 are performed in order.
[0062] Then, when the measurement for the last (final) sixth counterweight CW6 is completed (YES in step S17 in FIG. 11), the control unit 21 controls the sensor moving mechanism 29 to retract the distance sensor 29A to the standby position (step S18 in FIG. 11). Then, the control unit 21 controls the raw material rotation unit 25 to stop the rotation of the raw material holding and rotating mechanism 24 (S19 in FIG. 11: Stopping rotation of raw material crankshaft). The control unit 21 also controls the clamping unit 24C to release the clamping of the raw material crankshaft 9 (step S20 in FIG. 11).
[0063] Thereafter, a rotor transfer device (not shown) enters the outer shape data measuring machine 2 and carries the measured raw crankshaft 9 out of the machine. The carried-out raw crankshaft 9 is transferred to the hole machining machine 5 and delivered to the support portion 54 of the hole machining machine 5 (step S5 in FIG. 10). Meanwhile, while the distance sensor 29A is performing the outer periphery measurement (steps S15 and S16 in FIG. 11), the detection output from the distance sensor 29A and the detection output from the encoder 27 are sequentially transmitted from the control unit 21 to the main control device 4. The control unit 41 of the main control device 4 checks the change in the distance data accompanying the change in the rotational angle position of the material crankshaft 9 based on these detection outputs. By connecting the changes in the distance data in the outer periphery measurement of each of the counterweights CW1 to CW6, the graph shown in FIG. 12(A) is obtained. FIG. 12(B) shows an enlarged view of the change in the distance data in the outer periphery measurement of the first counterweight CW1.
[0064] As explained in FIG. 3(A) and FIG. 3(B), the shortest distance to the arc C1 centered on the axis C changes depending on the rotation angle position of the counterweights CW1-CW6 (the rotation angle position of the raw crankshaft 9). Therefore, as shown in FIG. 12(A) and FIG. 12(B), the distance data, which is the detection output of the distance sensor 29A, also changes depending on the rotation angle position of the counterweights CW1-CW6 (the rotation angle position of the raw crankshaft 9). The value of the distance data increases as the distance between the distance sensor 29A and the outer periphery of the counterweights CW1-CW6 increases, that is, as the distance from the axis C to the outer periphery of the counterweights CW1-CW6 decreases. As for the outer periphery measurement of the first counterweight CW1, as shown in FIG. 12(B), the distance data is largest when the reference positions A5 and A9 are measured, and is smallest when the reference positions A2 and A7 are measured. Therefore, these reference positions A2, A5, A7, and A9 can be easily identified from the distance data. In addition, the other reference positions A1, A3, A4, A6 to A8, and A10 are also set at inflection points on the outer periphery of the first counterweight CW1, so these reference positions A1, A3, A4, A6 to A8, and A10 can be easily identified from the distance data. The same can be said for the other counterweights CW2 to CW5.
[0065] Then, the control unit 41 of the main control device 4 obtains the distance data value at each of the reference positions A1 to A60 (step S6 in FIG. 10), and transmits the collected distance data string (position data at the 60 reference positions A1 to A60) to the center hole position determining device 3. The control unit 31 of the center hole position determining device 3 uses the trained model 104 stored in the trained model storage unit 322 to determine the positions of the center holes 10A and 10B corresponding to the acquired external shape data P (S7 in FIG. 10: estimation step). Specifically, as shown in FIG. 13, a data string of measured distance data (external shape data) (i.e., position data at 60 reference positions A1 to A60) is input to the trained model 104 (S21 in FIG. 13). As a result, the coordinates (H1, V1) of the first eccentricity data and the coordinates (H2, V2) of the second eccentricity data are output from the trained model 104 (S22 in FIG. 13). The output eccentricity data may be converted according to the position of the hole processing machine 5, or may be output by adding stationary offset data. Then, the control unit 31 of the center hole position determining device 3 determines the drilling positions of the center holes 10A and 10B (S23 in FIG. 13). Then, the control unit 31 transmits the drilling positions to the main control device 4.
[0066] Based on the drilling position received from the main control device 4, the control unit 51 of the hole drilling machine 5 drills a center hole 10A in the first end face 9a of the raw crankshaft 9 that has been loaded into the hole drilling machine 5, and drills a center hole 10B in the second end face 9b (S8 in Figure 10). As described above, according to this embodiment, a trained model 104 for estimating the positions of the center holes 10A, 10B corresponding to the outer shape of the material crankshaft 9 is generated by machine learning using the multiple pieces of outer shape data P as input data and the eccentricity data Q as teacher data.
[0067] In addition, using the trained model 104 generated in advance by such machine learning, the positions of the center holes 10A, 10B corresponding to the outer shape data of the measured material crankshaft 9 are determined (estimated). Therefore, the positions of the center holes 10A, 10B located on the principal axis of inertia can be determined with high accuracy. Moreover, this embodiment has the following advantages over the method of Patent Document 1 (Japanese Patent No. 5994184): The center hole position determining system 7 only measures the external data of the material crankshaft 9, and does not actually measure the eccentricity data. Therefore, there is no need to rotate the crankshaft at high speed as in Patent Document 1, and there is no risk of the drive motor or device becoming larger.
[0068] In addition, because there is little slippage in the rotational direction between the rotating platen 24A and the holding unit 24B, there is no need for two synchronous motors as in Patent Document 1, which allows for cost reduction. Furthermore, there is no need for a complex clamping mechanism as in Patent Document 1, and a vibration frame with a similarly complex structure. This allows for cost reduction. Therefore, the positions of the center holes 10A, 10B located on the principal axis of inertia can be determined with high accuracy while reducing costs.
[0069] In addition, since the outer shape data P includes distance data, it is possible to acquire the outer shape data P as one-dimensional data. In this case, the outer shape data P can be handled more easily than when the outer shape data P is three-dimensional data. In addition, the center of mass of the material crankshaft 9 depends greatly on the outer peripheral shape of each of the counterweights CW1-CW6. Therefore, by using data on the reference position set on the outer periphery of the counterweights CW1-CW6 as the outer peripheral data P, the positions of the center holes 10A, 10B located on the principal axis of inertia can be determined with even greater accuracy.
[0070] Furthermore, since the reference positions A1 to A60 are set at inflection points on the outer periphery of the counterweights CW1 to CW6, the reference positions A1 to A60 can be easily specified. In addition, since the distance sensor 29A is moved in the axial direction of the material crankshaft 9, measurement can be performed with only one distance sensor 29A. Therefore, costs can be reduced compared to the case where distance sensors are provided in one-to-one correspondence with the counterweights CW1 to CW6.
[0071] In addition, the positions of the center holes 10A, 10B are determined using the trained model 104, which is generated using the measured (actually measured) outer shape data P as input data and the measured (actually measured) eccentricity data Q as output data, so that the positions of the center holes 10A, 10B located on the principal axis of inertia can be obtained more accurately. This allows the positions of the center holes 10A, 10B to be determined with even greater precision.
[0072] FIG. 14 is a diagram for explaining another embodiment of the present invention. In the above embodiment, it has been described that the outer shape data P as a data set used in machine learning is obtained by measuring (actually measuring) a learning crankshaft using the mass centering machine 6 and the outer shape data measuring machine 2. Also, it has been described that the eccentricity data Q as a data set used in machine learning is obtained by measuring (actually measuring) a learning crankshaft in the same way.
[0073] The embodiment in FIG. 14 differs from the embodiments in FIGS. 1 to 13 described above in that machine learning is performed using provisionally determined eccentricity data Q and external shape data P calculated by the data set calculation device 200. The data set calculation device 200 is configured by a computer. The data set calculation device 200 includes a control unit, a storage unit, and a communication unit. The control unit includes a CPU, RAM, and ROM. The storage unit is configured by a storage device such as a non-volatile memory or a hard disk drive. The communication unit is communicatively connected to the main control device 4 via a communication line (not shown). The communication unit may be provided so as to be communicatively connected to the external shape data measuring device 2, the center hole position determining device 3, etc.
[0074] 14, the outer shape data P used in the machine learning is calculated data obtained by the data set calculation device 200. When the radius, the existence angle, the axial position, and the like of each of the counterweights CW1-CW6 of the material crankshaft 9 are known, the data set calculation device 200 calculates the outer shape data P for each of the reference positions A1-A60 based on these. The calculation of the outer shape data P is performed based on drawing data.
[0075] In the embodiment of FIG. 14, machine learning is carried out using external shape data P obtained by calculation and eccentricity data Q corresponding to the external shape data P. Specifically, the data set calculation device 200 provisionally determines eccentricity data Q (provisional eccentricity data). If, for example, 25 types each of the first eccentricity data and the second eccentricity data are prepared, 625 pieces of eccentricity data Q can be prepared. Then, the data set calculation device 200 determines, by calculation, outer shape data P (i.e., 60 pieces of position data) corresponding to each eccentricity data Q. Then, the data set calculation device 200 provides combinations of the outer shape data P and the eccentricity data Q to the center hole position determination device 3. The data set storage unit 321 of the center hole position determination device 3 stores multiple combinations of the outer shape data P and the eccentricity data Q (625 combinations when the number of eccentricity data Q is 625).
[0076] Then, the machine learning unit 371 reads out the outer shape data P (position data at reference positions A1 to A60) and the eccentricity data Q stored in the data set storage unit 321, and performs machine learning using these as input data and teacher data, respectively. This generates a trained model 104 for estimating the positions of the center holes 10A, 10B corresponding to the outer shape of the material crankshaft 9. This machine learning is equivalent to the above-mentioned machine learning described with reference to Figures 6, 7, and 9, and therefore a detailed description thereof will be omitted.
[0077] For example, when machine learning is performed using 625 pairs of contour data P and eccentricity data Q, the number of learning times is 25×25=625. In other words, by provisionally determining the eccentricity data Q and calculating the contour data P corresponding to the provisionally determined eccentricity data Q, it is possible to obtain many combinations of the contour data P and the eccentricity data Q, and thus it is possible to perform machine learning many times.
[0078] The control unit 31 of the center hole position determining device 3 inputs a data string of distance data (outer shape data) obtained by measuring the material crankshaft 9 (i.e., position data at 60 reference positions A1 to A60) to the trained model 104 obtained in this manner. Then, the coordinates (H1, V1) of the first eccentricity data and the coordinates (H2, V2) of the second eccentricity data output from the trained model 104 are regarded as approximate eccentricity data. Then, based on this approximate eccentricity data, the control unit 31 of the center hole position determining device 3 determines the drilling positions of the center holes 10A and 10B.
[0079] As described above, according to the embodiment of FIG. 14, by increasing the number of eccentricity data Q (temporary eccentricity data) to improve the resolution and increase the number of learning results, the positions of the center holes 10A, 10B located on the principal axis of inertia can be determined with high accuracy. Although two embodiments of the present invention have been described above, the present invention may be embodied in further other forms.
[0080] 14 may include both machine learning using provisionally determined eccentricity data Q and calculated contour data P (hereinafter referred to as "machine learning using calculated data") and machine learning using actually measured contour data P and actually measured eccentricity data Q (hereinafter referred to as "machine learning using measured data"). Specifically, machine learning using actually measured data may be executed as additional learning after machine learning using calculated data.
[0081] By combining machine learning using calculated data with machine learning using actual measurement data, machine learning is performed based not only on geometric data but also on actual measurement data, so the positions of center holes 10A, 10B that are located on the principal axis of inertia can be determined with even greater accuracy. In particular, when the number of learning crankshafts (samples) that are the subject of actual measurement data is small, if machine learning is performed based only on the actual measurement data, the contents of the generated trained model 104 may lack accuracy near the eccentricity data Q where no actual measurement data exists.
[0082] In machine learning using calculation data, a large number of data sets can be prepared, so that an accurate trained model 104 can be generated in all areas (areas of eccentricity data Q). In addition, the positions of the center holes 10A, 10B determined by the center hole position determining system 7 according to the embodiment of Figures 1 to 13 and the center hole position determining system 7 according to the embodiment of Figure 14 may be corrected. Specifically, the unbalance of the material crankshaft 9, the positions of the center holes 10A, 10B of which have been determined by the center hole position determining system 7, is measured by an unbalance measuring device. The positions of the center holes 10A, 10B determined by the center hole position determining system 7 may then be corrected using a coefficient calculated based on the unbalance measurement results, or may be corrected using a neural network that uses the unbalance measurement results as input data.
[0083] In addition, in each of the above-described embodiments, the external shape data has been described as being position data of the outer circumferences of the counterweights CW1 to CW6, but it may also include position data of the outer circumference of the shaft portion (crank journal). In addition, in each of the above-described embodiments, when the raw crankshaft 9 is a symmetrical raw crankshaft, the eccentricity data for either the first surface side or the second surface side is sufficient.
[0084] In addition, in each of the above-described embodiments, a non-contact type displacement sensor is taken as an example of the distance sensor 29A, but a contact type displacement sensor such as a differential transformer may also be used as the distance sensor 29A. In each of the above-described embodiments, the external measurement sensor 28 is not limited to the distance sensor 29A, and may be any other means capable of measuring the outer periphery of the counterweights CW1 to CW6. An example of the external measurement sensor 28 is an imaging means such as a CMOS sensor.
[0085] In each of the above-described embodiments, the main control device 4 may be omitted. In this case, the devices such as the external shape data measuring device 2 and the center hole position determining device 3 communicate directly with each other. In addition, in each of the above-described embodiments, the center hole position determining device 3 may be integrated with the outer shape data measuring machine 2. In this case, the center hole position determining system 7 is realized by the outer shape data measuring machine 2.
[0086] The present invention is not limited to the contents of the above-described embodiment, and various modifications are possible within the scope of the claims. [Explanation of symbols]
[0087] 2: External data measuring device (measurement method) 3: Center hole position determination device (estimation means) 7: Center hole position determination system 9: Material crankshaft 10A: First center hole (center hole) 10B: Second center hole (center hole) 104: Trained model A1 to A10: 1st to 10th reference positions A11 to A20: 11th to 20th reference positions A60: 60th reference position CW1: First counterweight CW2: Second counterweight CW3: 3rd counterweight CW4: 4th counterweight CW5: 5th counterweight CW6: 6th counterweight P: External data Q: Eccentricity data
Claims
1. an acquiring step of acquiring outer shape data including position information of a plurality of reference positions set in an entire angular range in the circumferential direction on the outer periphery of each of a plurality of counterweights included in a raw crankshaft having a first end face on which a first center hole is to be formed and a second end face opposite to the first end face on which a second center hole is to be formed, and a data set including eccentricity data on one surface side of the first end face and eccentricity data on two surfaces side of the second end face; and a learning step of generating, by machine learning, a trained model for estimating positions of the first center hole and the second center hole corresponding to the outer shape of the material crankshaft, using the outer shape data as input data and the eccentricity data of the first face side and the eccentricity data of the second face side as teacher data.
2. a measuring means for measuring outer shape data including position information of a plurality of reference positions set in an entire angular range in the circumferential direction on an outer periphery of each of a plurality of counterweights included in a raw crankshaft having a first end face on which a first center hole is to be formed and a second end face on the opposite side to the first end face on which a second center hole is to be formed; and an estimation means for estimating the positions of the first center hole and the second center hole corresponding to the measured outer shape data using a trained model generated in advance by machine learning using the outer shape data as input data and eccentricity data on one surface side of the first end face and eccentricity data on a second surface side of the second end face as teacher data.
3. 3. The center hole position determining system according to claim 2, wherein said external shape data includes distance data which is a distance of said reference position from a predetermined distance reference position.
4. 4. The center hole location system of claim 3, wherein the reference position is set at at least one of an inflection, a top, and a bottom on the outer periphery of the counterweight.
5. a measuring step of measuring outer shape data including position information of a plurality of reference positions set in an entire circumferential angle range on the outer circumference of each of a plurality of counterweights included in a raw crankshaft having a first end face on which a first center hole is to be formed and a second end face on the opposite side to the first end face on which a second center hole is to be formed; a center hole position determination method including an estimation step of estimating the positions of the first center hole and the second center hole corresponding to the measured outer shape data using a trained model generated in advance by machine learning using the outer shape data as input data and eccentricity data on one surface side of the first end face and eccentricity data on a second surface side of the second end face as teacher data.
Citation Information
Patent Citations
Pressure-sensitive tablet device
JP1984094184A
Crank shaft unbalance correcting device
JP1998009342A
Method and device for determining center hole of crankshaft
JP2010031987A
Machine learning device for learning set value of processing program in machine tool and processing system
JP2018065211A
Center hole determination device
JP2018189491A
