Learning device of processing condition

Abstract

To provide a learning device for learning a processing condition for joining and melt-cutting a part.SOLUTION: A learning device comprises: processing result measuring means for measuring a processing result; processing result evaluation means for evaluating the processing result as a reward; learning storage means for storing the processing result and the reward; processing condition learning means for reading the processing condition from the learning result obtained by learning the processing result and reward; processing specification monitoring means for monitoring a processing specification; status monitoring means; and processing condition reading means for reading the processing condition from the learning storage means. The learning device measures the processing results, evaluates the processing result as a reward, stores the processing results and reward, prepares to read the processing condition from the learning result learned by the processing condition learning means and reads the processing conditions that satisfy a latest and current processing specification from the learning storage means to output the read processing condition to the processing means. The processing means performs joining or fusion cutting of a part under the processing condition.SELECTED DRAWING: Figure 2

Description

Technical Field

[0001] The present invention relates to a learning device for processing conditions, a processing device with a learning device, and a processing system with a learning device. More specifically, it relates to a learning device for joining conditions of two or more parts, a joining device with a learning device, and a joining system with a learning device, a learning device for fusing conditions of one or more parts, a fusing device with a learning device, and a fusing system with a learning device.

Background Art

[0002] Conventionally, as methods for joining two or more parts such as plastic and metal, (1) a high-frequency welding method in which a high-frequency current is applied to two or more parts and welded by dielectric heating, (2) a non-contact hot plate joining method in which a hot plate is brought close to the surfaces of two or more parts without contact for joining, (3) a vibration welding method in which two or more parts are pressed together while applying a low-frequency vibration of about 100 Hz for welding, (4) a laser joining method in which two or more parts are irradiated with a laser beam for joining, (5) an ultrasonic welding method for welding two or more plastic parts, (6) an ultrasonic metal joining method for solid-phase joining of two or more metal sheets, (7) an ultrasonic joining method for joining a metal electrode to the surface of a glass plate, etc. have been used.

[0003] Also, as methods for fusing one or more parts such as plastic, food, and rubber sheets, (a) a laser fusing method in which a laser beam is irradiated for fusing, (b) an ultrasonic fusing method in which a cutter vibrating ultrasonically is pressed against one or more parts for fusing, etc. have been used.

[0004] In the above (1) to (7), (a) and (b), energy such as vibration energy and thermal energy is applied to the parts to heat, melt, or solid-phase bond the parts to perform joining and fusing processes to produce products. Therefore, by changing the magnitude of the energy applied to the parts and the supply time of the energy, the joining conditions and fusing conditions can be found.

[0005] However, the bonding conditions and fusing conditions vary depending on the material, shape, and dimensions of the parts to be bonded or fused. Conventionally, when bonding or fusing new parts, new bonding conditions and new fusing conditions that meet the bonding specifications and fusing specifications have been experimentally determined each time.

[0006] On the other hand, these processing apparatuses such as bonding apparatuses and fusing apparatuses are required to have versatility. These processing apparatuses have a wide range of processing targets and a wide variety of processing contents. Therefore, one state is selected from the multiple states of energy that can be supplied, and the energy supply time is determined.

[0007] The following describes how to determine processing conditions such as conventional bonding conditions and fusing conditions, taking the conventional high-frequency welding method in (1) above and the high-frequency welding apparatus using the same as representative examples. Note that in the conventional bonding methods in (2) to (7) above and the conventional fusing methods in (a) and (b), many contents are the same as or similar to (1), including how to determine processing conditions such as desired bonding conditions and fusing conditions by changing the magnitude and supply time of the applied energy, so the description is omitted.

[0008] In the high-frequency welding method, when a strong high-frequency electric field is applied to a plastic material, film, sheet, etc., the polarity of the electrodes continuously changes at the molecular level, and collisions, vibrations, and friction at the molecular level occur inside the material, generating self-heat and causing the plastic material, film, sheet, etc. to fuse and weld.

[0009] As a method for controlling the high-frequency current, a method of setting the supply time of the high-frequency current supplied to the object to be welded (hereinafter referred to as the "welding time") to a predetermined value is generally used. However, if the welding time is uniformly controlled for objects to be welded with different dimensions and shapes, over-welding may occur, sparks may be generated, or insufficient welding may occur.

[0010] Fig. 42 shows the relationship between the welding time (S) and the current (A) when high-frequency welding is controlled only by the conventional welding time and plastic materials, which are workpieces with different optimal welding times, are welded. In Fig. 42, the welding time (S) is taken on the X-axis and the current (A) is taken on the Y-axis.

[0011] In Fig. 42, for three types of plastic materials (Material A, Material B, and Material C), which are workpieces with different optimal welding times, when a uniform welding time, that is, when the workpiece is welded until the welding end time T2, (1) for Material A, the workpiece is over-welded, and (2) for Material C, the workpiece is under-welded, are exemplified. In Fig. 42, Material B exemplifies normal welding.

[0012] In Fig. 42, the high-frequency current is applied uniformly until the welding end time (T2). However, for Material A (the graph of the dashed-dotted line in Fig. 42), the welding of the workpiece is already completed at the welding time (T1) before the welding end time (T2) from the welding start time (T0). From the welding time (T1) to the welding end time (T2), an excessive high-frequency current is applied to the workpiece. As a result, the plastic material (Material A) is over-welded. This not only wastes energy but also damages the workpiece.

[0013] Conversely, for Material C (the graph of the dotted line in Fig. 42), although the welding time is originally required until T3, the high-frequency current is stopped at the welding end time T2 even though the welding is not yet completed, resulting in under-welding.

[0014] Therefore, as another conventional control method, it is known to detect the change in the distance between two molds sandwiching workpieces with different optimal welding times by the current value supplied to the high-frequency welding processing unit of the high-frequency welding apparatus, and control the amount of high-frequency energy supplied to the high-frequency welding processing unit by detecting the current value (see, for example, Patent Document 1).

[0015] In this other conventional control method, welding is started with the point where the workpiece contacts the molds on both sides as the welding start point. As the welding progresses and the distance between the molds decreases, the current value supplied to the high-frequency welding processing unit increases. The inflection point where the current value converges after reaching its maximum is regarded as the welding completion point. In other words, the time when the current value passes through the current value (L reference value) at the welding start point, reaches its maximum, and then reaches the current value (H reference value) at the inflection point is used as the welding completion point.

[0016] Fig. 43 shows the relationship between the welding time (S) and the current (A) when welding workpieces A, B, and C made of plastic material by another conventional high-frequency welding control method. Similar to Fig. 42, the welding time (S) is taken on the X-axis and the current (A) is taken on the Y-axis.

[0017] In the example shown in Fig. 43, the three types of workpieces A, B, and C have the same welding start point TS, but the welding completion points are different, namely TF’, TF’’, and TF’’’, and the optimal welding times TA, TB, and TC are assigned respectively. By this, compared with the conventional example shown in Fig. 42, (1) the over-welding of the workpiece with a short optimal welding time is reduced, and (2) the under-welding of the workpiece with a long optimal welding time is reduced.

[0018] In Fig. 43, for workpieces with predetermined dimensions and shapes in advance, the H reference value, which is the current value at the welding completion point, is measured and examined and then set in the welding apparatus. During the operation, the change of the current is detected, and when the current once exceeds the H reference value, shows its maximum, and then reaches the H reference value again, the high-frequency output is controlled to stop, and the workpiece is being welded.

[0019] However, the work of measuring and examining the welding conditions of workpieces with predetermined dimensions and shapes in advance is time-consuming and requires repeated trial and error even for skilled workers. To weld a new workpiece, it was necessary to measure and examine the welding conditions each time.

[0020] Recently, artificial intelligence technology (hereinafter abbreviated as AI technology) has attracted attention. AI technology learns from data accumulated in the past. If AI technology is used, it is expected that there is a possibility that it will not be necessary to find new joining conditions or welding conditions every time new joining or welding is performed.

[0021] However, in the conventional joining method of two or more parts and the welding method of one or more parts, no practical and specific AI technology has been disclosed, nor is there any suggestion. There is a known learning device that learns the arc welding conditions of metals using AI technology. In this document, the update formula of the action value function Q(a, s) shown in FIG. 44, which is known in Q-learning Q(s,a)← Q(s,a)+α(r+γmaxQ(Snext,anext)―Q(s,a)) is only conceptually disclosed for applying to one state of arc welding and performing arc welding, and the detailed content of practical and specific technology is not disclosed (see Patent Document 2).

[0022] Regarding conventional AI technologies, as shown in FIG. 45, research on reinforcement learning, Q-learning, statistical learning (neural network), deep learning (deep learning), and deep reinforcement learning that combines deep learning and reinforcement learning is progressing (see Known Document 1).

[0023] Reinforcement learning learns, in a state S in a certain environment, by trial and error of an action a, which action a should be taken to maximize the action value as a reward. A simple example will be explained using FIG. 46. In FIG. 46, in the state (St), when four actions (at1), (at2), (at3), and (at4) are performed, Q values (action values) obtained by evaluating the results of the actions as rewards, Q(St, at1), Q(St, at2), Q(St, at3), and Q(St, at4) are obtained respectively. Reinforcement learning is the idea of selecting the action (a) that gives the maximum reward by calculating backward from the magnitude of the reward. In FIG. 46, among the four rewards, Q(St, at3) is the maximum value, so by using the action (at3), the maximum reward (Q value) can be obtained.

[0024] In Q-learning, as shown in FIG. 44, using the action value function Q(a, s) as the update formula, online learning is continued in a state S in a certain environment, where a new action a is taken to obtain a reward, increasing the learning experience and updating the action value function Q(a, s), and quickly reaching an action that can obtain a large reward (see Non-Patent Documents 1 and 2).

[0025] In deep learning, a multi-layer neural network as shown in FIG. 47(b), which stacks multiple perceptrons shown in FIG. 47(a), is created. When an action is input to the input layer of the multi-layer neural network, each input is weighted, learning is performed in the hidden layer, and a policy obtained by multiplying the maximum reward by a probability (the probability to be adopted) is output from the output layer.

[0026] In deep reinforcement learning, as shown in FIG. 48, when an action is input to the input layer of a multi-layer neural network in a certain state, each input is weighted, and a policy obtained by multiplying the action by a probability is output from the output layer through the hidden layer. Then, the probability of an action leading to a high reward is increased, and learning is repeated with the multi-layer neural network to reach an action that can obtain a large reward. And when the policy obtained by multiplying the action by a probability changes and the environment or the state changes, learning is performed taking these into account.

[0027] However, in the conventional AI technology, (1) in processing methods such as the joining method of two or more parts or the fusing method of one or more parts, no practical and specific AI technology using a learning device is disclosed. Also, (2) in a general-purpose processing device that can perform processing in multiple states, when starting processing, the state of the energy given to the workpiece is checked, and no AI technology is disclosed for selecting specific processing conditions as an action that can obtain the maximum reward in that state, that is, an action that satisfies the processing specifications, and performing processing. (3) No AI technology is disclosed for monitoring the state of energy supply during processing of a processing device, judging whether the state of processing using a learning device is normal or abnormal using a learning device, and taking appropriate measures in the case of abnormality.

Prior Art Documents

Patent Documents

[0028] [Patent Document 1] Japanese Patent Application Laid-Open No. 2002-370283 [Patent Document 2] Japanese Patent Application Laid-Open No. 2017-30014 [Non-Patent Document]

[0029] [Non-Patent Document 1] "An Introduction to Artificial Intelligence Learned through Illustrations" by Tadahiro Taniguchi, published by Kodansha, December 2020 [Non-Patent Document 2] "Reinforcement Learning and Deep Learning: Simulation in C Language" by Tomohiro Odaka, published by Ohmsha, October 2017 [Summary of the Invention] [Problems to be Solved by the Invention]

[0030] In processing methods such as conventional joining methods and cutting methods, particularly in joining methods for two or more parts or cutting methods for one or more parts, when joining or cutting new parts, new joining conditions or new cutting conditions that satisfy the joining specifications or cutting specifications have been experimentally determined each time. New parts often have different materials, dimensions, shapes, etc. from conventional joining or cutting target parts, and it has been rare to directly use the joining or cutting conditions obtained in the past. Each time, it has been necessary for workers to conduct experiments to obtain desirable joining or cutting conditions, which has been a burden.

[0031]

[0032] ​Since conventional general-purpose processing devices have a wide range of processing targets and processing contents, the processing conditions frequently change. If it can be guaranteed that connecting a learning device that learns the processing conditions always supports processing under the processing conditions that meet the processing specifications, it would be convenient.

[0033] The second problem of the present invention is to provide a learning device, a processing device with a learning device, and a processing system with a learning device that monitor the processing specifications during processing from the start of processing under the processing conditions that meet the processing specifications and always support processing under the processing conditions that meet the processing specifications. For example, even when the processing specifications are updated from the previous processing operation or changed during processing, the processing operation is performed under the processing conditions that meet the latest and current processing specifications.

[0034] Also, even if the materials, shapes, and dimensions of the parts to be processed such as joining and welding are the same, if there are variations within the tolerance, the supply amount of energy applied to the parts will change slightly. Therefore, even if processing conditions such as joining conditions and welding conditions that are considered to be optimal obtained through experiments in advance are used, if the work is done with only one processing condition, variations in the adhesion force and variations in the appearance of the cut surface will occur due to variations within the tolerance of the parts to be processed. There was a problem that the quality after processing was not stable.

[0035] Also, in conventional joining devices, even when a part outside the tolerance is accidentally set as the object to be welded, joining is performed, so there was a case where defective products with parts outside the tolerance joined were produced.

[0036] The third problem of the present invention is to determine the state when the energy supply is started even if the materials, shapes, and dimensions of the parts to be processed such as joining and welding are substantially the same and there are variations within the tolerance, and read the energy supply time that meets the processing specifications such as joining and welding under the processing conditions such as joining conditions and welding conditions from the learning device. And when a part outside the tolerance is set, no joining is performed so as not to produce defective products with parts outside the tolerance joined.

[0037] Furthermore, it would be convenient if the learning device itself could indicate the need for updated learning when it fails to find processing conditions such as joining conditions and cutting conditions that meet the processing specifications from the learning device that has learned the processing conditions such as joining conditions and cutting conditions.

[0038] A fourth problem of the present invention is to determine whether processing conditions such as joining conditions and cutting conditions that meet the processing specifications can be found from the learned learning device, and to indicate that updated learning is necessary when processing conditions such as joining conditions and cutting conditions that meet the processing specifications cannot be found.

[0039] A fifth problem of the present invention is to perform updated learning on processing conditions such as joining conditions and cutting conditions that meet the processing specifications when it is determined that processing conditions such as joining conditions and cutting conditions that meet the processing specifications cannot be found from the learned learning device, and to provide a processing device with a learning device that can perform processing such as joining two or more parts or cutting one or more parts using the updated processing conditions such as joining conditions and cutting conditions.

[0040] A sixth problem of the present invention is to check the energy state given when a processing device such as a general-purpose joining device or cutting device starts processing such as joining or cutting, and to read the energy supply time that meets the processing specifications for joining or cutting in that state from the learning device for processing conditions such as joining conditions and cutting conditions, so as to perform processing such as joining work or cutting work.

[0041] A seventh problem of the present invention is to monitor the energy state during processing of the processing device, and when the energy state changes to an abnormal state, to indicate that the change to the abnormal state has occurred, stop the processing, and ensure the safety of the processing device.

Means for Solving the Problems

[0042] To achieve the above object, the learning device of the present invention includes at least (1) "processing result measurement means" for measuring a processing result, (2) "processing result evaluation means" for evaluating the processing result as a reward, (3) "learning storage means" for storing the processing result and the reward obtained by evaluating the processing result, (4) "processing condition learning means" for organizing the learning result evaluated based on the processing result so that the processing conditions can be read out, (5) "processing specification monitoring means" for monitoring the processing specification, and (6) "processing condition reading means" for reading the processing conditions from the learning storage means. In advance, the processing result is measured by the processing result measurement means of (1), the processing result is evaluated as a reward by the processing result evaluation means of (2), the processing result and the reward are stored in the learning storage means of (3), the learning result evaluated based on the processing result is organized by the processing condition learning means of (4) so that the processing conditions can be read out. When a processing specification is input to the processing means connected to the learning device, the processing specification is monitored by the processing specification monitoring means of (5), and the processing condition reading means of (6) reads out the processing conditions that satisfy the latest and current processing specification from the learning storage means and outputs the read processing conditions to the processing means. The learning device is configured as a processing condition learning device that outputs the read processing conditions to the processing means.

[0043] As a result, when a processing specification is input to the processing specification input means, the processing conditions such as the joining conditions or cutting conditions that satisfy the processing specification are read out from the learning storage means and output to the processing means. In the processing means, the processing conditions are stored in the "main storage means" and processing is performed using the stored processing conditions.

[0044] In addition, after the processing starts under the processing conditions that satisfy the processing specification, during the processing, the processing specification is monitored and it is constantly supported to perform the processing under the processing conditions that satisfy the processing specification.

[0045] In addition, when there are variations in the material, shape, and dimensions of the parts to be processed, it is configured to learn the processing conditions according to the variations and read out the processing conditions that satisfy the processing specification.

[0046] Furthermore, it is configured to determine whether processing conditions that satisfy the processing specifications can be found in the learning memory means, and when processing conditions that satisfy the processing specifications cannot be found, to present that update learning is necessary.

[0047] When it is determined that processing conditions that satisfy the processing specifications cannot be found in the learning memory means, it is configured to perform update learning on the processing conditions that satisfy the processing specifications and read out the updated processing conditions.

[0048] Furthermore, a "state monitoring means" for monitoring the state of the energy applied to the workpiece of the processing device connected to the learning device is provided, and the actions and rewards in the energy state when the processing device starts processing are learned. When the state when the processing device starts processing is confirmed, it is configured to read out the processing conditions that satisfy the processing specifications in that state.

[0049] Then, by the above state monitoring means, the state of the energy applied to the workpiece of the processing device after processing starts is monitored. When the state of the processing device becomes abnormal compared to the state assumed by the learning device, this is presented and the processing of the processing device is configured to be stopped.

[0050] Also, a learning device is configured in which the processing conditions to be learned are the joining conditions of two or more parts.

[0051] Also, a learning device is configured in which the processing conditions to be learned are the fusing conditions of one or more parts.

[0052] And, in the processing device with a learning device that connects any one of the above learning devices by a communication line, at least (a) a "processing specification input / output means" for inputting processing specifications, (b) a "main memory means", and (c) a "main control means" are provided. The processing conditions read from the learning memory means of the connected learning device are stored in the main memory means, and the processing operation using the stored processing conditions is made to be performed by the main control means, thereby configuring the processing device with a learning device.

[0053] In addition, the processing means constituting the processing system is provided with at least (a) "processing specification input / output means" for inputting processing specifications, (b) "main storage means", and (c) "main control means", and stores the processing conditions read from the learning storage means of the connected learning device in the main storage means, and causes the main control means to perform a processing operation using the stored processing conditions, thereby configuring a processing system with a learning device.

Effect of the Invention

[0054] In the present invention, a method and a learning device for learning processing conditions when joining two or more parts, cutting one or more parts, etc. are used to perform processing such as joining two or more parts or cutting one or more parts.

[0055] And the present invention monitors the processing specifications during processing from the start of processing under processing conditions that satisfy the processing specifications, and always supports processing under processing conditions that satisfy the processing specifications. Therefore, for example, even when the processing specifications are updated from the previous processing operation or changed during processing, the processing operation can be performed under processing conditions that satisfy the latest and current processing specifications.

[0056] In addition, the present invention uses a method and a learning device for learning processing conditions such as joining conditions and cutting conditions that satisfy the processing specifications when the materials, shapes, and dimensions of the parts to be joined are substantially the same and there are variations within the tolerance, and can perform processing such as joining two or more parts or cutting one or more parts. Also, when a part outside the tolerance is set as the object to be welded, joining can be prevented from being performed so as not to produce defective products with parts joined outside the tolerance.

[0057] Furthermore, the present invention can determine whether joining conditions or cutting conditions that satisfy the processing specifications are found in the learned learning device, and can present that update learning is necessary when joining conditions or cutting conditions that satisfy the processing specifications are not found.

[0058] When the processing conditions such as bonding conditions and fusing conditions that satisfy the processing specifications cannot be found from the learned learning device, the present invention updates and learns the processing conditions such as bonding conditions and fusing conditions that satisfy the processing specifications, and uses the updated processing conditions such as bonding conditions and fusing conditions to perform processing such as bonding two or more parts or fusing one or more parts.

[0059] In addition, when a processing device such as a general-purpose bonding device or fusing device starts processing such as bonding or fusing, the present invention checks the energy state applied at that time, and performs processing operations such as bonding operations and fusing operations under processing conditions that satisfy the processing specifications in that state.

[0060] Furthermore, the present invention monitors the energy state during the processing of the processing device. When the energy state changes to an abnormal state, it presents that the state has changed to an abnormal state, stops the processing, and can ensure the safety of the processing device.

Brief Description of the Drawings

[0061]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

Figure 11

Figure 12

Figure 13

Figure 14

Figure 15

Figure 16

Figure 17

Figure 18

Figure 19

Figure 20

Figure 21

Figure 22

Figure 23

Figure 24

Figure 25

Figure 26

Figure 27

Figure 28

Figure 29

Figure 30

Figure 31

Figure 32

Figure 33

Figure 34

Figure 35

Figure 36

Figure 37

Figure 38

Figure 39

Figure 40

Figure 41

Figure 42

Figure 43

Figure 44

Figure 45

Figure 46

Figure 47

Figure 48

Embodiments for Carrying Out the Invention

[0062] Regarding the embodiments for implementing the learning device for processing conditions, the processing device with a learning device, and the processing system with a learning device of the present invention, in the first to sixth embodiments, after explaining the learning device for joining conditions of two or more parts, the joining device with a learning device, and the joining system with a learning device, the embodiments for implementing the learning device for welding conditions of one or more parts, the welding device with a learning device, and the welding system with a learning device will be explained in the seventh embodiment.

[0063] (First Embodiment) In the first embodiment, a high-frequency welding method using a learning device for learning the joining conditions of two or more parts and a high-frequency welding device 301 with a learning device will be explained. In FIG. 1, an external view of the high-frequency welding device 301 with a learning device is shown. In FIG. 2, the internal configuration of the high-frequency welding device 301 with a learning device is shown. The high-frequency welding device 301 with a learning device arranges the learning device 100 to the right of the high-frequency welding means 1 and connects the two with a communication line 120 for integrated control.

[0064] In FIG. 1, the high-frequency welding means 1 has a table 70, an upper mold 2 and a lower mold 3 that sandwich the workpiece 4 in the vertical direction, a high-frequency current supply means 6 that passes a high-frequency current through the workpiece 4 sandwiched between the upper mold 2 and the lower mold 3, a state output means 13 that outputs to the learning device 100 the state in which the high-frequency current is supplied by the high-frequency current supply means 6, a main memory means 11 that stores information necessary for controlling welding conditions and the like, a welding specification input / output means 12 that inputs and outputs welding specifications, a main control means 10 that controls the entire high-frequency welding means 1, an air cylinder 5, and a lever 71 that operates the air cylinder 5.

[0065] In the high-frequency welding means 1, when the lever 71 is pushed down, the upper surface of the workpiece 4 placed on the lower mold 3 on the table 70 is pressed by the upper mold 2 with the force of the air cylinder 5. Then, a high-frequency current is passed through the workpiece 4 sandwiched between the upper mold 2 and the lower mold 3 by the high-frequency current supply means 6, and the workpiece 4 is dielectrically heated and welded.

[0066] As shown in FIG. 2, the learning device 100 is provided with a welding result measuring means 130 such as a thickness measuring means 101, an appearance inspection means 102, and a welding strength measuring means 103. And a welding result evaluation means 104 that evaluates the value measured by the welding result measuring means 130 as a reward, a learning memory means 105 that stores the welding result and the reward obtained by evaluating the welding result, a welding condition learning means 106 that organizes the learning result evaluated based on the welding result so that the welding conditions can be read out, an I / O means 107 that inputs and outputs the learning plan of the AI technology, specifically the software of the learning program, a welding condition reading means 108 that reads out the welding conditions that satisfy the welding specifications from the learning memory means 105, and a specification monitoring means 109A and a state monitoring means 109B.

[0067] As shown by the two-dot chain line frame in FIG. 2, the learning device 100 is divided into two parts: a lower part 100a and an upper part 100b. The lower part 100a includes a thickness measuring means 101 which is a welding result measuring means 130, an appearance inspection means 102, a welding strength measuring means 103, etc. The upper part 100b includes a welding result evaluation means 104, a learning memory means 105, a welding condition learning means 106, an I / O means 107, a welding condition reading means 108, and a specification monitoring means 109A and a state monitoring means 109B. The places indicated by the black dot ● on the communication line 120 connecting each means in FIG. 2 show that they are connected by connectors 101a, 102a, 103a, 110a, 110b. The connectors 101a, 102a, 103a, 110a, 110b are detachable, and for example, the configuration of the welding result measuring means 130 can be arbitrarily reconfigured.

[0068] Also, by connecting the high-frequency welding means 1 connected to the learning device 100 by the connector 110b to other processing devices, it can be configured as another processing device with a learning device. As will be described later with reference to FIG. 24, instead of the connectors 101a, etc., a configuration in which the learning device and the processing means are spatially separated using wireless transmission and reception means 601 to 609 can also be adopted.

[0069] A learning plan is input to the learning device 100 from the I / O means 107. As the learning plan to be input, software of learning methods such as reinforcement learning and Q-learning, which are AI technologies, is selected as necessary, and the software to be used is stored in the learning memory means 105. In the first embodiment, for the sake of understanding the invention, it will be described as an example that software of a simple reinforcement learning method is stored.

[0070] In the learning device 100, the welding result evaluation means 104 evaluates the welding result measured by the welding result measurement means 130 as a reward, and stores the information obtained by evaluating the welding result in the learning memory means 105. The welding condition learning means 106 learns the welding conditions from the stored information in the learning memory means 105, and generates the learning result as a learned model. Then, the welding condition reading means 108 is configured to be able to read out the welding conditions that satisfy the welding specifications from the learned model of the learning memory means 105.

[0071] Even when the welding specifications are updated from the previous welding operation or changed during welding, the specification monitoring means 109A monitors the latest and current welding specifications, and reads out the welding conditions that satisfy the latest and current welding specifications from the learned model of the learning memory means 105.

[0072] The state monitoring means 109B obtains information on the state of the high-frequency current that the high-frequency welding means 1 has started to pass from the state output means 13 of the high-frequency welding means 1, and transmits it to the welding condition reading means 108. The welding condition reading means 108 is configured to read out the welding conditions that satisfy the welding specifications in the state of the high-frequency current that has started to flow from the learned model of the learning memory means 105. In addition, the state monitoring means 109B monitors whether the state of the high-frequency current during high-frequency welding is normal or abnormal.

[0073] Fig. 3 shows the relationship between the transition of the anode current of the high-frequency welding apparatus 301 with a learning device according to the first embodiment and the welding strength after welding. The left vertical axis (left Y-axis) in Fig. 3 indicates the value of the high-frequency current flowing through the workpiece 4 sandwiched between the upper mold 2 and the lower mold 3, that is, the anode current value (A), and the right vertical axis (right Y-axis) indicates the welding strength (F) after welding, in other words, the force required to peel off the workpiece (peeling force). When the high-frequency current starts to flow from the high-frequency current supply means 6, an anode current starts to flow through the workpiece 4 sandwiched between the upper mold 2 and the lower mold 3. As the anode current flows through the workpiece 4, the current value gradually increases little by little. In other words, the current rises little by little. Even when the anode current flows through the workpiece 4, the heat generation amount is not sufficient until a certain measurement reference current value (It), so the workpiece 4 does not melt / fuse and no welding occurs. After that, when the anode current continues to flow, the workpiece 4 melts / fuses and welding starts. And as the welding time transitions to TA1, TA2, TA3, TA4, the welding strength of the welded workpiece 4 increases respectively like FA1, FA2, FA3, FA4.

[0074] As shown in Fig. 3, the curve of the current that supplies the high-frequency current reaching the target current value (IWA) after the high-frequency current, that is, the anode current, rises can be said to be one state "State A". In "State A", welding is performed along the current curve of "State A" with a small energy, so-called power.

[0075] Fig. 4 shows the current curve "State D" when the target current value (IWD) of the high-frequency current is made larger than that in Fig. 3. In "State D", welding is performed along the current curve of "State D" with a large energy, so-called power. In the case of "State D", compared with the case of "State A", the rise time (tD) is short, and the welding strength of the workpiece 4 becomes a large value like FD1, FD2, FD3, FD4.

[0076] The present invention, for example, for "State A", as shown in FIGS. 5(a) to (d), welds for each welding time (TA1, TA2, TA3, TA4), and individually measures the welding strength (FA1, FA2, FA3, FA4) after welding using the welding result measuring means 130. Then, the welding result evaluation means 104 evaluates the welding result as a reward, and stores it in the learning memory means 105. And the welding condition learning means 106 learns the welding conditions from the welding result evaluated as a reward, and generates the learning result as a learned model. And when a welding specification is input, the welding time can be read out from the learned model as a welding condition that satisfies the welding specification.

[0077] If the welding strength FA1 is input as a welding specification to the welding specification input / output means 12 of the high-frequency welding means 1, the learning device 100 calculates the welding time TA1 as an action by inverse calculation from the welding strength FA1 of the learned model. If the welding strength FA2 is input as a welding specification, the welding time TA2 as an action is read out by inverse calculation from the welding strength FA2 of the learned model. If the welding strength FA3 is input as a welding specification, the welding time TA3 as an action is read out by inverse calculation from the welding strength FA3 of the learned model. The same applies when the welding strength FA4 is input as a welding specification.

[0078] In addition, when values between the above four welding strengths FA1, FA2, FA3, and FA4 are input as welding specifications, by performing regression analysis, which is often used in AI technology for the relationship between welding time and welding strength, and formulating it into a mathematical formula, the welding time corresponding to the input welding strength is calculated and read out. Also, by increasing the number of data to be learned, the welding time that satisfies the specification can be read out even for a detailed welding specification.

[0079] For understanding the invention, FIG. 6 is shown as a simple example of reinforcement learning. In FIG. 6(a), a tree is shown when the curve of the anode current flowing through the high-frequency welding device 301 with a learning device according to the first embodiment is taken as one "state", the welding time is taken as an "action", and the obtained welding strength is taken as a Q value (action value) as a "reward".

[0080] In FIG. 6(a), the states of the high-frequency welding apparatus are "state A" with low energy, i.e., so-called power, and "state D" with high power. In "state A", if the welding times are TA1, TA2, TA3, and TA4 as "actions", Q-values, Q(A,TA1), Q(A,TA2), Q(A,TA3), and Q(A,TA4) are obtained as "rewards". Here, the Q-value is shown as Q(A,TA1). A in the parentheses indicates that the state is "state A". TA1 indicates that the action is "welding time TA1".

[0081] Similarly, in "state D", it is shown that if the welding times are TD1, TD2, TD3, and TD4 as "actions", Q-values, Q(D,TD1), Q(D,TD2), Q(D,TD3), and Q(D,TD4) are obtained as "rewards".

[0082] If the welding times as respective "actions" and the Q-values as "rewards" in "state A" and "state D" of FIG. 6(a) are measured corresponding to each other, as shown in FIG. 6(b), when it is desired to obtain a certain "reward" in a certain "state", the "action" can be found by back-calculating from the "reward".

[0083] FIGS. 7(a) to (d) show an image of measuring the welding strengths (FA1, FA2, FA3, FA4) when welding is performed by randomly setting the welding times (TA1, TA2, TA3, TA4) for four samples with the high-frequency welding apparatus 301 with a learning device according to the first embodiment of the present invention.

[0084] In FIGS. 7(a) to 7(d), an upper mold 2 with a ring-shaped end face is pressed against the upper surfaces of two stacked elongated plastic sheets 4a and 4b such as vinyl chloride sheets, and welded in a ring shape with random welding times. For example, in FIG. 7(a), four ring-shaped welded portions are respectively formed with the welding times being TA1, TA2, TA3, and TA4 from the right. In FIG. 7(b), the welding times are TA4, TA3, TA2, and TA1 from the right, in FIG. 7(c), the welding times are TA2, TA4, TA1, and TA3 from the right, and in FIG. 7(d), the welding times are TA3, TA1, TA4, and TA2 from the right, and four ring-shaped welded portions are respectively welded. Then, after welding, the upper plastic sheet 4a is vertically lifted and peeled off from the lower plastic sheet 4b to measure the welding strength. The welding result measurement data is stored in the learning storage means 105.

[0085] As shown in FIGS. 7(a) to 7(d), the learning device 100 stores the measurement results measured in the learning storage means 105, evaluates them as rewards by the welding result evaluation means 104, and the welding condition learning means 106 learns the relationship between the welding time as an action and the reward, summarizes the learning results in a table, and sets the learning results as a learned model. Then, when a welding specification is input, the welding condition reading means 108 can read out the welding conditions that satisfy the welding specification.

[0086] FIGS. 8(a) to 8(c) show the stored contents stored in the learning storage means 105. In FIG. 8, the welding time and the welding strength are stored as a pair. Also, in FIGS. 8(a) and 8(b), trapezoidal, round, and square frames are attached to each pair for easy understanding.

[0087] FIG. 8(a) stores the measured data in the order of measurement. For example, in data 1, it shows that from the right, a pair of welding strength FA11 and Q value, Q(A, TA11), a pair of welding strength FA21 and Q value, Q(A, TA21), a pair of welding strength FA31 and Q value, Q(A, TA31), and a pair of welding strength FA41 and Q value, Q(A, TA41) are stored. In data 2 to data 4, each pair is randomly arranged in the order of measurement.

[0088] In FIG. 8(b), the data, which is the stored content of the learning storage means 105 by the welding condition learning means 106, is arranged in a form where the magnitude of the welding force is sorted in ascending order from a random form. For example, in Data 1, compared with FIG. 8(a), the order is reversed, and from left to right, there are pairs of welding force FA11 and Q value (A, TA11), pairs of welding force FA21 and Q value (A, TA21), pairs of welding force FA31 and Q value (A, TA31), and pairs of welding force FA41 and Q value (A, TA41). For Data 2 to Data 4, each pair is arranged in the same order as Data 1.

[0089] And in FIG. 8(c), the welding condition learning means 106 calculates the average value of the welding force for each welding time for the data arranged in FIG. 8(b), and creates a table from which the welding time can be read by inverse calculation from the welding force. In the present invention, the content of FIG. 8(c), which is the learning result learned from Data 1 to Data 4, is used as the learned model. And the welding time can be read out as the welding condition that satisfies the welding force of the welding specification.

[0090] Figure 9 shows the details of the learning procedure in the learning device 100 as a flowchart. In the flowchart of Figure 9, a learning plan is input by the I / O means 107, stored in the learning storage means 105, and then learning is started (step ST1). A target current value is set so that the input welding specification is satisfied. Note that since a learning device that has not learned at all is ignorant, it is advisable to include and input, as an initial value, a target current value that is expected to satisfy the welding specification in the welding specification (step ST2). An anode current is started to flow (step ST3). Here, it is asked whether to check the state during welding (step ST4). This is because when the high-frequency welding means 1 has only a single energy supply state, there is only one energy state, so there is no need to check the state during welding. In that case, the next step ST5 is skipped (NO in step ST4). When the high-frequency welding means 1 is a general-purpose processing device that can select one state from a plurality of energy supply states, step ST5 is performed. In step ST5, the rise time until the anode current reaches the measurement reference current value is measured and stored in the learning storage means 105. Then, the anode current is flowed for each welding time to perform welding (step ST6). The welding force as a reward for the welding time as an action is measured, correspondence data between the welding time and the welding force is created, and stored in the learning storage means 105 (step ST7). The welding condition learning means 106 learns the welding conditions from the information stored in the learning storage means 105, generates the learning result as a learned model, and stores it in the learning storage means 105 in a form that allows the welding conditions to be read from the learned model (step ST8).

[0091] Then, it is asked whether data on the welding force corresponding to a predetermined number of welding times has been obtained for one state (i.e., one current curve) (step ST9). If the predetermined number of data has not been obtained, the process returns to step ST3 (NO in step ST9). If the predetermined number of data has been obtained (YES in step ST9), it is asked whether to update the target current value and learn about other states (step ST10). When updating the target current value based on the input learning plan, the process returns to step ST2 (YES in step ST10). When not updating the target current value (NO in step ST10), it is asked whether learning has been performed as planned (step ST11). If learning has not been performed as per the learning plan, the process returns to step ST2 (NO in step ST11). If learning as per the plan has been completed (YES in step ST11), the process proceeds to step ST12 to end the learning.

[0092] Note that steps ST8 and ST9 may be interchanged so that after obtaining a predetermined number of data, learning is performed and the learning results are stored in the learning storage means 105.

[0093] When performing update learning, starting from step ST34, the process enters step ST1 and performs the steps below step ST1. By performing the procedure of the flowchart in FIG. 9, for one or more states, data on the welding force for each welding time, which is the respective action, is obtained, the data on the welding force is stored as a reward, the stored data is learned, the learning results are made in a form that can be read out as welding conditions, and stored in the learning storage means 105.

[0094] Figure 10 shows the procedure of the welding operation as a flow chart. In Figure 10, first, welding specifications are input and the welding operation is started (step ST20). The workpiece to be welded is set (step ST21). It is asked whether to check the state (step ST22). As also explained in the flow chart of Figure 9, when the high-frequency welding means 1 has only a single energy supply state, since the energy state is one, there is no need to check the state during welding. At that time, steps ST23 and ST24 are skipped for this reason (NO in step ST22). When the high-frequency welding means 1 can select one state from a plurality of energy supply states, steps ST23 and ST24 are performed. In step ST23, the rise time of the anode current is measured. Then, the state of the energy that the high-frequency welding means 1 has started to supply is checked (step ST24). And in the state of the energy that has started to be supplied, it is asked whether there are welding conditions that satisfy the latest and current welding specifications confirmed by the specification monitoring means 109A among the learned models of the learning storage means 105 (step ST25). If there are welding conditions, they are read from the learning storage means 105 (step ST26), stored in the main storage means 11, and welding is performed by the high-frequency welding means 1 (step ST27). The state monitoring means 109B monitors whether the state during welding is abnormal (step ST28). When the state is normal (YES in step ST28), it is confirmed that the welding time for ending the welding operation has been reached (step ST29), and the welding operation is ended (step ST31). In addition, when the state is abnormal in step ST28, an abnormality is presented and welding is stopped (step ST30). Then, the welding operation is ended (step ST31).

[0095] Also, when no welding conditions are found in the learned models of the learning storage means 105 in step ST25 (NO in step ST25), "Update learning is required" is presented (step ST32). Then, it is asked whether to perform update learning (step ST33). When performing update learning, starting from step ST34, it jumps to the flow chart of Figure 9, and update learning is performed according to the procedure below step ST1 of Figure 9.

[0096] The content of the learning memory means 105 when the welding conditions are read in step ST26 is shown in FIG. 11. Note that FIG. 11 is shown in a simplified manner so that the image can be grasped for understanding the invention.

[0097] The case of FIG. 11(a) will be described. First, when the welding force is input as "FA2" as the welding specification (step ST20). When the rise time is measured (step ST23), since the rise time is "TA", it is confirmed that the state is "state A" (step ST24). Since the welding force that satisfies the welding specification in "state A" is "FA2" in the table, the welding time "TA2" paired with the welding force "FA2" is read as the action for obtaining the welding force "FA2" of the welding specification, that is, the welding time of the welding conditions.

[0098] Similarly, in the case of FIG. 11(b), when the welding force is input as "FD3" as the welding specification (step ST20). When the rise time is measured (step ST23), since the rise time is "TD", it is confirmed that the state is "state D" (step ST24). Since the welding force that satisfies the welding specification in "state D" is "FD3" in the table, the welding time "TD3" paired with the welding force "FD3" is read as the action for obtaining the welding force "FD3" of the welding specification, that is, the welding time of the welding conditions.

[0099] In the case of Fig. 11(c), when the welding force is input as the upper limit value "FA3" and the lower limit value "FA2" as the welding specification (step ST20). When the rise time is measured (step ST23), since the rise time is "TA", it is confirmed that the state is "state A" (step ST24). Since the welding forces that satisfy the welding specification in "state A" are "FA3" and "FA2" in the table, the welding time "TA3" paired with the welding force "FA3" and the welding time "TA2" paired with the welding force "FA2" are read as the actions for obtaining the welding forces "FA3" and "FA2" of the welding specification, that is, the upper limit welding time "TA3" and the lower limit welding time "TA2" of the welding conditions. When the upper limit welding time "TA3" and the lower limit welding time "TA2" are determined, the welding time within the range sandwiched between the upper limit welding time "TA3" and the lower limit welding time "TA2" is used as the welding condition. Fig. 12 shows the range of the welding time (TA2~TA3) specified as the welding condition when the upper limit welding force and the lower limit welding force are input as the welding specification in the case of Fig. 11(c).

[0100] The update learning will be described with reference to Fig. 13. For example, when learning has already been performed from the welding forces FA1 to FA4 and FD1 to FD4, when the welding force input in the welding specification is a value smaller than FA1 or a value larger than FD4, the welding conditions that satisfy the specification cannot be read from the learned model. When the welding force input as the welding specification is in step ST25 of Fig. 10 and the welding conditions are not in the learning storage means 105 (NO in step ST25), "update learning is required" is displayed in step ST32. And when performing update learning (YES in step ST33), it jumps from step ST34 of Fig. 10 to step ST1 of Fig. 9.

[0101] In step ST1 of Fig. 9, since the learning plan has already been input, update learning is started and it proceeds to step ST2. At this time, as the learning device 100, in light of the welding specification, the target current value is set in the direction in which the welding force that satisfies the welding specification can be obtained (step ST2). Then, according to the flowchart of Fig. 9, steps ST3 and below are performed to perform a learning operation to obtain the welding conditions under which the welding force that satisfies the processing specification can be obtained.

[0102] In FIG. 13, as indicated by FH1 to FH4, an example of updated learning of welding conditions for obtaining a welding force smaller than FA1 was shown. From the result of the updated learning in FIG. 13, the welding time as the welding conditions for obtaining welding forces FH1 to FH4 smaller than FA1 is obtained.

[0103] After the updated learning, if returning to step ST20 in the flowchart of FIG. 10 and resuming the welding operation, assuming that there are welding conditions (YES in step ST25), the welding conditions updated through learning are read from the learned model of the learning storage means 105 (step ST26). Then, a new welding condition that did not exist before the updated learning, for example, the welding time TH1 corresponding to FH1, is read and welding is performed (step ST27), and the welding operation is terminated (step ST31).

[0104] Thus, in the first embodiment of the present invention, learning is performed using a learning device for the joining conditions when joining two or more parts. When a welding specification is input, the welding conditions that satisfy the input welding specification are read from the learned model of the learning device, and two or more parts are joined under the read welding conditions.

[0105] Thereby, the first problem of the present invention, that is, a method and a learning device for learning the processing conditions when joining two or more parts, or cutting or processing one or more parts, are used to perform processing such as joining two or more parts or cutting one or more parts, is solved.

[0106] And the second problem of the present invention, that is, starting from the processing with the processing conditions that satisfy the processing specification and monitoring the processing specification during the processing, is realized to always support processing with the processing conditions that satisfy the processing specification.

[0107] Also, the fourth problem of the present invention, that is, determining whether joining conditions or cutting conditions that satisfy the processing specification can be found from the learned learning device, and presenting that updated learning is necessary when joining conditions or cutting conditions that satisfy the processing specification cannot be found, is solved.

[0108] And, with regard to the fifth problem of the present invention, that is, when processing conditions such as joining conditions and fusing conditions that satisfy the processing specifications cannot be found from the learned learning device, the processing conditions such as joining conditions and fusing conditions that satisfy the processing specifications are updated and learned, and two or more parts are joined or one or more parts are fused, etc. using the updated processing conditions such as joining conditions and fusing conditions. This solves the problem of performing processing.

[0109] And, with regard to the sixth problem of the present invention, that is, a general-purpose processing device such as a joining device or a fusing device checks the energy state when starting processing such as joining or fusing, and performs processing operations such as joining operations and fusing operations using processing conditions that satisfy the processing specifications in that state. This solves the problem.

[0110] Furthermore, with regard to the seventh problem of the present invention, that is, the energy state during processing of the processing device is monitored, and when the energy state changes to an abnormal state, it is presented that the state has changed to an abnormal state and the processing is stopped. This solves the problem.

[0111] Note that the solution to the third problem of the present invention will be described in the second embodiment and the third embodiment.

[0112] (Second Embodiment) In the second embodiment of the present invention, a high-frequency welding device with a learning device for welding a synthetic resin sheet such as a vinyl chloride sheet will be described.

[0113] Generally, when welding parts 4a and 4b to produce a product 4 which is a welded object with a certain thickness, a tolerance is defined for the thickness of each of parts 4a and 4b. If parts 4a and 4b are within the tolerance, it is ideal to be able to produce a product that satisfies the welding specifications when welded under one welding condition.

[0114] However, when the thickness of each of parts 4a and 4b is the combination of the maximum thickness within the tolerance or the combination of the minimum thickness within the tolerance, the way the anode current flows changes due to the difference in thickness.

[0115] For understanding the invention, Fig. 14 illustrates the thickness when the components 4a and 4b of the object to be welded 4 are combined. In Fig. 14, when the minimum value of the thickness within the tolerances of the components 4a and 4b is D1 and the maximum value is D2, the intermediate combination of the minimum and maximum values is Da (= D1 + D2). The combination of the thinnest minimum values is Db (= 2 × D1). The combination of the thickest maximum values is Dc (= 2 × D2).

[0116] Fig. 15 shows the transition of the anode current of the high-frequency welding apparatus with a learning device according to the second embodiment of the present invention. At the median value Da (= D1 + D2) within the tolerance of the thickness of the object to be welded, the anode current flows as in "State A". When the thickness is at the minimum value Db (= D1 + D1) within the tolerance, the anode current flows as "State B". And when the thickness is at the maximum value Dc (= D2 + D2) within the tolerance, the anode current flows as "State C".

[0117] In Fig. 15, "State B" where the thickness of the object to be welded is at the minimum value Db within the tolerance rises earlier than "State A" which is at the median value Da within the tolerance. Also, "State C" which is at the maximum value Dc within the tolerance rises later than "State A".

[0118] Fig. 16 shows a comparison of the transition of the anode current and the welding strength after welding in the cases of "State A" and "State B". In Fig. 16, when in "State A", the line connecting the anode current flowing through the components 4a and 4b and the welding strengths FA1, FA2, FA3, FA4 is shown as a solid line. Also, when in "State B", the line connecting the anode current flowing through the components 4a and 4b and the welding strengths FB1, FB2, FB3, FB4 after welding is shown as a dashed-dotted line. Looking at Fig. 16, it can be seen that the welding strengths FB1, FB2, FB3, FB4 in "State B" increase earlier than the welding strengths in "State A" and are shifted to the left.

[0119] Fig. 17 shows a comparison of the transition of the anode current and the relationship of the welding strength after welding in the "state A" and "state C". In Fig. 17, when in the "state A", the line connecting the anode current flowing through the components 4a and 4b and the welding strengths FA1, FA2, FA3, and FA4 is shown as a solid line. Also, when in the "state C", the line connecting the anode current flowing through the components 4a and 4b and the welding strengths FC1, FC2, FC3, and FC4 after welding is shown as a dashed line. Looking at Fig. 17, it can be seen that the welding strengths FC1, FC2, FC3, and FC4 in the "state C" increase more slowly and shift to the right compared to the welding strengths in the "state A".

[0120] If the welding time in state A is directly used for state B or state C, problems such as excessive welding or insufficient welding will occur as in the conventional example shown in Fig. 42, and the welding specifications will not be met.

[0121] In the second embodiment of the present invention, for the "state A", "state B", and "state C" of the tree shown in Fig. 18, the welding strengths after welding when welding at a plurality of welding times are measured as Q values, and learning is performed using a learning device.

[0122] The tree in Fig. 18 shows that there are three states in a high-frequency welding device with a learning device: "state A" where an anode current is passed through the median value Da (= D1 + D2) within the tolerance of the workpiece to be welded, "state B" where an anode current is passed through the minimum value Db (= D1 + D1) within the tolerance of the workpiece to be welded, and "state C" where an anode current is passed through the maximum value Dc (= D2 + D2) within the tolerance of the workpiece to be welded. And Fig. 18 shows that a plurality of actions and Q values as rewards corresponding to the actions are connected in each of the three states.

[0123] The learning procedure is the same as the procedure shown in the flowchart of Fig. 9 of the first embodiment. Since the explanation of the learning procedure is repetitive, it is omitted. However, when learning is executed according to the flowchart of Fig. 9, the learning result shown in Fig. 19 is obtained as a learned model. In Fig. 19, the materials of the components 4a and 4b, the rise time, and the pairs of the target welding current value, the welding strength, and the welding time are stored in the learning storage means 105.

[0124] The welding procedure is the same as the one shown in the flowchart of FIG. 10 of the first embodiment. Although the description of the overlapping part of the welding procedure is omitted, when welding along the flowchart of FIG. 10, as the main procedure, at step ST23 of FIG. 10, the rise time of the anode current is measured, and at step ST24, it is confirmed which of "state A", "state B", and "state C" it is. If "state A" is confirmed, then at step ST25, it is asked whether there are welding conditions that satisfy the welding specifications in state A. Similarly, when "state B" is confirmed based on the rise time of the anode current at step ST24, at step ST25, it is asked whether there are welding conditions that satisfy the welding specifications in state B. When "state C" is confirmed based on the rise time of the anode current at step ST24, at step ST25, it is asked whether there are welding conditions that satisfy the welding specifications in state C. And if there are welding conditions in the learned model stored in the learning storage means 105 of the learning device, then at step ST26, they are read out from the welding storage means 105, stored in the main storage means 11 of the high-frequency welding means 1, and the parts 4a and 4b are welded under the welding conditions stored in the main storage means 11 to produce the product which is the welded object 4.

[0125] In this way, even if the thickness of the welded object varies within the tolerance, after confirming the state of the anode current flow from the rise time, the welding time as the action that satisfies the welding specifications in that state is read out as the welding condition and welding is performed by the high-frequency welding means 1.

[0126] Using FIG. 20, the state of reading out the welding time as the welding condition that satisfies the welding specifications will be described. In FIG. 20(a), when "state A" is confirmed by the rise time (TA), the welding time (TA3) is read out from the learned model of the learning storage means 105 by inverse calculation from the welding force (FA3) of the welding specifications as the action, that is, the welding condition.

[0127] In FIG. 20(b), when "state B" is confirmed by the rise time (TB), the welding time (TB3) is read out from the learned model of the learning storage means 105 by inverse calculation from the welding force (FB3) of the welding specifications as the action, that is, the welding condition.

[0128] In FIG. 20(c), when "State C" is confirmed based on the rise time (TC), the welding time (TC3) as an operation, i.e., welding condition, is read from the learned model of the learning storage means 105 by back-calculating from the welding strength (FC3) of the welding specification.

[0129] In the present invention, the rise time when starting to flow the anode current is measured to confirm the state of the anode current that has started to flow through the high-frequency welding apparatus. Then, the welding time as a welding condition that provides a welding strength satisfying the welding specification in the confirmed state is found from the learned model of the learning storage means 105 of the learning apparatus 100. If found, the welding condition is read from the learning storage means 105 of the learning apparatus 100 and stored in the main storage means of the high-frequency welding means 1, and welding is performed by the main control means.

[0130] Thus, even if the thickness of the workpiece varies within the tolerance, it is possible to read and perform welding the welding conditions that satisfy the welding specification in each state from among the welding conditions learned by the learning apparatus.

[0131] In the second embodiment, when the materials, shapes, and dimensions of the parts to be welded in the third problem of the present invention are substantially the same and there are variations within the tolerance, it has been described that the state when starting to apply the anode current is confirmed from the rise time when starting to apply the anode current, and the welding time that satisfies the welding specification in the confirmed state is learned as a welding condition, read from the learning storage means, and used for welding.

[0132] Note that since the second embodiment of the present invention is based on the first embodiment, as already described in the first embodiment, it goes without saying that the first problem, the second problem, the fourth problem to the seventh problem of the present invention are solved.

[0133] (Third Embodiment) In the third embodiment, the latter half of the third problem of the present invention, i.e., "when a part with a thickness outside the tolerance is set, no joining is performed so as not to produce defective products with parts joined outside the tolerance" is explained.

[0134] Note that the configuration of the high-frequency welding apparatus 301 with a learning device according to the third embodiment is basically the same as that of the first and second embodiments, so the description thereof is omitted.

[0135] Fig. 21 shows the transition of the anode current of the high-frequency welding apparatus with a learning device according to the third embodiment. In the third embodiment, the anode current flowing through the parts 4a and 4b sandwiched between the upper mold 2 and the lower mold 3 flows in five states, namely, "state B", "state E", "state A", "state F", and "state C" from the left in Fig. 21. In each state, the welding strength after welding when welding is performed at a plurality of welding times is measured and stored in the learning storage means 105 of the learning device 100, and learning is performed so that the welding conditions satisfying the welding specifications can be read out. This is the same as the first and second embodiments already described.

[0136] In the third embodiment, with respect to the range of the rise time from the start of energization of the anode current to reaching a predetermined current value (measurement reference current value), rise time ranges (tB~tE), (tE~tA), (tA~tF), and (tF~tC) are set. The state of the anode current is determined based on which rise time range the rise time (t) during the welding operation is included in. The welding conditions satisfying the welding specifications are read out from the learned model of the learning storage means 105 based on the determined state. Then, the parts 4a and 4b sandwiched between the upper mold 2 and the lower mold 3 are welded under the read welding conditions.

[0137] Fig. 22 shows a flowchart of the welding procedure according to the third embodiment. In Fig. 22, the welding specifications are input and the welding operation is started (step ST20). The workpiece to be welded is set (step ST21). In the third embodiment, since the state is always confirmed, the rise time (t) of the anode current is measured (step ST23). Therefore, step ST22 in Fig. 10 is not present in Fig. 22. The state is confirmed based on which rise time range the rise time (t) is included in (step ST24).

[0138] In the third embodiment, detailed procedures are provided in step ST24. In Fig. 22, the detailed procedures of step ST24 are shown within the dashed frame.

[0139] As the detailed procedure of step ST24, it is determined whether the rise time (t) is equal to or greater than (tB) and equal to or less than (tE) (step ST40). If YES (YES in step ST40), it is determined that the state is E (step ST41). Then, the welding time as the welding condition satisfying the welding specification in state E is read out (step ST26), and welding is performed (step ST27). If NO in step ST40, it is determined whether the rise time (t) is greater than (tE) and equal to or less than (tA) (step ST42). If YES (YES in step ST42), it is determined that the state is A (step ST43). Then, the welding time as the welding condition satisfying the welding specification in state A is read out (step ST26), and welding is performed (step ST27). If NO in step ST42, it is determined whether the rise time (t) is greater than (tA) and equal to or less than (tF) (step ST44). If YES (YES in step ST44), it is determined that the state is F (step ST45). Then, the welding time as the welding condition satisfying the welding specification in state F is read out (step ST26), and welding is performed (step ST27). If NO in step ST44, it is determined whether the rise time (t) is greater than (tF) and equal to or less than (tC) (step ST46). If YES (YES in step ST46), it is determined that the state is C (step ST47). Then, the welding time as the welding condition satisfying the welding specification in state C is read out (step ST26), and welding is performed (step ST27).

[0140] The procedure for ending the welding operation from step ST27 to step ST31 following step ST26 is the same as that in FIG. 10 described in the first embodiment, so the description is omitted.

[0141] When the answer is NO in step ST46, that is, when the state is not "state E", "state A", "state F", or "state C", it means that the thickness of the stacked parts 4a and 4b is outside the tolerance range. In this case, proceed to step ST48 (NO in step ST46). Then, it is determined whether the rise time (t) is less than (tB) (step ST48). If YES (YES in step ST48), it is indicated that the rise time is too short (step ST49). If the answer is NO in step ST48, it is determined whether the rise time (t) is greater than (tC) (step ST50). If YES (YES in step ST50), it is indicated that the rise time is too long (step ST51). If the answer is NO in step ST50, proceed to step ST32. The procedure from step ST32 to the end of the welding operation (step ST31) is the same as that shown in FIG. 10 described in the first embodiment, so the description is omitted.

[0142] By performing the procedure shown in FIG. 22, when starting to flow the anode current, the rise time of the anode current is measured to confirm the state of what state the high-frequency welding apparatus is trying to weld in. Then, welding conditions that satisfy the welding specifications are read out in the confirmed state.

[0143] By performing the above procedure, welding is performed only when the rise time of the anode current is within a certain range, that is, only when the thicknesses of parts 4a and 4b are both within the tolerance range, and welding is not performed when outside the certain range, that is, when the thicknesses of parts 4a and 4b are both outside the tolerance range. In the third embodiment, when the thickness of the parts of the workpiece to be welded is determined by the tolerance, only the workpieces within the tolerance are welded to produce good products, and the workpieces outside the tolerance are not welded. It has been explained that no defective products are produced.

[0144] Note that, as described in the second embodiment, providing a learning device that determines the state of energy supply according to the variation when the materials, shapes, and dimensions of the parts to be joined are substantially the same and there are variations within the tolerance, and learns the energy supply time at which the reward is maximized in that state as the welding conditions or fusing conditions, which is the first half of the third problem of the present invention.

[0145] Also, since the third embodiment of the present invention is already based on the first embodiment, solving the first, second, fourth to seventh problems of the present invention is as described in the first embodiment.

[0146] (Fourth Embodiment) In the fourth embodiment of the present invention, a learning method using a neural network will be described. In the learning method using a neural network, a multi-layer neural network with multiple stacked perceptrons is created. When an action is input to the input layer of the multi-layer neural network, each input is weighted, learning is performed in the hidden layer, and a policy obtained by multiplying the maximum reward from the output layer by a probability (the probability to be adopted) is output.

[0147] In the fourth embodiment, as shown in FIG. 23, learning is performed in a form where "material" is added to "thickness" and "adhesion strength". When "thickness", "adhesion strength", and "material" are input on the input side, "anode current state" and "welding time" are output as welding conditions on the output side.

[0148] In the fourth embodiment of the present invention, a learning plan using the multi-layer neural network shown in FIG. 23 is input and stored in the learning storage means 105 in advance from the I / O means of the learning device. The welding condition learning means 107 performs learning by comprehensively considering "thickness", "adhesion strength", and "material", and outputs "anode current state" and "welding time" to generate a learned model. Then, the welding conditions that satisfy the welding specification are read from the learned model in the learning storage means 105. After reading the welding conditions that satisfy the welding specification, they are stored in the main storage means 11 of the high-frequency welding means 1, and the workpiece 4 is welded by the main control means 10.

[0149] As described above, in the first to fourth embodiments, a learning device for joining conditions of two or more parts, a joining device using the same, and a joining system using the same have been described.

[0150] (Fifth Embodiment) In the fifth embodiment of the present invention, a learning device for processing conditions, a processing device using the same, and a processing system using the same will be described. In this specification, processing is described as a superordinate concept such as joining and cutting.

[0151] Fig. 24 shows a schematic configuration diagram of the fifth embodiment of the present invention. In Fig. 24, three processing devices, a first processing device 501, a second processing device 502, and a third processing device 503 are arranged, and transmission / reception means 601, 602, and 603 are respectively attached to the three processing devices, and transmission / reception means 604 for communicating with these transmission / reception means is attached to the learning device 200, showing a system in which one learning device 200 operates the three processing devices as processing devices with a learning device.

[0152] The transmission / reception means 605 of the learning device 200 takes in and learns processing result measurement data by communicating with transmission / reception means 606, 607, 608, and 609 to which a thickness measurement means 201, an appearance inspection means 202, an adhesive force measurement means 203, and a cutting depth measurement means 204 are respectively attached.

[0153] For example, one learning device 200 can be installed in one factory, and a plurality of processing devices 501, 502, 503, etc. and the learning device 200 can be installed in other factories within the same site so as to be individually communicable, and the learning device 200 can be shared. It can be operated as a joining system in which the learning data increases and the learning progresses.

[0154] Instead of the first processing device 501, the second processing device 502, and the third processing device 503, a system in which transmission / reception means 601, 602, and 603 are respectively attached to the first welding device 371, the second welding device 372, and the third welding device 373 may be used. Also, a system in which transmission / reception means 601, 602, and 603 are respectively attached to the first cutting device 471, the second cutting device 472, and the third cutting device 473 may be used.

[0155] Also, the Internet may be used as the communication means and used as an IOT processing system.

[0156] (Sixth Embodiment) Using FIG. 25, an example of a bonding apparatus 300 with a learning device 150 of the present invention attached thereto and a fusing apparatus 400 with a learning device 150 attached thereto is shown as a system diagram. Note that a learning device for the fusing method of one or more components, a fusing apparatus using the same, and a fusing system using the same will be described later as the seventh embodiment of the present invention.

[0157] The bonding apparatus 350 includes a "high-frequency welding apparatus", a "non-contact hot plate welding apparatus", a "vibration welding apparatus", a "laser welding apparatus", a "ultrasonic welding apparatus", a "ultrasonic metal bonding apparatus", a "ultrasonic bonding apparatus", etc. These bonding apparatuses apply energy such as vibration energy and heat energy to components to heat, melt, or solid-phase bond the components to perform bonding processing to produce products. Therefore, they are common in that by changing the magnitude of the energy applied to the components and the supply time of the energy, bonding conditions that satisfy the processing specifications can be found.

[0158] Therefore, those with a learning device attached to these bonding apparatuses can perform the bonding method using the learning device of the present invention as a "high-frequency welding apparatus 301 with a learning device", a "non-contact hot plate welding apparatus 302 with a learning device", a "vibration welding apparatus 303 with a learning device", a "laser welding apparatus 304 with a learning device", a "ultrasonic welding apparatus 305 with a learning device", a "ultrasonic metal bonding apparatus 306 with a learning device", and a "ultrasonic bonding apparatus 307 with a learning device", respectively.

[0159] And those with the learning device 150 attached to the fusing apparatus 450 can perform the fusing method using the learning device of the present invention as a "laser fusing apparatus 401 with a learning device" and a "ultrasonic fusing apparatus 402 with a learning device", respectively.

[0160] Although not shown in FIG. 25, it can be applied to a "spin welder" and an "impulse welder". These will be described later with reference to FIGS. 30 and 31.

[0161] Fig. 26 shows an external view of the non-contact hot plate welding apparatus 302 with a learning device according to the present invention. In the non-contact hot plate welding apparatus 352, the welding conditions are determined by the magnitude of the thermal energy heated to a predetermined temperature of the hot plates 352a and 352b and the time for applying the thermal energy.

[0162] For example, taking the surface temperature for holding the hot plates 352a and 352b as the "state" of the apparatus, the heating time for heating while approaching the workpiece 4W1 as the "action", and the welding strength or the aesthetic level as the "reward", welding can be performed by learning.

[0163] Note that the learning procedure and the welding procedure are the same as those in the first embodiment. Therefore, by applying the bonding method of the present invention already described in the first to fifth embodiments, the learning device 150 can be attached and welding can be performed using the learning device of the present invention as the "non-contact hot plate welding apparatus 302 with a learning device".

[0164] Fig. 27 shows an external view of the vibration welding apparatus 303 with a learning device according to the present invention. In the vibration welding apparatus 353, low-frequency vibration of about 100 Hz is applied to one component 4W2a by the upper mold 353a, and the surface of the other component 4W2b fixed to the lower mold 353b is rubbed to generate frictional heat, and it is a bonding apparatus that welds the contact surfaces of the components 4W2a and 4W2b with the frictional heat. The welding conditions are determined by the generated frictional force and the time for applying the frictional force.

[0165] For example, taking the magnitude of the frictional force applied to the components 4W2a and 4W2b of the workpiece as the "state", the heating time for heating by rubbing the components 4W2a and 4W2b of the workpiece, that is, the time for applying the frictional force as the "action", and the welding strength or the aesthetic level as the "reward", welding can be performed by learning.

[0166] Note that the learning procedure and the bonding / welding procedure are the same as those in the first embodiment. Therefore, by applying the bonding method of the present invention already described in the first to fifth embodiments, the learning device 150 can be attached and welding can be performed using the learning device of the present invention as the "vibration welding apparatus 303 with a learning device".

[0167] Fig. 28 shows an external view of the laser welding apparatus 304 with a learning device according to the present invention. In the laser welding apparatus 354, the welding conditions are determined by the magnitude of the energy of the laser beam 354a and the time for applying the energy.

[0168] For example, by taking the magnitude of the laser light output as the "state", the laser light irradiation time for irradiating a pair of workpieces 4W3 with the laser light as the "action", and the welding strength or the aesthetic level as the "reward" for learning, welding can be performed.

[0169] Note that the learning procedure and the joining / welding procedure are the same as those in the first embodiment. Therefore, by applying the joining method of the present invention already described in the first to fifth embodiments, the learning device 150 can be attached and welding can be performed using the learning device of the present invention as the "laser welding apparatus 304 with a learning device".

[0170] Fig. 29 shows an external view of the ultrasonic welding apparatus 305 with a learning device. In the ultrasonic welding apparatus 355, a tool horn 355a that vibrates ultrasonically in the vertical direction is pressed against a pair of workpieces 4W4 for welding.

[0171] The welding conditions are determined by the magnitude of the energy of the ultrasonic vibration and the time for applying the energy.

[0172] For example, by taking the frequency and amplitude of the ultrasonic vibration as the "state", the time for applying the ultrasonic vibration as the "action", and the welding strength or the aesthetic level as the "reward" for learning, welding can be performed.

[0173] Note that the learning procedure and the joining / welding procedure are the same as those in the first embodiment. Therefore, by applying the joining method of the present invention already described in the first to fifth embodiments, the learning device 150 can be attached and welding can be performed using the learning device of the present invention as the "ultrasonic welding apparatus 305 with a learning device".

[0174] Although the appearance of the device is not illustrated, the ultrasonic metal bonding device 306 is a bonding device that applies ultrasonic vibration parallel to the contact surface between metals to solid-phase bond the contact surfaces between metals. The welding conditions are determined by the magnitude of the energy of the ultrasonic vibration and the time for applying the energy.

[0175] Note that the learning procedure and the bonding / welding procedure are the same as those in the first embodiment. Therefore, by applying the bonding method of the present invention already described in the first to fifth embodiments, the learning device 150 can be attached and bonding can be performed using the learning device of the present invention as the "ultrasonic metal bonding device 306 with learning device".

[0176] Similarly, although the appearance of the device is not illustrated, the ultrasonic bonding device 307 is a bonding device that places a metal sheet on the surface of glass and applies ultrasonic vibration parallel to the contact surface between them to solid-phase bond the metal sheet to the surface of the glass. The welding conditions are determined by the magnitude of the energy of the ultrasonic vibration and the time for applying the energy.

[0177] Note that the learning procedure and the bonding / welding procedure are the same as those in the first embodiment. Therefore, by applying the bonding method of the present invention already described in the first to fifth embodiments, the learning device 150 can be attached and bonding can be performed using the learning device of the present invention as the "ultrasonic bonding device 307 with learning device".

[0178] Figures 30(a) to (c) are cross-sectional views showing the welding process of a spin welder. As shown in Figure 30(a), the spin welder has the end faces of two cylindrical plastic parts 4W6a and 4W6b facing each other and in contact. As shown in Figure 30(b), with the lower part 4W6b fixed, the upper part 4W6a is rotated at high speed, and as shown in Figure 30(c), it is a welding device that welds the contact surfaces together with frictional heat. The upper part 4W6a is rotated at high speed to impart kinetic energy, generating frictional heat at the contact surface with the lower part 4W6b to melt and bond the parts together. Therefore, the welding conditions are determined by the magnitude of the rotational kinetic energy applied to the upper part 4W6a and the time for applying the rotational kinetic energy. The present invention learns and welds these welding conditions with a learning device.

[0179] Note that the learning procedure and the joining / welding procedure are the same as those in the first embodiment. Therefore, the joining method of the present invention already described in the first to fifth embodiments can be applied to join two parts as a spin welder with a learning device.

[0180] FIG. 31 is a diagram showing the transition of the temperature of the workpiece when a large current is passed through an impulse welder. The impulse welder is a welding device that instantaneously passes a large current through two or more parts to generate heat and then cools them to melt and join the parts together. As shown in FIG. 31, the welding conditions are determined by the amount of heat generated by passing a large current and then cooling, and the heating / cooling time. The present invention learns these welding conditions with a learning device and then performs welding.

[0181] Note that the learning procedure and the joining / welding procedure are the same as those in the first embodiment. Therefore, the joining method of the present invention already described in the first to fifth embodiments can be applied to join two parts as an impulse welder with a learning device.

[0182] (Seventh Embodiment) As the seventh embodiment of the present invention, a learning device for a method of fusing one or more parts, a fusing device with a learning device, and a fusing system with a learning device will be described.

[0183] The fusing device 450 includes a "laser fusing device", an "ultrasonic fusing device", etc. These fusing devices irradiate the workpiece with a laser beam or press a cutter vibrating ultrasonically against it to fuse one or more parts. These fusing devices determine the fusing conditions based on the magnitude of the energy of the laser beam or the ultrasonic vibration and the time for applying the energy. In the tree of FIG. 25, "laser fusing device 401 with a learning device" and "ultrasonic fusing device 402 with a learning device" are shown at the lower right.

[0184] Although the "laser cutting device 401 with a learning device" is not shown in the figures, it has a similar appearance to the laser welding device in Fig. 28. The laser beam is irradiated from the laser cutting means of the "laser cutting device 401 with a learning device" onto the object to be cut, and cutting is performed. The learning procedure and cutting procedure of the cutting conditions are substantially the same as those in the first embodiment. Therefore, by applying the method of the present invention already described in the first to fifth embodiments, those with the learning device 150 attached can perform cutting using the learning device of the present invention as the "laser cutting device 401 with a learning device" and the "ultrasonic cutting device 402 with a learning device", respectively.

[0185] The ultrasonic cutting device 402 with a learning device will be described in detail with reference to Figs. 32 to 41. Fig. 32 shows an external perspective view of the ultrasonic cutting device 402 with a learning device for cutting one or more components of the present invention. The ultrasonic cutting means 452 shown in Fig. 32 has an ultrasonic cutter 412 attached to the lower end of an ultrasonic vibrator 411 that vibrates ultrasonically up and down, and the ultrasonic cutter 412 that is ultrasonically vibrating by an air cylinder 414 attached to a support column 410 presses the object to be cut 4W5 placed on a table 413 for cutting.

[0186] As the object to be cut 4W5, as shown in Fig. 33, relatively soft foods such as sponge cake, strawberry cake, mochi, and relatively hard mochi, sanshoku diamond mochi, rubber sheet, old tire, etc. are used as the objects to be cut. Fig. 33 shows the type of the object to be cut, the necessity of aesthetics after cutting, and the magnitude of the pressing force. Some of the objects to be cut require aesthetics after cutting, while others do not. Also, some objects can be cut with a small pressing force, while others require a large pressing force.

[0187] Among the items in Fig. 33, it is required that the food does not lose its shape even after cutting. On the other hand, for rubber sheets and old tires, etc., it is sufficient if they can be cut, and there are few problems even if the surface state changes or the shape collapses.

[0188] The ultrasonic cutting device 402 with a learning device determines the cutting conditions based on the magnitude of ultrasonic energy and the cutting time according to the cutting specifications of the object to be cut. Although there is a difference between "cutting" and "welding" in the first embodiment, they are common in controlling the magnitude of ultrasonic energy and the energy supply time. Therefore, the content of the present invention described in the first to fifth embodiments can be used as a cutting device with a learning device.

[0189] In the following description, the parts that overlap with the descriptions of the first to fifth embodiments already described will be omitted, and the description will focus on the key points.

[0190] FIG. 34 is a schematic configuration diagram of the ultrasonic cutting device 402 with a learning device according to the seventh embodiment of the present invention. In terms of content, it is similar to the schematic configuration diagram of FIG. 2. Instead of the high-frequency welding means 1, an ultrasonic cutting means 452 is connected to the learning device 150 by a communication line 170 and is integrally controlled.

[0191] Regarding each means of the learning device 150, although the names are different from those of each means of the learning device 100 described in the first embodiment, the functions are the same or similar. That is, the learning device 150 is provided with a thickness measurement means 151, an appearance inspection means 152, a cutting depth measurement means 153, etc. as the cutting result measurement means 180. And a cutting result evaluation means 154 that evaluates the value measured by the cutting result measurement means 180 as a reward, a cutting learning storage means 155 that stores the information obtained by evaluating the cutting result, a cutting condition learning means 156 that learns the cutting conditions from the information stored in the cutting learning storage means 155, an I / O means 157 for inputting and outputting the learning plan of the AI technology, a cutting condition reading means 158 that reads out the cutting conditions that satisfy the cutting specifications from the cutting learning storage means 155, and a specification monitoring means 159A and a state monitoring means 159B are provided.

[0192] Even when the processing specifications are updated from the previous cutting operation or changed during cutting, the specification monitoring means 159A monitors the latest and current cutting specifications, and reads out the cutting conditions that satisfy the latest and current cutting specifications from the learned model of the cutting learning storage means 155.

[0193] The state monitoring means 159B obtains information on the state of the current that the ultrasonic cutting means 452 has started to flow from the state output means 463 of the ultrasonic cutting means 452, and transmits it to the cutting condition reading means 158. The cutting condition reading means 158 reads out the cutting conditions that satisfy the cutting specifications based on the state of the current that has started to flow from the learned model of the cutting learning storage means 155. Further, the state monitoring means 159B monitors whether the state of the current during ultrasonic cutting is normal or abnormal.

[0194] Fig. 35 shows the transition of the current value and the value of the cutting depth when ultrasonic energy is applied by the cutting device with a learning device according to the seventh embodiment of the present invention. The X-axis is the time axis of the cutting operation, the left Y-axis is the current value flowing through the ultrasonic driving means, and the right Y-axis is the value of the cutting depth. When the tip of the cutting edge of the ultrasonic cutter is pressed against the object to be cut and the current starts to flow, cutting starts after a relatively short rise time.

[0195] Fig. 36 separately shows the cutting time as several actions (TG) in one state (G), the reward when cutting for that cutting time, and the cutting depth as the Q value (Q(G,TG)) in the cutting device according to the seventh embodiment of the present invention.

[0196] Here, for the "state G", as shown in FIGS. 36(a) to (d), the learning device 150 performs fusing for each fusing time (TG1, TG2, TG3, TG4), individually measures the values of the fusing depths after fusing (LG1, LG2, LG3, LG4) using the fusing depth measuring means 153 of the fusing result measuring means 180, evaluates the fusing result as a reward by the fusing result evaluating means 154, and stores it in the fusing learning storage means 155. Then, the fusing condition learning means 156 learns the fusing conditions from the fusing results evaluated as rewards so that the fusing time can be read out as the fusing condition that satisfies the fusing specification when the fusing specification is input.

[0197] For example, if the fusing depth LG1 is input as the fusing specification to the fusing specification input / output means 462 that inputs the fusing specification of the ultrasonic fusing means 452, the learning device 150 calculates the fusing time TG1 as an action by inverse calculation from the fusing depth LG1 and reads it out. If the fusing depth LG2 is input as the fusing specification, the fusing time TG2 as an action is read out by inverse calculation from the fusing depth LG2. If the fusing depth LG3 is input as the fusing specification, the fusing time TG3 as an action is read out by inverse calculation from the fusing depth LG3. The same applies when the fusing depth LG4 is input as the fusing specification.

[0198] In addition, when values between the above four fusing depths LG1, LG2, LG3, LG4 are input as the fusing specification, by performing regression analysis, which is often used in AI technology for the relationship between the fusing time and the fusing depth, and formulating it into a mathematical formula, the fusing time corresponding to the input fusing depth is calculated and read out. Also, by increasing the number of data to be learned, the fusing time that satisfies the specification can be read out even for a fine fusing specification. This is the same as in the case of the welding device of the first embodiment.

[0199] FIG. 37 shows, as an image, the procedure for measuring the fusing depths when the fusing times (TG1, TG2, TG3, TG4) are randomly set for four samples in the ultrasonic fusing device 402 with a learning device according to the seventh embodiment of the present invention.

[0200] In FIG. 37, a rubber sheet is used as the workpiece 4W5, and an ultrasonic cutter 412 with a linear cutting edge on its end face (not shown) is pressed against the upper surface of the rubber sheet, showing the state of cutting with random cutting times. FIG. 37 shows cases where the cutting depth L is shallow and cutting is not possible, cases where the cutting depth L becomes deeper, and cases where complete cutting is possible.

[0201] In FIG. 37(a), the cutting times are TG1, TG2, TG3, TG4 from the right; in FIG. 37(b), the cutting times are TG4, TG3, TG2, TG1 from the right; in FIG. 37(c), the cutting times are TG2, TG4, TG1, TG3 from the right; and in FIG. 37(d), the cutting times are TG3, TG1, TG4, TG2 from the right. The measurement of the cutting depth L uses the value measured by the cutting depth measuring means 153 for the descending amount of the ultrasonic cutter 411. The cutting time that satisfies the cutting specification is stored in the cutting learning storage means 155 as the cutting condition that satisfies the cutting specification.

[0202] Fig. 38 shows a flowchart of the learning procedure according to the seventh embodiment of the present invention. In the flowchart of Fig. 38, the learning procedure of reinforcement learning is used as a learning plan, input by the I / O means 157, stored in the fusion learning storage means 155, and then learning is started (step ST51). Based on the fusion specifications, a target current value is set so as to satisfy the fusion specifications. Note that since a learning device that has not learned at all is ignorant, it is advisable to include, as an initial value, a target current value that is expected to satisfy the fusion specifications in the welding specifications and input it (step ST52). A current is passed through the ultrasonic driving means (step ST53). Here, it is asked whether to check the state during fusion (step ST54). This is because when the ultrasonic fusion means 452 has only a single energy supply state, since there is only one energy state, there is no need to check the state during fusion. In that case, the next step ST55 is skipped (NO in step ST54). When the ultrasonic fusion means 452 is a general-purpose processing device and can select one state from a plurality of energy supply states, step ST55 is performed. In step ST55, the rise time until the reference current value is reached is measured and stored in the fusion learning storage means 155. Then, a current is passed for each fusion time to perform fusion (step ST56). The fusion depth L as a reward for the action is measured, correspondence data with the fusion time is created, and stored in the fusion learning storage means 155 (step ST57). The fusion condition learning means 156 learns the fusion conditions from the information stored in the fusion learning storage means 155, and stores the learning result as a learned model in the fusion learning storage means 155 in a form that can be read out as the fusion conditions (step ST58).

[0203] Then, it is asked whether data corresponding to a predetermined number of fusing times has been obtained for one state (i.e., one current curve) (step ST59). If the predetermined number of data has not been obtained, the process returns to step ST53 (NO in step ST59). If the predetermined number of data has been obtained (YES in step ST59), it is asked whether to update the target current value and learn about other states (step ST60). When updating the target current value, the process returns to step ST52 (YES in step ST60). When not learning about other states (NO in step ST60), it is asked whether learning has been done as planned (step ST61). If learning has not been done as per the learning plan, the process returns to step ST52 (NO in step ST61). If learning as per the plan has been completed (YES in step ST61), the process proceeds to step ST62 and learning is terminated. Note that steps ST58 and ST59 may be interchanged so that after obtaining the predetermined number of data, learning is performed and the learning result is stored in the fusing learning storage means 155.

[0204] Note that when performing update learning, from step ST84, the process enters step ST51 and performs the procedures below step ST51. By performing the procedures in the flowchart of FIG. 38, for the state where the ultrasonic cutter 412 is vibrating, data on the fusing depth L for each fusing time, which is each action, is obtained, the data is stored as a reward, the stored data is learned, the learning result is made into a form that can be read out as fusing conditions, and the result is stored in the fusing learning storage means 155.

[0205] Thus, if the fusing depth L is input according to the fusing specification, the fusing time as the fusing condition that satisfies the fusing specification can be read out from the fusing learning storage means 155.

[0206] When performing a fusing operation, the fusing time as the fusing condition that satisfies the fusing specification is read out from the learned model of the learning device, stored in the main storage means of the ultrasonic fusing means 452, and fusing is performed.

[0207] FIG. 39 shows the content of the learning result stored in the storage means according to the seventh embodiment of the present invention. In FIG. 39, the fusing time and the fusing depth L are stored as a pair. Also, trapezoidal, round, and square frames are added to each pair to make it easier to understand.

[0208] In FIG. 39(a), the measured data are stored in the order in which they were measured. For example, in Data 1, from the right, the pair of the fusing depth LG11 and the Q value (G, TG11), the pair of the fusing depth LG21 and the Q value (G, TG21), the pair of the fusing depth LG31 and the Q value (G, TG31), and the pair of the fusing depth LG41 and the Q value (G, TG41) are shown as being stored. For Data 2 to Data 4, each pair is arranged randomly.

[0209] In FIG. 39(b), the data, which is the stored content of the fusing learning storage means 155 by the fusing condition learning means 156, is arranged in a form sorted by the fusing depth L. For example, in Data 1, compared with FIG. 39(a), the order is reversed, and from the left, the pair of the fusing depth LG11 and the Q value (G, TG11), the pair of the fusing depth LG21 and the Q value (G, TG21), the pair of the fusing depth LG31 and the Q value (G, TG31), and the pair of the fusing depth LG41 and the Q value (G, TG41) are arranged. For Data 2 to Data 4, each pair is arranged in the same order as Data 1.

[0210] And in FIG. 39(c), the fusing condition learning means 156 obtains the average value of the fusing depth L for each fusing time for the data arranged in FIG. 39(b), and creates a table from which the fusing time can be read by inverse calculation from the fusing depth L. And this is used as the learned model. Since Data 1 to Data 4 are arranged, when the fusing depth L is input as the fusing specification, the fusing time as the fusing condition that satisfies the fusing specification can be read from FIG. 39(c).

[0211] In the fusing operation, the fusing operation is performed according to the fusing operation procedure shown in the flowchart of FIG. 40. In FIG. 40, first, the fusing specifications are input to start the fusing operation (step ST70). The object to be fused is set (step ST71). It is asked whether to check the state (step ST72). As also described in the flowchart of FIG. 38, when the ultrasonic fusing means 452 has only a single energy supply state, since there is only one energy state, it is not necessary to check the state during fusing. At that time, steps ST73 and ST74 are skipped (NO in step ST72). When the ultrasonic fusing means 452 can select one state from a plurality of energy supply states, steps ST73 and ST74 are performed. In step ST73, the rise time is measured. Then, the state of the energy that the ultrasonic fusing means 452 has started to supply is checked (step ST74). It is asked whether there are fusing conditions that satisfy the latest and current fusing specifications in the confirmed state (step ST75). If there are fusing conditions, they are read from the fusing learning storage means 155 (step ST76) and stored in the main storage means 461. Fusing is performed by the ultrasonic fusing means 452 (step ST77). The state monitoring means 159B monitors whether the state during fusing is abnormal (step ST78). When the state is normal (YES in step ST78), it is confirmed that the fusing time for ending the fusing operation has been reached (step ST79), and the fusing operation is ended (step ST81). When the state is abnormal in step ST78, an abnormality is presented and fusing is stopped (step ST80). Then, the fusing operation is ended (step ST81).

[0212] Also, when no fusing conditions that satisfy the fusing specifications are found in the fusing learning storage means 155 in step ST75 (NO in step ST75), "Update learning is required" is presented (step ST82). Then, it is asked whether to perform update learning (step ST83). When performing update learning, starting from step ST84, it jumps to the flowchart of FIG. 38, and update learning is performed according to the procedure of step ST50 and below in FIG. 38.

[0213] The content of the fusing learning storage means 155 when reading the fusing conditions in step ST76 is shown in FIG. 41. In the case of FIG. 41(a), "LG2" is input as the fusing depth L of the fusing specification, and the fusing operation is started (step ST70). When measuring the rise time (step ST73), since the rise time is "TG", it is confirmed that the state is "state G" (step ST74). Since the fusing depth L that satisfies the fusing specification in "state G" is "LG2" in the table, the fusing time "TG2" paired with the fusing depth "LG2" is read as the action for obtaining the fusing depth "FD3" of the fusing specification, that is, the fusing time as the fusing condition is "TG2".

[0214] Similarly, in the case of FIG. 41(b), "LH3" is input as the fusing depth L of the fusing specification, and the fusing operation is started (step ST70). When measuring the rise time (step ST73), since the rise time is "TH", it is confirmed that the state is "state H" (step ST74). Since the fusing depth L that satisfies the fusing specification in "state H" is "LH3" in the table, the fusing time "TH3" paired with the fusing depth "LH3" is read as the action for obtaining the fusing depth "FD3" of the fusing specification, that is, the fusing time as the fusing condition is "TH3". Regarding the update learning, since it is the same as the content described in the first embodiment, the description is omitted.

[0215] As described above, in the present invention, processing conditions such as joining conditions or fusing conditions are learned by learning means using AI technology, and welding conditions that realize processing specifications such as welding specifications or fusing specifications that satisfy the processing specifications are extracted from the learning results, and the workpieces are welded, fused, etc. using the extracted processing conditions such as welding conditions and fusing conditions.

[0216] By using the learning device of the processing method of the present invention and the processing device using the same, in the present invention, it is possible to learn processing conditions such as joining conditions and fusing conditions when joining two or more parts or processing one or more parts by fusing, etc., and to join two or more parts or process one or more parts by fusing, etc. using the learning method and learning device.

[0217] In addition, even when the processing specifications are updated from the immediately preceding processing operation or changed during the processing, by monitoring the latest and current processing specifications with the specification monitoring means, the processing conditions such as the joining conditions or the fusing conditions that satisfy the latest and current processing specifications can be read from the learning storage means, and processing operations such as the joining operation and the fusing operation can be performed. Therefore, for example, when the hardness of the object to be fused changes midway, such as in an old tire containing steel wires, or when a hard fruit is contained in a soft sponge cake such as a strawberry cake, the fusing operation with complex fusing specifications can also be dealt with by switching the learned model after learning.

[0218] And when the materials, shapes, and dimensions of the parts to be joined are substantially the same and there are variations within the tolerance, the present invention can stably join two or more parts or fuse one or more parts by using a method and a learning device for learning the joining conditions and the fusing conditions.

[0219] Furthermore, the present invention can provide a method and a device for determining whether or not joining conditions or fusing conditions that satisfy the processing specifications are found in the learning storage means, and presenting that update learning is necessary when the joining conditions or the fusing conditions that satisfy the processing specifications are not found.

[0220] And when it is determined that the joining conditions or the fusing conditions that satisfy the processing specifications are not found in the learning storage means, the joining conditions or the fusing conditions that satisfy the processing specifications are updated and learned, and two or more parts can be joined or one or more parts can be fused with the updated and learned joining conditions or fusing conditions.

[0221] In addition, when there are multiple energy level application states applied to the workpiece by one joining device or one fusing device, the present invention can confirm the state of the energy that has actually started to be supplied, and read the joining conditions or the fusing conditions, which are actions that satisfy the processing specifications in the confirmed state, from the learning device, and perform the joining operation and the fusing operation.

[0222] Furthermore, the present invention can monitor the state of the processing apparatus after processing starts, and when the state of the processing apparatus changes and becomes abnormal, it can present this fact and stop the processing of the processing apparatus.

[0223] Also, in the present invention, by checking the state of the actually supplied energy, reading out the processing conditions, which are actions that satisfy the processing specifications in the confirmed state, from the learning device and performing processing, and having the learning device monitor whether the state of the processing apparatus after processing starts is normal or abnormal, is useful for a processing method with a learning device, a processing apparatus using this, and a processing system using this.

[0224] In the description of the present invention, for the sake of understanding the invention, as an AI technology, a simple example of reinforcement learning was shown. The AI technology will continue to evolve in the future. Therefore, in the learning device 100 of the present invention, by using the I / O means 107, any one of complex reinforcement learning, Q-learning, statistical learning (neural network), deep learning (deep learning), and deep reinforcement learning that combines deep learning and reinforcement learning can be used as a learning plan (learning algorithm, learning software), it is possible to provide a more advanced learning device for processing conditions, a processing apparatus using this, and a processing system using this.

Industrial Applicability

[0225] The present invention can be widely applied not only to high-frequency welding apparatuses such as vinyl chloride sheets, but also to learning devices for joining methods of two or more other parts, joining apparatuses using this, and joining systems using this, learning devices for fusing methods of one or more parts, fusing apparatuses using this, and fusing systems using this.

Explanation of Signs

[0226] 1 High-frequency welding means 2 Upper mold 3 Lower mold 4 Welded object 4a, 4b Parts 5 Air cylinder 6 High-frequency current supply means 10 Main control means 11 Main memory means 12 Welding specification input / output means 13 State output means 100 Learning device 101 Thickness measurement means 102 Appearance inspection means 103 Weld strength measurement means 104 Welding result evaluation means 105 Learning memory means 106 Welding condition learning means 107 I / O means 108 Welding condition reading means 109A Specification monitoring means 109B State monitoring means 120 Communication line 130 Welding result measurement means

Claims

[Claim 1] A learning device that learns machining conditions, At least (1) a "processing result measuring means" for measuring the processing result; (2) a "processing result evaluation means" that evaluates the processing result as a reward; (3) a "learning storage means" that stores the processing results and the rewards for evaluating the processing results; (4) A "machining condition learning means" that organizes the learning results evaluated based on the machining results and enables the machining conditions to be read out; (5) a "processing specification monitoring means" for monitoring the processing specifications; (6) A "machining condition reading means" that reads out the machining conditions from the learning storage means, (7) Further, a "state monitoring means" is provided to monitor the state of energy applied to the workpiece by the processing device connected to the learning device, The processing result is measured in advance by the processing result measuring means of (1), the processing result is evaluated as a reward by the processing result evaluation means of (2), the processing result and the reward are stored in the learning storage means of (3), and the learning result evaluated based on the processing result is organized by the processing condition learning means of (4), so that the processing conditions can be read out. The machining condition learning means of (4) is configured to read out the machining conditions from the output of the trained model using a trained model that has been trained to output machining conditions that satisfy the desired machining specifications when at least the desired machining specifications are input based on the information stored in the learning storage means of (3), When machining specifications are input to the machining means connected to the learning device, the machining specifications are monitored by the machining specification monitoring means (5), machining conditions that satisfy the latest and current machining specifications are read from the learning storage means by the machining condition reading means (6), and the read machining conditions are output to the machining means, A learning device that learns processing conditions, configured to use the state monitoring means (7) to monitor the state of energy applied to the workpiece of a processing device connected to the learning device after starting processing under the read processing conditions, and if the state of the processing device becomes abnormal compared to the state assumed by the learning device, to notify this and stop processing by the processing device.

Images