JP7531127B2 - Production data generation device, production data generation method, and program - Google Patents

Description

The present disclosure relates to an apparatus, method, and program for generating production data for producing mounting boards.

A component mounting line including at least one component mounting device produces mounted boards by mounting components on the board. At this time, the component mounting line mounts the components on the board based on production data. The production data includes identification information for each component to be mounted on the board and the mounting order of those components. The production data may also include component data for each component to be mounted on the board. The component data includes information indicating the shape of the component to be mounted, and the operating parameters of the component mounting device that handles that component. The operating parameters include, for example, the suction speed or mounting load of the mounting head of the component mounting device.

For example, in the mounting board manufacturing system of Patent Document 1, the control parameters (or machine parameters) corresponding to the operating parameters are corrected based on the results of the component mounting work. This allows the component data to be corrected appropriately and efficiently.

JP 2019-4129 A

However, the mounting board manufacturing system of Patent Document 1 above has the problem that it can be difficult to set appropriate operating parameters.

Therefore, the present disclosure provides a production data generation device that can set appropriate operating parameters.

A production data generation device according to one aspect of the present disclosure includes a model selection unit that selects at least one learning model from a plurality of mutually different learning models each indicating a relationship between an operating condition of a component mounting device for mounting a component on a board and the component, a parameter estimation unit that estimates operating parameters, which are operating conditions of a component mounting device for mounting the target component on a board, based on the at least one selected learning model and component information relating to the target component to be mounted on the board, and a data generation unit that generates production data including component data having the component information and the operating parameters.

These comprehensive or specific aspects may be realized by a system, a method, an integrated circuit, a computer program, or a recording medium such as a computer-readable CD-ROM, or by any combination of a system, a method, an integrated circuit, a computer program, and a recording medium. The recording medium may also be a non-transitory recording medium.

The production data generation device disclosed herein is capable of setting appropriate operating parameters.

Further advantages and effects of one aspect of the present disclosure will become apparent from the specification and drawings. Such advantages and/or effects are provided by the features described in some embodiments and the specification and drawings, respectively, but not necessarily all of them are required to obtain one or more identical features.

FIG. 1 is a diagram showing an example of a configuration of a production system according to the first embodiment. FIG. 2 is a diagram showing an example of the configuration of the component mounting apparatus according to the first embodiment. FIG. 3 is a diagram partially showing an example of a cross section taken along line AA in FIG. FIG. 4 is a block diagram showing the functional configurations of the production data generating device and the component mounting line according to the first embodiment. FIG. 5 is a diagram showing an example of a part library according to the first embodiment. FIG. 6 is a diagram showing an example of production data according to the first embodiment. FIG. 7A is a diagram showing an example of multiple learning models stored in a learning model storage unit in embodiment 1 and managed on a time basis. FIG. 7B is a diagram showing another example of multiple learning models stored in the learning model storage unit in embodiment 1 and managed on a time basis. FIG. 8 is a diagram showing an example of multiple learning models stored in the learning model storage unit in embodiment 1 and managed on a production facility basis. FIG. 9A is a diagram showing an example of multiple learning models stored in a learning model storage unit in embodiment 1 and managed by production type. FIG. 9B is a diagram showing an example of multiple learning models stored in the learning model storage unit in embodiment 1 and managed on a production type basis and a production equipment basis. FIG. 10A is a diagram for explaining an overview of the estimation process of the motion parameters in the first embodiment. In FIG. FIG. 10B is a diagram for explaining an overview of the learning process of the motion parameter model in the first embodiment. FIG. 11 is a diagram illustrating an example of an overall process in the first embodiment. FIG. 12 is a diagram showing another example of the overall process in the first embodiment. In FIG. FIG. 13 is a flowchart showing the processing operation of the production data generating device in the first embodiment. FIG. 14 is a diagram showing an example of a configuration of a production system in the second embodiment. FIG. 15 is a block diagram showing the functional configurations of the production management device, the component mounting line, and the processing device according to the second embodiment. FIG. 16 is a diagram showing an example of an overall process in the second embodiment. FIG. 17A is a diagram showing an example of filtering information generated based on the inspection result of a mounting board according to the second embodiment. FIG. FIG. 17B is a diagram showing an example of filtering information generated based on the mounting performance of a component mounting line in the second embodiment. FIG. 18A is a diagram showing an example of filtering information generated by selecting a part in the second embodiment. FIG. FIG. 18B is a diagram showing an example of filtering information generated by selection of a board in the second embodiment. FIG. 19 is a flowchart showing the processing operation of the production management device in the second embodiment.

In order to solve the above-mentioned problems, a production data generation device according to one aspect of the present disclosure includes a model selection unit that selects at least one learning model from a plurality of mutually different learning models that indicate a relationship between the operating conditions of a component mounting device for mounting a component on a board and the component, a parameter estimation unit that estimates an operating parameter that is an operating condition of the component mounting device for mounting the target component on the board based on the at least one selected learning model and component information on the target component to be mounted on the board, and a data generation unit that generates production data including component data having the component information and the operating parameter. For example, the component information may indicate at least one of the dimensions, shape, appearance, and type of the component corresponding to the component information, and the supply form for supplying the component. The operating parameter may also be a parameter related to at least one of the transport, recognition, pickup, and mounting of the component by the component mounting device.

As a result, at least one learning model is selected from a plurality of mutually different learning models and used to estimate the operating parameters, thereby increasing the possibility that appropriate operating parameters will be estimated for the components to be mounted. Therefore, appropriate operating parameters can be set. Furthermore, when component data having such operating parameters and component information is included in production data, and the production data is used to mount components on a board by a component mounting device, a high-quality mounted board can be produced. In other words, the quality of the mounted board can be improved.

In addition, the operation parameter may be not only one parameter but also a set of multiple parameters. In this case, when multiple learning models are selected by the model selection unit, the parameter estimation unit may estimate each of the multiple parameters included in the operation parameter from the selected multiple learning models.

The production data generating device may further include a data acquisition unit that acquires actual production data used by the component mounting device, the actual production data including component information on the mounted components and actual component data having operating parameters used to mount the mounted components, and a learning unit that updates the relationship indicated by a learning model among the multiple learning models that corresponds to the acquired actual production data by learning using the actual component data as training data.

The operating parameters of the actual component data included in the actual production data are used to mount the mounted components, and are modified during this process. In other words, the operating parameters are modified so that mounted boards of better quality are produced. Therefore, by using actual component data having such operating parameters as training data for training the learning model, the learning model can be further optimized. As a result, when the learning model is selected by the model selection unit, the estimation accuracy of the operating parameters can be improved.

In addition, each of the multiple learning models may be associated with a different period, and the learning unit may perform learning on the learning model corresponding to the period during which the actual production data was acquired.

For example, one of the multiple learning models is associated with the entire period (e.g., the entire period from the past to the present), and each of the remaining at least one learning model is associated with a different era. The different eras are, for example, the 1990s, 2000s, 2010s, etc. As a result, from these learning models, a learning model associated with the entire period or any of the eras is selected and used to estimate the operating parameters. Therefore, appropriate operating parameters according to the period can be estimated for the components to be mounted.

In addition, each of the multiple learning models may be associated with a different piece of production equipment, and the learning unit may perform learning on a learning model corresponding to production equipment including a component mounting device using the actual production data.

For example, the production equipment may be a component mounting device, a component mounting line including a component mounting device, or equipment including multiple component mounting lines. In this case, for example, one of the multiple learning models is associated with the production equipment including all component mounting lines arranged in the factory, and each of the remaining at least one learning model is associated with a different component mounting line. As a result, from these learning models, a learning model associated with all component mounting lines or any one of the component mounting lines is selected and used to estimate the operating parameters. Therefore, it is possible to estimate appropriate operating parameters for the components to be mounted according to the production equipment.

In addition, each of the multiple learning models may be associated with a different type of mounting board, and the learning unit may learn from the learning model corresponding to the type of mounting board produced using the actual production data.

For example, the type of mounting board is a mass production type or a prototype type. In this case, for example, one of the multiple learning models is associated with the mass production type, and the remaining learning model is associated with the prototype type. As a result, from these learning models, the learning model associated with the mass production type or the prototype type is selected and used to estimate the operating parameters. Therefore, it is possible to estimate appropriate operating parameters for the mounting target components according to the type of mounting board.

The following describes the implementation form in detail with reference to the drawings.

Note that the embodiments described below are all comprehensive or specific examples. The numerical values, shapes, materials, components, component placement and connection forms, steps, and order of steps shown in the following embodiments are merely examples and are not intended to limit the present disclosure. Furthermore, among the components in the following embodiments, components that are not described in an independent claim that indicates the highest concept are described as optional components.

The figures are schematic diagrams and are not necessarily strict illustrations. In each figure, the same components are denoted by the same reference numerals.

(Embodiment 1)
[Production System]
FIG. 1 is a diagram showing an example of a configuration of a production system according to the present embodiment.

The production system 1 in this embodiment has three component mounting lines L1 to L3 and a production data generation device 100.

Each of the component mounting lines L1 to L3 is an example of a production facility for mounted circuit boards, and produces mounted circuit boards by performing solder printing, component mounting, reflow, and other operations on boards brought in from the upstream side, and then transports the produced mounted circuit boards downstream.

The production data generating device 100 generates and outputs production data for producing mounted boards for each of the component mounting lines L1 to L3. The production data generating device 100 may communicate with the component mounting lines L1 to L3 wirelessly or via wires. The wireless communication may be Wi-Fi (registered trademark), Bluetooth (registered trademark), ZigBee, or a specific low-power radio.

The component mounting line L1 includes a line management device 200, a board supply device M1, a board transfer device M2, a solder printing device M3, component mounting devices M4 and M5, a reflow device M6, and a board recovery device M7. The devices other than the line management device 200 included in the component mounting line L1 are arranged in the order of the board supply device M1, the board transfer device M2, the solder printing device M3, the component mounting devices M4 and M5, the reflow device M6, and the board recovery device M7, and are connected in series. The devices other than the line management device 200 are hereinafter referred to as working devices. The component mounting line L1 does not need to include all of the above working devices as long as it includes the board supply device M1, at least one component mounting device, and the board recovery device M7. In addition to the above working devices, the component mounting line L1 may also include a solder application device that applies solder to a board, a component insertion machine that mounts radial or axial components on a board, and the like.

The line management device 200 acquires the production data generated by the production data generating device 100 from the production data generating device 100, and causes each work device included in the component mounting line L1 to produce a mounted board based on the production data.

The board supply device M1 supplies the boards used for the mounted boards produced on the component mounting line L1 to the solder printing device M3 via the board delivery device M2. The solder printing device M3 performs the solder printing work described above. In other words, the solder printing device M3 screen-prints solder on the boards delivered from the board delivery device M2.

Each of the component mounting devices M4 and M5 performs the above-mentioned component mounting work of mounting at least one component on a board. Although the component mounting line L1 is equipped with two component mounting devices M4 and M5, the number of devices is not limited to two, and may be one, or three or more. It can also be said that the mounted board is essentially produced by the component mounting work performed by these component mounting devices M4 and M5.

The reflow device M6 performs the reflow operation described above. That is, the reflow device M6 heats the boards on which components are mounted, which have been brought in from the component mounting devices M4 and M5, hardening the solder on the board and bonding the electrodes of the board to the components. Specifically, the reflow device M6 melts and solidifies the solder used to bond the components by heating according to a specified heating profile. This solders the components to the board. The board recovery device M7 recovers the board after solder bonding has been performed from the reflow device M6.

Component mounting lines L2 and L3 have the same configuration as component mounting line L1. In this embodiment, component mounting lines L1 to L3 each have the same configuration, but may have different configurations. Also, in this embodiment, component mounting lines L1 to L3 are equipped with a line management device 200, but the line management device 200 may be provided independently of each of component mounting lines L1 to L3, or may be incorporated into each of component mounting lines L1 to L3.

[Component mounting equipment]
FIG. 2 is a diagram showing an example of the configuration of the component mounting device M4. In this embodiment, the component mounting device M5 also has a similar configuration to the component mounting device M4. In this embodiment, the transport direction of the board B is referred to as the X-axis direction, and the direction perpendicular to the X-axis direction is referred to as the Y-axis direction. The X-axis direction and the Y-axis direction are directions along a horizontal plane. Furthermore, the direction perpendicular to the X-axis direction and the Y-axis direction is referred to as the Z-axis direction. The positive and negative sides of the X-axis direction are the downstream and upstream sides in the transport direction of the board B, respectively, and the positive and negative sides of the Y-axis direction are the rear side (or the back side) and the front side (or the near side) in the front-rear direction, respectively. The positive and negative sides of the Z-axis direction are the upper side and the lower side in the up-down direction, respectively. In FIG. 2, the upper surface of the component mounting device M4 is shown.

The component mounting device M4 comprises a base 4, a board transport mechanism 5, two component supply units 6, two X-axis beams 9, a Y-axis beam 8, two mounting heads 10, two component recognition cameras 11, and two board recognition cameras 12.

The board transport mechanism 5 has two rails along the X-axis direction and is disposed in the center of the base 4. The board transport mechanism 5 transports the board B brought in from the upstream side, and positions and holds the board B in a position for performing component mounting work.

The two component supply units 6 are arranged on either side of the board transport mechanism 5 in the Y-axis direction. In each component supply unit 6, multiple feeders 7 are arranged in parallel along the X-axis direction. The feeders 7 feed the component tape containing the components at a pitch in the tape feed direction, thereby supplying the components to a position where the components are picked up by the mounting head 10 (hereinafter referred to as the component pick-up position).

Note that the part supply unit 6 may be equipped with a tray feeder, stick feeder, bulk feeder, or the like. A tray feeder supplies parts from a tray that stores the parts. A stick feeder supplies parts from a stick case that stores the parts. A bulk feeder supplies parts from a bulk case that stores the parts.

The Y-axis beam 8 is disposed along the Y-axis direction at one end in the X-axis direction (the right side in FIG. 2) on the top surface of the base 4. The two X-axis beams 9 are connected to the Y-axis beam 8 along the X-axis direction so as to be freely movable in the Y-axis direction.

The mounting head 10 is attached to each of the two X-axis beams 9 so as to be movable in the X-axis direction. The mounting head 10 has multiple suction units 10a that can move up and down while suctioning and holding components. A suction nozzle 10b is provided at the tip of each of the suction units 10a (see Figure 3).

Each of the two mounting heads 10 moves in the X-axis direction and the Y-axis direction by driving the Y-axis beam 8 and the X-axis beam 9. As a result, each of the two mounting heads 10 picks up a component from the component pick-up position of the feeder 7 arranged in the component supply unit 6 corresponding to that mounting head 10 by suction nozzle 10b, and mounts the component at the mounting point (or mounting position) of the board B positioned by the board transport mechanism 5.

Each of the two component recognition cameras 11 is disposed between one of the two component supply units 6 and the board transport mechanism 5. The component recognition camera 11 captures an image of a component when the mounting head 10, which has picked up a component from the component supply unit 6, moves above the component recognition camera 11. In other words, the component recognition camera 11 captures an image of the component held by the mounting head 10, thereby recognizing the holding posture of the component.

The board recognition camera 12 is attached to the plate 9a to which the mounting head 10 is attached. Therefore, the board recognition camera 12 moves integrally with the mounting head 10. As the mounting head 10 moves, the board recognition camera 12 moves above the board B positioned by the board transport mechanism 5, and captures an image of a board mark (not shown) provided on the board B to recognize the position of the board B. When the mounting head 10 mounts components on the board B, the mounting position is corrected based on the recognition results of the components by the component recognition camera 11 and the recognition results of the position of the board B by the board recognition camera 12.

Figure 3 is a diagram partially showing an example of a cross section taken along line A-A in Figure 2. The component mounting device M4 has the function of mounting a component P onto a board B.

As shown in FIG. 3, the parts supply unit 6 comprises a feeder base 13a, a plurality of feeders 7 mounted on the feeder base 13a, and a trolley 13 supporting the feeder base 13a.

The cart 13 is configured to be detachable from the component mounting devices M4 and M5, and further includes a cassette holder 15. The cassette holder 15 is configured to be able to hold a plurality of component reels C. The component reels C store the component tape 14 in a wound state. Each of the plurality of component reels C is held at an upper holding position Hu or a lower holding position Hd of the cassette holder 15. The component tape 14 pulled out from the component reel C held by the cassette holder 15 is attached to the feeder 7. The feeder 7 may be disposed on a feeder base 13a provided on the base 4 without using the cart 13. Also, the cart 13 may hold the component reel C instead of the cassette holder 15.

In this embodiment, as described above, component mounting devices M4 and M5 each have the same configuration, but may have different configurations.

[Functional configuration of production data generation device and component mounting line]
FIG. 4 is a block diagram showing the functional configuration of the production data generating device 100 and each of the component mounting lines L1 to L3.

The production data generation device 100 includes a control unit 101, a data generation unit 102, a model selection unit 103, a learning unit 104, a parameter estimation unit 105, a display unit 106, an input/output unit 107, a data acquisition unit 108, a production data storage unit DB1, a learning model storage unit DB2, and a part library storage unit DB3.

The model selection unit 103 selects at least one learning model from the multiple learning models stored in the learning model storage unit DB2.

The learning model storage unit DB2 stores the above-mentioned multiple learning models. Each of these multiple learning models is a different model and indicates the relationship between the operating conditions of the component mounting device M4 or M5 for mounting the component P on the board B and the component P.

The parameter estimation unit 105 estimates operating parameters that are operating conditions of the component mounting device M4 or M5. In other words, the parameter estimation unit 105 estimates operating parameters for mounting the target component P on the board B based on at least one learning model selected by the model selection unit 103 and component information related to the target component P to be mounted on the board B.

The data generating unit 102 generates production data including component data having the above-mentioned component information and operation parameters. Here, the production data indicates, for example, the mounting order of at least one component P to be mounted on the board B and the positions at which the components P are mounted on the board B (i.e., the above-mentioned mounting positions), and includes component data for each of the at least one component P. In addition, the component data for each component P is held in the component library holding unit DB3. That is, the component library holding unit DB3 holds a component library including component data for each of a plurality of types of components P. Therefore, if the component data for the component P to be mounted on the board B is included in the component library, the data generating unit 102 selects the component data for the component P from the component library and generates production data including the selected component data. On the other hand, if the component data for the component P to be mounted on the board B is not included in the component library, the data generating unit 102 generates production data including the component data for the component P having the component information and the operation parameters estimated as described above.

The data generation unit 102 generates and outputs such production data for each of the component mounting lines L1 to L3, and stores the production data in the production data storage unit DB1.

The data acquisition unit 108 acquires actual production data used by the component mounting device M4 or M5, including actual component data having component information regarding the mounted component P and operating parameters used to mount the mounted component P.

The actual production data is data that has been modified or tuned, for example, from the production data generated by the data generation unit 102. That is, the component mounting device M4 or M5 included in each of the component mounting lines L1 to L3 mounts the component P on the board B based on the production data, but the mounting may result in the production of a defective mounted board. In such a case, the component data included in the production data is modified or tuned in each of the component mounting lines L1 to L3 so as to reduce the frequency of occurrence of defective products. By such modification or tuning, actual production data including actual component data is generated. The component mounting device M4 or M5 of each of the component mounting lines L1 to L3 uses the actual production data to mount the component P on the board B. The data acquisition unit 108 acquires the actual production data generated in this manner from each of the component mounting lines L1 to L3.

The learning unit 104 generates or updates the learning model through machine learning. Machine learning will hereinafter be simply referred to as learning. For example, the learning unit 104 generates a learning model through learning, and stores the generated learning model in the learning model holding unit DB2. The learning unit 104 also selects one learning model from the multiple learning models held in the learning model holding unit DB2, and updates the selected learning model through learning. For such learning by the learning unit 104, actual production data used in each of the component mounting lines L1 to L3 and acquired by the data acquisition unit 108 is used.

That is, the learning unit 104 updates the relationship indicated by the learning model corresponding to the actual production data acquired by the data acquisition unit 108, among the multiple learning models stored in the learning model storage unit DB2. At this time, the learning unit 104 performs the update by learning using the actual part data included in the actual production data as teacher data. Note that the learning model may be, for example, a neural network, a decision tree, or another model.

The display unit 106 displays the production data stored in the production data storage unit DB1, the part library stored in the part library storage unit DB3, etc. Specific examples of the display unit 106 include, but are not limited to, a liquid crystal display, a plasma display, or an organic EL (Electro-Luminescence) display.

The input/output unit 107 receives input data based on, for example, operations by an operator of the production system 1, and outputs the input data to the control unit 101. Such an input/output unit 107 has, for example, a keyboard, a touch sensor, a touchpad, or a mouse. The input/output unit 107 also outputs data to the component mounting lines L1 to L3, and inputs data from the component mounting lines L1 to L3. The production data generated by the data generation unit 102 may be output to each of the component mounting lines L1 to L3 via this input/output unit 107. The input/output unit 107 may also acquire the above-mentioned component information based on operations by the operator, and output the information to the parameter estimation unit 105.

The control unit 101 controls each component other than the control unit 101 included in the production data generating device 100. For example, the control unit 101 controls each component based on operator input data received by the input/output unit 107.

The production data storage unit DB1, the learning model storage unit DB2, and the part library storage unit DB3 are recording media for storing the production data, the learning model, and the part library. For example, such recording media may be a hard disk, a ROM (Read Only Memory), a RAM (Random Access Memory), or a semiconductor memory. Note that such recording media may be volatile or non-volatile.

The component mounting line L1 includes a work control unit 211, an input/output unit 212, a display unit 213, a work mechanism 214, and a production data storage unit DB4. Each component included in the component mounting line L1 other than the work mechanism 214 may be included in the line management device 200, or may be included in any work device other than the line management device 200.

The input/output unit 212, like the input/output unit 107 of the production data generating device 100, accepts input data based on operations by the operator of the production system 1, for example, and outputs the input data to the work control unit 211. Such an input/output unit 212 may have, for example, a keyboard, a touch sensor, a touchpad, or a mouse. In addition, the input/output unit 212 outputs data to the production data generating device 100 and inputs data from the production data generating device 100. For example, the input/output unit 212 acquires production data from the production data generating device 100 and stores the production data in the production data storage unit DB4.

The display unit 213 displays the production data stored in the production data storage unit DB4. Examples of the display unit 213 include, but are not limited to, a liquid crystal display, a plasma display, or an organic EL display.

The working mechanism 214 consists of mechanisms such as a mounting head 10 and a feeder 7 for producing mounted substrates.

The work control unit 211 controls each component other than the work control unit 211 included in the component mounting line L1. For example, the work control unit 211 controls each component based on the operator's input data received by the input/output unit 212. For example, the work control unit 211 causes the work mechanism 214 to perform at least one of the above-mentioned solder printing work, component mounting work, and reflow work based on the production data stored in the production data storage unit DB4. In addition, the work control unit 211 corrects or tunes the production data stored in the production data storage unit DB4 according to the operator's input data received by the input/output unit 212. This generates the above-mentioned actual production data. The work control unit 211 controls the input/output unit 212 to output the actual production data from the input/output unit 212 to the production data generation device 100.

The production data storage unit DB4 is a recording medium for storing production data. For example, such a recording medium is a hard disk, a RAM, a ROM, or a semiconductor memory. Note that such a recording medium may be volatile or non-volatile.

[Parts Library]
FIG. 5 is a diagram showing an example of a part library.

The component library consists of multiple component data Dc. Each of the multiple component data Dc is data for one type of component P and is associated with a component code for identifying the type of component P. Such component data Dc has component information d related to the component P and operating parameters m which are the operating conditions of the component mounting device M4 or M5 for mounting the component P on the board B. Note that images, numerical values, terms, etc. are shown in the blank spaces of each item in the component data Dc shown in Figure 5.

The part information d includes, for example, a shape drawing d1 of part P, size data d2 and part parameters d3.

The shape diagram d1 illustrates the external shape of the component P corresponding to the component data Dc. The size data d2 numerically indicates information about the size of the component P, such as the external dimensions, number of leads, lead pitch, lead length, lead width, and component height.

The component parameters d3 are attribute information about the component P. Such component parameters d3 include component attributes d31, which are information about the component P itself, and tape information d32, which is information about the component tape 14 for supplying the component P by the feeder 7. The component attributes d31 indicate, for example, the polarity, polarity mark, mark position, component type, and shape type of the component P. The tape information d32 includes, for example, the tape material of the component tape 14, the tape width indicating the width dimension of the component tape 14, the feed interval indicating the tape feed pitch of the component tape 14 by the feeder 7, and information about the color and material of the component tape 14.

In this way, the part information d in this embodiment indicates at least one of the dimensions, shape, appearance, and type of the part P corresponding to the part information d, and the supply form for supplying the part P. The supply form corresponds to, for example, tape information d32.

The operation parameter m is a machine parameter that specifies the operation mode of the component mounting device M4 or M5 when mounting the component P on the board B. In the example shown here, the operation parameter m includes model information m1 indicating the type of the component mounting device M4 or M5, and nozzle setting information m2 indicating the type of suction nozzle 10b to be used. Furthermore, the operation parameter m includes a speed parameter m3, recognition information m4, gap information m5, suction information m6, and mounting information m7.

The speed parameter m3 includes the lifting speed when the suction nozzle 10b picks up the component P, the mounting speed when the mounting head 10 transports the component P, and the tape feed speed when the feeder 7 feeds the component tape 14. The recognition information m4 is a parameter that specifies the mode of component recognition. Specifically, the recognition information m4 includes a camera type indicating the type of component recognition camera 11 used, an illumination mode indicating the illumination mode when the component recognition camera 11 captures an image, and a recognition speed indicating the speed of movement of the mounting head 10 when capturing an image. The gap information m5 includes a suction gap when the suction nozzle 10b picks up the component P, and a mounting gap when the picked-up component P is mounted on the board B.

The suction information m6 includes a suction position offset indicating the offset amount when the suction nozzle 10b picks up the component P, and a suction angle. The mounting information m7 indicates the pressure load when the component P picked up by the suction nozzle 10b is mounted on the board B as a mounting load.

Thus, the operational parameter m in this embodiment is a parameter related to at least one of the transport, recognition, pickup, and mounting of component P by component mounting device M4 or M5.

Note that the part information d and operation parameters m included in the part data Dc in FIG. 5 are each just an example, and may indicate information other than the information shown in FIG. 5, may indicate both the information shown in FIG. 5 and other information, or may indicate only a portion of the information shown in FIG. 5. Furthermore, the number of pieces of information included in each of the part information d and operation parameters m may be one or more.

In this embodiment, when the data generation unit 102 generates production data, it selects component data Dc corresponding to the component P to be mounted on the board B from the component library held in the component library holding unit DB3, and generates production data including the component data Dc.

Furthermore, if the component library does not contain component data Dc corresponding to the component P to be mounted on the board B, the data generation unit 102 generates production data using the component data Dc that is not included in the component library. For example, the data generation unit 102 generates production data using component data Dc having component information d acquired by the input/output unit 107 and operation parameters m estimated by the parameter estimation unit 105 for the component information d. The component information d acquired by the input/output unit 107 may be, for example, information acquired from CAD (Computer Aided Design) information regarding the component P to be mounted on the board B, or may be information input by an operator's operation.

In addition, a default operation parameter m may be set in the part data Dc of the part library. If a default operation parameter m is set in the part data Dc selected from the part library, the data generation unit 102 may cause the parameter estimation unit 105 to estimate the operation parameter m corresponding to the part information d included in the part data Dc. When the operation parameter m is estimated by the parameter estimation unit 105, the data generation unit 102 replaces the default operation parameter m included in the part data Dc with the operation parameter m estimated by the parameter estimation unit 105. Then, the data generation unit 102 generates production data using the part data Dc in which the operation parameter m has been replaced.

Note that all information (i.e., parameters) included in the operation parameters m of the part data Dc may be default, or only some of the parameters may be default. If only some of the parameters are default, the parameter estimation unit 105 may estimate parameters to replace some of the default parameters. The data generation unit 102 replaces some of the operation parameters m included in the part data Dc with parameters estimated by the parameter estimation unit 105. The data generation unit 102 then generates production data using the part data Dc in which some of the parameters have been replaced.

[Production data]
FIG. 6 is a diagram showing an example of the production data Dp.

In the production data Dp, for example, the component names and component codes of the multiple components P to be mounted on the board B are arranged in the order in which the multiple components P are mounted. The component code of the component P is a code for identifying the component data Dc of the component P from the component library. In addition, the production data Dp indicates the mounting coordinates, mounting angle, identification information of the feeder 7, and identification information of the mounting head 10 or suction nozzle 10b for each of the multiple components P. The mounting coordinates of the component P are the position on the board B where the component P is mounted or mounted, and are also called the mounting point, mounting position, or mounting position. The mounting angle of the component P is the angle at which the suction nozzle 10b that picks up the component P rotates around the central axis of the suction nozzle 10b as the rotation axis in order to mount the component P on the board B. The identification information of the feeder 7 corresponding to the component P is information for identifying the feeder 7 that supplies the component P. The identification information of the mounting head 10 corresponding to the component P is information for identifying the mounting head 10 used to mount the component P on the board B.

For example, the production data Dp indicates, for the component P to be first mounted on the board B, the component name of that component P, "Component A", the component code "C001", the mounting coordinates "x1, y1", the mounting angle "θ1", the identification information of the feeder 7, "F2", and the identification information of the mounting head 10, "H3".

Moreover, the production data Dp in this embodiment includes component data Dc for each of the multiple components P to be mounted on the board B. For example, the production data Dp includes component data Dc associated with the component code "C001" of the component P with the component name "Part A."

[Learning model]
7A and 7B are diagrams showing an example of a plurality of learning models stored in the learning model storage unit DB2 and managed on a time basis. The learning models in this embodiment are also referred to as motion parameter models hereinafter.

For example, each of the multiple operation parameter models Pm11 to Pm14, which are multiple learning models stored in the learning model storage unit DB2, is managed in time units, as shown in FIG. 7A. Specifically, the operation parameter model Pm11 is generated by learning using actual production data used between 1990 and 1999. Similarly, the operation parameter model Pm12 is generated by learning using actual production data used between 2000 and 2009, and the operation parameter model Pm13 is generated by learning using actual production data used between 2010 and the present. Moreover, the operation parameter model Pm14 is generated by learning using actual production data used between 1990 and the present.

In addition, each of the multiple operation parameter models Pm21 to Pm25, which are multiple learning models stored in the learning model storage unit DB2, may be managed on a time basis, as shown in Figure 7B, and may also be managed in chronological order.

For example, the operation parameter model Pm21 is generated by learning using actual production data used on July 1, 2019. The operation parameter model Pm22 is generated by learning using the operation parameter model Pm21 and actual production data used on July 2, 2019. In other words, the operation parameter model Pm22 is generated by learning using actual production data used between July 1 and 2, 2019.

Similarly, the operation parameter model Pm23 is generated by learning using the operation parameter model Pm22 and the actual production data used on July 3, 2019. In other words, the operation parameter model Pm23 is generated by learning using the actual production data used between July 1 and 3, 2019. The operation parameter model Pm24 is generated by learning using the operation parameter model Pm22 and the actual production data used on July 4, 2019. In other words, the operation parameter model Pm24 is generated by learning using the actual production data used between July 1 and 2, 2019 and on the 4th. Therefore, the actual production data used on July 3, 2019 is not reflected in the operation parameter model Pm24.

The operation parameter model Pm25 is generated by learning using the operation parameter model Pm24 and the actual production data used on July 5, 2019. In other words, the operation parameter model Pm25 is generated by learning using the actual production data used between July 1 and 2, and between July 4 and 5, 2019. Therefore, like the operation parameter model Pm24, the operation parameter model Pm25 does not reflect the actual production data used on July 3, 2019.

Thus, in this embodiment, each of the multiple learning models (i.e., operation parameter models) is associated with a different period. In other words, the multiple operation parameter models are managed in time units. In this case, when learning is performed using the actual part data included in the actual production data as teacher data, the learning unit 104 learns from the learning model corresponding to the period during which the actual production data was acquired.

For example, in the example shown in Fig. 7A, when the data acquisition unit 108 of the production data generating device 100 acquires actual production data in 2011, the learning unit 104 learns the operation parameter models Pm13 and Pm14 corresponding to the year 2011. Also, in the example shown in Fig. 7B, when the data acquisition unit 108 acquires actual production data on July 3, 2019, the learning unit 104 learns the operation parameter model Pm22 or Pm23 corresponding to the period during which the actual production data was acquired.

Figure 8 is a diagram showing an example of multiple learning models stored in the learning model storage unit DB2 and managed on a production equipment basis.

For example, each of the multiple operation parameter models Pm31 to Pm34, which are multiple learning models stored in the learning model storage unit DB2, may be managed for each production facility as shown in FIG. 8. Specifically, the operation parameter model Pm31 is generated by learning using the actual production data used in the component mounting line L1. Similarly, the operation parameter model Pm32 is generated by learning using the actual production data used in the component mounting line L2, and the operation parameter model Pm33 is generated by learning using the actual production data used in the component mounting line L3. Moreover, the operation parameter model Pm34 is generated by learning using the actual production data used in each of all the component mounting lines L1 to L3.

In this embodiment, the multiple learning models (i.e., operation parameter models) are associated with different production equipment. In other words, the multiple operation parameter models are managed on a production equipment basis. As described above, the production equipment may be a component mounting line or a collection of multiple component mounting lines. The production equipment may be one or multiple component mounting devices, a floor on which multiple component mounting devices or component mounting lines are arranged, or a factory. In this case, when learning is performed using the actual component data included in the actual production data as teacher data, the learning unit 104 learns the learning model corresponding to the production equipment including the component mounting device M4 or M5 that used the actual production data.

For example, in the example shown in Figure 8, when the data acquisition unit 108 of the production data generating device 100 acquires actual production data from component mounting line L2, the learning unit 104 learns the operation parameter models Pm32 and Pm34 corresponding to the component mounting line L2.

Figure 9A is a diagram showing an example of multiple learning models stored in the learning model storage unit DB2 and managed by production type.

For example, each of the multiple operation parameter models Pm41 to Pm44, which are multiple learning models stored in the learning model storage unit DB2, may be managed by production type as shown in FIG. 9A. The production types include, for example, a prototype type and a mass production type. The prototype type is a type of mounting board produced as a prototype, and the mass production type is a type of mounting board produced as a mass-produced product. In the prototype type, operation parameters that emphasize quality are set compared to the mass production type, and in the mass production type, operation parameters that emphasize productivity tend to be set compared to the prototype type. Therefore, since the prototype type and the mass production type have different operation parameters set even for the same mounting board, learning by production type improves the estimation accuracy. Specifically, the operation parameter model Pm41 is generated by learning using the actual production data used in the production of the mounting board of the prototype type T1. The operation parameter model Pm42 is generated by learning using the actual production data used in the production of the mounting board of the prototype type T2, which is different from the prototype type T1. The operation parameter model Pm43 is generated by learning using the actual production data used in the production of mass-produced type mounting boards, while the operation parameter model Pm44 is generated by learning using the actual production data used in the production of each of all production types of mounting boards.

Figure 9B is a diagram showing an example of multiple learning models stored in the learning model storage unit DB2 and managed by production type and by production equipment.

For example, each of the multiple operation parameter models Pm51 to Pm54, which are multiple learning models stored in the learning model storage unit DB2, may be managed for each combination of production type and production equipment, as shown in Figure 9B. Note that, in order to manage each of the multiple learning models for each production type, an item for setting the production type may be provided in the part data Dc.

Specifically, the operation parameter model Pm51 is generated by learning using actual production data used in the production of prototype type T1 mounted boards by component mounting line L1. The operation parameter model Pm52 is generated by learning using actual production data used in the production of prototype type T2 mounted boards by component mounting line L2. The operation parameter model Pm53 is generated by learning using actual production data used in the production of mass-produced type mounted boards by component mounting line L3. Moreover, the operation parameter model Pm54 is generated by learning using actual production data used in the production of all production types of mounted boards by all component mounting lines L1 to L3.

Thus, in this embodiment, each of the multiple learning models (i.e., operation parameter models) is associated with a different production type of the mounting board. In other words, the multiple learning models are managed on a production type basis. In this case, when learning is performed using the actual component data included in the actual production data as training data, the learning unit 104 performs learning on the learning model corresponding to the type of mounting board produced using the actual production data.

For example, in the example shown in Figure 9A, when the data acquisition unit 108 of the production data generating device 100 acquires actual production data for prototype type T1, the learning unit 104 learns the operation parameter models Pm41 and Pm44 corresponding to that prototype type T1. Also, in the example shown in Figure 9B, when the data acquisition unit 108 acquires actual production data for prototype type T2 from component mounting line L2, the learning unit 104 learns the operation parameter models Pm52 and Pm54 corresponding to that prototype type T2 and component mounting line L2.

[Processing overview and flow]
FIG. 10A is a diagram for explaining an outline of the estimation process of the motion parameter m in this embodiment.

The parameter estimation unit 105 acquires, for example, component information d of the component P from the input/output unit 107. The model selection unit 103 selects, for example, one operation parameter model Pm from the multiple operation parameter models Pm held in the learning model holding unit DB2. Each of the multiple operation parameter models Pm may be any of the operation parameter models Pm11 to Pm14, Pm21 to Pm25, Pm31 to Pm34, Pm41 to Pm44, and Pm51 to Pm54 shown in Figures 7A to 9B. The parameter estimation unit 105 uses the acquired component information d and the selected operation parameter model Pm to estimate an operation parameter m, which is an operation condition of the component mounting device M4 or M5 for mounting the component P indicated by the component information d on the board B. Then, the parameter estimation unit 105 outputs component data Dc including the estimated operation parameter m and the component information d.

For example, when multiple operation parameter models Pm are managed in time units as shown in FIG. 7A, the model selection unit 103 may select one operation parameter model Pm according to the manufacturing date of the part P indicated by the part information d. For example, if the manufacturing date is in the 1990s, the model selection unit 103 may select the operation parameter model Pm11 shown in FIG. 7A. Also, if the manufacturing date is unknown, the model selection unit 103 may select the operation parameter model Pm14 shown in FIG. 7A. This makes it possible to select an appropriate operation parameter model Pm for estimating the operation parameter m for the part P.

In addition, when multiple operation parameter models Pm are managed in chronological order as shown in FIG. 7B, the model selection unit 103 may select one operation parameter model Pm updated on the most recent date. For example, the model selection unit 103 may select the operation parameter model Pm25 shown in FIG. 7B. In addition, when many mounting boards using a component similar to the component P were produced on July 3, 2019, the model selection unit 103 may select the operation parameter model Pm23 shown in FIG. 7B. This makes it possible to select an appropriate operation parameter model Pm for estimating the operation parameter m for the component P.

In addition, when multiple operation parameter models Pm are managed on a production facility basis as shown in FIG. 8, the model selection unit 103 may select one operation parameter model Pm associated with the production facility that produces the mounted board using the component P. For example, when the production facility that produces the mounted board using the component P is the component mounting line L2, the model selection unit 103 may select the operation parameter model Pm32 shown in FIG. 8. In addition, when the production facility that produces the mounted board using the component P is not only the component mounting line L2, the model selection unit 103 may select the operation parameter model Pm34 shown in FIG. 8. This makes it possible to select an appropriate operation parameter model Pm for estimating the operation parameter m for the component P.

In addition, when multiple operation parameter models Pm are managed by production type as shown in FIG. 9A, the model selection unit 103 may select one operation parameter model Pm associated with the production type of the mounting board produced using the component P. For example, when the production type of the mounting board produced using the component P is mass production type, the model selection unit 103 may select the operation parameter model Pm43 shown in FIG. 9A. In addition, when the production type of the mounting board produced using the component P is not only mass production type, the model selection unit 103 may select the operation parameter model Pm44 shown in FIG. 9A. In this way, it is possible to select an appropriate operation parameter model Pm for estimating the operation parameter m for the component P.

In this embodiment, the model selection unit 103 selects one operation parameter model Pm, but may select multiple operation parameter models Pm for estimating different operation conditions. For example, the operation parameter m includes different parameters such as a speed parameter m3 and recognition information m4 as shown in FIG. 5. Therefore, the model selection unit 103 may select, for example, an operation parameter model Pm for estimating the speed parameter m3 and an operation parameter model Pm for estimating the recognition information m4. In this case, the parameter estimation unit 105 may estimate the speed parameter m3 using the part information d and the operation parameter model Pm for the speed parameter m3, and may estimate the recognition information m4 using the part information d and the operation parameter model Pm for the recognition information m4.

In addition, the model selection unit 103 may automatically select the operation parameter model Pm as described above, or may make the selection in response to an operator's operation of the input/output unit 107.

Figure 10B is a diagram for explaining an overview of the learning process of the operation parameter model Pm in this embodiment.

The learning unit 104 acquires actual production data from one of the component mounting lines L1 to L3 via the data acquisition unit 108, for example. The actual production data includes actual component data Dcu, as described above. That is, the learning unit 104 acquires actual component data Dcu. The actual component data Dcu is component data Dc that was used in mounting the component P on the board B by the component mounting device M4 or M5, and that has been modified through such use. For example, this actual component data Dcu includes an operation parameter mu as a modified operation parameter m, and in the operation parameter mu, the pickup speed, which is an operation condition, has been modified from V1 to V2.

Next, the learning unit 104 selects an operation parameter model Pm corresponding to the actual component data Dcu from the multiple operation parameter models Pm stored in the learning model storage unit DB2. For example, as shown in FIG. 7A and FIG. 7B, the learning unit 104 selects an operation parameter model Pm corresponding to the period during which the actual production data including the actual component data Dcu was acquired. Alternatively, as shown in FIG. 8, the learning unit 104 selects an operation parameter model Pm corresponding to the production equipment including the component mounting device M4 or M5 that used the actual production data including the actual component data Dcu. Alternatively, as shown in FIG. 9A and FIG. 9B, the learning unit 104 selects an operation parameter model Pm corresponding to the production type of the mounting board produced using the actual production data including the actual component data Dcu.

Then, the learning unit 104 updates the selected operation parameter model Pm by learning using the acquired actual part data Dcu as teacher data. In other words, the relationship between the part information d and the operation conditions indicated by the operation parameter model Pm is updated. As a result, the operation parameter model Pmu after learning is generated. The learning unit 104 replaces the operation parameter model Pm selected as described above and held in the learning model holding unit DB2 with the learned operation parameter model Pmu. As a result, the learned operation parameter model Pmu is stored in the learning model holding unit DB2 as a new operation parameter model Pm.

Figure 11 is a diagram showing an example of the overall processing in this embodiment. In the example shown in Figure 11, multiple operation parameter models Pm stored in the learning model storage unit DB2 are managed on a time basis or a production type basis.

As in the example shown in FIG. 10A, the parameter estimation unit 105 estimates the operation parameter m of the part P using the operation parameter model Pm selected by the model selection unit 103, and generates part data Dc including the part information d of the part P and the operation parameter m. Then, the data generation unit 102 generates production data Dp including the part data Dc, and outputs the production data Dp to, for example, the part mounting line L1 via the input/output unit 107. Note that in the example shown in FIG. 11, the production data Dp is output to the part mounting line L1, but may be output to another part mounting line L2 or L3.

When the component mounting devices M4 and M5 included in the component mounting line L1 acquire production data Dp from the input/output unit 107 of the production data generating device 100, they mount at least one component P on the board B based on the production data Dp to produce a mounted board. At this time, in the component mounting line L1, the component data Dc included in the production data Dp is corrected so that, for example, the defective occurrence rate of the mounted board is reduced. As a specific example, the pickup speed V1 included in the operation parameter m of the component data Dc is corrected to V2. As a result, actual production data including actual component data Dcu having the operation parameter mu is generated. The actual component data Dcu is stored in the component library holding unit DB3 as new component data Dc. Specifically, the data acquisition unit 108 of the production data generating device 100 acquires actual production data from the component mounting line L1, and stores the actual component data Dcu included in the actual production data in the component library holding unit DB3 as new component data Dc. It should be noted that correction of the component data Dc in the component mounting line L1 is not necessarily performed, and if no correction is made, the component mounting line L1 generates actual production data that includes the component data Dc acquired from the production data generation device 100 as actual component data Dcu.

When the data acquisition unit 108 acquires the actual production data, the learning unit 104 selects an operation parameter model Pm corresponding to the actual part data Dcu included in the actual production data from the learning model holding unit DB2, as in the example shown in FIG. 10B. Then, the learning unit 104 performs learning on the selected operation parameter model Pm using the actual part data Dcu, and stores the learned operation parameter model Pmu in the learning model holding unit DB2.

Figure 12 is a diagram showing another example of the overall processing in this embodiment. In the example shown in Figure 12, multiple operation parameter models Pm stored in the learning model storage unit DB2 are managed on a production facility basis.

As in the example shown in FIG. 10A, the parameter estimation unit 105 estimates the operation parameter m of the part P for each production facility using the operation parameter model Pm selected by the model selection unit 103, and generates part data Dc including part information d of the part P and its operation parameter m.

For example, the parameter estimation unit 105 estimates the operation parameter m based on the operation parameter model Pm for the component mounting line L1, i.e., the operation parameter model Pm31 shown in FIG. 8, to generate the component data Dc. Then, the data generation unit 102 generates production data Dp including the component data Dc, and outputs the production data Dp to, for example, the component mounting line L1 via the input/output unit 107. Furthermore, the parameter estimation unit 105 estimates the operation parameter m based on the operation parameter model Pm for the component mounting line L2, i.e., the operation parameter model Pm32 shown in FIG. 8, to generate the component data Dc. Then, the data generation unit 102 generates production data Dp including the component data Dc, and outputs the production data Dp to, for example, the component mounting line L2 via the input/output unit 107. Furthermore, the parameter estimation unit 105 estimates the operation parameter m based on the operation parameter model Pm for the component mounting line L3, i.e., the operation parameter model Pm33 shown in FIG. 8, to generate the component data Dc. Then, the data generation unit 102 generates production data Dp including the component data Dc, and outputs the production data Dp via the input/output unit 107 to, for example, the component mounting line L3.

Alternatively, the parameter estimation unit 105 estimates the operation parameter m based on the operation parameter model Pm for the component mounting lines L1 to L3, i.e., the operation parameter model Pm34 shown in Fig. 8, to generate the component data Dc. Then, the data generation unit 102 generates production data Dp including the component data Dc, and outputs the production data Dp via the input/output unit 107 to, for example, each of the component mounting lines L1 to L3.

In each of the component mounting lines L1 to L3, the component mounting devices M4 and M5 acquire production data Dp from the input/output unit 107 of the production data generating device 100, and mount at least one component P on the board B based on the production data Dp. This produces a mounted board. At this time, in each of the component mounting lines L1 to L3, the component data Dc included in the production data Dp is corrected so that, for example, the defective occurrence rate of the mounted board is reduced. As a result, actual production data including actual component data Dcu is generated. The actual component data Dcu is stored in the component library holding unit DB3 as new component data Dc. Specifically, the data acquisition unit 108 of the production data generating device 100 acquires actual production data from each of the component mounting lines L1 to L3, and stores the actual component data Dcu included in the actual production data in the component library holding unit DB3 as new component data Dc. Note that the component data Dc in each of the component mounting lines L1 to L3 is not necessarily corrected. If there is no modification, each of the component mounting lines L1 to L3 generates actual production data including the component data Dc acquired from the production data generating device 100 as actual component data Dcu.

When the data acquisition unit 108 acquires the actual production data of the component mounting line L1, the learning unit 104 selects the operation parameter model Pm corresponding to the actual component data Dcu included in the actual production data from the learning model holding unit DB2, as in the example shown in FIG. 10B. Specifically, the learning unit 104 selects the operation parameter model Pm for the component mounting line L1, that is, the operation parameter model Pm31 shown in FIG. 8. Then, the learning unit 104 performs learning for the selected operation parameter model Pm using the actual component data Dcu of the component mounting line L1. Note that this actual component data Dcu is the component data Dc acquired from the component mounting line L1 by the data acquisition unit 108 and stored in the component library holding unit DB3 as described above. As a result, the learning unit 104 updates the operation parameter model Pm for the selected component mounting line L1 stored in the learning model holding unit DB2 to the learned operation parameter model Pmu for the component mounting line L1.

The learning unit 104 also updates the operation parameter model Pm for each of the component mounting lines L2 and L3 in the same manner as for the component mounting line L1 described above. That is, the learning unit 104 updates the operation parameter model Pm for the component mounting line L2 stored in the learning model holding unit DB2 to the learned operation parameter model Pmu for the component mounting line L2. Furthermore, the learning unit 104 updates the operation parameter model Pm for the component mounting line L3 stored in the learning model holding unit DB2 to the learned operation parameter model Pmu for the component mounting line L3.

In addition, when the data acquisition unit 108 acquires the actual production data of any of the component mounting lines L1 to L3, the learning unit 104 may select the operation parameter model Pm for all the component mounting lines L1 to L3, that is, the operation parameter model Pm34 shown in FIG. 8. In this case, the learning unit 104 performs learning for the selected operation parameter model Pm using the actual component data Dcu of any of the component mounting lines L1 to L3. Note that this actual component data Dcu is the component data Dc acquired from any of the component mounting lines L1 to L3 by the data acquisition unit 108 as described above and stored in the component library holding unit DB3. As a result, the learning unit 104 updates the operation parameter model Pm for the selected component mounting lines L1 to L3 stored in the learning model holding unit DB2 to the learned operation parameter model Pmu for the component mounting lines L1 to L3.

In addition, the data generation unit 102 in this embodiment may import production data Dp from production equipment such as other factories other than the factory having the production system 1, or may export production data Dp to production equipment such as other factories.

Figure 13 is a flowchart showing the processing operation of the production data generating device 100 in this embodiment.

The input/output unit 107 of the production data generating device 100 receives the part information d (step S11). This part information d may be generated and received by an operator operating the input/output unit 107, or may be received by selection from multiple pieces of part information d. The input/output unit 107 may also receive the part information d by selecting part data Dc having a default operation parameter m from multiple pieces of part data Dc included in a part library and extracting the part information d from the part data Dc. Note that the part information d includes, for example, size data d2 and part attributes d31, as shown in FIG. 5.

Next, the model selection unit 103 selects at least one operation parameter model Pm from the multiple operation parameter models Pm stored in the learning model storage unit DB2 (step S12).

Next, the parameter estimation unit 105 estimates an operation parameter m based on at least one operation parameter model Pm selected in step S12 and the component information d received in step S11 (step S13). This operation parameter m is an operation condition of the component mounting device M4 or M5 for mounting the component P identified by the component information d on the board B. Then, the parameter estimation unit 105 generates component data Dc having the component information d and the operation parameter m (step S14).

Next, the data generation unit 102 generates production data Dp including the component data Dc generated in step S14 (step S15). The data generation unit 102 then outputs the production data Dp to each of the component mounting lines L1 to L3. That is, each of the component mounting lines L1 to L3 downloads the production data Dp from the data generation unit 102 and starts production of the mounted board using the production data Dp (step S16).

Then, the learning unit 104 re-learns the operation parameter model Pm using the component data Dc (i.e., the actual component data Dcu) included in the production data Dp used in each of the component mounting lines L1 to L3 as teacher data (step S17). The operation parameter model Pm to be re-learned is, for example, the operation parameter model Pm corresponding to the period during which the used production data Dp (i.e., the actual production data) was acquired.

As described above, in the production data generating device 100 in this embodiment, at least one operation parameter model Pm is selected from the multiple operation parameter models Pm. Then, based on the selected at least one operation parameter model Pm and the component information d of the component P to be mounted, the operation parameter m for mounting the component P to be mounted on the board B is estimated.

As a result, at least one operation parameter model Pm is selected from the multiple operation parameter models Pm and used to estimate the operation parameter m, which increases the possibility of estimating an appropriate operation parameter m for the mounting target component P. Therefore, an appropriate operation parameter m can be set. Furthermore, when component data Dc having such operation parameter m and component information d is included in production data Dp and the production data Dp is used to mount the component P on the board B by the component mounting device M4 or M5, a high-quality mounting board can be produced. In other words, the quality of the mounting board can be improved.

In addition, in the production data generating device 100 in this embodiment, actual production data including actual component data Dcu used by the component mounting device M4 or M5 is acquired. Then, among the multiple operation parameter models Pm, the operation parameter model Pm corresponding to the acquired actual production data is updated by learning using the actual component data Dcu as teacher data.

The operation parameters mu of the actual component data Dcu included in the actual production data are used to mount the mounted components P, and are modified during the process. In other words, the operation parameters mu are modified so that a mounting board of better quality is produced. Therefore, by using the actual component data Dcu having such operation parameters mu as teacher data for learning the operation parameter model Pm, the operation parameter model Pm can be further optimized. As a result, when the operation parameter model Pm is selected by the model selection unit 103, the estimation accuracy of the operation parameters m can be improved.

In addition, in the production data generating device 100 in this embodiment, each of the multiple operation parameter models Pm is associated with a different period, and learning is performed on the operation parameter model Pm corresponding to the period for which the actual production data was acquired.

For example, as shown in FIG. 7A, one of the operation parameter models Pm11 to Pm14, operation parameter model Pm14, is associated with the entire period (for example, the entire period from 1990 to the present). The remaining operation parameter models Pm11 to Pm13 are associated with different decades. The different decades are, for example, the 1990s, 2000s, 2010s, etc. As a result, from the operation parameter models Pm11 to Pm14, an operation parameter model Pm associated with the entire period or any of the decades is selected and used to estimate the operation parameter m. Therefore, an appropriate operation parameter m according to the period can be estimated for the mounting target component P.

In the production data generating device 100 in this embodiment, as shown in Fig. 8, each of the operation parameter models Pm31 to Pm34 is associated with a different piece of production equipment. Learning is then performed on the operation parameter model Pm corresponding to the production equipment including the component mounting device M4 or M5 using the actual production data.

As a result, from among the operation parameter models Pm31 to Pm34, the operation parameter model Pm associated with all or any of the component mounting lines is selected and used to estimate the operation parameter m. Therefore, it is possible to estimate an appropriate operation parameter m for the mounting target component P according to the production equipment.

In the production data generating device 100 according to the present embodiment, as shown in FIG. 9A, each of the operation parameter models Pm41 to Pm44 is associated with a different type of mounting board. Learning is then performed on the operation parameter model Pm that corresponds to the type of mounting board produced using the actual production data.

As a result, from among the operation parameter models Pm41 to Pm44, an operation parameter model Pm corresponding to, for example, a mass production type or a prototype type is selected and used to estimate the operation parameter m. Therefore, it is possible to estimate an appropriate operation parameter m for the mounting target component P according to the type of mounting board.

In this manner, in this embodiment, an operating parameter model Pm specialized for a period, a type of production equipment, or a type of mounting board can be utilized, and as a result, an operating parameter m appropriate for that period, a type of production equipment, or a type of mounting board can be estimated.

(Modification of the first embodiment)
In the above embodiment, as shown in Figures 7A to 9B, a plurality of operation parameter models Pm are managed in units of time, production equipment, or production type. However, the manner of management is not limited to these, and a plurality of operation parameter models Pm may be managed in other units. In the example shown in Figure 9B, a plurality of operation parameter models Pm are managed in units of combinations of production equipment and production types, but the combinations are not limited to this and may be any combination.

In the example shown in FIG. 7B in the above embodiment, the model selection unit 103 selects one operation parameter model Pm updated on the most recent date. Here, if the defect rate of the mounting board produced by the production data Dp including the operation parameter m estimated based on the operation parameter model Pm is high, the model selection unit 103 may reselect the previously selected operation parameter model Pm to the operation parameter model Pm updated before the most recent date. Also, even in the examples shown in FIG. 7A, FIG. 8, FIG. 9A, and FIG. 9B, the model selection unit 103 may reselect the operation parameter model Pm according to, for example, the defect rate. The reselection may be performed randomly or according to a predetermined procedure.

In addition, the input/output unit 107 of the production data generating device 100 in the above embodiment may export at least one operation parameter model Pm stored in the learning model storage unit DB2 to another facility other than the facility having the production system 1. The facility may be a factory or a floor. Furthermore, the input/output unit 107 may import at least one operation parameter model Pm from another facility and store it in the learning model storage unit DB2. This allows further optimization of the operation parameter model Pm. In addition, the input/output unit 107 may import and export the production data Dp stored in the production data storage unit DB1, and may import and export the part data Dc stored in the part library storage unit DB3.

Furthermore, the learning unit 104 of the production data generating device 100 in the above embodiment learns the operation parameter model Pm corresponding to the actual production data acquired by the data acquisition unit 108, among the multiple operation parameter models Pm held in the learning model holding unit DB2. However, the learning unit 104 may switch the operation parameter model Pm to be learned, in response to an operation by the operator accepted by the input/output unit 107. This allows learning of the operation parameter model Pm specified by the operator.

In addition, the motion parameter m estimated by the parameter estimation unit 105 included in the part data Dc of the part library may be associated with identification information of the motion parameter model Pm used to estimate the motion parameter m and the date and time when the estimation was performed. This allows the motion parameter m to be appropriately managed.

The parameter estimation unit 105 may estimate the operation parameter m using only a part of the information, without using all the information included in the part information d shown in FIG. 5. For example, the input/output unit 107 may accept a part of the information used to estimate the operation parameter m from the part information d in response to an operation by an operator. When such a part of the information is accepted, the parameter estimation unit 105 estimates the operation parameter m using only the accepted part of the information. Furthermore, the parameter estimation unit 105 may estimate only a part of the parameters, without estimating all the parameters included in the operation parameter m shown in FIG. 5. For example, the input/output unit 107 may accept the designation of a part of the parameters to be estimated among the operation parameters m shown in FIG. 5 in response to an operation by an operator. When such a part of the parameters is designated, the parameter estimation unit 105 estimates only the designated part of the operation parameters m. Furthermore, the parameter estimation unit 105 may perform a principal component analysis on all the information included in the part information d, and estimate the operation parameter m according to the analysis result.

(Embodiment 2)
In this embodiment, filtering is performed on the operation parameter mu contained in the actual production data output from each of the component mounting lines L1 to L3.

[Production System]
FIG. 14 is a diagram showing an example of the configuration of a production system according to the present embodiment.

The production system 2 in this embodiment includes three component mounting lines L1 to L3, a production management device 100a, a data management device 300, and three inspection devices 401 to 403. In other words, the production system 2 in this embodiment includes component mounting devices M4 and M5 that produce mounted boards by mounting components P on board B, and processing devices such as the data management device 300 or the inspection devices 401 to 403 that perform processing related to the production of the mounted boards.

In addition, among the components in this embodiment, the components that are the same as those in embodiment 1 are given the same symbols as in embodiment 1 and detailed explanations are omitted.

The three component mounting lines L1 to L3 are identical to the three component mounting lines L1 to L3 of production system 1 in embodiment 1.

The production management device 100a manages the production of mounting boards in the production system 2. Specifically, the production management device 100a has the same functions as the production data generating device 100 in the first embodiment, and further has a function of filtering the operation parameter mu included in the actual production data.

The data management device 300 is connected to the production management device 100a and each of the component mounting lines L1 to L3, and manages the production data Dp of each of the component mounting lines L1 to L3. This production data Dp may be actual production data used in each of the component mounting lines L1 to L3. In addition, the data management device 300 in this embodiment generates filtering information used for filtering by the production management device 100a, and outputs the filtering information to the production management device 100a.

Inspection devices 401-403 inspect the mounted boards produced by component mounting lines L1-3, respectively. That is, inspection device 401 inspects the mounted boards of component mounting line L1, inspection device 402 inspects the mounted boards of component mounting line L2, and inspection device 403 inspects the mounted boards of component mounting line L3. In addition, each of inspection devices 401-403 in this embodiment is connected to production management device 100a, generates the above-mentioned filtering information based on the inspection results of the mounted boards, and outputs the filtering information to production management device 100a.

In this embodiment, the data management device 300 and the inspection devices 401 to 403 are processing devices that perform processing related to the production of mounted boards.

[Functional configuration of production management device, component mounting line, and processing device]
15 is a block diagram showing the functional configurations of the production management device 100a, the component mounting lines L1 to L3, and the processing devices. In this embodiment, as shown in FIG. 15, a processing device 500 is configured from a data management device 300 and inspection devices 401 to 403.

The production management device 100a includes a control unit 101, a data generation unit 102, a model selection unit 103, a learning unit 104, a parameter estimation unit 105, a display unit 106, an input/output unit 107, a data acquisition unit 108, a production data storage unit DB1, a learning model storage unit DB2, and a part library storage unit DB3, similar to the production data generation device 100 of the first embodiment. The production management device 100a also includes a filtering unit 109 that filters the operation parameter mu included in the actual production data.

Specifically, the data acquisition unit 108 of the production management device 100a in this embodiment acquires production data Dp used in the production of mounted boards by component mounting device M4 or M5, i.e., actual production data, from each of the component mounting lines L1 to L3. This actual production data includes, for each of at least one type of component P, operating parameters mu that are operating conditions of component mounting device M4 or M5 for mounting the component P on the board B. Furthermore, the data acquisition unit 108 acquires filtering information from the processing device 500.

The filtering unit 109 selects one or more operation parameters mu by filtering at least one operation parameter mu contained in the acquired actual production data using filtering information obtained from the processing device 500.

The learning unit 104 generates or updates an operation parameter model Pm, which is a learning model stored in the learning model storage unit DB2, by learning using one or more selected operation parameters mu as teacher data. This operation parameter model Pm indicates the relationship between the component P and the operating conditions of the component mounting device M4 or M5 for mounting the component P on the board B. In the above learning, specifically, for each of the one or more selected operation parameters mu, actual component data Dcu including the operation parameter mu is used as teacher data.

In addition, the parameter estimation unit 105 in this embodiment, as in the first embodiment, estimates an operation parameter m, which is an operation condition of the component mounting device M4 or M5 for mounting the mounting target component P that has not yet been mounted on the board. The estimation of this operation parameter m is performed based on the operation parameter model Pm stored in the learning model storage unit DB2 and the component information d regarding the mounting target component P to be mounted on the board B.

The processing device 500 includes inspection devices 401 to 403 and a data management device 300.

The inspection device 401 includes an inspection control unit 411, an input/output unit 412, a display unit 413, an inspection mechanism 414, and an inspection data storage unit DB5.

The input/output unit 412 receives input data based on, for example, an operation by an operator of the production system 2, and outputs the input data to the inspection control unit 411. Such an input/output unit 412 may have, for example, a keyboard, a touch sensor, a touch pad, or a mouse. In addition, the input/output unit 412 outputs data to the production management device 100a and inputs data from the production management device 100a.

The inspection mechanism 414 consists of a mechanism including, for example, a camera for inspecting the mounting board, and stores inspection data indicating the inspection results in the inspection data storage unit DB5.

The inspection data storage unit DB5 is a recording medium for storing the inspection data. For example, such a recording medium is a hard disk, a RAM, a ROM, or a semiconductor memory. Note that such a recording medium may be volatile or non-volatile.

The display unit 413 displays the test data stored in the test data storage unit DB5. Examples of the display unit 413 include, but are not limited to, a liquid crystal display, a plasma display, or an organic EL display.

The inspection control unit 411 controls each of the input/output unit 412, the display unit 413, the inspection mechanism 414, and the inspection data storage unit DB5. For example, the inspection control unit 411 causes the inspection mechanism 414 to start inspection of the mounting board in response to an operation by an operator received by the input/output unit 412. In addition, the inspection control unit 411 in this embodiment generates filtering information and outputs it to the data acquisition unit 108 of the production management device 100a via the input/output unit 412.

Inspection devices 402 and 403 also have a configuration similar to that of inspection device 401 described above.

The data management device 300 has a data control unit 311, an input/output unit 312, a display unit 313, and a data storage unit DB6.

The input/output unit 312 receives input data based on operations by an operator of the production system 2, for example, and outputs the input data to the data control unit 311. Such an input/output unit 312 may have, for example, a keyboard, a touch sensor, a touchpad, or a mouse. The input/output unit 312 also outputs data to the production management device 100a and the component mounting lines L1 to L3, and inputs data from the production management device 100a and the component mounting lines L1 to L3.

The data storage unit DB6 is a recording medium for storing data. For example, the data is filtering information. Such a recording medium may be a hard disk, RAM, ROM, or semiconductor memory, and may be volatile or non-volatile.

The display unit 313 displays data stored in the data storage unit DB6. Examples of the display unit 313 include, but are not limited to, a liquid crystal display, a plasma display, or an organic EL display.

The data control unit 311 controls each of the input/output unit 312, the display unit 313, and the data storage unit DB6. In addition, the data control unit 311 in this embodiment may generate filtering information and output it to the data acquisition unit 108 of the production management device 100a via the input/output unit 312, similar to the above-mentioned inspection control unit 411.

For example, the data control unit 311 in this embodiment may generate board identification information for identifying board B as filtering information. In this case, multiple actual production data Dpu corresponding to multiple different types of mounting boards are filtered by the filtering information. Therefore, the data management device 300 can be said to be a device that manages multiple actual production data Dpu corresponding to multiple different types of mounting boards. Alternatively, the data control unit 311 may generate one or more pieces of component identification information for identifying the type of each component P as filtering information.

[Processing Overview]
FIG. 16 is a diagram showing an example of the overall process in this embodiment.

In this embodiment, as in embodiment 1, at least one production data Dp is generated, and actual production data Dpu is output from each of the component mounting lines L1 to L3 based on the at least one production data Dp.

The filtering unit 109 of the production management device 100a filters at least one operation parameter mu included in the actual production data Dpu. At this time, the filtering unit 109 obtains filtering information Df from the processing device 500, and selects one or more operation parameters mu by filtering based on the filtering information Df. Then, for each of the selected one or more operation parameters mu, the filtering unit 109 stores actual part data Dcu including the operation parameter mu as new part data Dc in the part library holding unit DB3.

[Filtering information]
FIG. 17A is a diagram showing an example of filtering information Df generated based on the inspection results of a mounting board.

For example, each inspection control unit 411 of the inspection devices 401 to 403 included in the processing device 500 generates filtering information Df shown in FIG. 17A and outputs it to the filtering unit 109.

That is, the inspection control unit 411 generates, as filtering information Df, information indicating the quality index of each of at least one type of component P mounted on the mounting board through inspection of the mounting board by the inspection mechanism 414. The quality index indicated by the filtering information Df is also called a mounting quality index, and indicates, for example, a larger value when the mounting state of the component P corresponding to the quality index is better.

More specifically, when the inspection mechanism 414 includes a camera, the inspection control unit 411 calculates a mounting quality index indicating the positional deviation of the mounted component P based on an image of the mounting board obtained by imaging with the camera. The positional deviation of the component P is the difference between the mounting position of the component P on the board B shown by the image and the mounting coordinates (or mounting position) of the component P shown by the production data Dp. For example, the inspection control unit 411 calculates a mounting quality index that is closer to 1 the smaller the positional deviation of the component P is, and conversely, calculates a mounting quality index that is closer to 0 the larger the positional deviation of the component P is. In other words, the mounting quality index may be normalized as a number in the range of 0 to 1. The mounting quality index may also be referred to as a score or evaluation value.

By calculating the mounting quality index in this manner, the inspection control unit 411 generates filtering information Df indicating the mounting quality index of each of multiple types of components P, as shown in Fig. 17A. This filtering information Df indicates the mounting quality index for each component name and component code of the component P. For example, the filtering information Df indicates "0.95" as the mounting quality index for the type of component P identified by the component name "Component A" and component code "C001".

When the filtering unit 109 of the production management device 100a acquires the filtering information Df shown in FIG. 17A, the filtering unit 109 performs filtering using the filtering information Df. That is, the filtering unit 109 selects one or more operation parameters mu corresponding to the type of part P whose mounting quality index is equal to or greater than the threshold value by the filtering. For example, the filtering unit 109 selects one or more operation parameters mu corresponding to the type of part P whose mounting quality index is equal to or greater than the threshold value "0.85". In the example shown in FIG. 17A, the filtering unit 109 selects the operation parameter mu corresponding to the part P with the part name "Part A" and the part code "C001", and the operation parameter mu corresponding to the part P with the part name "Part G" and the part code "C034". That is, the filtering unit 109 selects the actual part data Dcu of the part code "C001" and the actual part data Dcu of the part code "C034" from each of the multiple actual production data Dpu.

In this manner, in this embodiment, the operation parameter mu corresponding to the type of component P having a large mounting quality index is selected by filtering. As a result, one or more operation parameters mu corresponding to the type of component P with a good mounting state are selected by filtering and used for learning, and the operation parameters mu corresponding to the type of component P with a poor mounting state are not used for learning. Therefore, it is possible to generate an operation parameter model Pm for estimating an appropriate operation parameter m for achieving a good mounting state.

Figure 17B shows an example of filtering information Df generated based on the mounting performance of component mounting lines L1 to L3.

In addition, the data control unit 311 of the data management device 300 included in the processing device 500 may generate filtering information Df shown in FIG. 17B and output it to the filtering unit 109.

That is, the data control unit 311 acquires information indicating the operating status of component mounting devices M4 and M5 included in each of the component mounting lines L1 to L3 via the input/output unit 312. Based on the information indicating the operating status, the data control unit 311 generates information indicating the mounting performance index of each of at least one type of component as filtering information. The mounting performance index is an index related to errors that occurred in component mounting devices M4 and M5 for component P due to operations based on the performance production data Dpu by component mounting devices M4 and M5, and indicates a smaller numerical value, for example, as the number of errors decreases. The mounting performance index may also be referred to as a score or evaluation value.

More specifically, the mounting performance index is an index related to errors such as failure to pick up components P by component mounting devices M4 and M5, dropping of components P, or failure to supply components from feeder 7 to mounting head 10. For example, data control unit 311 indicates the mounting performance index as a percentage. In other words, the fewer the errors, the closer to 0% the data control unit 311 calculates the mounting performance index, and conversely, the more the errors, the closer to 100% the data control unit 311 calculates the mounting performance index.

By calculating the mounting performance index in this manner, the data control unit 311 generates filtering information Df indicating the mounting performance index for each of multiple types of components P, as shown in FIG. 17B. This filtering information Df indicates the mounting performance index for each component name and component code of the component P. For example, the filtering information Df indicates "0.5%" as the mounting quality index for the type of component P identified by the component name "Component A" and component code "C001".

When the filtering unit 109 of the production management device 100a acquires the filtering information Df shown in FIG. 17B, it performs filtering using the filtering information Df. That is, the filtering unit 109 selects one or more operation parameters mu corresponding to the type of part P whose mounting performance index is equal to or less than the threshold value by the filtering. For example, the filtering unit 109 selects one or more operation parameters mu corresponding to the type of part P whose mounting performance index is equal to or less than the threshold value "1%". In the example shown in FIG. 17B, the filtering unit 109 selects the operation parameter mu corresponding to the part P whose part name is "Part A" and whose part code is "C001", and the operation parameter mu corresponding to the part P whose part name is "Part B" and whose part code is "C102". That is, the filtering unit 109 selects the actual part data Dcu whose part code is "C001" and the actual part data Dcu whose part code is "C102" from the multiple actual production data Dpu.

In this manner, in this embodiment, the operation parameter mu corresponding to the type of component P having a small mounting performance index is selected by filtering. As a result, one or more operation parameters mu corresponding to the type of component P with few errors in the component mounting device M4 or M5 are selected by filtering and used for learning, and the operation parameters mu corresponding to the type of component P with many errors are not used for learning. Therefore, it is possible to generate an operation parameter model Pm for estimating an appropriate operation parameter m for reducing the occurrence of errors.

Figure 18A shows an example of filtering information Df generated by selecting part P.

The data control unit 311 of the data management device 300 included in the processing device 500 may generate filtering information Df shown in (b) in accordance with the result of an operation by an operator shown in (a) of Figure 18A and output it to the filtering unit 109.

For example, the data control unit 311 displays a part selection screen shown in (a) of FIG. 18A on the display unit 313. This part selection screen shows the part name, part code, and at least a part of the part information d (e.g., external dimensions and number of leads) of each part P handled in the production system 2. The operator operates the input/output unit 312 while looking at this part selection screen to input a learning flag for the desired part P. For example, if the mounting state of the part P on a mounting board produced by mounting the part P is good, the operator inputs a learning flag for the part P. Or, if the part P used in the past is an exceptional special part, the operator does not input a learning flag for the part P. Also, if the operation parameter mu of the part P included in the actual production data Dpu was set in another factory, the operator does not input a learning flag for the part P. In the example shown in (a) of FIG. 18A, the operator inputs a learning flag for each of the parts P with the part names "Part A", "Part B", "Part D" and "Part F". Then, the operator further selects the decision button shown on the part selection screen by operating the input/output unit 312. As a result, the data control unit 311 generates the filtering information Df shown in (b) of Fig. 18A in accordance with the learning flag input on the part selection screen.

For each learning flag input by the operator, the filtering information Df indicates the part name and part code of the part P corresponding to that learning flag as part identification information of that part P. For example, the filtering information Df indicates the part name "Part A" and part code "C001", the part name "Part B" and part code "C002", the part name "Part D" and part code "C003", and the part name "Part F" and part code "C005" as part identification information for each of four parts P. In this way, the data management device 300 outputs filtering information Df including one or more part identification information for identifying the type of each part P.

When the filtering unit 109 of the production management device 100a acquires the filtering information Df shown in (b) of FIG. 18A, the filtering unit 109 performs filtering using the filtering information Df. That is, in filtering, the filtering unit 109 selects, for each of one or more pieces of part identification information indicated by the filtering information Df, an operation parameter mu corresponding to the type of part P identified by the part identification information. In the example shown in (b) of FIG. 18A, the filtering unit 109 selects, from the multiple pieces of actual production data Dpu, actual part data Dcu of part code "C001", actual part data Dcu of part code "C002", actual part data Dcu of part code "C003", and actual part data Dcu of part code "C005".

In this manner, in this embodiment, the operation parameters mu corresponding to the type of part P specified by the operation by the operator are selected by filtering. As a result, one or more operation parameters mu corresponding to the type of part P identified by the part identification information specified by the operator are selected by filtering and used for learning, and operation parameters mu corresponding to other types of part P are not used for learning. Therefore, it is possible to generate an operation parameter model Pm for estimating an appropriate operation parameter mu for a specific part P.

Figure 18B shows an example of filtering information Df generated by selecting substrate B.

The data control unit 311 of the data management device 300 included in the processing device 500 may generate filtering information Df shown in (b) in accordance with the result of an operation by an operator shown in (a) of Figure 18B and output it to the filtering unit 109.

For example, the data control unit 311 displays the board selection screen shown in (a) of FIG. 18B on the display unit 313. This board selection screen shows the board name, board code, and auxiliary information of each board B handled by the production system 2. While looking at this board selection screen, the operator operates the input/output unit 312 to input a learning flag for the desired board B. For example, the operator inputs a learning flag for each of the boards B with the board names "A board", "B board", "D board", and "F board". The operator then further operates the input/output unit 312 to select the decision button shown on the board selection screen. As a result, the data control unit 311 generates filtering information Df shown in (b) of FIG. 18B according to the learning flag input to the board selection screen.

For each learning flag input by the operator, the filtering information Df indicates the board name and board code of the board B corresponding to that learning flag as the board identification information of that board B. For example, the filtering information Df indicates the board identification information of each of the four boards B as follows: board name "A board" and board code "B001", board name "B board" and board code "B002", board name "D board" and board code "B004", and board name "F board" and board code "B006". In this way, the data management device 300 outputs filtering information Df including one or more board identification information for identifying the type of each board B.

When the filtering unit 109 of the production management device 100a acquires the filtering information Df shown in (b) of FIG. 18B, the filtering unit 109 performs filtering using the filtering information Df. In other words, when a result production data group consisting of a plurality of result production data Dpu is acquired by the data acquisition unit 108, the filtering unit 109 selects one or more operation parameters mu from at least one operation parameter mu included in the result production data group in the filtering. Each of the one or more selected operation parameters mu corresponds to the type of component P to be mounted on the type of board B identified by the board identification information included in the filtering information Df.

In the example shown in (b) of FIG. 18B, the filtering unit 109 selects from the group of actual production data actual component data Dcu corresponding to the type of component P mounted on each of the boards B having board codes "B001", "B002", "B004", and "B006". Note that the production data Dp and actual production data Dpu may indicate the board code of the board B used in the production of the mounted board. In this case, the filtering unit 109 selects from the group of actual production data actual production data Dpu indicating each of the board codes "B001", "B002", "B004", and "B006", and extracts the actual component data Dcu from the selected actual production data Dpu.

In this manner, in this embodiment, the operation parameters mu corresponding to the type of component P to be mounted on the board B specified by the operator are selected by filtering. In other words, one or more operation parameters mu corresponding to the type of component P to be mounted on the board B specified by the operator are selected by filtering and used for learning, and operation parameters mu corresponding to the types of component P to be mounted on other boards B are not used for learning. Therefore, an operation parameter model Pm can be generated for estimating an appropriate operation parameter mu for a specific board B.

[Processing flow]
FIG. 19 is a flowchart showing the processing operation of production management device 100a in this embodiment.

The data acquisition unit 108 of the production management device 100a acquires actual production data Dpu from the component mounting lines L1 to L3 (step S21). Furthermore, the data acquisition unit 108 acquires filtering information Df from the processing device 500 (step S22).

Next, the filtering unit 109 uses the filtering information Df acquired in step S22 to filter each operation parameter mu included in the actual production data Dpu acquired in step S21 (step S23). As a result, the operation parameter mu to be used for learning is selected from the actual production data Dpu.

Next, the learning unit 104 generates or updates the operation parameter model Pm by learning using the operation parameters mu selected by the filtering in step S23 (step S24). In this learning, the selected operation parameters mu and the part information d included in the actual part data Dcu together with the operation parameters mu are used as teacher data. In addition, in updating the operation parameter model Pm, the operation parameter model Pm corresponding to the actual production data Dpu stored in the learning model storage unit DB2 is updated.

Then, when a part P whose operating parameter m is undetermined is selected by the input/output unit 107, the parameter estimation unit 105 estimates the operating parameter m of the selected part P using the operating parameter model Pm generated or updated in step S24 (step S25).

As described above, in the production management device 100a in this embodiment, the actual production data Dpu is acquired, and at least one operation parameter mu included in the acquired actual production data Dpu is filtered. In other words, one or more operation parameters mu are selected from the actual production data Dpu using the filtering information Df obtained from the processing device 500. Then, the operation parameter model Pm is generated or updated by learning using the selected one or more operation parameters mu as teacher data.

As a result, one or more operation parameters mu selected by filtering are used for learning, and the operation parameters mu that were not selected are not used for learning, so that the operation parameter model Pm can be optimized. As a result, by using this operation parameter model Pm, the appropriate operation parameter mu can be estimated and set in the production data Dp used thereafter in the component mounting device M4 or M5. Therefore, it is possible to produce a high-quality mounting board. In other words, it is possible to improve the quality of the mounting board.

In other words, in this embodiment, the operating parameter mu used as training data in learning can be controlled, allowing the user to perform the learning intended.

In addition, in the production management device 100a in this embodiment, the operation parameter m for mounting the target component P on the board B is estimated based on the generated or updated operation parameter model Pm and the component information d. This makes it possible to estimate and set an appropriate operation parameter m for the target component P.

(Modification of the second embodiment)
In the above embodiment, as shown in Figures 18A and 18B, filtering is performed based on the type of the specified component P or the type of the board B, but filtering may also be performed based on the product series of the mounting board. Filtering may also be performed based on the date on which the actual production data Dpu was acquired. For example, only the operation parameters mu included in the actual production data Dpu acquired on the most recent date may be selected by filtering. Filtering may also be performed based on the specified component mounting line. For example, when a component mounting line L1 is specified, only the operation parameters mu included in the actual production data Dpu acquired from that component mounting line L1 may be selected by filtering.

In the examples shown in Figures 17A and 17B, the filtering information Df indicates the implementation quality index or the implementation performance index, but these are merely examples and may indicate other indices. In addition, these indices may be expressed in PPM (parts per million).

In addition, in this embodiment, it is not necessary to select at least one motion parameter model Pm from the multiple motion parameter models Pm as in embodiment 1. In other words, only one motion parameter model Pm may be held in the learning model holding unit DB2.

(Other Modifications)
Although the production data generating device and the production management device according to one or more aspects have been described based on the embodiments and their modifications, the present disclosure is not limited to these embodiments and their modifications. As long as it does not deviate from the gist of the present disclosure, various modifications conceived by a person skilled in the art to the embodiments or their modifications, and forms constructed by combining the components of the embodiments and their modifications may also be included within the scope of the present disclosure.

In each of the above embodiments and their variations, each component may be configured with dedicated hardware, or may be realized by executing a software program suitable for each component. Each component may be realized by a program execution unit such as a CPU (Central Processing Unit) or a processor reading and executing a software program recorded on a recording medium such as a hard disk or semiconductor memory. Here, the software that realizes the devices of each of the above embodiments and their variations is a program that causes a computer to execute each step included in the flowchart shown in FIG. 13 or FIG. 19.

The following cases are also included in this disclosure:

(1) Specifically, each of the above devices is a computer system consisting of a microprocessor, ROM, RAM, a hard disk unit, a display unit, a keyboard, a mouse, etc. A computer program is stored in the RAM or hard disk unit. Each device achieves its function when the microprocessor operates in accordance with the computer program. Here, a computer program is composed of a combination of multiple instruction codes that indicate commands to a computer to achieve a specified function.

(2) Some or all of the components constituting each of the above devices may be composed of a single system LSI (Large Scale Integration). A system LSI is an ultra-multifunctional LSI manufactured by integrating multiple components on a single chip, and specifically, is a computer system comprising a microprocessor, ROM, RAM, etc. A computer program is stored in the RAM. The system LSI achieves its functions by the microprocessor operating in accordance with the computer program.

(3) Some or all of the components constituting each of the above devices may be composed of an IC card or a standalone module that can be attached to each device. The IC card or the module is a computer system composed of a microprocessor, ROM, RAM, etc. The IC card or the module may include the above-mentioned ultra-multifunction LSI. The IC card or the module achieves its functions by the microprocessor operating in accordance with a computer program. This IC card or this module may be tamper-resistant.

(4) The present disclosure may be the methods described above. It may also be a computer program for implementing these methods by a computer, or a digital signal comprising the computer program.

The present disclosure may also be the computer program or the digital signal recorded on a computer-readable recording medium, such as a flexible disk, a hard disk, a CD-ROM, an MO, a DVD, a DVD-ROM, a DVD-RAM, a BD (Blu-ray (registered trademark) Disc), a semiconductor memory, etc. The present disclosure may also be the digital signal recorded on such a recording medium.

The present disclosure may also involve transmitting the computer program or the digital signal via a telecommunications line, a wireless or wired communication line, a network such as the Internet, data broadcasting, etc.

The present disclosure may also provide a computer system having a microprocessor and a memory, the memory storing the above-mentioned computer program, and the microprocessor operating in accordance with the computer program.

The program or the digital signal may also be implemented by another independent computer system by recording it on the recording medium and transferring it, or by transferring the program or the digital signal via the network, etc.

(5) The above embodiments and the above variations may be combined with each other.

The present disclosure can be used in systems for producing mounted boards by mounting components onto a board.

1, 2 Production system 7 Feeder 10 Mounting head 10a Suction unit 10b Suction nozzle 11 Component recognition camera 12 Board recognition camera 14 Component tape 100 Production data generation device 100a Production management device 101 Control unit 102 Data generation unit 103 Model selection unit 104 Learning unit 105 Parameter estimation unit 106, 213, 313, 413 Display unit 107, 212, 312, 412 Input/output unit 108 Data acquisition unit 109 Filtering unit 200 Line management device 211 Work control unit 214 Work mechanism 300 Data management device 311 Data control unit 401-403 Inspection device 411 Inspection control unit 414 Inspection mechanism d Component information DB1, DB4 Production data storage unit DB2 Learning model storage unit DB3 Component library storage unit DB5 Inspection data storage unit DB6 Data storage unit Dc Component data Dcu Actual component data Dp Production data Dpu Actual production data L1 to L3 Component mounting line m, mu Operation parameters M4, M5 Component mounting device

Claims (15)

a model selection unit that selects at least one learning model from a plurality of learning models that correspond to different time periods each of which indicates a relationship between an operating condition of a component mounting device for mounting a component on a board and the component;
a parameter estimation unit that estimates operating parameters, which are operating conditions of a component mounting device for mounting the target components on a board, based on the at least one selected learning model and component information related to the target components to be mounted on a board;
a data generating unit that generates production data including part data having the part information and the operation parameters,
Production data generation device. The production data generation device further includes:
a data acquisition unit that acquires past production data used by the component mounting device, the past production data including component information related to mounted components and operational parameters used in mounting the mounted components;
a learning unit that updates the relationship indicated by a learning model corresponding to the acquired actual production data among the plurality of learning models by learning using the actual parts data as teacher data.
The production data generating device according to claim 1 . The part information indicates at least one of a size, a shape, an appearance, and a type of the part corresponding to the part information, and a supply form for supplying the part.
3. The production data generating device according to claim 1 or 2. The operation parameter is a parameter related to at least one of transportation, recognition, pickup, and mounting of a component by the component mounting device.
The production data generating device according to any one of claims 1 to 3. The learning unit performs learning on a learning model corresponding to a period during which the actual production data was acquired.
The production data generating device according to claim 2 . Each of the plurality of learning models corresponds to a different piece of production equipment,
The learning unit performs learning on a learning model corresponding to the production equipment including the component mounting device using the actual production data.
6. The production data generating device according to claim 2 or 5. Each of the plurality of learning models corresponds to a different type of mounting board;
the learning unit performs learning on a learning model corresponding to a type of mounting board produced using the actual production data;
7. The production data generating device according to claim 2, 5 or 6. Selecting at least one learning model from a plurality of learning models associated with different time periods each of which indicates a relationship between an operating condition of a component mounting device for mounting a component on a board and the component;
estimating operating parameters, which are operating conditions of a component mounting device for mounting the target components on a substrate, based on the at least one selected learning model and component information related to the target components to be mounted on a substrate;
generating production data including part data having the part information and the operating parameters;
Production data generation methods. The production data generation method further comprises:
Acquire actual production data used by the component mounting device, the actual production data including component data having component information related to the mounted components and operational parameters used to mount the mounted components;
updating the relationship indicated by a learning model corresponding to the acquired actual production data among the plurality of learning models by learning using the actual parts data as teacher data;
The production data generating method according to claim 8. The part information indicates at least one of a size, a shape, an appearance, and a type of the part corresponding to the part information, and a supply form for supplying the part.
The production data generating method according to claim 8 or 9. The operation parameter is a parameter related to at least one of transportation, recognition, pickup, and mounting of a component by the component mounting device.
The production data generating method according to any one of claims 8 to 10. In the learning, learning is performed on a learning model corresponding to a period during which the actual production data was acquired.
The production data generating method according to claim 9 . Each of the plurality of learning models corresponds to a different piece of production equipment,
In the learning, learning is performed on a learning model corresponding to a production facility including a component mounting device using the actual production data.
The production data generating method according to claim 9 or 12. Each of the plurality of learning models corresponds to a different type of mounting board;
In the learning, learning is performed on a learning model corresponding to a type of mounting board produced using the actual production data.
14. The production data generating method according to claim 9, 12 or 13. A program for causing a computer to execute the production data generation method according to any one of claims 8 to 14.

Images